运用遗传规划法进行电力系统中长期负荷预测

Mid-long term load forecasting in power system by genetic programming

运用遗传规划法进行中长期负荷预测,将预测模型作为遗传规划中的个体,根据"优胜劣汰"的原则,运用复制、变异和交叉三个主要的遗传算子操作,搜索最优预测模型。它根据历史样本数据自动生成负荷预测模型,包括模型的函数形式以及模型参数。同时在模型的实现上对遗传个体进行Read线性编码,用十进制编码来代替个体树,通过对编码的操作来实现各种遗传操作,极大地提高了程序运算效率。通过对某地的年用电量进行预测,同时与传统的多元线性回归模型进行比较,结果表明,GP模型可以显著提高预测精度。

Genetic programming(GP) is introduced to solve mid-long term load forecasting. Forecasting models are taken as the individuals of GP, which searches the optimal forecasting model by reproduction, mutation and crossover according to the rule of good kept and bad eliminated. It can create automatically load forecasting model including the function form and the numerical coefficients. To realize the model, Read linear code is used to code genetic individually. The tree-like individuals are replaced by decimal codes and the genetic operation is implemented by the operation of the linear code, which improve the computing efficiency. The results of annual forecasting of electric power for some region and comparison with the convential regression model show that GP model can improve forecasting precision obviously.