Electrical Load Forecasting Using a Neural-Fuzzy Approach

Electrical load forecasting plays a vital role in the operation, scheduling, maintenance and planning of electrical power generation and distribution companies. A novel hybrid system based on the neural implementation of fuzzy logic models of the nonlinear and chaotic electrical load time series is presented in this chapter. Being a universal function approximator, the hybrid neural-fuzzy system can very accurately capture the nonlinearity and chaotic behavior of electrical load series. A hybrid network’s performance is further largely improved from the application of moving average filter and fuzzy clustering on electrical load data along with least square error estimation of rules’ consequents. This, in turn, makes the short-term and mid-term forecasting of electrical load feasible. The results presented at the end confirm the efficacy of the neural-fuzzy system as a good electrical load forecaster.