Short-term Forecasting Model of Regional Power Load Based on Neural Network

Short-term forecasting of regional power load is the basic work of planning and dispatching for distribution network system, which is quite challenging due to the characteristics of strong uncertainty and time dependence. In this paper, taking an region as an example, the correlation of power load and its influence factors, such as the 24-hours- average load before the forecast time, the load at the same time of the previous day (24-hours-ago), the load at the same time of the previous week (168-hours-ago), the dry bulb temperature and dew point temperature at the forecast time, etc., is analyzed based on the three-year historical data of power load and meteorological information, then a short-term forecasting model of regional power load is established using BP neural network method, and the forecasting results under different number of neurons, activation functions and hidden layers are compared. The conclusion shows that BP neural network has strong nonlinear fitting ability, can well realize the short-term prediction of regional power load, and is easy to be realized in engineering.