Improved Elman Neural Network Short-Term Residents Load Forecasting Considering Human Comfort Index

• Published in: Journal of electrical engineering & technology 2019, 14(6), , pp.2315-2322
• Main Authors: Yu, Yunjun, Wang, Xianzheng, Bründlinger, Roland
• Format: Article
• Language: English
• Subjects: Electrical Engineering; Electrical Machines and Networks; Electronics and Microelectronics; Engineering; Instrumentation; Original Article; Power Electronics; 전기공학
• ISSN: 1975-0102; 2093-7423
• DOI: 10.1007/s42835-019-00289-5

The massive access of distributed power in distribution network increases the complexity of user’s power consumption mode. It puts higher requirements on the accuracy and stability of load forecasting. The forward neural network has limitation in dynamic performance, and the prediction accuracy needs to be improved. This paper considers the influence of daily feature correlation factors on residential load. Then a method for improved Elman neural network short-term residential load forecasting considering human comfort index is designed. Using the human comfort index overcomes the shortcomings of low accuracy in load prediction when meteorological factor as a direct input. The Elman neural network’s incentive function is improved. The softmax function serves as an incentive function for hidden layer. Short-term load forecasting model was established for the load of Nanchang, Jiangxi, China. In order to reduce the impact of residents’ load characteristics, the samples of load are divided into weekend load, seasonal load and typical weather type load. Experiments show that the improved Elman neural network has higher prediction accuracy under three load types, compared with Elman neural network and RBF neural network.