The development of a fuzzy neural system for load forecasting

In order to design an aggregate domestic load control system, a controller requires accurate predictions of load curves to make decisions about which loads should be connected to the grid. This paper presents a 24-hour load forecaster to be used by the controller. The forecaster will employ an Artificial Neural Network (ANN) structure with one input provided by a fuzzy weather controller. The use of fuzzy logic will enhance the performance of the system as well as make it more transparent and adaptable. A unique method is introduced to efficiently incorporate a larger number of inputs into the fuzzy controller without the problem of having an unmanageable rule base. The results show that the fuzzy neural system performs better than the artificial neural network load forecaster with further gains possible by fine-tuning the fuzzy logic block.