You are on page 1of 16

Sustainable Energy, Grids and Networks 32 (2022) 100915

Contents lists available at ScienceDirect

Sustainable Energy, Grids and Networks


journal homepage: www.elsevier.com/locate/segan

An Energy Management System for microgrids including costs, exergy,


and stress indexes

Martina Caliano a ,1 , Federico Delfino b ,1 , Marialaura Di Somma a ,1 , Giulio Ferro c , ,1 ,
Giorgio Graditi a ,1 , Luca Parodi d ,1 , Michela Robba c ,1 , Mansueto Rossi b ,1
a
Italian National Agency for New Technologies, Energy and Sustainable Economic Development (ENEA),
Department of Energy Technologies and Renewable Sources–Portici Research Centre, P. E. Fermi, 1, IT-80055 Portici (Naples), Italy
b
Department of Naval, Electrical, Electronic and Telecommunication Engineering, University of Genova, Genova 16145, Italy
c
Department of Informatics, Bioengineering, Robotics and Systems Engineering, University of Genoa, Genoa, Italy
d
Università degli studi di Genova, Department of Informatics, Bioengineering, Robotics and Systems Engineering, Genova, Italy

article info a b s t r a c t

Article history: In this paper, a multi-objective Energy Management System (EMS) for polygeneration microgrids
Received 11 February 2022 is presented. The proposed tool has been developed within the LIVING GRID project, funded by
Received in revised form 19 August 2022 the Italian Ministry of Research (actions related to the Italian Technology Cluster on Energy), and
Accepted 20 August 2022
it is characterized by a detailed representation of generation units and flexible loads, as well as
Available online 5 September 2022
electric/thermal networks and storage systems that can be present in microgrids and sustainable
Keywords: districts. A new optimization model has been developed in which the objective function is related
Microgrids to the minimization of costs, the maximization of the overall exergy efficiency of the system, and the
Energy management system minimization of stress indexes. To solve efficiently the optimization problem, a novel algorithm based
Optimization on an accelerated version of the Bregman Alternating Direction Method of Multipliers (with Heavy Ball
Exergy dual acceleration term) has been developed and discussed. The decision model has been applied to a
real case study represented by the Savona Campus of the University of Genova in Italy. A sensitivity
analysis has been performed for evaluating the effect of each term of the objective function on the
solution.
© 2022 Elsevier Ltd. All rights reserved.

1. Introduction Union climate action & Green Deal [2,3] will require decommis-
sioning a large amount of traditional fossil fuel-powered gen-
Three major trends will characterize the evolution of the en- eration, to be replaced by renewable, non-dispatchable ones, as
ergy system in the next decades: increasing rate of electrification, this is recognized as the more effective strategy for decarboniza-
decarbonization, and increasing penetration of renewables. As tion [1]. Some of the proposed scenarios for the evolution of
an example, in Europe the share of electricity in final energy the energy system set the share of renewables in 2050 as high
consumptions, according to long-term predictions, is expected to as 85% (2 ◦ C scenario, Remap case [4]), of which more than
grow to about 30% in 2030 and to reach a value approximately 20% photovoltaic. This will have an impact on the management
between 40% and 50% (depending on the considered scenario) in of the entire power network, due to issues such as generation
2050 [1]; the transition affecting sectors such as transportation, unpredictability and lack of rotating inertia, but it will particularly
as well as residential heating and conditioning, is expected to affect the distribution network, as a considerable amount of the
accelerate, with the more and more widespread use of electric photovoltaic generation will likely consist of many medium and
vehicles and electric heat pumps, replacing internal combustion small-scale installations connected to the distribution grid. This
engine-based vehicles and traditional, natural gas-fired heating will dramatically change the role played by Distribution System
systems. Furthermore, the challenging goals set by the European Operators (DSOs) in the context of the energy infrastructure:
from the management of a mostly passive network to that of an
∗ Corresponding author. active network with myriads of potentially active users, with an
E-mail addresses: martina.caliano@enea.it (M. Caliano), increasing impact on the stability of the whole network. This will
federico.delfino@unige.it (F. Delfino), marialaura.disomma@enea.it
(M. Di Somma), giulio.ferro@edu.unige.it (G. Ferro), giorgio.graditi@enea.it
require a closer and more complex interaction between Distri-
(G. Graditi), luca.parodi@edu.unige.it (L. Parodi), michela.robba@unige.it bution System Operators and the Transmission System Operators
(M. Robba), mansuseto.rossi@unige.it (M. Rossi). (TSOs), these latter responsible for the balance between genera-
1 Authors provide the same contributions to the paper. tion and load that ensures the stable operation of the power grid.

https://doi.org/10.1016/j.segan.2022.100915
2352-4677/© 2022 Elsevier Ltd. All rights reserved.
M. Caliano, F. Delfino, M. Di Somma et al. Sustainable Energy, Grids and Networks 32 (2022) 100915

A key concept that will be crucial in dealing with such a complex to provide a more complete representation of the impact of the
scenario is flexibility [4]: not only the potential of the network to microgrid operation and to analyse which are the objectives that
adapt to rapidly changing generation and load profiles, as well as are more significant for a specific case study, making it possible
to reconfigure in response to a contingency, but also the ability to analyse which are the effects of giving a higher weight to a
to exploit the flexibility offered by loads and decentralized gen- specific objective for the others. In particular, the inclusion of
eration to better guarantee the network stability. Consequently, exergy modelling in the objective function allows improving the
already well-established solutions such as microgrids and virtual exploitation of the available resources, and vulnerability indexes
power plants, are expected to gain more and more attention, as are crucial to mitigate the risk of failures in the grid operation.
they represent some of the most promising ways to help System Finally, a suitable model for the thermal storage in buildings
Operators interact with dispersed generation and active loads [5], equipped with heat pumps is essential to correctly account for
taking full advantage of their flexibility. In turn, this calls for the building flexibility [13,14], and for this reason it is included
in the overall optimization problem.
the development of suitable Energy Management Systems (EMSs)
A novel algorithm based on an accelerated version of the Breg-
able to manage these complex energy infrastructures to maximize
man Alternating Direction Method of Multipliers (BADMM) [15]
or minimize a particular objective function [6], which can take
has been developed to support the embedded implementation
into account requests of ancillary services from DSOs and TSO.
of the proposed EMS. The main advantages of this algorithm
This objective function is generally related to the minimization of are the non-use of mathematical programming tools during the
operating costs or environmental emissions, but, when needed, iterations and the accelerated feature, that leads to a fast and easy
the EMS can take into account demand response requirements, to implement solution. The EMS has been developed within the
power quality indicators, or constraints on power flows [7], im- LIVING GRID project (2017–2020), funded by the Italian Ministry
posed to meet the service requests from the System Operators. In of Research (actions related to the Italian Technology Cluster on
this way, the microgrid becomes an asset the System Operators Energy), which aimed at investigating solutions for the man-
can exploit to help manage the network during normal operating agement of smart grids, by developing and testing in field new
conditions or when a contingency occurs that threaten the safe tools and technologies for demand response and the manage-
operation of the network, ultimately contributing to face the new ment of local systems such as microgrids or sustainable districts.
challenges outlined before. For instance, this will help the DSO Specifically, during the LIVING GRID project, the possibility for
avoiding congestions (due, by way of example, to an unusually dispersed generation and microgrids to offer ancillary services to
very high or very low renewable generation) by imposing con- System Operators was investigated and tested by developing an
straints on the active power the microgrid injects to the network. architecture whose main components are, first, an ICT platform
The microgrid can be also asked to perform voltage support via by which DSO and TSO can request services (by imposing con-
the generation of reactive power. By imposing constraints on the straints on power fluxes on connections, or on absorbed/injected
power exchanged at the point of common coupling of the micro- power at buses), and, second, an EMS which can receive from
grid, the System Operator can ask the microgrid to compensate the platform those requests, translated in terms of constraints
the variability of its own renewable generation by exploiting elec- on the active and reactive power at the point of common cou-
tric storage devices, or flexible loads, thus mitigating renewables pling of the microgrid. The present paper focuses on this second
unpredictability. component of the architecture, i,e, the new multi-objective EMS
The possibility to manage and combine the flexibility offered for polygeneration microgrids. The proposed optimization model
has been applied to a real test case represented by the Smart
by multiple microgrids can even enable the DSO deliver ancillary
Polygeneration Microgrid (SPM) at the Savona Campus of the
services to the TSO.
University of Genova (Liguria Region, north-west Italy) [12]. The
Considering that the optimization of microgrid operation may
considered system is a polygeneration microgrid characterized
involve multiple and conflicting objectives [8], a multi-objective
by photovoltaic plants (PV), cogeneration microturbines coupled
approach can be better suited to identify balancing solutions with absorption chillers, electrical and thermal storage systems,
to promote stakeholders’ participation in the decision-making boilers, and heat pumps. A Smart Energy Building (SEB), equipped
process and facilitate collective decisions [9]. Exergy objectives with PV, thermal solar collectors, and a geothermal heat pump, is
can be also included in the multi-objective optimization pro- connected to the microgrid. Thermal (heating and cooling) and
cesses of microgrids [10,11] to promote the rational use of energy electrical loads are present too.
resources and increase the sustainability of microgrids, by taking In summary, the main contributions of the present paper are:
into account the different energy quality levels of energy supply
and demand. • The development of a new EMS, tested in the field as a joint
The main aim of the present paper is to present a new EMS for effort of companies and research centers (within the LIVING
microgrids and sustainable districts, in which the objective func- GRID project), for polygeneration microgrids and sustainable
tion is related to the minimization of costs, the maximization of districts;
the overall exergy efficiency of the system, and the minimization • The modelling and integration of new objective functions
of indexes to mitigate grid vulnerabilities. The decision variables related to exergy aspects, vulnerability indexes, and heat
are the scheduling of the different production plants and storage production, storage and distribution;
systems. Constraints consist in the modelling of plants, storage • The development of an accelerated optimization algorithm
systems, and distribution networks; demand response constraints based on BADMM for embedded applications.
are here included (i.e., bounds on active and reactive power can • The application to a real case study (the Savona Campus
be imposed by a DSO.) The EMS is customizable for general case Smart Polygeneration Microgrid).
studies and can work both in the day-ahead and in real time The paper is organized as follows: in Section 2, a state-of-the-
under a Model Predictive Control (MPC) scheme. This EMS is an art discussion is reported. The system model and the optimiza-
evolution of the one proposed in [12], presenting several new tion problem are formalized in Sections 3 and 4, respectively. In
features such as the inclusion of new terms in the objective Section 5, the design and discussion of the optimization algorithm
function (exergy and vulnerability indexes) and new constraints are given. The results of the application of the proposed energy
related to the modelling of heat pumps, temperature modelling management system to the pilot are presented in Section 6.
and thermal storage systems. The different objectives are normal- Finally, in Section 7, conclusions are drawn and some directions
ized through the calculation of utopia and nadir points; this helps for future developments are outlined.
2
M. Caliano, F. Delfino, M. Di Somma et al. Sustainable Energy, Grids and Networks 32 (2022) 100915

2. State of the art and a detailed representation of the electrical network. In [24],
the upper problem determines the optimal configuration of the
A microgrid is an integration of distributed energy resources microgrid, while the lower problem optimizes the output of the
and loads that can be controlled either in grid-connected or distributed energy resources through the implementation of an
islanded mode and that can provide greater energy stability and EMS. Our paper corresponds to a microgrid already designed that
resilience within a project site or community [16]. Generally, mi- needs an optimal operational management and thus, concern-
crogrids are equipped with Energy Management Systems (EMSs), ing [25], corresponds to the lower level. However, both the MPC
i.e., software tools (integrated with the hardware in field), able architecture, the system model and the multiple objectives are
to manage complex energy infrastructures, which include opti- different.
mization models to maximize or minimize a particular objective In fact, in the present paper attention is focused on multi-
function under the system model constraints [6,12]. ple objectives (and, differently from literature, costs, exergy and
In literature, there are many papers on EMSs with differ- stress indexes are combined) and detailed thermal and electrical
ent capabilities and applications. Some examples can be found models. A detail on electrical models allows using the EMS for
in [17–22]. The approach given in [17] proposes a novel EMS that general case studies and portions of the distribution grid [12],
performs distributed energy resources (DER) schedule combined while thermal models are very useful to analyse the thermal flex-
with a single time interval optimal power flow. The DER schedule ibility, to reduce energy costs and decrease CO2 emissions [25]. As
enables the microgrid to fully utilize the latest updated renew- regards the new terms in the objective function, their inclusion
able and load forecasts, and the optimal power flow (OPF)-based is considered relevant in the current literature. Stress indexes
integrates realistic microgrid operating constraints in the real- are very important to guarantee a safe operation of the grid,
time control. In addition, a switching control function is included preventing damages and faults. As regards exergy, Annex 49
to re-dispatch DERs in the microgrid for the planned microgrid Low Exergy Systems for High-Performance Buildings and Com-
switching from grid-connected mode to islanded mode. In con- munities [13] reports that the application of exergy principles
trast to our approach, this paper does not consider polygenerative in the context of energy supply systems is useful to achieve a
units (with the thermal balance), demand response programs, rational use of energy resources by considering the energy quality
grid stress indexes, and multiperiod optimization. In [18], the levels of energy resources and users’ energy demand. The thermal
resilience of the consensus-based distributed economic dispatch demand for heating and cooling in buildings accounts for more
in a cyber–physical DC microgrid EMS is proposed. Differently than one-third of the final global energy consumption and is
from our work, there is no consideration of the network’s stress typically satisfied by fossil fuel energy with consequent green-
and demand response. The work in [19] presents a stochas- house gas (GHG) emissions. Evaluation of energy use in buildings
tic mixed-integer nonlinear programming model for the optimal is generally based on quantitative assessments with the First
energy management system of unbalanced three-phase of alter- Law of Thermodynamics, without considering the degradation of
nating current microgrids. The proposed mixed-integer nonlinear energy quality occurring when high-quality energy resources as
programming model is transformed into a mixed-integer linear fossil fuels or electricity are used to satisfy low-quality thermal
programming model through a set of linearization that can be demands as heating, domestic hot water, or cooling. By applying
solved via off-the-shelf convex programming solvers. However, the concept of exergy with the Second Law of Thermodynamics,
the MPC (Model Predictive Control) framework is not applied it is possible to consider the quality of energy, that is defined
considering electrical and thermal coupling. The authors of [20] as the maximum amount of work that can be obtained from
propose an EMS for microgrids of buildings’ prosumers. It is based an energy flow as it reaches the equilibrium with a reference
on a hierarchical multi-agent system, aiming to minimize the environment [13,26–29] and can be considered as the potential
operation cost of a microgrid comprising a group of buildings, of an energy carrier. Through matching the energy quality levels
renewable energy sources, energy storage systems, and plug-in of supply and demands with the exergy analysis, low-quality
electric vehicles. The optimization problem is solved by particle thermal demands are met by low exergy sources as solar thermal
swarm optimization. However, the optimality of the solution is or waste heat from combined heat and power (CHP) thereby re-
not guaranteed like in our paper and polygenerative characteris- ducing the waste of fossil fuels with consequent reduction of GHG
tics are not addressed. In [21], a double-layer coordinated control emissions [28]. In the literature, several works are investigating
approach for microgrid energy management is proposed, which the concept of exergy in energy management systems. In [30],
consists of two layers: the scheduling layer and the dispatch though attention is focused on general energy hubs, exergy was
layer. The schedule layer obtains an economic operation scheme considered in the framework of multi-carrier energy systems to
based on forecasting data, while the dispatch layer provides the maximize the overall exergy efficiency defined as the ratio of
power of controllable units based on real-time data. Although this the total exergy required to meet the given energy demands
approach is interesting, it does not consider typical features of to the total primary exergy input to the system. In [10] exergy
microgrids like polygeneration, demand response, and maximiza- was investigated in the operation optimization of a multi-carrier
tion of the overall exergy efficiency like in the proposed work. energy system with multiple energy devices converting a set of
Finally, the study in [22] proposes an efficient energy manage- primary energy carriers with different energy quality levels to
ment method to systematically manage the energy consumption meet given time-varying user demands at different energy quality
in the residential area to alleviate the peak to average ratio and levels. In [11] exergy and costs objectives are considered, together
mitigate electricity cost along with user comfort maximization. with a detailed modelling of the thermal system, but, differently
There is no polygeneration modelling and the exergy feature is from our work, vulnerability indexes and a detailed representa-
not considered. Bi-level optimization is present in some papers tion of the electrical grid are not present. The influence of energy
applied to microgrids. In [23], the upper-level problem repre- quality management on CO2 emissions in operation optimization
sents a microgrid planner whose goal is to minimize planning of a multi-carrier energy system was investigated in [31], demon-
and operational costs, while the lower-level problem represents strating that more efficient use of energy resources and reduction
a Distribution System Operator whose primary duty is to en- in CO2 emissions can be achieved through exergy analysis. In [32],
sure reliable power supply. The main differences from our work exergy analysis is applied to the natural gas-based multi-energy
are that we consider operational management for tertiary level microgrid. The overall optimization problem considers originative
control under a MPC approach considering multiple objectives units in a multiperiod optimization framework; the problem has
3
M. Caliano, F. Delfino, M. Di Somma et al. Sustainable Energy, Grids and Networks 32 (2022) 100915

been solved thanks to the particle swarm method; however, this The energy balance constraints for cool in Eq. (2) includes the
paper does not model the power distribution grid as well as cooling demand, Dcool,t [kW], the cooling energy generated by the
demand response and stress indexes as we do in the present heat pumps, Pc,HP ,i,t and by the absorption chillers, PCHI ,t . Heat
paper. pumps produce thermal power from electrical power PHP ,i,t :

3. The system model PHP ,i,t COPi = Pth,HP ,i,t i ∈ N , t = 0, . . . , T − 1 (4)

The considered system is characterized by co-generative gas


microturbines, production plants from renewable resources, ab-
MAX
PHP ,i,t ≤ PHP ,i i ∈ N , t = 0, . . . , T − 1 (5)
sorption chiller, storage systems, thermal storage, deferrable and MAX
where PHP [kW] and COPi are the maximum rate and the co-
not deferrable loads, and a geothermal heat pump. In the fol-
efficient of performance of the heat pumps, respectively. Similar
lowing the system model is described highlighting parameters,
decision variables, and sets. constraints are fixed for the absorption chillers. Finally, aMIN and
aMAX are parameters that set how much the power request can
3.1. Sets definition be violated. It is important to note that Eq. (4) is valid also during
summer substituting Pth,HP ,i,t with Pc,HP ,i,t .
Before describing the mathematical model the following sets The SEB is heated and cooled with a geothermal heat pump
have to be introduced: integrated with a thermal energy storage system, which provides
thermal power for the building through fan coils. Electrical power
• N = {1, .{. . , n} : the}set of all nodes of the power network taken from the microgrid Pel,HP ,SEB,t is used to produce thermal
• HF ,i = 1, . . . , hf ,i is the set of controllable generation power Pth,HP ,SEB,t :
plants at{ node i; }
Pth,HP ,SEB,t = COPHP ,SEB Pel,HP ,SEB,t t = 0, . . . , T − 1 (6)
• HR,i = 1, . . . , hr ,i is the set of renewable generators at
node{i; where, COPHP ,SEB is the heat pump coefficient of performance.
• Si = 1, . . . , ks,i set of storage systems at node i;
}
The thermal behaviour of the energy storage system is mod-
• Ai = {1, . . . , ni }set of nodes directly linked to node i elled according to the following relation:
Pth,HP ,SEB,t − Pth,S ,out ,t
3.2. The thermal system TS ,t +1 = ∆ + T S ,t t = 0, . . . , T − 1 (7)
CS
The thermal demands to be satisfied are referred to two main where, TS [K] is the temperature of the thermal storage, CS
sub-systems: (a) buildings in the Campus fed by a district heating [kWh/K] is the thermal capacity of thermal storage, and Pth,S ,out ,t
connected to a boiler, a renewable plant, some heat pumps and [kW] is the power exiting the thermal energy storage system that
some microturbines, as well as a chiller (which represents a load is related to the thermal power of the ventilation system Pth,FC ,t
for the heating power balance); (b) the smart energy building [kW] by
(SEB) with a dedicated controllable heat pump.
Pth,FC ,t
For the Campus (except the SEB), the heating and cooling Pth,S ,out ,t = t = 0, . . . , T − 1 (8)
energy balances are given by ηFC
∑ where ηFC is the efficiency of the ventilation system.
aMIN Dheat ,t ≤ Pth,B,t + Pth,RES ,t − Pth,CHI ,t + Pth,HP ,i,t
The SEB has been modelled as a single equivalent room. Thus,
i∈N
∑ ∑ the temperature trend is given by:
+ Pth,h,i,t ≤ a MAX
Dheat ,t t = 0, . . . , T − 1 ( )
Text ,t −Tbuild,t
i∈N h∈HF ,i
Rbuild
+ αFC Pth,FC ,t
(1) Tbuild,t +1 = ∆+Tbuild,t t = 0, . . . , T −1
Cbuild
∑ (9)
aMIN Dcool,t ≤ Pc,HP ,i,t + PCHI ,t ≤ aMAX Dcool,t t = 0, . . . , T − 1
where, Rbuild [K/kW] and Cbuild [kWh/K] are the thermal resistance
i∈N
and the capacity of the building, respectively, Text ,t [K] is the
(2) external temperature, and αFC is the efficiency of the fan coil.
Text ,t −Tbuild,t
In particular, the energy balance for heat in Eq. (1) includes: In the summer scenario, Rbuild
is positive, whereas Pth,FC ,t is
the heat demand, Dheat ,t [kW]; the thermal power generated by negative. Moreover, Pth,HP ,SEB,t is negative.
the boiler, Pth,B,t [kW]; the thermal power generated by renew- Finally, the following bound constraints should be considered:
able plants, Pth,RES ,t [kW]; the thermal power entering the chillers,
0 ≤ Pth,FC ,t ≤ ηFC mair cp,air TS ,t − Tbuild,t t = 0, . . . , T − 1 (10)
( )
Pth,CHI ,t [kW]; the thermal power generated by the heat pumps,
Pth,HP ,i,t [kW]; the thermal power generated by the cogeneration
plants (microturbines), Pth,h,i,t [kW]. Note that the thermal power
from cogeneration plants and from the heat pumps depend on TSMIN ≤ TS ,t ≤ TSMAX t = 0, . . . , T − 1 (11)
the node of the electric network because they are linked with
their electric power production and consumption, respectively;
this relation will be presented in the following equations (see (4)
MIN
Tbuild MAX
≤ Tbuild,t ≤ Tbuild t = 0, . . . , T − 1 (12)
and (19)).
where: mair [kg/h] is the air mass flow rate passing through the
It is important to note that Pth,B,t depends on the primary
fan coils, cp,air [kWh/(kg K)] is the specific heat of air at constant
energy per unit of time entering the boiler, PPE ,B,t [kW], and the
boiler’s efficiency, ηB,i , according to the following relation: pressure, Tbuild,t [K] is the temperature of the building at time
t, respectively; TSMIN and TSMAX are lower and upper temperature
PPE ,B,t ηB = Pth,B,t t = 0, . . . , T − 1 (3) bounds for TS ,t .
4
M. Caliano, F. Delfino, M. Di Somma et al. Sustainable Energy, Grids and Networks 32 (2022) 100915

3.3. The electrical system parameters. That is:

Pel,h,i,t = µh PPE ,h,i,t i ∈ N , h ∈ HF ,i , t = 0, . . . , T − 1 (18)


The electrical system is characterized by electrical storage
systems, loads, deferrable demand, production plants, and a dis-
tribution system in which power losses are considered. In the fol-
Pel,h,i,t = µh Pth,h,i,t i ∈ N , h ∈ HF ,i , t = 0, . . . , T − 1 (19)
lowing, the mathematical formulation is expressed with reference
to a generic number of the aforementioned devices, connected to To quantify costs, it is necessary to derive the quantity of con-
different nodes of a generic network. To better clarify the layout sumed gas Qgas,h,i,t [m3 /s] as a function of primary energy per unit
of the system, in Section 6 a simplified version of the one-line of time.
diagram of the Savona Campus, which will be presented as an
PPE ,h,i,t = Qgas,h,i,t LHV i ∈ N , h ∈ HF ,i , t = 0, . . . , T − 1 (20)
application example, will be provided. The active and reactive
power balances are given by [33]. 3
where LHV [kJ/m ] is the Lower Heating Value of the gas.
∑ ∑ ∑ The storage systems are described by (21)–(22) where SOCk,i,t
Pel,h,i,t + PRES ,l,i,t − PS ,k,i,t − PD,i,t + Pgrid,i,t − PHP ,i,t [kWh] is the state of charge of the storage, bounded by SOCkMIN ,i,t
h∈HF ,i l∈HR,i k∈Si
and SOCkMAX ,i,t . The parameters relevant to the charge and the dis-
charge are ηc ,k and ηd,k , i.e., the charging and discharging efficien-

− Pveh,i,t − Pel,HP ,SEB,i,t + LOLi,t = Pi,j,t
j∈Ai
cies, ak,i,t is a loss coefficient, CAPS ,k [kWh] is the capacity of the
storage system, and ∆[h] is the length of the time interval.
i ∈ N , t = 0, . . . , T − 1 (13)
SOCk,i,t +1 = ak,i,t SOCk,i,t
ηk,i,t PS ,k,i,t ∆
∑ ∑ ∑
QRES ,l,i,t − QS ,k,i,t − QD,i,t + Qgrid,i,t = Qi,j,t + i ∈ N , k ∈ Si , t = 0, . . . , T − 1
l∈HR,i k∈Si j∈Ai CAPS
(21)
i ∈ N , t = 0, . . . , T − 1 (14)

In particular, PRES ,l,i,t [kW] represents the active powers from ηc ,k,i if PS ,k,i,t > 0
{
renewables, PS ,k,i,t [kW] is the active power absorbed by storage ηk,i,t = i ∈ N , k ∈ Si , t = 0, . . . , T − 1
systems, PD,i,t [kW] is the load, Pgrid,i,t [kW] is the power ex- 1/ηd,k,i other w ise
changed with the grid (this term will be non-null only for the (22)
node to which the external grid is connected), PHP ,i,t [kW] and
Pv eh,i,t [kW] are power consumptions of heat pumps and electric Constraint (23) is relevant to the deferrable load energy in the
vehicles respectively, Pel,HP ,SEB,i,t [kW] is the power of the heat optimization horizon:
pumps in the SEB (this term will be non-null only for the node to T
∑ ∑
which the SEB is connected). Moreover, QRES ,l,i,t , QS ,k,i,t , QD,i,t and PDiff ,i,t ∆ ≥ DDiff ,i i ∈ N , t = 0, . . . , T − 1 (23)
Qgrid,i,t [kVAR] are the corresponding reactive powers. Finally, Pi,j,t t =1 i∈N
[kW] and Qi,j,t [kVAR] are the power flows on link (i, j). Since we
Finally, all variables have upper and lower bounds, but, for
are using the active sign convention, all loads are with minus sign.
the sake of brevity, these constraints are not reported. It is worth
It is important to note that PD,i,t is composed of two terms,
mentioning that either maximum and minimum bounds or equal-
PDiff ,i,t and PNDiff ,i,t [kW], which represent deferrable and non-
ity constraints can be set on both the active and reactive power
deferrable portions of the load, i.e., loads that can be modulated exchanged with the grid in given periods to account for requests
for demand response purposes, as: of ancillary services from the System Operators. That is,
PD,i,t = PDiff ,i,t + PNDiff ,i,t i ∈ N , t = 0, . . . , T − 1 (15) P grid,i,t ≤ Pgrid,i,t ≤ P grid,i,t i ∈ N , t = 0, . . . , T − 1 (24)
Power flow equations in per unit values are represented by the
general nonlinear formalization that includes power losses [12,
33], that is: Q grid,i,t ≤ Qgrid,i,t ≤ Q grid,i,t i ∈ N , t = 0, . . . , T − 1 (25)

pi,j,t = Gi,j (vi,t )2 − vi,t vj,t Gi,j cos(δi,t − δj,t ) + Bi,j sin(δi,t − δj,t ) For instance, a maximum value for the power absorbed by the
( )
microgrid from the network in a given period can be specified, as
i, j ∈ N , i ̸ = j, t = 0, . . . , T − 1 (16) a response to a request from the DSO, issued in order to solve a
congestion. A given injection of reactive power can be prescribed
qi,j,t = −Bi,j (vi,t )2 − vi,t vj,t −Bi,j cos(δi,t − δj,t ) + Gi,j sin(δi,t − δj,t )
( )
as a voltage support action, or a minimum injection of active
power to the grid can be required to support the network in an
i, j ∈ N , i ̸ = j, t = 0, . . . , T − 1 (17)
emergency condition. These latter scenarios where tested during
where, Gi,j and Bi,j are the conductance and susceptance of the the LIVING GRID project, as one of the goals of the project was
lines, vi,t and δi,t are voltage and phase at node i. Note that the to test possibility for dispersed generation to offer services to the
power flow equation are written using the p.u. system and the network in emergency situations.
power flows pi,j,t and qi,j,t are the p.u. equivalent of Pi,j,t and Qi,j,t
presented in (13) and (14) divided by the base Sb . 4. The optimization problem
The electrical system includes the modelling of production
plants and storage systems. The model of the cogeneration plants The optimization problem is characterized by different objec-
(microturbines) is described through (18)–(19) where, PPE ,h,i,t tives: minimization of operational costs of the microgrid, max-
[kW] is the primary energy per unit of time used to feed (with imization of the overall exergy efficiency, and minimization of
natural gas) the microturbines, Pel,h,i,t [kW], and Pth,h,i,t [kW] indexes for vulnerability assessment. In the following, the dif-
are the electric and thermal power produced by the microtur- ferent objectives are formalized, and the overall optimization
bines, respectively. In these expressions, µh and µh efficiency problem is defined.
5
M. Caliano, F. Delfino, M. Di Somma et al. Sustainable Energy, Grids and Networks 32 (2022) 100915

4.1. Exergy The linear objective function for the exergy contribution (to be
minimized) is given by
The exergetic objective is to maximize the overall exergy 1
efficiency of the system, ψ , defined as the ratio of the total daily f1 =
ψ
(35)
exergy output, Exout [kWh], to the total daily exergy input, Exin
[kWh] [34–38] to the system. That is, 4.2. Costs: fuel costs and emissions
Exout
ψ= (26) As regards the minimization of operational costs the objective
Exin
is:
The total daily exergy output required to meet the thermal T −1
demand for space heating (SH) and space cooling (SC) can be
∑ ∑
f2 =
formulated as:
t =0
T −1 ⎧ [i ]
(Ct + fe CCO2 ,em )Pgrid,IN ,i,t − Bt Pgrid,OUT ,i,t
∑ ∑
Exj,t ∆

Exout = (27)
× ∆
j∈{e,SH ,SC } t =0 ⎩ + PPE ,B,i,t (CCO2 ,em Ẽf −ng + TESpp ) + CCO2 ,em PPE,h,i,t Ef −ng
where, Exj,t is the total exergy rate required to meet the elec- H F ,i

tricity, and space heating and space cooling demand that are ∑ ⎬
+ Qgas,h,i,t Cgas (36)
assigned. In detail, the exergy rate required to meet the electricity ⎭
h=1
demand can be calculated as:
∑ with:
Exe,t = PD,i,t + Pv eh,i,t t = 0, . . . , T − 1 (28)
i∈N Pgrid,i,t = Pgrid,IN ,i,t − Pgrid,OUT ,i,t i ∈ N , t = 0, . . . , T − 1 (37)
where, PD,i,t [kW] is the power load (assigned), and Pv eh,i,t [kW] is where: ∆ is the time discretization interval, Pgrid,IN ,t and Pgrid,OUT ,t
the power consumption of the electric vehicles (assigned). As for [kW] represent the power injected into the microgrid and the
thermal demand, the energy quality depends on the temperature power sold to the main grid, Ct [e/kWh] is the unitary cost for
required. The exergy rate can be calculated, depending on the purchasing energy, Bt [e/kWh] is the unitary energy selling price,
Carnot factors, Fqheat
,t , as [10]: PPE ,B,i,t [kW] is the primary energy per unit of time consumed
by boilers, TESpp [e/kWh] is the unitary price for purchasing the
Exheat ,t = Dheat ,t Fqheat
,t t = 0, . . . , T − 1 (29) primary energy used to feed boilers, Cgas [e/m3 ] is the price
of natural gas, fe [tCO2 /kWh] is the emission factor relevant to
the power purchased from the grid, CCO2 ,em [e/tCO2 ] is the cost
Excool,t = Dcool,t Fqcool
,t t = 0, . . . , T − 1 (30) relevant to the emission of one ton of CO2 , according to the
most recent prices it is set equal to 50 [e/tCO2 ], Ẽf −ng and Ef −ng
where, Dheat ,t [kW] and Dcool,t [kW] are the total thermal and cool-
[tCO2 /kWh] are emission factor relevant to the fuel, PPE ,h,i,t [kW]
ing demand (assigned).
is the primary power consumed by the cogenerative plant h. This
On the supply side, the input energy carriers are grid power,
objective function is linear in the optimization variables.
natural gas, and solar energy. The total daily primary exergy input
is formulated as:
4.3. Stress indexes
T −1
∑ ∑
Exin = Exj,t ∆ (31) The objective function includes stress indexes to reduce vul-
j∈{grid,gas,solar } t =0 nerability and risks. Their mathematical formalization is taken
from the literature [39,40] and includes: the current level over
with:
∑( links (SIC ), the normalized voltage drop at each node of the dis-
Pgrid,IN ,i,t /εgen t = 0, . . . , T − 1
)
Exgrid,t = (32) tribution network (SIV ), and produced power (SIG ). Another term,
i∈N detailed in the following, is included in the formalization that
represents the maximum loss of load (LoLmax ) that may happen
∑ in the system. The objective is given by:
ςgas PPE ,B,i,t + PPE ,h,i,t t = 0, . . . , T − 1
( )
Exgas,t = (33)
i∈N f3 = αC SIC + αV SIV + αG SIG + αL LoLmax (38)
∑∑ with:
ExRES ,t = PRES ,l,i,t t = 0, . . . , T − 1 (34) T −1
i∈N l∈HR,i
∑ ∑∑( )2
SIC = Ii,j,t /Iimax
,j (39)
where: εgen and ςgas are fixed to 0.4 and 1.04, respectively, and t =0 i∈N j∈N
j̸ =i
represent the exergy efficiency of the power plants, and the
exergy factor of the natural gas; Pgrid,IN ,i,t , is the power injected T −1
into the microgrid; PRES ,l,i,t [kW] is the power produced by the l-th
∑ ∑ ( vi,t − 1 )2
SIV = (40)
renewable plant. Since energy demand, as well as the tempera- ∆vimax
t =0 i∈N
tures required for the evaluation of the Carnot factor for space
heating and space cooling, are assumed known, the total exergy T −1
( )
required to meet the demand is also known. Therefore, being the
∑ ∑ ∑ ∑ Pel,h,i,t PRES ,l,i,t
SIG = + (41)
objective the maximization of the overall exergy efficiency of the PelMAX
,h,i P RES ,l,i,t
t =0 i∈N h∈HF ,i l∈HR,i
microgrid, as reported in Eq. (24), and being Exout assigned, the
exergetic objective is formulated as the total daily primary exergy where: Iimax
,j is the current limit over the link (i, j), Ii,j,t is the cur-
input to the system, Exin , to be minimized (18). rent over link (i, j), αC , αV , αG , αL are known weight coefficients, vb
6
M. Caliano, F. Delfino, M. Di Somma et al. Sustainable Energy, Grids and Networks 32 (2022) 100915

Fig. 1. The MPC scheme.

is the base voltage, ∆vimax is the maximum allowable variation of ci,j,t = cos(δi,t − δj,t ) i, j ∈ N , i ̸ = j, t = 0, . . . , T − 1 (47)
voltage concerning the nominal value (given by standards), PelMAX ,h,i di,j,t = sin(δi,t − δj,t ) i, j ∈ N , i ̸ = j, t = 0, . . . , T − 1 (48)
is the rated power for the h-th production plant at node i, P RES ,l,i,t
is the maximum forecasted power from renewables. ei,j,t = ai,j,t ci,j,t i, j ∈ N , i ̸ = j, t = 0, . . . , T − 1 (49)
The current over the link (i, j) in [p.u.] is given by: fi,j,t = ai,j,t di,j,t i, j ∈ N , i ̸ = j, t = 0, . . . , T − 1 (50)
i2i,j,t = p2i,j,t + q2i,j,t i ∈ N , t = 0, . . . , T − 1 (42) For the sake of brevity, we do not report the convex relaxation
of each auxiliary variables, but we refer to [41] for a complete
To explain the term related to the maximum loss of load
discussion and about the quality con the overall relaxation with
(LoLmax ) in the objective function, it is important introduce the
respect to the complete nonlinear formulation.
variable LOLi,t , which is representative of a situation in which
there is a load shedding and a fee incurs (see [38]). In this paper,
the loss of load is associated with each node i, and the objective 4.5. The overall optimization problem
is either the one of minimizing loss of load and or the one of
avoiding that a single node is in a very bad situation with respect The overall objective function to be minimized is given by
to the other nodes. For this reason, we have developed a robust [37,38]:
min max approach in which we minimize the maximum loss of
3 utopia
load under a constraint that says that at each node the loss of load ∑ fp − fp
should be less or equal to the maximum value. Thus, the following J = utopia
(51)
fpnadir − fp
constraint must be added for bounding the loss of load: p=1

utopia
0 ≤ LOLi,t ≤ LoLmax i ∈ N , t = 0, . . . , T − 1 (43) where: fp is the pth objective function, and fp and fpnadir are
utopia and nadir reference values for the pth objective func-
4.4. Problem convexification tion (such points are calculated minimizing and maximizing,
respectively, each objective subject to the constraints of the opti-
In this section we present a model of convexified power mization problem). The constraints of the optimization problem
flow Eq. (16)–(17) that are strictly nonlinear and nonconvex. As are given by (1)–(15) and (18)–(50).
regards the objective function we do not have any convexity con-
cerns since the total aggregated function is quadratic. We adopt 4.6. The MPC scheme
the convex relaxation known as QC relaxation [41]. To introduce
the model, first we need to define the McCormick envelopes for a
The MPC optimization scheme is depicted in Fig. 1. Optimiza-
bilinear form w = xy as M(w, x, x, y, y) = {w = xy : x ∈ [x, x], y ∈
tion (T ) and simulation horizons (S) can be chosen according to
[y, y]} that corresponds to the following set of convex constraints: the available information about uncertain parameters/forecasts.

w ≥ xy + xy − xy The optimization horizon is the length over which the opti-

⎪ mization problem is run. In this MPC scheme, only the solution
⎨w ≤ xy + xy − xy

related to the first-time interval (first run) is employed for the

M(w, x, x, y, y) = (44) plants. Then, new in field measurements and new forecasts are


⎪ w ≥ xy + xy − xy added, and a new optimization problem is run (second run); until

w ≤ xy + xy − xy the simulation horizon S. This approach allows updating new

information when available and to reduce uncertainties related
where, x, y and x, y are upper and lower bounds of x and y. to renewables and load forecasting.
In Eq. (16)–(17) the nonconvexities are due to bilinear, trilin-
ear (i.e. a combination of two bilinear forms) and trigonometric
terms. In this regard we define six auxiliary variables that are 5. The optimization algorithm
made convex thanks to (44)
Thanks to the convexification process for the power flow equa-
ai,j,t = vi,t vj,t i, j ∈ N , i ̸ = j, t = 0, . . . , T − 1 (45) tions, the overall optimization problem described by (1)–(51) is
a quadratic program (QP) that can be rewritten in standard form,
bi,t = vi2,t i ∈ N , t = 0, . . . , T − 1 (46) (defining x as the vector that includes all the decision variables)
7
M. Caliano, F. Delfino, M. Di Somma et al. Sustainable Energy, Grids and Networks 32 (2022) 100915

as: Proof. We start introducing a slack variable z to the inequality


1 constraints of problem (52), we obtain the following problem
min x′ Qx + q′ x
x∈{Cx≤d} 2 which becomes separable:
(52)
1
s.t Ax = b min x′ Qx + q′ x + IΩ (z )
where the objective function is defined by a positive semidefinite
x
{2
Ax + Bz = c (64)
matrix Q and a vector q, and the constraints by matrices A, C and s.t
vectors b and d. Ω {z : z ≥ 0}
To solve problem efficiently, we propose an extended ver-
where, IΩ (z ) is the indicator function of Ω : IΩ (z ) = 0if z ∈ Ω
sion of the BADMM [15], including and Heavy ball (HB) dual
and IΩ (z ) = +∞ otherwise. Choosing f (x) = 21 x′ Qx + q′ x and
acceleration term (BADMM-HB). BADMM generalizes the ADMM
g(z) = IΩ (z ) we can apply the BADMM-HB steps (56)–(59). Sub
method [42] allowing the choice of different Bregman diver-
problem (56) thanks to (64) becomes:
gences to exploit the structure of problems. To introduce the
algorithm the following definition is needed. xk+1 = argmin
x
ρ
{ }
Definition 1. For a convex function φ : R → R, the associated
n 1 ′
× x Qx + q x + λk Ax +
′ ′ 2
∥Ax + Bz − c ∥ + ∥x −
2 xk 2Zx

Bregman distance is defined as 2 2
Bφ (x, y) = φ (x) − φ (y) − ⟨φ (y) , (x − y)⟩ ∀x, y ∈ Rn (53) (65)
that is an unconstrained QP whose unique solution is (60). In-
The BADMM-HB algorithm solves a 2-block separable problem
stead, sub problem (57) becomes:
in the form:
min f (x) + g(z)
{ ′ ρ }
x∈X ,z ∈Ω zk+1 = argmin IΩ (z ) + λk Bz + ∥Ax + Bz − c ∥22 + ∥z − zk ∥2Zz
(54) z 2
s.t . Ax + Bz = c (66)
with variables x ∈ Rn and z ∈ Rm , where A ∈ Rp×n , B ∈ Rp×m , Using some algebra, the z-update becomes the proximal map-
and c ∈ Rp . We will assume that f and g are convex functions. ping of the indicator function IΩ (z):
To solve problem, we form the augmented Lagrange function
IΩ (z ) + ∥z − ζk ∥2Zz
{ }
ρ zk+1 = argmin (67)
L(x, z , y) = f (x) + g(z) + λ (Ax + Bz − c) +
T 2
∥Ax + Bz − c ∥ (55)
2
z
2
with ζ = (I + Zz )−1 −Gxk+1 + c − λk (ρ B)−1 + Zz zk
[( ) ]
that is
where, λ is the Lagrange multiplier associated to the equality
equivalent to the projection over the set Ω ,
constraints, and ρ is a step size parameter.
zk+1 = argmin ∥z − ζk ∥2Zz
{ }
For this specific problem we employed the Mahalanobis [43] (68)
distance Bφ (x, y) = (x − y)′ Z (x − y) = ∥x − y∥2Z , with Z ≻ z ≥0

0, as Bregman divergence. BADMM-HB performs the following which is a projection over a box set that can be explicitly solved
iterations, until convergence: obtaining (61). This completes the proof. ■
L(x, zk , λk ) + Bφ (x, xk )
{ }
xk+1 = argmin (56)
x∈X
6. Results
= argmin L(xk , z , λk ) + Bφ (z , zk )
{ }
zk+1 (57)
z ∈Ω
The developed EMS is applied to the Savona Campus poly-
generation microgrid. The SPM is a cogenerative low-voltage
λk+1 = λk + ρ (Axk+1 + Bzk+1 − c ) (58)
microgrid that has been operating since February 2014. It has
λk+1 = λk+1 + θ (λk+1 − λk ) (59) been funded by the Italian Ministry of research.
A simplified version of the one-line diagram of the network is
where, λk+1 is the accelerated multiplier (from HB update) with sketched in Fig. 2. The components included in the SPM are two
as θ step size parameter. cogeneration units (indicated as C651 and C652 in the following
In the following, we exploit the structure of problem (52) to and characterized by a rated electrical power of 65 kW); one
obtain an efficient explicit QP algorithm that does not need any NaNiCl2 battery storage system (ST), with a rated capacity of 141
mathematical programming tool to perform the iterations. kWh; one boiler system (B) with a maximum thermal power of
about 1000 kW; two photovoltaic fields (PV1 and PV2) charac-
Result 1. For problem (52) the BADMM-HB iterations are: terized by peak powers of 80 kWp and 15 kWp respectively;
)−1 [ one prosumer (i.e., the SEB, Smart Energy Building) equipped
xk+1 = − Q + ρ G′ G + Z q + ρ G′ (Bzk + λk − c ) − Zx xk
( ]
(60)
with a geothermal heat pump and a third PV field (PV3, peak
power: 21 kW). The main Campus loads, representing the main
electrical demand, are connected at node 1 and include buildings
zk+1 = max 0, (I + Zz )−1 −Gxk+1 + c − λk (ρ B)−1 + Zz zk
{ [( ) ]}
for students, classes, a canteen, a bar, offices, and laboratories. In
(61) the presented case study, analyses for different periods of the year
have been performed. In the next sub-sections, firstly, data are
reported; then the procedure to obtain utopia and nadir points is
λk+1 = λk + ρ (Axk+1 + Bzk+1 − c ) (62) detailed, and, finally, optimal results are shown.
As regards data in the considered scenarios, it is important
to note that: PV fields have on the whole a peak power of
50 kW (winter scenario) and 90 kW (summer scenario) at noon;
λk+1 = λk+1 + θ (λk+1 − λk ) (63)
electrical power demand has a peak power of 180 kW (winter
where G = col [A; C ], B = col [I ; 0], c = col [b; d], Zx and Zz are scenario) and 210 kW (summer scenario) at noon; external tem-
weight matrices, and I is a suitable identity matrix. perature is between 0 ◦ C and 10 ◦ C in winter and 20 ◦ C and
8
M. Caliano, F. Delfino, M. Di Somma et al. Sustainable Energy, Grids and Networks 32 (2022) 100915

Fig. 2. Savona Campus and SPM one-line diagram.

utopia
Fig. 3. Electric power balance — f1 calculation.

30 ◦ C in summer scenarios; thermal power has a peak value f3 ) we get 786.8775 e and 847.3812, respectively. The electric
of 530 kW (winter scenario) and 150 kW (summer scenario). power schedule is reported in Fig. 3, where the profiles of the
It is important to note that the thermal power demand for the powers generated by the various sources are shown, along with
summer period is represented by the chiller demand that will the total demand (PD ) and the power absorbed by the heat pump
satisfy the campus cooling service during the working hour (from of the SEB PHP ,SEB . The largest part of the power requested is
8 am to 5 pm). The unitary cost of the power purchased from the purchased from the external grid. Instead, as regards the thermal
grid Ct is equal to 0.198/0.220/0.256 [e/kWh] according to the part, the thermal balance is not exceeded but only satisfied. This
low/intermediate/high power request time intervals. The other behaviour is mainly since here we try to reduce the overall exergy
parameters are the same as in [12]. in input and a larger production of power from the microturbines
would reflect also in higher production of thermal power (whose
6.1. Derivation of utopia points and discussion demand is already satisfied), and therefore in a thermal power
loss.
In this subsection, every utopia point of each element of the
objective function is discussed.
6.1.2. Costs and emissions — f2
6.1.1. Exergy - f1 In this case, the objective to be minimized is f2 , i.e., the one
In this case, minimizing only objective f1 , we obtain a value relevant to the costs for the energy purchase and for the CO2
utopia
of 0.3570 (f1 ) while evaluating the other two terms (f2 and emission. The value obtained through the minimization is 712.96
9
M. Caliano, F. Delfino, M. Di Somma et al. Sustainable Energy, Grids and Networks 32 (2022) 100915

utopia
Fig. 4. Electric power balance — f2 calculation.

utopia
Fig. 5. Electric power balance - f3 calculation.

utopia
e(f2 ). In this case, the other objectives (not included in the the storage is discharged during the morning and then charged
objective function), result to be equal to 0.3729 (f1 ), and 832.37 taking advantage of the power generated by the renewables.
(f3 ). Fig. 4 shows the electrical plants (active power) scheduling. The design on the nadir points has been carried out with the
In this case, microturbines produce more power with respect same approach adopted for the utopia points but maximizing the
to the previous one (exceeding also the thermal balance), espe- objectives f1 , f2 and f3 . Proper dissertations on multi objective
cially in the central part of the day. A huge difference is also optimization are in [44,45].
present for the storage schedule, as it is charged and discharged Finally, Tables 1 and 2 show the values of the nadir and utopia
multiple times along the day. objective functions for each objective function considered.

6.1.3. Stress indexes — f3


utopia
Finally, the third objective minimization (f3 ) is carried out. 6.2. Optimal results
The resulting optimal value is 568.26 while the other terms result
to be 0.405 and 758.551 e(f1 and f2 ). Fig. 5 shows the active
power schedule of the plants. In this case, the result obtained is In the following, the results of the overall optimization prob-
utopia
similar to the one in the f1 result. The microturbines supply lem described in Section 4.5 are reported. Two different scenarios
only to a small portion of the power demand but, in this case, are detailed: winter and summer scenarios.
10
M. Caliano, F. Delfino, M. Di Somma et al. Sustainable Energy, Grids and Networks 32 (2022) 100915

Fig. 6. Electric power balance result, summer scenario.

Fig. 7. Thermal power balance result, summer scenario.


Table 1
Comparison between utopia and nadir points summer scenario. Table 2
Comparison between utopia and nadir points winter scenario.
utopia
p fp fpnadir
utopia
p fp fpnadir
1 0.3570 0.435
1 0.3302 0.4552
2 712.96 e 810.63 e
2 998.65 e 16712.98 e
3 568.26 5983.3
3 821.50 65829.56

6.2.1. Summer scenario


The optimal schedules obtained solving the optimization prob- The storage is discharged to face the rapid increase of the
lem during the summer scenario are presented from Figs. 6 to 9. demand while the renewable production is still contained, and it
The value of the objective function is 0.4529 (with f1 = 0.3628, f2 is charged in the next hours when the PV production is maximum.
= 747.97 and f3 = 675.43). The thermal power balance in Fig. 7 shows that the thermal
The electric power balance is presented in Fig. 6. It is evident production is larger than the power request and consequently
that microturbines satisfy about one-half of the power demand some of the power is dispersed in the atmosphere.
while the remaining portion is provided by the grid in the first The other Figures present the stress indexes considered in the
and last hours, and by renewables in the central part of the day. optimization problem. In Fig. 8 it is possible to see the index

11
M. Caliano, F. Delfino, M. Di Somma et al. Sustainable Energy, Grids and Networks 32 (2022) 100915

Fig. 8. Currents Stress Indexes on the different lines, summer scenario.

Fig. 9. Voltage Stress Indexes for each node, summer scenario.

relevant to the currents on each line. The values reached are thermal request, while the microturbines are operated at their
always far from the maximum value (equal to 1) but for one line, maximum thermal setpoint.
namely from node 5 to node 1, which is the most stressed in the As regards the stress indexes, from Fig. 12 it is possible to see
central hours of the day. The last index (Fig. 9) is relevant to the that the most stressed lines are line 5–1 and 0–1, recalling the
voltage at each node. In this case, the voltage magnitude is very same behaviour of the summer scenario. Finally, Fig. 13 shows
close to the base value (SIV = 0 when vi,t = vb ). again that as regards voltages the power grid nodes are not
deeply stressed since the value of this index is quite low for all
of them (with maximum value 0.15).
6.2.2. Winter scenario
The optimal schedules obtained solving the optimization prob- 6.3. Sensitivity analysis over the different objectives
lem during the winter scenario are presented from Figs. 10 to 13.
The value of the objective function is 0.0037166 (with f1 = 0.3305, The proposed objective function (54) includes three not com-
f2 = 1001.5 and f3 = 879.033). The electric power scheduling is mensurable terms: exergy, costs and emissions and stress in-
given in Fig. 10. The microturbines are activated only when are dexes. It is interesting to investigate how the different terms of
needed to satisfy the thermal request. The storage is mainly used cost are sensible to the others. Thus, a sensitivity analysis has
during the PV peak hours to be recharged and discharged in night been carried out to evaluate the influence of the different terms of
hours. The thermal sources scheduling is presented in Fig. 11. It cost by considering all the possible pairs, namely f1 -f 2 , f1 -f 3 and
is clearly shown the need of the boiler plant to satisfy the overall f2 -f 3 . This analysis has been performed through the weighted sum
12
M. Caliano, F. Delfino, M. Di Somma et al. Sustainable Energy, Grids and Networks 32 (2022) 100915

Fig. 10. Electric power balance result, winter scenario.

Fig. 11. Thermal power balance result, winter scenario.

Table 3 f1 -f 2 case):
Sensitivity analysis for objectives f1 and f3 .

f1 f3 w min [(1 − w) f1 + w f2 ] (69)

0.3999 993.098 0.9 To this aim the winter scenario has been chosen. Specifically,
five possible values of w have been chosen to obtain the tradeoff
0.4178 934.2785 0.7
between the two objectives. After this analysis is evident that
0.4305 926.6137 0.5 costs f1 -f 2 are not conflictual so the results are not presented. The
0.4368 925.6651 0.3 obtained values for costs f1 -f 3 and f2 -f 3 are reported in Tables 3
and 4.
0.4382 922.4597 0.1
The analysis shows that costs f1 and f3 are conflictual, this
means that a little increase of exergy gives a great decrease of the
stress indexes objective. The same consideration can be done for
costs f2 and f3 namely the economic and stress indexes objectives.
method, in which a sensitivity parameter w ∈ [0, 1] is assigned From an operational perspective, it is possible to find a trade-
to one of the elements of the cost pair properly normalized. Thus, off between the different objectives based on the applicative
the following minimization is performed (the example is for the framework (economic savings or increasing resilience).
13
M. Caliano, F. Delfino, M. Di Somma et al. Sustainable Energy, Grids and Networks 32 (2022) 100915

Fig. 12. Currents Stress Indexes on the different lines, winter scenario.

Fig. 13. Voltage Stress Indexes for each node, winter scenario.

Table 4 13:00 to 15:00. The demand response constraint modifies the


Sensitivity analysis for objectives f2 and f3 . scheduling of the two microturbines.
f2 f3 w Figs. 14 and 15 show the optimal results regarding the electric
scheduling without and with demand response, respectively.
1001.3465 1140.5685 0.9 As it can be seen in Fig. 15, in the two hours in which the
constraint applies, the power produced by the microturbines is
1114.7 1019.8792 0.7
reduced to follow the demand response constraints.
1124.1 959.5519 0.5

1233.4 908.0143 0.3 7. Conclusion


1244.9 827.0002 0.1
In this paper, a new EMS for microgrids, developed within the
LIVING GRID project (2017–2020), funded by the Italian Ministry
of Research (actions related to the Italian Technology Cluster
6.4. Demand response scenario on Energy), has been presented. A new optimization model has
been developed in which the objective function is related to the
Demand response events consist of additional constraints minimization of costs, the maximization of the overall exergy
(namely (24) and (25)) whose validity must be met for certain efficiency of the system, and the minimization of stress indexes.
time intervals. In this case, we have that Pgrid,i,t = 70 kW from Moreover, a novel accelerated algorithm based on BADMM has
14
M. Caliano, F. Delfino, M. Di Somma et al. Sustainable Energy, Grids and Networks 32 (2022) 100915

Fig. 14. Scheduling of electrical power for the 24-hour horizon without demand response.

Fig. 15. Scheduling of electrical power for the 24-hour horizon with demand response.

been developed for the solution of the EMS problem. The EMS References
is customizable for general case studies and can work both in
the day-ahead and in real time under a Model Predictive Control [1] I. Kustova, C. Egenhofer, The EU electricity sector will need reform, again,
(MPC) scheme. The EMS is applied to a real case study (Savona Interecon.: Rev. Eur. Econ. Policy 54 (6) (2019) 332–338.
[2] European Commission, EU Climate Strategies & Targets. Available online:
Campus SPM) and several tests and sensitivity analysis have been
https://ec.europa.eu/clima/policies/strategies_en.
carried out during summer and winter scenarios. The obtained [3] European Commission, 2030 Climate Target Plan. Available online: https:
results are consistent with the modelled objectives and con- //ec.europa.eu/clima/policies/eu-climate-action/2030_ctp_en.
straints proposing an effective tool also at industrial level. Future [4] International Renewable Energy Agency, Power system flexibility for the
developments regard a detailed modelling of the district heating energy transition part 1: Overview for policy makers [internet], 2018,
network and the application of different solution methods such Available from: https://irena.org/-/media/Files/IRENA/Agency/Publication/
2018/Nov/IRENA_Power_system_flexibility_1_2018pdf.
as distributed optimization.
[5] I. Oleinikova, E. Hillberg, Micro vs MEGA: Trends influencing
the development of the power system, 2020, Available online:
Declaration of competing interest https://ntnuopen.ntnu.no/ntnu-xmlui/bitstream/handle/11250/2678100/
ISGAN_DiscussionPaper_Annex6_microVsMEGA_2020.pdf?sequence=1.
[6] M.F. Zia, E. Elbouchikhi, M. Benbouzid, Microgrids energy management
The authors declare that they have no known competing finan- systems: A critical review on methods, solutions, and prospects, Appl.
cial interests or personal relationships that could have appeared Energy 222 (2018) 1033–1055.
to influence the work reported in this paper. [7] D. Thomas, G. D’Hoop, O. Deblecker, K.N. Genikomsakis, C.S. Ioakimidis,
An integrated tool for optimal energy scheduling and power quality
improvement of a microgrid under multiple demand response schemes,
Acknowledgements Appl. Energy 260 (2020) 1–16.
[8] A. Alarcon-Rodriguez, G. Aul, G. Galloway, Multi-objective planning of
distributed energy resources: A review of the state-of-the-art, Renew.
This work was supported in part by the LIVING GRID project
Sustain. Energy Rev; 14 (2010) 1353–1366.
(2017–2020) from Italian Ministry of research. A portion of the [9] H. Ren, W. Zhou, K. Nakagami, W. Gao, Q. Wu, Multi-objective optimization
work has been funded by RLOF18ASSRIC/3/1 project from Liguria for the operation of distributed energy systems considering economie and
Region, Italy. environmental aspects, Appl. Energy 87 (12) (2010) 3642–3651.

15
M. Caliano, F. Delfino, M. Di Somma et al. Sustainable Energy, Grids and Networks 32 (2022) 100915

[10] M. Di Somma, B. Yan, N. Bianco, P.B. Luh, G. Graditi, L. Mongibello, V. Naso, [29] M. Molinari, Exergy and Parametric Analysis: Methods and Concepts for a
Operation optimization of a distributed energy system considering energy Sustainable Built Environment, Royale Institute of Technology, Stockolm,
costs and exergy efficiency, Energy Convers. Manag. 103 (2015) 739–751. Sweden, 2012.
[11] B. Yan, M. Di Somma, N. Bianco, P.B. Luh, G. Graditi, L. Mongibello, V. [30] L.M. Ramirez-Elondo, G.C. Paap, R. Ammerlaan, R.R. Negenborn, R. Toon-
Naso, Exergy-based operation optimization of a distributed energy system ssen, On the energy, exergy and cost optimization of multi-energy-carrier
through the energy-supply chain, Appl. Thermal Eng. 101 (2016) 741–751. power systems, Int. J. Exergy (2013) 364–385.
[12] F. Delfino, G. Ferro, M. Robba, M. Rossi, An energy management platform [31] M. Di Somma, Y. Bing, Bianco, et al., Influence of energy quality manage-
for the optimal control of active and reactive power in sustainable ment on CO2 emissions in operation optimization of a distributed energy
microgrids, IEEE Trans. Ind. Appl. (2019). system, in: 2015 International Conference on Clean Electrical Power, ICCEP,
[13] ECBCS – Annex 49 – Low Exergy Systems for High Performance Buildings pp. 297–304.
and Communities, homepage. Available: http://www.ecbcs.org/annexes/ [32] J. Chen, J. Liu, Q. Wang, J. Zeng, N. Yang, A multi-energy microgrid
annex49.htm. modelling and optimization method based on exergy theory, in: 2018
[14] J. Szargut, International progress in second law analysis, Energy 5 (1980) Chinese Automation Congress, CAC, 2018, pp. 483–488.
709–718. [33] G. Ferro, M. Robba, D. D’Achiardi, R. Haider, A. Annaswamy, A distributed
[15] H. Wang, A. Banerjee, Bregman alternating direction method of multipliers, approach to the optimal power flow problem for unbalanced and mesh
Adv. Neural Inf. Process. Syst. 4 (January) (2014) 2816–2824. networks, IFAC-PapersOnLine 53 (2) (2020) 13287–13292.
[16] W. Feng, M. Jin, X. Liu, Y. Bao, C. Marnay, C. Yao, J. Yu, A review of [34] G. Bianco, S. Bracco, F. Delfino, G. Ferro, L. Parodi, M. Robba, M.
microgrid development in the United States – A decade of progress on Rossi, A demand response energy management system (DR-EMS) for
policies, demonstrations, controls, and software tools, Appl. Energy 228 sustainable district, in: 7th International Conference on Control, Deci-
(2018) 1656–1668. sion and Information Technologies, CoDIT), Prague, Czech Republic, 2020
[17] F. Yang, X. Feng, Z. Li, Advanced microgrid energy management system for pp. 551–556.
future sustainable and resilient power grid, IEEE Trans. Ind. Appl. 55 (6) [35] M. A. Rosen Dincer, Energy, environment and sustainable development,
(2019) 7251–7260. Appl. Energy 64 (1) (1999) 427–440.
[18] Z. Cheng, M.Y. Chow, Resilient collaborative distributed energy manage- [36] G. Wall, M. Gong, On exergy and sustainable development—Part 1:
ment system framework for cyber–physical DC microgrids, IEEE Trans. Conditions and concepts, Exergy, Int. J. 1 (3) (2001) 128–145.
Smart Grid 11 (6) (2020) 4637–4649. [37] M.A. Rosen, I. Dincer, M. Kanoglu, Role of exergy in increasing efficiency
[19] J.A.A. Silva, J.C. López, N.B. Arias, M.J. Rider, L.C. da Silva, An optimal and sustainability and reducing environmental impact, Energy Policy 36
stochastic energy management system for resilient microgrids, Appl. (1) (2008) 128–137.
Energy 300 (2021) 117435. [38] L.M. Ramirez-Elizondo, G.C. Paap, R. Ammerlaan, R.R. Negenborn, R. Toon-
[20] G.K. Farinis, F.D. Kanellos, Integrated energy management system for ssen, On the energy, exergy and cost optimization of multi-energy-carrier
microgrids of building prosumers, Electr. Power Syst. Res. 198 (2021) power systems, Int. J. Exergy 13 (2013) 364–385.
107357. [39] M. Vahedipour-Dahraie, H. Rashidizadeh-Kermani, A. Anvari-Moghaddam,
[21] Q. Jiang, M. Xue, G. Geng, Energy management of microgrid in grid- J.M. Guerrero, Stochastic risk-constrained scheduling of renewable-
connected and stand-alone modes, IEEE Trans. Power Syst. 28 (3) (2013) powered autonomous microgrids with demand response actions: Relia-
3380–3389. bility and economic implications, IEEE Trans. Ind. Appl. 56 (2) (2019)
[22] S. Aslam, A. Khalid, N. Javaid, Towards efficient energy management in 1882–1895.
smart grids considering microgrids with day-ahead energy forecasting, [40] G. Ferro, R. Minciardi, L. Parodi, M. Robba, M. Rossi, A multi-objective and
Electr. Power Syst. Res. 182 (2020) 106232. multi-decision maker approach for the balancing market in distribution
[23] M. Quashie, C. Marnay, F. Bouffard, G.A. Joós, Optimal planning of mi- grids in presence of aggregators, in: 2020 7th International Conference
crogrid power and ì operating reserve capacity, Appl. Energy 210 (2018) on Control, Decision and Information Technologies, Vol. 1, CoDIT, 2020
1229–1236. pp. 784–789.
[24] M. Quashie, F. Bouffard, C. Marnay, R. Jassim, G. Joós, On bilevel program- [41] C. Coffrin, H. Hijazi, P. Van Hentenryck, The QC relaxation: Atheoretical
ming of advanced microgrids, Int. J. Electr. Power Energy Syst. 96 (2018) and computational study on optimal power flow, IEEE Trans. Power Syst.
422–431. 31 (4) (2016) 3008–3018.
[25] D. Steen, M. Stadler, G. Cardoso, M. Groissböck, N. DeForest, C. Marnay, [42] S. Boyd, N. Parikh, E. Chu, Distributed Optimization and Statistical Learning
Modeling of thermal storage systems in MILP distributed energyresource Via the Alternating Direction Method of Multipliers, Now Publishers Inc,
models, Appl. Energy (2015) 782–792. 2011.
[26] M.J. Moran, Availability Analysis: A Guide to Efficient Energy Use. Revised [43] M. Chao, Z. Deng, J. Jian, Convergence of linear Bregman ADMM for non-
Ed, ASME, New York, 1990. convex and nonsmooth problems with nonseparable structure, Complexity
[27] F. Björk, Ş. Kilkiş, M. Molinari, Energy quality management and low energy 2020 (2020) 1–14.
architecture, in: ASES National Solar Conference, North Carolina, USA, 2011. [44] D. Anghinolfi, M. Paolucci, M. Robba, Optimal planning of door-to-door
[28] B. Yan, M. Di Somma, G. Graditi, Multiobjective operation optimization multiple materials separated waste collection, IEEE Trans. Autom. Sci. Eng.
of DER for short-and long-run sustainability of local integrated energy 14 (3) (2016) 1448–1457.
systems, in: Distributed Energy Resources in Local Integrated Energy [45] K. Il Yong, O.L. De Weck, Adaptive weighted sum method for multiob-
Systems, Elsevier, 2021, pp. 89–123. jective optimization: A new method for Pareto front generation, Struct.
Multidiscip. Optim. 31 (2) (2006) 105–116.

16

You might also like