摘要
根据数值气象预报的风速和风向数据,采用粒子群优化的前向神经网络模型进行短期风电功率预测。分析了数据相关性对预测结果的影响,比较了粒子群优化前后的前向神经网络模型的性能。研究结果表明,经过粒子群优化后的前向神经网络模型比没有优化的前向神经网络有更好的性能,且根据数据相关性较好的测试样本可得到较好的预测结果。风速风向数值预报与输出功率的相关系数越高,基于粒子群优化前向神经网络模型的预测精度越高于没有优化的前向神经网络模型。
According to the wind speed and wind direction data from numerical weather prediction, a short-term wind power forecasting is performed by BP neural network model with particle swarm optimization (PSO). The influences of data dependency on forecasting results are analyzed and the performances of BP neural network model with and without PSO are compared. Research results show that the BP neural network model with PSO possesses better performance that that without PSO, and based on the testing samplings with better data dependency a better forecasting result can be achieved. The higher the correlation coefficient between the numerically predicted wind data and output power, the higher the forecasting accuracy by BP neural network model with PSO will be than that without PSO.
出处
《电网技术》
EI
CSCD
北大核心
2011年第5期159-164,共6页
Power System Technology
基金
国家863高技术基金项目(2007AA05Z428)~~
关键词
风电功率
预测
粒子群
神经网络
wind power
prediction
particle swarm
neural network