Neuro-expert system for planning and load forecasting of distribution systems

This work presents a neuro-expert system for medium term load forecasting and expansion planning of distribution systems. The planning module uses optimization technique, with heuristic rules to minimize the total loss and total annual cost. The constraints like overloading capacity and thermal ratings of conductors and transformers, voltage regulation, etc are satisfied. Concentric circle relaxation technique is proposed for feeder configuration. The trained network is used to forecast the yearly peak loads for a lead time of ten years. The back propagation algorithm is slightly modified and is used to train the artificial neural network. The modifications are done to reduce the training time and memory requirements. The proposed algorithms were tested on the practical 66/11 kV primary distribution system of Mysore City, Karnataka State, South India. The software was developed using the C++ language.