摘要
分别介绍了采用BP神经网络模型和线性回归模型进行电价预测的方法和结果。方法的突出特点是在预测模型中引入了一个衡量市场力的新指标———发电容量必须运行率 (MRR) ,从而充分考虑了市场力对电价的影响 ,提高了电价预测的精度 ,特别是增强了短期预测模型对最高限价的预测能力。文中对MRR指标进行了简单的介绍 ,并针对电价预测的不同特点 ,对预测模型和预测变量的选择进行了探讨 ,提出了自己的观点。基于浙江电力市场实际运营数据的初步预测结果表明 ,所建预测模型是适用的 ,选择的预测输入变量是恰当的 ,电价预测精度能够满足电力市场实际运营的需要。
A backpropagation (BP) forecast model and a linear regression forecast model are described, and their prediction results are reported. A salient feature of the reported methods is that the forecast models take into account the influence of market power on the fluctuation of electricity price. This is achieved by using a market power index, namely, must run ratio (MRR), as an input to the price-forecast model. Particularly, the capability of the BP model in forecasting when price reaches price caps is evidently enhanced. A detailed discussion on the choice of forecast models and forecast variables is reported. The suggested method has been used to forecast short-term, medium-term, and long-term prices in Zhejiang Electricity Market. The results show that the proposed forecast models work reasonably well. Using the proposed forecast models, the price-forecast errors can be limited within a range that meets the requirement of actual electricity market operation.
出处
《电力系统自动化》
EI
CSCD
北大核心
2003年第22期16-22,共7页
Automation of Electric Power Systems
关键词
电价预测
电力市场
市场力
BP网络模型
线性回归模型
相关性分析
Backpropagation
Correlation methods
Electric utilities
Forecasting
Mathematical models
Regression analysis
