Short-Term Electricity Load Forecasting Based On Extreme Learning Machine

Short-term electricity load forecasting is an important part of modern power system,and it plays an important role in ensuring the stable,safe and efficient running of the power system.Based on Extreme Learning Machine（ELM）,which has simple structure and fast learning speed,batch and online algorithms are studied in this paper more systematically.and a short-term electricity load forecasting system based on MATLAB platform is designed at the end of the paper.The main work of this paper and the research results are as follows:（1）A batch training algorithm（ELM-ISA）for short-term electricity load forecasting based on improved Simulated Annealing algorithm（ISA）and Extreme Learning Machine is proposed in this paper.It improved the accuracy and stability of the prediction model by using Simulated Annealing algorithm to optimize input weights and hidden layer bias of ELM instead of random generation.Furthermore,the improved Simulated Annealing algorithm can adaptively set the initial parameters according to different training data,which improved the generalization ability of prediction model.Finally,In order to further improve the prediction accuracy,K-means clustering is added to the prediction model according to the idea of classification prediction,and an improved forecasting algorithm（ELM-ISA-K）based on ELM is proposed.Simulation results show that ELM-ISA-K algorithm has higher prediction accuracy and stronger stability.（2）The short-term electricity load forecasting mainly focuses on off-line forecasting at present,.However,electricity load forecasting is time-effective,and the model parameters need to be updated continuously according to the passage of time.A large amount of historical data is required for every training if off-line training mode is adopted.Therefore,an online training algorithm（ELM-SRUKF）based on square root unscented Kalman Filter（SRUKF）and Extreme Learning Machine is proposed in this paper.It used SRUKF to update the output weights of ELM and realized the online prediction of short-term electricity load.Simulation results show that compared with Online Sequential Extreme Learning Machine（OS-ELM）and ELM based on Kalman Filter,the ELM-SRUKF algorithm has higher prediction accuracy.（3）Taking the two algorithms proposed in this paper as the core,a short-term electricity load forecasting system based on MATLAB platform is designed,which supports offline and online forecasting modes.