Short-term load forecasting based on improved extreme learning machine

Short-term load forecasting is the basis of power system regulation, and it affects many decisions of power system. In order to deal with the challenge of decline in prediction accuracy caused by reduction of cost, and improve the forecasting accuracy and speed, an improved extreme learning machine algorithm, which combines prior knowledge of residential electricity consumption habits is proposed to automatically select the number of hidden layer neurons and improve prediction effect. The purpose is to achieve a better short-term load forecasting effect with less manpower and material resources by mining information between data deeply. The experimental results show that the proposed algorithm has better performance on prediction accuracy and operation speed, when compared with the traditional prediction algorithm, and it has high practical value.