An evolutionary fuzzy neural network approach for Short-Term Electric Power Load Forecasting

A Hybrid Evolutionary Programming and Fuzzy Neural Network (HEFNN) for load forecasting is presented in this paper. A Fuzzy Hyper-Rectangular Composite Neural Networks (FHRCNNs) was used for the initial load forecasting. Evolutionary Programming (EP) was then used to find the optimal solution for the parameters of FHRCNNs (including parameters such as synaptic weights, biases, membership functions, sensitivity factor in membership functions and adjustable synaptic weights). The EP generates a set of feasible solution parameters The EP has good global optimal search capabilities. The HEFNN was used to see if it could improve the solution quality and reduce the load forecasting error.