基于遗传程序设计方法的短期电力负荷预测

Short-term load forecasting based on Genetic Programming

对遗传程序设计算法进行介绍,尝试将其应用到短期电力负荷预测中。先对负荷样本进行常规和滤波处理,消除伪数据,然后运用遗传程序设计的复杂表达能力,把不同日同一时刻的负荷序列作为样本,对未来负荷进行分时短期预测。该遗传程序设计方法无需考虑如温度等因素,随机常数能自动平衡与负荷有关因素的影响,并自动生成相应的数学模型,通过模型可以计算出未来时刻的负荷,有效避免了人为造成的误差,简单可行。经过和时间序列方法的预测结果进行比较,表明该方法效果较好。

The Genetic Programming is introduced and applied to short-term load forecasting in this paper. In order to eliminate the pseudo-data, the load samples were filtered and processed generally first, then the load series of the same time but different days were chosen as the training samples. And by means of the complex expressive capacity of Genetic Programming, the future short-term load of different time point was forecasted according to time-sharing. It doesn＇ t need to take the temperature into account, the random constants would balance the influence of related factors effectively and automatically. And the corresponding forecasting models can be created automatically as well, so the artificial errors can be avoided effectively. Using the models created, the future load value of each point can be calculated. The method of Genetic Programming is simple and feasible in short-term load forecasting. According to comparison with the results forecasted by means of time series, it proves that the method of Genetic Programming in short - term load forecasting is better.