基于PSO训练BP神经网络的短期电价预测  

Short-term Price Forecasting Based on PSO Train BP Neural Network

作　　者：王立晶[1] 刘丛[1] 

机构地区：[1]华北电力大学电气与电子工程学院,河北保定071003

出　　处：《电力科学与工程》2008年第10期21-23,共3页Electric Power Science and Engineering

摘　　要：以美国PJM电力市场为背景,利用历史负荷、系统剩余容量百分比和清算电价对未来时段电价的影响来进行短期电价预测,给出了一个发电侧竞价模型中利用PSO训练BP神经网络进行市场出清电价预测的实例。与使用传统BP神经网络预测的方法进行比较,结果表明,该方法具有更高的预测精度,并能收敛于全局最优解。Based on American PJM electricity market, this paper use the influence which the historical load, percent of system reserve capacity and clearing electrovalence take to the future electrovalence to forecast short-term electrovalence. The paper provides an example of bidding model to forecast market clearing electrovalence using BP neural network trained by PSO. Compared with the result of traditional BP neural network, the proposed method has higher forecasting precision and can converge to global optimal solution at all times.

关 键 词：BP神经网络 粒子群优化算法(PSO) 电力市场 市场出清价 

分 类 号：TM743[电气工程—电力系统及自动化]