摘要
电力系统短期负荷作为电网运行的重要指标,因其具有非线性和时序性而难以精准预测。针对传统高斯过程(Gaussianprocess,GP)对初始值依赖性强、预测易陷入局部最优解的问题,提出了一种改进的高斯过程短期负荷预测模型。在估算GP参数时,使用分布估计算法(Estimation of distribution algorithm,EDA)描述GP的真实解分布。首先对种群维度进行降维处理,优化高斯分布采样;然后添加遗传操作加速种群的收敛,并构建自适应混合机制模型;根据适应值函数的变化率对多代群体混合,更新种群信息,经多次迭代直到获得最优超参数。改进的EDA不仅可以包含全局解的信息,而且经此算法优化的GP在搜索能力方面也具有一定优势。实验结果表明,基于混合机制EDA的GP算法能够针对性地避免算法陷入局部最优解、增强模型的稳定性和预测的精准性,且能够准确反映电力负荷的变化趋势。
As an important index of power system operation,short-term load of power system is difficult to predict precisely because of its non-linearity and time sequence.An improved short-term load forecasting model based on Gaussian process is proposed to solve the problem that the traditional Gaussian process(GP)is highly dependent on the initial value and easy to fall into the local optimal solution.When estimating GP parameters,estimation of distribution algorithm(EDA)is used to describe the real distribution of GP solutions.Firstly,the population dimension is reduced to optimize the Gaussian distribution sampling,then the genetic operation was added to accelerate the convergence of the population,and an adaptive mixing mechanism model was constructed,which mixed the multi-generation population according to the rate of change of the fitness function,updated the population information,and iterated many times until the optimal super parameters were obtained.The improved EDA can not only contain the information of the global solution,but also the optimized GP has some advantages in the search ability.The experimental results show that GP based on EDA can keep the algorithm from falling into local optimal solution,enhance the stability and accuracy of the model,and reflect the changing trend of power load accurately.
作者
高宏宇
徐宁
GAO Hongyu;XU Ning(Huaian Power Supply Company,State Grid Jiangsu Electric Power Co.,Ltd.,Huaian 211600,China;College of Internet of Things Engineering,Hohai University,Changzhou 213000,China)
出处
《电力科学与工程》
2023年第12期51-59,共9页
Electric Power Science and Engineering
基金
中国高校基本科研业务费专项资金资助项目(B210202083)。
关键词
电力负荷预测
高斯过程
分布估计算法
自适应
power load forecasting
Gaussian process
estimation of distribution algorithm
self-adaption
