摘要
提出一种基于反向传播 ( BP)网络预测下一交易日无约束市场清算价格的方法。该方法去除了原始数据中的季节性趋势和增长趋势 ,用相关性分析技术选择输入量 ,遵循“重近轻远”的原则重新设计了拟合误差的代价函数 ,利用最新数据在线修正 BP网络 ,并用美国加州电能交易所公布的真实数据进行仿真 ,结果令人满意。
An innovative method to forecast the day-ahead unconstrained market clearing price (UMCP) based on BP network is presented. The seasonal and increasing tendency of raw data is wiped off. Correlation analysis technique is used to choose the input variables of BPN. An error performance function is redesigned to put more emphasis on the latest data. A tracking mechanism is developed by on-line adjusting weights. Finally, a simulation on the retrospective real data acquired from California Power Exchange web site shows considerable promising results.
出处
《电力系统自动化》
EI
CSCD
北大核心
2001年第19期11-14,33,共5页
Automation of Electric Power Systems