Type-2 Fuzzy Wavelet Neural Network for Estimation Energy Performance of Residential Buildings

This paper proposes a type-2 fuzzy wavelet neural network (T2FWNN) for modelling the energy
performance of residential buildings. Based on the type-2 fuzzy rule base the multi-input multi-output
T2FWNN is proposed. The construction of T2FWNN structure has been accomplished using a cross-
validation approach and gradient descent learning algorithm. During construction, the adaptive learning
procedure was developed in order to stabilize and speed up the learning process. The proposed model is
used for the solution of two problems. At the first stage, based on statistical data, the T2FWNN model
has been designed for modelling the cooling and heating load of residential buildings. In the second
stage, using T2FWNN the prediction model was designed for the energy consumption of residential
buildings in Northern Cyprus. Comparative results have been provided in order to prove the efficiency of
using the designed model in prediction of the energy load of residential buildings. The obtained results
indicated the suitability of using T2FWNN system for estimation of the energy performance and
prediction energy consumption of residential buildings.

Empirical mode decomposition with support vector regression model [18], regression model based
on hybrid empirical mode decomposition and support vector regression with back propagation neural
network [19] were developed for mid-short term load forecasting. Hybrid models that integrate
different machine learning techniques were developed in order to improve the performance of the
prediction systems [20-25]. The effects of various parameters of residential buildings on energy
consumption were considered in [3,23]. Cooling and heating loads were evaluated using various
variables of buildings. Different research works have investigated the relationships between the
building’s parameters and the energy consumption related to cooling and heating loads. Improving the
performance characteristics of the designed models are the primary concern of most research works.
As we mentioned above the factors affecting energy utilization in buildings are different. These are
physical characteristics of buildings, weather conditions of the region, social and demographic location
of pupils, etc. As can be seen, the factors affecting energy utilization in buildings are different. The
influences of these factors on the energy load of buildings are imprecise and uncertain. Therefore, it is
difficult to evaluate accurately the energy performance and energy load of the building based on
uncertain information. One effective approach to solve these problems is the use of a fuzzy set theory
that allows handling the uncertainties and imprecise information in designed models.

These research studies demonstrated good performances of designed models. However, most of these
research studies are devoted to the solution of particular problems and sometimes did not adequately
describe some special characteristics such us uncertainty, high nonlinearity of considered problems. To
improve the performances of constructed models in the paper we propose the design of a hybrid system
based on type-2 fuzzy sets, neural networks and wavelet technology for modelling energy performances
of residential buildings. The integration of these paradigms in system design offsets the demerits of one
paradigm by the merits of another. Wavelet functions allow us to analyze non-stationary signals and
discover their local details. Neural networks have self-learning characteristics that increase the accuracy
of the model. The localization properties of wavelets used in NN structures allow modelling the local
detail of nonlinear processes. Fuzzy logic allows to reduce the complexity of the data and to handle
uncertainty and imprecision. The combination of fuzzy logic and wavelet neural networks automate the
construction of the fuzzy rule-based models, decrease their design time and develop trainable fuzzy
systems that can describe nonlinear systems characterized with uncertainties. The fast and accurate
design of fuzzy rules increases the accuracy of the model. In the paper, these methodologies are
integrated into the T2FWNN structure for modelling the energy performances of residential buildings.
The contribution of the paper is summarised as follows: The structure of a multi-input and multi-output
type-2 fuzzy wavelet neural network (T2FWNN) is proposed for modelling energy performance of
residential buildings, in particular, prediction of cooling and heating loads of buildings; The learning
algorithm of T2FWNN is designed using cross-validation technique, fuzzy clustering and gradient descent
algorithm with an adaptive learning rate; using T2FWNN the energy consumption prediction model is
designed for residential buildings in North Cyprus. Comparative results are provided to demonstrate the
effectiveness of designed T2FWNN models used for modelling energy performances and prediction
energy consumption of residential buildings.

The design stages of the T2FWNN for estimation of energy performances of residential buildings are
presented below.

Enter all X data set. Arbitrarily define cluster centers c for the X dataset.
for all X data set do
Apply fuzzy clustering determine cluster centers of input data sets Xi
Assign cluster centers to the centers of membership functions
end for
for all membership functions do
Determine ij width of membership functions using distances between the centers of membership
functions ij=max (abs(cij-ci,j-1), abs(cij+1-cij))
(where i=1,..,m, j=1,..,r, k=1,..,n, m is the number of input signals, r is the number of rules, n is the
number of outputs)
end for
Randomly initialize the values of the parameters of antecedent parts c1, c2, s and consequent parts-
w, a, b of T2FWNN.
Set maximum epoch number max_ep, learning rate , momentum rate , number of input and output
signal of networks, learning tag for antecedent part learn_antecedant to 0 or 1.
Set current epoch number ep_no = 1;
Set initial values of p = 0.5, q = 0.5
while ep_no<=max_ep do
 for each fold
determine input-output training and testing samples
for all input training data pairs do
For the given input data calculate the output of T2FWNN using (2-6)
Calculate output error as e(t) = UT2FWN-Ud ; (where UT2FWN is T2FWN output, Ud is the
desired output)
if abs(e(t)> Δ) then
Update w,a,b,p,q parameters using (8,10-13). (where Δ is an acceptable small value)
end if
if learn_antecedant = 1 then
Update c1, c2 and  parameters of membership functions
end if
end for
Use formula (16) to calculate MSE and RMSE values for training data (for training stage)
Use formulas (14) and (15) to adjust the learning rate
for all input testing, data pairs do
For the given input data calculate the output of T2FWNN using (2-6)
Calculate output error as e(t) = UT2FWN-Ud ;
end for
Use formula (16) to calculate MSE and RMSE values for testing data (evaluation stage)
end for
If the current RMSE value for training data is less than the one obtained in the previous epoch then
save parameters of T2FWNN
end while

In this paper, a novel T2FWNN model is proposed for the estimation of the energy performance of
residential buildings. The integration of fuzzy clustering, gradient descent algorithm and cross-validation
technique was used for the design of T2FWNN. The T2FWNN was proposed for solving two problems,
the first one is the determination of associations between building parameters and energy
consumptions for accurate prediction of the buildings’ energy load, and the second one is the prediction
of energy consumption in residential buildings in Northern Cyprus. Using statistical data and training
algorithm the development of the T2FWNN prediction models has been performed. The special adaptive
learning algorithm is developed in order to stabilize and speed up the training of the T2FWNN model.
After the design of prediction models, comparative results have been provided in order to show the
efficiency of the designed T2FWNN models. The obtained simulation results demonstrate that the
T2FWNN has obtained a better accuracy value than the existing models used for modelling energy
performances of residential buildings as presented in Table 4. The designed T2FWNN structure is also
applied for the prediction of energy consumptions of residential buildings of North Cyprus. Experimental
results indicate that the T2FWNN prediction model outperforms neural networks (NN) and type-1 fuzzy
neural networks based models in terms of prediction accuracy indices MSE and RMSE as presented in
Table 6. Comparative results with the different machine learning models demonstrated the efficiency of
the proposed T2FWNN model in prediction the energy consumption in residential buildings. Future
research is based on the improvement of the learning algorithm of the T2FWNN model and the
application of the presented model for solution other prediction and classification problems in
engineering.

Comments

The corrections have been done in the abstract in order to reflect this argument. In the paper type-2 fuzzy
wavelet neural network (T2FWNN) model is proposed for the solutions of two problems: for modelling
energy performance (the cooling and heating loads) of residential buildings and for the prediction energy
consumption of residential buildings.

The corrections have been done in order to reflect these arguments. In section 1 we have
presented a short review of research works for solving the indicated problems. We also add some
references and extended the review section. We add the following paragraph in order to reflect
the reason for using this model and the contribution of the paper.
These research studies demonstrated good performances of designed models.
However, most of these research studies are devoted to the solution of particular problems and
sometimes did not adequately described some special characteristics such us uncertainty, high
nonlinearity of considered problems. To improve the performances of constructed models in the
paper we propose the design of a hybrid system based on type-2 fuzzy sets, neural networks and
wavelet technology for modelling energy performances of a residential building. The integration
of these paradigms in system design offsets the demerits of one paradigm by the merits of
another. Wavelet functions allow us to analyze non-stationary signals and discover their local
details. Neural networks have self-learning characteristics that increase the accuracy of the
model. The localization properties of wavelets used in NN structures allow modelling the local
detail of nonlinear processes. Fuzzy logic allows to reduce the complexity of the data and to
handle uncertainty and imprecision. The combination of fuzzy logic and wavelet neural networks
automate the construction of the fuzzy rule-based models, decrease their design time and develop
trainable fuzzy systems that can describe nonlinear systems characterized with uncertainties.
The fast and accurate design of fuzzy rules increases the accuracy of the model. At the same
time, this hybrid structure allows approximate complex functions more effectively. In the paper,
these methodologies are integrated into the T2FWNN structure for modelling the energy
performances of residential buildings. The contribution of the paper is summarised as follows:
The structure of a multi-input and multi-output type-2 fuzzy wavelet neural network (T2FWNN)
is proposed for modelling energy performance of residential buildings, in particular, prediction
of cooling and heating loads of buildings; The learning algorithm of T2FWNN is designed using
cross-validation technique, fuzzy clustering and gradient descent algorithm with an adaptive
learning rate; using T2FWNN the energy consumption prediction model is designed for
residential buildings in North Cyprus. Comparative results are provided to demonstrate the
effectiveness of designed T2FWNN models used for modelling energy performances and
prediction energy consumption of residential buildings.
References [42] and [43] (in the current version of the paper they are numbered as [40] and
[47]) are now cited in the text.

In section 2 based on type-2 TSK fuzzy rule base presented by the formula (1) the structure of T2FWNN
was designed. The structure and mathematical formulas used for determining T2FWNN outputs signal are
presented in section 2. The presented T2FWNN structure realizes compositional rule inference using
type-2 fuzzy sets. In section 3 learning algorithms of T2FWNN is given. Parameters’ update rules are
presented.
We presented the basic design stages of the T2FWNN using the following pseudocodes.

Enter all X data set. Arbitrarily define cluster centers c for the X dataset.
for all X data set do
Apply fuzzy clustering determine cluster centers of input data sets Xi
Assign cluster centers to the centers of membership functions
end for
…..

We look at many journals that consider the modelling of energy performances of residential
buildings and find few references. Of course, there are too many papers related to the solution
of the prediction problem, but modelling of energy performances of residential buildings is a new
topic and only a few papers are published to solve this problem. We have given a comparison of
the models that use the same data set in Table 4. The comparisons have been done at the
same initial conditions. We have evaluated the performance characteristics of the considered
models. Comparative results demonstrate that the best performance is obtained by T2FWNN
model. We included the papers related to the solution of the given problem in the reference list.
The papers from the above list are included in the reference list.
2,

The novelty of the paper is added to the end of section 1. In the paper, a multi-input and multi-output
type-2 fuzzy wavelet neural network (T2FWNN) model is proposed for the estimation of the energy
performances of residential buildings. The learning algorithm based on cross-validation technique, fuzzy
clustering and gradient descent algorithm with an adaptive learning rate is proposed for the design of
T2FWNN; using T2FWNN the energy consumption prediction model is designed for residential buildings
in North Cyprus.

In section 2 based on type-2 TSK fuzzy rule base presented by the formula (1), the structure of
T2FWNN was designed. The structure and mathematical formulas used for determining
T2FWNN outputs are presented in detail. The presented T2FWNN structure realizes
compositional rule inference using type-2 fuzzy sets. In section 3 learning algorithms of
T2FWNN is given. Parameters’ update rules are presented. In section 4.1 we applied the
structure for determination energy performances of residential buildings and in section 4.2 we used
T2FWNN for the prediction of energy consumption of residential buildings in Northern Cyprus.
-Introduction section needs to be re-written to improve its quality and readability.
We have done corrections in the introduction section to fix the issue.

-Improve the quality of figures and explain those properly.

We have done some correction

Thank You for the comment. We have extended the review section. Some new references
including the above one are reviewed and added to the reference list.

Thank You for the comment. We format the reference list according to journal guidelines. We
have checked the references and have done some corrections
-There are many English and grammatical issues in the paper which needs to be rectified. 
Thank You for the comment. We have done corrections to reflect this argument.

The corrections have been done to reflect this argument.

The estimation of the energy performance of residential buildings has gained importance because of the
significant consumption of electricity in housing estate areas. Optimal control of energy utilization is
important for decreasing energy waste and its adverse impact on the environment. For this aim, different
approaches were utilized for robust and accurate prediction of the energy load in buildings. The use of
different kind of construction materials, timely change in building parameters lead to imprecise and vague
evaluation of energy consumption. As can be seen, the factors affecting energy utilization in buildings are
different. The influences of these factors on the energy load of buildings are imprecise and uncertain.
Therefore, it is difficult to evaluate accurately the energy performance and energy load of the building
based on uncertain information. One effective approach to solve these problems is the use of fuzzy set
theory that allow to handle uncertainties and imprecise information in designed models

The estimation energy performance of residential buildings as well as for prediction of their energy load.

We add a contribution to the end of section 1. The contribution of the paper is summarised as follows:
The structure of a multi-input and multi-output type-2 fuzzy wavelet neural network (T2FWNN) is
proposed for modelling energy performance of residential buildings, in particular, prediction of cooling
and heating loads of buildings; The learning algorithm of T2FWNN is designed using cross-validation
technique, fuzzy clustering and gradient descent algorithm with an adaptive learning rate; using
T2FWNN the energy consumption prediction model is designed for residential buildings in North Cyprus.
Comparative results are provided to demonstrate the effectiveness of designed T2FWNN models used for
modelling energy performances and prediction energy consumption of residential buildings.

We have added the reason why we implemented the research.

Most of the existing research studies are devoted to the solution of particular problems and sometimes did
not adequately described some special characteristics such us uncertainty, high nonlinearity of considered
problems. To improve the performances of constructed models in the paper we propose the design of a
hybrid system based on type-2 fuzzy sets, neural networks and wavelet technology for modelling energy
performances of residential buildings. The integration of these paradigms in system design offsets the
demerits of one paradigm by the merits of another. Wavelet functions allow us to analyze non-stationary
signals and discover their local details. Neural networks have self-learning characteristics that increase
the accuracy of the model. The localization properties of wavelets used in NN structures allow to model
of the local detail of nonlinear processes. Fuzzy logic allows to reduce the complexity of the data and to
handle uncertainty and imprecision. The combination of fuzzy logic and wavelet neural networks
automate the construction of the fuzzy rule-based models, decrease their design time and develop
trainable fuzzy systems that can describe nonlinear systems characterized with uncertainties. The fast and
accurate design of fuzzy rules increases the accuracy of the model. In the paper, these methodologies are
integrated into the T2FWNN structure for modelling the energy performances of residential buildings.
Thank You for the comment. The systems constructed on the base of type-1 fuzzy systems cannot directly
handle the uncertainties associated with information or data in the knowledge base of the process. Type-2
Fuzzy systems can handle such uncertainties and have the capability to model more complex
relationships. We add some new papers to the reference list and did some corrections to reflect the
comment.

- Mathematical formulation of proposed FLS type 2 - All variables used in equations should be
explained in text.

The definitions of symbols used in formulas were presented.

- Why you use Gaussian type-2 MFs? Please explain benefits of using G MFs over other types
of MFs.
There are other forms (triangle, trapezoid etc) for representation type-2 MF forms. But, for example,
triangle form needs 3 parameters (left, middle, right), trapezoid form needs 4 parameters for the
representation of MFs. In Gaussian, we are using 2 parameters (center and width) to represent MFs. In the
paper, the design of T2FWNN with a small number of parameters is important. A number of parameters
is affecting learning time. We selected the Gaussian MF function because of less number of parameters.

Thank You for the comment. Output fuzzy signals are reduced before defuzzification. A set of
algorithms have been developed for type-reduction. They can be divided into iterative and non-
iterative algorithms. The more used algorithm is Karnik-Mendel iterative algorithm that takes
time. In the paper, we implemented Beglarian-Melek-Mendel (BMM) algorithm for reduction
operation. The algorithm is presented in the following reference

M.B. Biglarbegian, W.W. Melek, and J.M. Mendel, “On the Stability of Interval Type-2 TSK Fuzzy
Logic Control Systems”, IEEE Trans. Syst., Man, Cybern. B, Cybern., vol. 40, no. 3, 2010, pp. 798–
818

- Training set - how many training parts are in it?

Thank You for the comment. We used crossvalidation technique for the training of T2FWNN. Using
cross-validation two independent data sets were generalised: training and evaluation. During the training
of the T2FWNN model, 10-fold cross-validation was used for separating the dataset into training and
testing sets. Here the original data samples are randomly partitioned into 10 groups of equal size. A single
group was used as a validation group for testing, and the remaining 9 groups were used for training. The
cross-validation is repeated 10 times (number of folds). The training is continued for 10 epochs set by the
programmer. In each epoch, one set of data (group) is used for evaluation of the model, and the remaining
other part is used for training. In each epoch, the testing data are changed and moved to the next group in
order to test all data sets. Final classification accuracy is determined using averaged values of accuracies
obtained from the folds.
Thank You for the comment. We add some fragments to the Conclusion in order to reflect this
argument.