Maximum electric power demand prediction by neural network

This paper presents a maximum electric load prediction method using a neural network. The proposed prediction system learns 2-past-weeks data, consisting of the temperature at peak load, its difference from the previous day, the weather, and peak load on each day. Then it forecasts the rate of change in peak load for the following day, inputting the temperature, its difference, the weather and so on. Simulation results show that the average prediction error of the method is about 3%. The prediction error can be further reduced by, for example, changing the number of hidden layers and neural network parameters, such as the system temperature.