基于自评判自学习机制的短期电力负荷预报  

Critic Self-learning Approach to Forecast Short-term Load

作　　者：林志玲[1] 李鸿儒[2] 

机构地区：[1]天津理工大学自动化学院,天津300191 [2]东北大学信息科学与工程学院,辽宁沈阳110004

出　　处：《系统仿真学报》2007年第23期5406-5410,共5页Journal of System Simulation

基　　金：国家自然科学基金(60674063);辽宁省自然科学基金(20062024)

摘　　要：以RBF神经网络为基础,提出了一种具有自评判自学习能力的短期电力负荷预报方案,该方案包含预报器、评估器、探测器和学习机四个组成部分。预报器用来预测未来时段的电力负荷,评估器用来对预报结果进行评估,探测器用来确定预报的有效步长,学习机用于预报器的自我学习。预报器根据评估结果通过学习机制能够自动适应电力负荷的变化,从而保持一种良好的预报状态。仿真实验表明该方法在电力负荷规律不明确的环境下预报精度比传统方法高。A critic self-learning method based on RBF neural network was introduced to predict the power loads. The system consists of four elements, which are a predictor, an estimator, an explorer and a learning machine. The predictor was used to forecast the future power loads. The estimator was used to evaluate this prediction＇s validity. The explorer was used to determine the predictive step length. And the learning machine was used to keep the predictor self-learning. So the predictor could conform to the power loads by self-learning and be in a good forecasting state. The simulation shows the proposed method has higher forecasting accuracy in irregular power loads cases than the conventional method has.

关 键 词：电力负荷 神经网络 预报 评估 学习 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]