Professional Documents
Culture Documents
To cite this article: Dhananjay Kumar , H. D. Mathur , S. Bhanot & Ramesh C. Bansal
(2020): Forecasting of solar and wind power using LSTM RNN for load frequency
control in isolated microgrid, International Journal of Modelling and Simulation, DOI:
10.1080/02286203.2020.1767840
Article views: 6
ARTICLE
Forecasting of solar and wind power using LSTM RNN for load frequency control
in isolated microgrid
a
Dhananjay Kumar , H. D. Mathura, S. Bhanota and Ramesh C. Bansalb
a
Department of Electrical & Electronics Engineering, BITS Pilani, Pilani, India; bDepartment of Electrical Engineering, University of Sharjah,
Sharjah, United Arab Emirates
CONTACT Ramesh C. Bansal rcbansal@ieee.org Department of Electrical Engineering, University of Sharjah, Sharjah, United Arab Emirates
© 2020 Informa UK Limited, trading as Taylor & Francis Group
2 D. KUMAR ET AL.
With growing penetration of wind and solar power Forecasting methods of wind and solar power can be
to the main grid, transmission, and distribution com- categorized as physical methods, statistical methods,
panies are experiencing instability concerns related to and machine learning (ML) methods [14]. Physical
the voltage and frequency and the power sharing methods are based upon the numerical weather predic-
dynamics and its economic aspects are required to be tion (NWP) since it is related to the process of trans-
reanalyzed [6,7]. This concern is mainly due to inter- forming the wind and solar in to electric power [15].
mittent nature of renewable generation sources that is Statistical methods applied for forecasting are time-
dependent upon uncertain weather conditions. Better series-based approach, where predicted values can be
prediction of wind and solar power will pave the way expressed as a linear or nonlinear function of the pre-
for efficient and optimum utilization of renewable vious data [16,17]. The ML network models can map
energy sources [8]. The forecasted patterns of genera- a nonlinear and complex relationship between input
tion and load power for upcoming days are used to and output by applying various algorithms. Several
formulate the policy for scheduling of generators and structures of artificial neural networks (ANNs) such as
planning the transactions in the electricity market to feed-forward neural networks (FFNNs), support vector
balance the supply and demand of energy ensuring machines (SVMs), and extreme learning machines
stable operation of grid. The trends obtained after (ELMs) have been used by researchers to forecast wind
forecast are utilized by transmission/distribution com- and solar power. However, deep learning has been used
panies, energy solution companies, power trade mar- rarely for multi-step sequential forecasting. The possible
kets, and isolated power grids so that their power reason behind it may be the large number of forecasting
transaction can be effectively scheduled, dispatched, steps that eventually reduces the accuracy. The classifi-
and monitored. Hence, there is a need for cation and sub classification of machine learning tech-
a forecasting method with better accuracy and intelli- niques for various applications is given in [18]. The
gence and also forecasting based study of MG model conventional shallow neural network (SNN) models
for stability analysis under varying load conditions. suffer with slower convergence and overfitting problem
The speed and direction of wind depend upon the and they can be easily trapped in to the local minimum
natural parameters like the Coriolis force, pressure- [19]. To solve this issue, researchers have applied the
gradient force, and frictional force to the Earth surface deep neural network (DNN), which has many variants
[4]. The forecasted wind speed can be expressed in like the deep belief network (DBN), convolutional
terms of different wind components such as base neural network (CNN), and recurrent neural network
speed, gust speed, ramp speed, and noise [5]. The (RNN) [20,21]. RNNs are more powerful as compared
work presented in [9] briefs about the three different to ANNs in terms of feature extraction and finding the
power forecast horizons, i.e., for an hour, for a day, hidden and high level invariant structures from data.
and day before. Several grid managing activities such The output of RNN depends upon the present input and
as unit commitment, day ahead markets, ramping previous step observation and sometimes while learning
events, transmission scheduling, variability assessment, it faces the difficulty of long-term dependency. This is
planning, hedging, and asset optimization are related to referred as vanishing/exploding gradient problem [22].
the forecast horizons. In [10] the different category of To overcome this problem, an advance RNN has been
time horizons has been defined which are considered in proposed which is termed as long short-term memory
forecasting, i.e., very short-term (up to 30 minutes), (LSTM) [23,24]. The LSTM network gives better result
short-term (few minutes to 6 hours), medium-term for recursive multistep forecasting by eliminating the
(few hours to 24 hours), and long-term (more than vanishing gradient and gradient explosion problem. It
1 day). As per the regulation enacted by National completely utilizes the large amount of training data for
Energy Bureau (NEB) of China in 2011, for a wind forecasting, clustering and classification.
plant to plan the dispatch preparation the hourly pre- The variational mode decomposition (VMD) and sin-
diction of day-ahead forecasting is needed and also the gular spectrum analysis (SSA) techniques have been used
maximum error in daily forecast curve is limited to 25% in [25] for recursive small multistep wind speed forecast-
and the root mean square error (RMSE) of the all-day ing. The work reported in [26,27] predicts the upcoming
forecast should be less than 20% [11]. The scheduling of workload on the server using ML approaches like self-
wind and solar power is at a nascent stage in India and organizing feature map (SOFM), support vector machines
the forecast error margin is higher than ±30% [12]. For (SVMs), k Nearest Neighbors (kNN) neural networks, and
variable speed wind turbines the forecasting of wind nature inspired algorithms. In [28] two-tier architecture
speed has been presented in [13] for maximum power using kNN has been presented for financial time series
point tracking and control. prediction. Forecasting of mean hourly wind speed data
INTERNATIONAL JOURNAL OF MODELLING AND SIMULATION 3
using time series analysis has been presented in [29] where The survey of literature in above paragraph clearly indi-
the forecasting have been accomplished using linear cates the strength of deep learning (DL) for sequential time
ARIMA models and feed forward artificial neural net- series forecasting. Since, the RNN suffers with the difficulty
works. These models were tested on two different datasets of long term dependency, i.e., vanishing/exploding gradi-
and the ANN model outperformed the ARIMA model. ent problem, hence to solve this problem an advanced
A local recurrent neural network (RNN) model trained RNN, i.e., long short-term memory recurrent neural net-
using Recursive Prediction Error (RPE) algorithm for fore- work (LSTM-RNN) has been applied in this work to fore-
casting of wind power and speed been presented in [30]. In cast the solar irradiance and wind speed. The predicted
[31] review of forecasting methods for solar irradiance has solar radiation and wind speed have been utilized to calcu-
been summarized. Several models were introduced late the forecasted power from renewable energy sources.
for hour ahead forecast in [32], where ANN-based model In the survey of literatures it has been also found that the
was optimized by genetic algorithm. To estimate the solar forecast-based load frequency controller analysis is rarely
radiation by ANNs, review of several techniques has been investigated. Forecasted renewable power based load fre-
presented in [33] that discuss the nature of the ANNs and quency study is an effective way of studying the stability in
the comparison of different models for solar radiation MGs. The forecasted power is injected to small signal
estimation. An approach for forecasting power generation model of microgrid (MG) for load frequency control
24 hours ahead using a radial basis function network study. The varying load demand in MG results into fre-
(RBFN) is proposed in [34] for online training of experi- quency excursion from nominal value that may destabilize
mental roof top PV system. A hybrid iterative method for the MG power system. This paper introduces a novel
short-term wind speed forecast is presented in [35]. Based approach of renewable power prediction based load fre-
on variational mode decomposition method a multi-scale quency control analysis to study the stability of MG in the
wind speed forecasting is presented in [36]. presence of intermittent renewable energy sources and
To ensure the stability of MG power system in dynamic load demand. For forecasting the power from
response to the disturbance such as unbalancing renewable energy sources the hourly data of solar radiation
between demand and supply the controller play a very and wind speed have been taken from [49] for training the
significant role by regulating the frequency within a safe LSTM-RNN. Section 2 discusses the architecture of LSTM-
operating range. In [37] a neural network based upon RNN utilized for training the network with sequential data
the Gaussian mixture model has been proposed to fore- of wind speed and solar irradiance. The section 3 presents
cast the short-term wind power generation. Using fuzzy the different steps involved in training the LSTM-RNN
PI controllers in [38] the FC and AE hybrid system are network with hourly data of solar irradiance and wind
being controlled dynamically to achieve active power speed. The real data is obtained from a Solar Park located
balancing. A control system design has been presented in Charanka village of Patan district under Gujarat Energy
in [39] for under frequency load shedding in microgrid Transmission Corporation Limited. Section 4 describes the
utilizing the power generated from renewable energy MG model for small signal analysis that integrates the
resources. Various optimization methods for load fre- power obtained from solar and wind sources, auxiliary
quency control in autonomous MG is presented such as sources like DEG, MT, FCG, BESS, and FESS. The simula-
grasshopper optimization algorithm for fractional-order tion results of load frequency control for regulating the
PI-D controller [40], reinforcement learning-based con- deviation in frequency of MG caused due to the change
trol [41], and invasive weed optimization for automatic in power of load and renewable generation (wind and
generation control [42]. The work in [43] presents an solar) is presented in section 4. The section 5 briefs the
intelligent tuning method of fuzzy logic controller conclusion drawn from simulation results.
(FLC) using particle swarm optimization (PSO). The
frequency control in an AC microgrid using online
PSO-based fuzzy tuning approach has been presented 2. LSTM RNN-based forecasting
in [44]. For energy management in smart grids a firebase The property to learn the long-term dependency by
cloud messaging (FCM) system has been presented in RNNs makes it the more powerful network as compared
[45]. In [46], literature survey has been presented to other neural networks that have been applied for time
describing different intelligent forecasting techniques series forecasting. However it suffers with vanishing/
for load demand in power system. Design of two degree exploding gradient. In deep neural network the hidden
of freedom-internal model control [47] and perfor- layers and corresponding time steps are related to each
mance analysis of a cascade controller [48] has been other through multiplication so the derivatives/gradi-
presented for load-frequency control purpose. ents/errors are more sensitive toward vanishing or
4 D. KUMAR ET AL.
utilized as an input to the function for each prediction Figure 9. Forecasted and observed values (test data) for wind
step. Firstly, standardize the test data to have zero mean speed and RMSE (3 days).
and unity variance. To initialize the network state, first
predict on the training data. Next, make the first pre-
diction using the last time step of the training response network state is then updated with current step
and loop over the remaining predictions and input the input and predicted value of previous step. As we
previous prediction to predictAndUpdateState. have access of the observed data hence to improve
Afterward, de-standardize the prediction using the the training accuracy, instead of updating the net-
mean and variance calculated earlier. The RMSE is work with predicted values the observed values are
calculated and the training data with the forecasted feed to the network. To execute this step the net-
ones are plotted in Figures 6 and 7 for solar and wind work is reset first and then the network state is
time series respectively. initialized by predicting on the training data. The
The RMSE calculated after training for solar irradi- advantage of network state reset is that it does not
ance data is 11.62 and for wind speed data is 1.15. The allow the past prediction to affect the predictions
higher value of RMSE for solar is due to the fact that for done on the new set of data. The prediction is
half of the day time the solar has zero output while the destandardized using mean and standard deviation
wind is having nonzero output for the complete day calculated earlier and then RMSE is recalculated.
(24 hours). The observed values of solar irradiance and The updated forecast for solar irradiance and wind
wind speed is compared with forecasted values, i.e., the speed is shown in Figures 10 and 11 with reduced
test (observed) data which is 10 % of the complete dataset RMSE values of 4.45 and 0.17 for solar irradiance
(last 3 days or 72 hours out of 31 days or 744 hours of and wind speed respectively.
January month) is plotted with the forecasted values as To analyze the improvement in the accuracy of fore-
shown in Figures 8 and 9 with RMSE values. casting, a comparison is made between RNN LSTM
based forecasting and other techniques applied earlier.
The comparison parameter is the RMSE. Table 2 sum-
3.2. Updating the network state marizes the RMSEs obtained while forecasting the wind
While training the LSTM-RNN the predicted values speed and solar irradiance using proposed and existing
of previous step is feedback to the hidden layer. The techniques.
INTERNATIONAL JOURNAL OF MODELLING AND SIMULATION 7
Figure 15. Scenario 1 when a step load rise of 0.2 p.u is applied Figure 16. Scenario 2 when a step load fall of 0.2 p.u. is applied
at t = 21,600 seconds (a) step load rise (p.u.), (b) frequency at t = 72,000 seconds (20th hour) (a) step load fall (p.u.), (b)
deviation (p.u. Hz), (c) micro turbine power (p.u.), (d) fuel cell frequency deviation (p.u. Hz), (c) microturbine power (p.u.), (d)
power (p.u.), (e) diesel engine generator power (p.u.), and (f) fuel cell power (p.u.), (e) diesel engine generator power (p.u.),
BESS/FESS power output (p.u.). and (f) BESS/FESS power output (p.u.).
When there is rise in load demand as shown in Figure sequence forecasting of solar and wind power. This
15(a) the frequency falls out of nominal range as shown method is based on training, predicting and updating
in Figure 15(b). To mitigate the frequency deviation the the RNN states using memory cells in LSTM network.
controller responds and commands the artificial sources The LSTM based forecasting is used to eliminate the
to increase their power share to MG as shown in Figure vanishing/exploding gradient problems that may occur
15(c-e) and during this period the storage sources with RNN based forecasting. The RMSE obtained with
(BESS/FESS) are discharged as shown in 14(f). LSTM based forecasting of wind speed is found least
Scenario 2: A step fall in load at t = 72,000 seconds among the other reviewed techniques while for solar
(20:00:00 Hrs.) is applied to the MG irradiance forecast the RMSE obtained with RNN-LSTM
A step fall in load has been applied at 20th hour of is also found very less. Also, the feature of updating the
the day. The simulation results for scenario 2 in Figure LSTM-RNN with observed values instead of predicted
16 show that when there is fall in load demand of 0.2 p. values further reduces the RMSE in forecast process to
u. as shown in Figure 16(a) the frequency shoots up yield better prediction. The RMSE for solar forecasting is
from its nominal value as shown in Figure 16(b). To high in comparison to wind speed forecasting since the
mitigate the frequency deviation the controller responds mean and the standard deviation of the solar irradiance
and sends control signals to the artificial sources to data is high for solar. A more accurate forecasting for
reduce their power share as shown in Figure 16(c-e), wind speed and solar radiation is obtained with LSTM-
and during this period the storage sources (BESS/FESS) RNN. The day ahead forecasted wind and solar power is
are charged as shown in Figure 16(f). This leads to fed to mathematical model of MG and load frequency
restoration of frequency in MG power system. behavior has been studied. The close loop behavior of
auxiliary sources in MG is also studied to achieve the
demand supply balance after a step load disturbance of
5. Conclusion 20% in MG. A proportional and integral (PI) controller is
The implementation of LSTM-RNN based deep learning applied for actuating the secondary sources to restore the
method has been applied for multistep sequence to frequency deviations in MG caused due to sudden change
INTERNATIONAL JOURNAL OF MODELLING AND SIMULATION 11
in power demand. The forecasting error (RMSE) of wind integration, energy management in V2G etc. funded by Govt.
speed and solar irradiance is well within the acceptable agencies.
range. The proposed forecasting based MG model can be S. Bhanot obtained B.E. (Hons) Mechanical Engineering and
utilized for analyzing the wind and solar integration to M.Phil (instrumentation) from BITS, Pilani, and PhD from
MG and load frequency control study. IIT Roorkee (then University of Roorkee). She has 17 years of
teaching experience in BITS Pilani, and 19 years in Thapar
University. She is presently Professor at EEE Department
BITS Pilani. Her teaching and research areas are sensors,
Disclosure statement industrial instrumentation & automation, biomedical instru-
No potential conflict of interest was reported by the authors. mentation, application of AI techniques in process modelling,
control, image processing. She has been guiding projects/
theses at first degree, higher degree and PhD level and has
published research articles in international, national journals
Funding and conferences. She has published a book “Process Control,
This research work is supported by the Department of Science Principles and applications” with Oxford University press.
and Technology, New Delhi under the ICPS scheme through She is reviewer of many books, research papers and sponsored
letter no. [DST/CPS/CLUSTER/IoT/2018/General]. projects. She received best teacher award at Thapar university
and nominated by globally dispersed BITS alumni as a teacher
that they would especially like to recognise.
Notes on contributors Ramesh C. Bansal has more than 25 years of teaching,
research, academic leadership, and industrial experience.
Dhananjay Kumar received the B.E. degree from Rajiv Currently he is Professor in the Department of Electrical
Gandhi Proudyogiki Vishwavidyalay (RGPV), Bhopal, India Engineering at University of Sharjah, Sharjah, United Arab
in 2009 and M.E. (Power Electronics) degree from Samrat Emirates. In previous postings, he was Professor and Group
Ashok Technological Institute, Vidisha under RGPV Bhopal head (Power) in the Department of Electrical, Electronic and
in 2013. He is currently a full-time research scholar and Computer Engineering at the University of Pretoria, South
pursuing Ph.D. degree at department of Electrical and Africa and worked with the University of Queensland,
Electronics Engineering, Birla Institute of Technology and Australia; University of the South Pacific, Fiji; Birla Institute
Science, Pilani, Pilani campus, Rajasthan, India. His research of Technology and Science, Pilani, India; and Civil
interests include small signal modelling and stability control Construction Wing, All India Radio. Prof. Bansal has pub-
in microgrids, renewable energy integration to MG, forecast- lished over 300 journal articles, presented papers at confer-
ing of renewable power, robust control synthesis for fre- ences, and has published several books and chapters in books.
quency control in microgrids. He has authored 3 research He has Google citations of over 9000 and h-index of 44. He
articles in peer-reviewed international journals of repute and has supervised 20 PhD and 4 Post Docs and currently super-
his 7 papers have been presented in IEEE sponsored interna- vising several PhD students. He is an Editor/Associate Editor
tional and top national conferences. of member many reputed journals including IEEE Systems
Journal, IET-Renewable Power Generation, Technology and
H. D. Mathur received B.E. degree from Nagpur University,
Economics of Smart Grids and Sustainable Energy. He is a
Nagpur, India, in 1998; M.E. degree from Malaviya Regional
Fellow, and CP Engg. IET-UK, Fellow Institution of Engineers
Engineering College, Jaipur, India, in 2000; and the Ph. D.
(India), and Senior Member IEEE. He has diversified research
degree from Birla Institute of Technology and Science (BITS),
interests in the areas of Renewable Energy, Power Systems and
Pilani, India in 2007 and He was Post-Doctoral Fellow in
Smart Grid.
Supelec, Paris, France in 2013. He was also visiting scientist
to Centralesupelec, France in May- June of 2015 and May-
June of 2019. Currently, He is Associate Professor in the
Department of Electrical and Electronics Engineering, Birla ORCID
Institute of Technology and Science, Pilani. He is having
teaching and research experience of more than 20 years. Dhananjay Kumar http://orcid.org/0000-0002-5793-1418
Prof. Mathur is Associate Editor of IET Renewable Power
Generation as well as in editorial board and reviewer of
various journals mainly in field of electrical power system, References
renewable energy. He is a Senior member IEEE, Fellow of
Institution of Engineers (India), Chartered Engineer and life [1] Jose J, Kowli A, Bhalekar V, et al. Dynamic
member of Indian Society of Technical Education. He has microgrid-based operations: A new operational para-
published more than 65 papers in national/international jour- digm for distribution networks. IEEE Innov Smart Grid
nals and conference proceedings. His research interests Technol. 2016 Nov;28:564–569.
include power system control of isolated and interconnected [2] Bansal RC, Editor. Handbook of distributed generation:
power systems, power system optimization, automatic gen- electric power technologies, economics and environ-
eration control and artificial intelligence techniques applica- mental impacts. Cham, Switzerland: Springer; 2017.
tions in power systems and distributed generation (DG) with [3] Parhizi S, Lotfi H, Khodaei A, et al. State of the art in
grid interconnection issues. He is handling research projects research on microgrids: a review. IEEE Access.
related to area of electrical power system, renewable energy 2015;3:890–925.
12 D. KUMAR ET AL.
[4] Luigi F, Nunnari G, Nunnari S. Prediction models for [19] Niu T, Wang J, Zhang K, et al. Multi-step-ahead wind
solar radiation and wind speed time series. In: Fortuna speed forecasting based on optimal feature selection
L, Nunnari G, Nunnari S, editors. Nonlinear modeling and a modified bat algorithm with the cognition
of solar radiation and wind speed time series. Cham: strategy. Renewable Energy. 2018;118:213–229.
Springer; 2016. p. 41–46. [20] Yuan X, Tan Q, Lei X, et al. Wind power prediction
[5] Lee DJ, Wang L. Small-signal stability analysis of an using hybrid autoregressive fractionally integrated
autonomous hybrid renewable energy power genera- moving average and least square support vector
tion/energy storage system part I: time-domain machine. Energy. 2017;129:122–137.
simulations. IEEE Trans Energy Convers. 2008;23 [21] Aydin O, Guldamlasioglu S. Using LSTM networks to
(1):311–320. predict engine condition on large scale data processing
[6] Bansal RC, Bhatti TS, Kothari DP. Some aspects of grid framework. Proceedings of the 4th International
connected wind electric energy conversion systems. Conference on Electrical and Electronic Engineering
Interdiscip J Inst Eng. 2001;82(1):25–28. (ICEEE), Ankara, Turkey, 2017 Apr 8–10. pp. 281–285.
[7] Adefarati T, Bansal RC. Reliability, economic and [22] Srivastava S, Lessmann S. A comparative study of
environmental analysis of a microgrid system in the LSTM neural networks in forecasting day-ahead global
presence of renewable energy resources. Appl Energy. horizontal irradiance with satellite data. Solar Energy.
2019;236:1089–1114. 2018;162:232–247.
[8] Diagne M, David M, Lauret P, et al. Review of solar [23] Liu H, Mi X, Li Y. Wind speed forecasting method
irradiance forecasting methods and a proposition for based on deep learning strategy using empirical wavelet
small-scale insular grids. Renew Sust Energ Rev. transform, long short term memory neural network
2013;27:65–76. and Elman neural network. Energy Convers Manage.
[9] Kostylev V, Pavlovski A, “Solar power forecasting per- 2018;156:498–514.
formance–towards industry standards,” 1st International [24] Qu X, Kang X, Zhang C, et al. Short-term prediction of
workshop on the integration of solar power into power wind power based on deep long short-term memory.
systems, Aarhus, Denmark, 14 Oct. 2011. Proceedings of the IEEE PES Asia-Pacific Power and
[10] China Energy Storage Network News Center. [cited Energy Engineering Conference (APPEEC), Xi’an,
2018 Oct 9]. Available from: www.escn.com.cn. (In China, 2016 Oct 25–28.
Chinese). [25] Liu H, Mi X, Li Y. Smart multi-step deep learning model
[11] De Queiroz AR. Stochastic hydro-thermal scheduling for wind speed forecasting based on variational mode
optimization: an overview. Renew Sust Energ Rev. decomposition, singular spectrum analysis, LSTM network
2016;62:382–395. and ELM. Energy Convers Manage. 2018;159:54–64.
[12] Central Electricity Regulatory Commission (CERC). [26] Cao L. Support vector machines experts for time series
2012. New Delhi, Statement of objects and reasons forecasting. Neuro comput. 2003;51:321–339.
(terms and conditions for tariff determination from [27] Kumar J, Goomer R, Singh AK. Long short term mem-
renewable energy sources). 2019 Jul 25. Available ory recurrent neural network (LSTM-RNN) based
from: http://www.cercind.gov.in/2012/regulation/RE- workload forecasting model for cloud datacenters.
Tariff-Regulations-2012-SOR%206-2-2012.pdf Procedia Comput Sci. 2018;125:676–682.
[13] Achouri F, Mendil B. Wind speed forecasting techni- [28] Ban T, Zhang R, Pang S, et al. Referential kNN regres-
ques for maximum power point tracking control in sion for financial time series forecasting. International
variable speed wind turbine generator. Int J Model Conference on Neural Information Processing, Daegu,
Simulat. 2019;39(4):246–255. Korea: Springer Berlin Heidelberg; 2013 Nov 3. pp.
[14] Zhao X, Liu J, Yu D, et al. One-day-ahead probabilistic 601–608.
wind speed forecast based on optimized numerical [29] Steftos A. A novel approach for the forecasting of mean
weather prediction data. Energy Convers Manag. hourly wind speed time series. J Renewable Energy.
2018;164:560–569. 2002;27:163–174.
[15] Wang J, Song Y, Liu F, et al. Analysis and application of [30] Barbounis TG, Theocharis JB. Locally recurrent neural
forecasting models in wind power integration: A review networks for wind speed prediction using spatial
of multi-step-ahead wind speed forecasting models. correlation. Inf Sci. 2007 Dec 15;177:5775–5797. .
Renew Sust Energ Rev. 2016;60:960–981. [31] da Cunha Nunes RM. Big data techniques for solar
[16] Tascikaraoglu A, Uzunoglu M. A review of combined power forecasting [Dissertation of Master in Data
approaches for prediction of short-term wind speed Analytics and Decision Support Systems]. University
and power. Renew Sust Energ Rev. 2014;34:243–254. of Porto; 2017.
[17] Ahmed A, Khalid M. Multi-step ahead wind forecasting [32] Bao A, Fei S, Zhong M. Short-term solar irradiance
using nonlinear autoregressive neural networks. Energy forecasting using neural network and genetic
Procedia. 2017;134:192–204. algorithm. Chinese Intelligent Systems Conference.
[18] Lorberfeld A. Machine learning algorithms in layman’s Singapore: Springer, 2016 Oct 22. pp. 619–627.
terms, Part-1. 2019 Mar 2. 2019 Aug 03. Available from: [33] Kumar R, Aggarwal RK, Sharma JD. Energy analysis of
https://towardsdatascience.com/machine-learning- a building using artificial neural network: A review.
algorithms-in-laymans-terms-part-1-d0368d769a7b Energy Build. 2013;65:352–358.
INTERNATIONAL JOURNAL OF MODELLING AND SIMULATION 13
[34] Mellit A, Pavan AM. A 24-h forecast of solar irradiance frequency control using model approximation.
using artificial neural network: application for perfor- Int J Model Simulat. 2019;39(1):27–37.
mance prediction of a grid-connected PV plant at [48] Acharyulu BS, Mohanty B, Hota PK. Performance ana-
Trieste, Italy. Solar Energy. 2010;84(5):807–821. lysis of a cascade controller for conventional and
[35] Amjady N, Keynia F, Zareipour H. A new hybrid itera- deregulated power systems. Int J Model Simulat.
tive method for short-term wind speed forecasting. Eur 2019;1–19. DOI:10.1080/02286203.2019.1610690
Trans Electr Power. 2011;21(1):581–595. [49] Kumar D, Mathur HD, Bhanot S, et al. Frequency
[36] Ali M, Khan A, Rehman NU. “Hybrid multiscale wind regulation in islanded microgrid considering stochastic
speed forecasting based on variational mode decompo- model of wind and PV. Int Trans Electr Energy Syst.
sition. Int Trans Electr Energy Syst. 2018;28(1):e2466. 2019;29(9):e12049.
[37] Chang GW, Lu HJ, Wang PK, et al. Gaussian mixture [50] Long Short-Term Memory (LSTM). 2019 May 30.
model-based neural network for short-term wind https://in.mathworks.com/discovery/lstm.html
power forecast. Int Trans Electr Energy Syst. 2017;27 [51] Flores JJ, Graff M, Rodriguez H. Evolutive design of
(6):e2320. ARMA and ANN models for time series forecasting.
[38] Das S, Akella AK. A fuzzy logic-based frequency con- Renewable Energy. 2012;44:225–230.
trol scheme for an isolated AC coupled PV-wind- [52] Laopaiboon T, Ongsakul W, Panyainkaew P, et al.
battery hybrid system. Int J Model Simulat. 2019;1–13. Hour-ahead solar forecasting program using back pro-
DOI:10.1080/02286203.2019.1613063 pagation artificial neural network. IEEE International
[39] Madiba T, Bansal RC, Justo J Optimal control system of Conference and Utility Exhibition on Green Energy for
under frequency load shedding in microgrid system Sustainable Development, 2018 Oct 24; Phaket,
with renewable energy resources. In: Smart energy Thailand. pp. 1–7
grid design for Island countries-challenges and oppor- [53] Syed MS, Suresh CV, Sivanagaraju S. Short term solar
tunities. Cham, Switzerland: Springer; 2017. p. 71–96. insolation prediction: P-ELM approach. Int J Parallel
[40] Guha D, Roy PK, Banerjee S. Grasshopper optimization Emergent Distrib Syst. 2018;33(6):663–674.
algorithm-scaled fractional-order PI-D controller [54] Tena García JL, Cadenas Calderón E, González
applied to reduced-order model of load frequency con- Ávalos G, et al. Forecast of daily output energy of
trol system. Int J Model Simulat. 2019;1–26. wind turbine using sARIMA and nonlinear autoregres-
DOI:10.1080/02286203.2019.1596727 sive models. Adv Mech Eng. 2019;11(2):1–15.
[41] Mahmoud MS, Abouheaf M, Sharaf A. Reinforcement [55] Cadenas E, Rivera W, Campos-Amezcua R, et al. Wind
learning control approach for autonomous microgrids. speed prediction using a univariate ARIMA model and
Int J Model Simulat. 2019;1–10. DOI:10.1080/ a multivariate NARX model. Energies. 2016;9(2):109.
02286203.2019.1655701 [56] Cadenas E, Rivera W. Wind speed forecasting in three
[42] Mishra S, Barisal AK, Babu BC. Invasive weed different regions of Mexico, using a hybrid ARIMA–
optimization-based automatic generation control for ANN model. Renewable Energy. 2010;35
multi-area power systems. Int J Model Simulat. (12):2732–2740.
2019;39(3):190–202. [57] Nakagawa S, Tokoro T, Nakano T, et al. An effect of
[43] Li X, Song YJ, Han SB. Frequency control in micro-grid snow for electric energy generation by 40 kW PV
power system combined with electrolyzer system and system. IEEE Proc. 3rd World Conference on
fuzzy PI controller. J Power Sources. 2008;180 Photovoltaic Energy Conversion, 2003 May; Osaka,
(1):468–475. Japan. vol. 3, pp. 2447–2450.
[44] Bevrani H, Habibi F, Babahajyani P, et al. Intelligent [58] Matsuo H, Kobayashi K, Sekine Y, et al. Novel solar cell
frequency control in an ac microgrid: online PSO-based power supply system using the multiple-input DC-DC
fuzzy tuning approach. IEEE Trans Smart Grid. 2012;3 converter. INTELEC - Twentieth International
(4):1935–1944. Telecommunications Energy Conference, IEEE San
[45] Sujil A, Kumar R, Bansal RC, “FCM clustering based Francisco, CA, USA, 1998 Oct 4-8. pp. 797–8022.
adaptive neuro-fuzzy inference system based PV and [59] Shimizu T, Hirakata M, Kamezawa T, et al. Generation
wind generation forecasting agent for energy manage- control circuit for photovoltaic modules. IEEE Trans
ment in a smart microgrid”, 7th IET-RPG conf., Power Electron. 2001;16(3):293–300.
Copenhagen, Denmark, Sep. 26-27, 2018. [60] Senjyu T, Nakaji T, Uezato K, et al. A hybrid power system
[46] Bansal RC, Pandey JC. Load forecasting using artificial using alternative energy facilities in isolated island. IEEE
intelligence techniques: a literature survey. Trans Energy Convers. 2005;20(2):406–414.
Int J Comput Appl Technol. 2005;22(2/3):109–119. [61] Bevrani H, Feizi MR, Ataee S. Robust Frequency Control
[47] Sonker B, Kumar D, Samuel P. Design of two degree of in an Islanded Micro-grid: and μ -Synthesis Approaches.
freedom-internal model control configuration for load IEEE Trans Smart Grid. 2016;7(2):706–717.