| Load forecasting is the most fundamental job for power system when dealing with scheduling and planning problems, and it helps the power system to be operated safely and stably. At the same time, it relates to the economic benefit of power corporations. With the further reform of electric power system in China, the research on load forecasting is becoming increasingly important. Without an accurate load forecasting report, the deployment plan for power plants cannot be designed reasonably. Meanwhile, grid’s security and economy can’t be improved, nor does the power quality.When dealing with short-term load forecasting problems, the key is to understand and analyze the influence of the load correctly. Therefore, in this paper, features of load data in Nanchang has been analyzed, regarding meteorological factors, time factors and historical data factors as the factors that will influence the short-term load within the area. First, the application of artificial neural network algorithm in load forecasting has been introduced in this paper. Then, the architecture and establishment of BP neural network as well as Elman neural network which combines BP neural network with an undertake layer have been focused. After that, the standard particle swarm intelligence algorithm and its improved edition has been introduced briefly. Finally, in order to improve the accuracy of load forecasting, an algorithm “IPSO-Elman†based on the improved particle swarm intelligent algorithm and Elman neural network has been analyzed.The multi-characteristic vectors which will influence the short-term load have been regarded as a part of sample data in this paper for neural network algorithm to be quantized. Another part of sample data would be the historical load data which relevant to the predictive day mostly. The load forecasting procedures by neural network algorithm are as follows. First, pretreatment should be carried on for the sample data to obtain the dataset without abnormal points. Then, with the data obtained, BP, Elman and IPSO-Elman neural network can be trained respectively. At last, calculate the load on the predictive day by those networks. By analyzing the error indicators of the example, it can be concluded that the accuracy of short-term load forecasting by BP is higher than that by Elman, and the accuracy by Elman is higher that by IPSO-Elman. The result indicates that the IPSO-Elman short-term load forecasting model in this paper is effective and practical, and this model can improve the accuracy of short-term load forecasting. |
