Recurrent Neural Net works-based User Side Load Situation Perception of Distribution Networks
This paper proposes a load situation awareness method based on recurrent neural networks to address the issue of MAPE errors in user side load analysis of distribution networks.The method uses denoising autoencoder to process his-torical load data,obtains low-frequency and high-frequency component sequences through singular spectrum analysis to describe the characteristics of load situation,constructs load situation perception model by using recurrent neural networks as the core and long short-term memory networks as assistance to predict independently low-frequency and high-frequency component sequences,and utilizes modified whale algorithm to optimize model parameters and improve prediction accura-cy.The experimental results show that the MAPE value of the perception results obtained by the new method is less than 5％,which meets the requirements of load situation perception for distribution network users.
NETL
NSTL
万方数据
