Abstract— This paper presents a new electrical load forecasting (ELF) model based on constructive approach using feed-forward neural network (FFNN). The vital aspect of this model is to determine the FFNN architecture automatically during training in order to forecast the electrical load. Thus, the strength of standard FFNN increases in forecasting the electrical load. Furthermore, the proposed model overcomes efficiently the existing shortcomings of FFNN to predict loads of holidays and fast load changes. We call this model as constructive approach for electrical load forecasting (CAELF) as per short term basis. In order to evaluate the performance of CAELF, the daily electrical load demand data of Spain has been used. Experimental result shows that CAELF has a significant capability to forecast the electrical load compared to the other standard FFNN models. Keywords— Electrical load forecasting, Short-term, neural network, constructive technique, partial training, and CAELF. I. INTRODUCTION HE economy of the operation and control of power systems is sensitive to system demand; large savings can be obtained by increasing the accuracy of demand forecast. The effect of a large forecast error is reflected in terms of over conservative or over risky operation. It implies that, over estimation leads to the startup of too many units or excessive energy purchase, thereby supplying an unnecessary level of reserve. On the other hand, under estimation persuades insufficient preparation of spinning reserve and causes the system to operate at a risk region to the disturbance. Thus, improvement in load forecasting accuracy leads to the cost savings and increases in the system security [1]. A number of approaches exist in the literature (e.g., [1]- [11]), where they try to solve the short term electrical load forecasting (STELF) problem using neural networks (NNs). It has been confirmed that, the usage of NN in STELF always outperforms any human-based computational analysis in terms of accuracy, easy maintenance for users. Because, NN has a good capability for mapping between input and output although load (i.e., output) is being increased day by day [12]. Feed forward NN (FFNN) has been used in [7]-[11] to solve the ELF problem for different regions with a reasonable Kazi Rafiqul Islam is with the Dhaka University of Engineering and Technology,Gazipur-1700, Bangladesh (Email: rafiqul@duet.ac.bd). Md. Monirul Kabir is with the Dhaka University of Engineering and Technology, Gazipur-1700, Bangladesh(Email: munir@duet.ac.bd) Kazuyuki Murase is with the University of Fukui, Fukui 910-8507, Japan (Email: murase@u-fukui.ac.jp) computational cost. It is noted that, FFNNs are much suitable for mapping static relationships between inputs and outputs and ultimately providing good results in ELF. However, FFNNs need large historical data and have a limited capability to predict loads of holidays and fast load changes [13]. To overcome the shortcomings of FFNN, a number of efforts have been done in [2], [4]-[6] recently, among which echo state NN, radial basis function NN, recurrent NN, and nonlinear autoregressive NN are used, respectively. It is noted here that, the performances of aforementioned NN models are satisfactory in predicting the electrical load comparing to the FFNN, but computationally expensive. Thereby, huge requirements are necessary for the hardware setups as well as experts are needed for maintenances. This paper describes a new single-stage online ELF approach using FFNN, called constructive ELF approach (CAELF). This approach differs from previous works in a way that, CAELF determines the appropriate NN architecture in advance before the ELF starts using constructive NN training. In contrast to the previous approaches (e.g., [7]-[11]), they generally use a fixed NN architecture with randomly selecting the hidden neuron in the hidden layer during training before the ELF starts. It well known that, the random selection of hidden neurons affects the generalization performance of NNs. The reason is that, the performance of any NN is greatly dependent on its architecture [14][15].Thus determining hidden neurons’ number automatically provide a novel approach in building learning models using NNs for ELF. The remainder of this paper is organized as follows. Section II describes about the feed-forward neural network, whereas, a detailed description about CAELF has been presented in Section III. Section IV discusses the results of our experimental study. Finally, Section V concludes the paper with a brief summary and a few remarks. II. NEURAL NETWORK Artificial Neural Network is massively parallel inter- connected networks of simple adaptive elements and their hierarchical organizations [16]. These networks are intended to interact with the objects of the real world in the same way as biological nervous systems do. They constitute an alternative knowledge representation paradigm for artificial intelligence. Particularly, NNs are made up of processing units, called neurons. This neuron performs a weighted summation over the outputs of the neurons that are connected to its inputs. Then, Short-Term Electrical Load Forecasting using Constructive Feed-Forward Neural Network Kazi Rafiqul Islam, Md. Monirul Kabir, and Kazuyuki Murase T 2nd International Conference on Machine Learning and Computer Science(IMLCS'2013) August 25-26, 2013 Kuala Lumpur (Malaysia) 31