Application of network structure method based on UD-IGA in short-term load forecasting system

Electric power system short term load forecasting is not only the basis for the scheduling of generating sets,but also the basis to work out the transaction schedule in electricity market.It has great influence on the operating,controlling and planning of electric power system.Due to the complicacy and uncertainty of load forecasting,electric power load is difficult to be forecasted precisely.In order to improve the precision of electric power system short term load forecasting,according to the features of power load and considering the combined influence of historical load data,weather and day type,a neural network structure method which combines uniform design with improved genetic algorithm is used to short-term load forecasting in this paper.The training and testing results show that the model can not only avoid convergence to the local minimum,but also improve the training speed for the neural network and accuracy for load forecasting.