Artificial Neural Network based Short-Term Residential Load Forecasting

With the increase in people's standardization of living, there has been a great increase in the demand for electricity which enhances the conventional power as well as renewable energy. However, due to uncertainty of the demand as well as renewable power generation, the availability of spinning reserve or energy storage in the grid is essential for power backup which increases the cost of electricity. In this direction, the short-term load forecasting (STLF) model acts an important role in handling demand response management in smart grids extensively for the residential consumers to manage power delivery from the grid to the load in exact count. Now, the STLF model formulation can be achieved by applying an artificial neural network (ANN) to forecast the exact model with improving accuracy considering different constraints. In this work, weekly load data of a small residential consumers in the Indian power grid for five consecutive years are collected, and an ANN-based STLF model is formulated in MATLAB-2015a to solve the STLF problem. Further, the load factors for both the actual and the forecasted load model are calculated which shows improvement leading to an economically efficient model. As input data days, the type of day, hour, load demand and other environmental factors are considered. While solving the STLF using ANN, the training, testing and validation of the inputs are procured utilizing a multilayer perceptron feed forward neural network with a back propagation algorithm. To verify the performance of the ANN classifier, mean square error (MSE) is determined in addition to the graphical data in terms of the gradient, momentum parameter (mu) and validation checks. Furthermore, the accuracy of the ANN-based STLF model is justified by obtaining the regression (0.990) followed by the 99.4% accuracy with a self-comparative study.