Short term load forecasting using a robust novel Wilcoxon Neural Network

Short term load forecasting is essential to the operation of electricity companies. It enhances the energyefficient and reliable operation of power system. Artificial Neural Networks are employed for short term load forecasting owing to their powerful non-linear mapping capabilities. These are generally trained through backpropagation, genetic algorithm (GA), particle swarm optimization (PSO) and artificial immune system (AIS). All these algorithms have specific benefits in terms of accuracy, speed of convergence and historical data requirement for training. Load data is collected from remote locations through remote terminal units (RTU) over a communication channel that introduces noise which be Gaussian or non-Gaussian in nature. This paper provides the comparative study between Wilcoxon neural network (WNN) with Wilcoxon norm cost function and a Multi layer perceptron neural network (MLPNN) with least mean square (LMS) cost function. It is found that in case of regression or forecasting problem, similar to this containing few data sets, MLPNN provides better performance than WNN in terms of mean absolute percentage error (MAPE). Then a novel WNN is proposed to improve the MAPE of forecasting and to reduce computational complexity.