摘要
提高风电出力的预测精度可降低含高渗透率风电电力系统调度、优化、规划等策略的保守性和控制策略的复杂性。该文在分析风电出力历史数据与气象因素关系的基础上,建立了基于风电出力数据驱动的短期风电功率预估–校正预测模型。采用具有较高精度的小波神经网络预测模型实现预估环节,以自适应动态规划作为附加优化结构,利用风电出力实测数据及时更新预估模型中的参数,实现校正环节,使得预估模型能够适应风机在额定风速以下运行区域内多变的运行点。测试结果表明,该方法在风机出力变化频繁时,能获得比BP、GABP预测模型更高的精度。
Improving the prediction accuracy of wind power is of great value in reducing the complexity and conservative of dispatch, optimization and control of power system with high percentage penetration of wind power. Motivated by the idea of data-driven, a short-time predictor-corrector model with an additional learning architecture was proposed based on the analysis of the coupled relationship between wind power historical data and meteorological factors. Wavelet neural network was taken to perform as a predictor while additional learning structure with adaptive dynamic programming that makes use of the wind power data to update the state parameters of predictor was also introduced to form the corrector. Consequently, this model could adapt to a great variety of the operating points of wind system. The test results indicate that this method could both adapt to the frequent change of environment and get better forecast accuracy than BP and GABP methods.
出处
《中国电机工程学报》
EI
CSCD
北大核心
2015年第11期2645-2653,共9页
Proceedings of the CSEE
基金
国家863高技术基金项目(2015AA050104)
清华大学电力系统国家重点实验室资助项目(SKLD14KM02)~~
关键词
风电出力预测
数据驱动
预估–校正
自适应动态规划
小波神经网络
wind power forecast
data-driven
predictor-corrector
direct heuristic dynamic programming （HDP）
wavelet neural network
