基于信赖域法改进的BP网络在新能源并网方面的研究  

Research on improved BP network in new energy integration based on trust region method

作　　者：张凤林 陈峦 姚亮 鲁尔洁 杨云聪 张珺 

机构地区：[1]电子科技大学能源科学与工程学院,四川成都611731 [2]国网四川省电力公司天府新区供电公司,四川成都610041

出　　处：《可再生能源》2018年第1期43-50,共8页Renewable Energy Resources

基　　金：国家自然科学基金项目(61304177);四川省教育厅科研项目(15ZB0345)

摘　　要：传统大电网调频以PID结合智能算法为主要调节方式,系统存在迟滞性,不能有效控制较大的频率波动。为此引入神经网络模型,从负荷预测的角度对频率实施超前控制。通过信赖域法对现有BP神经网络模型加以改进,提高了其学习速度与预测精度,进而设计了一种基于负荷预测系统的大电网频率控制策略,通过预测负荷的分级实现机组优化调节,减少了不必要的旋转备用容量。通过搭建3种新能源-传统能源互补式发电机组仿真模型,对负荷预测频率控制的效果进行了仿真分析。仿真结果表明,与传统频率控制相比较,负荷预测控制下的大电网频率波动更小,调节时间进一步缩短。Frequency regulation of traditional large power grid is mainly regulated by PID combined the intelligent algorithms, these systems have hysteresis, which can＇t effectively control the larger frequency fluctuations. Therefore, the neural network model is introduced to implement the advanced control of frequency from the point of view of load forecasting. Through the trust region method, the existing BP neural network model is improved, the learning speed and prediction accuracy are improved. Then, a frequency control strategy of large power grid based on load forecasting system is designed. The optimal adjustment of the unit is realized by the classification of predicted loads, and the unnecessary spinning reserve capacity is reduced. Three kinds of new energy-traditional energy simulation models of complementary generators are built, and the effect of frequency control on load forecasting is simulated and analyzed. Simulation results show that compared with traditional frequency control, the frequency fluctuation of large power grid under load forecasting control is smaller and the regulation time is further shortened.

关 键 词：大电网 神经网络 信赖域 频率调节 负荷预测 新能源 

分 类 号：TK01[动力工程及工程热物理]