摘要
建立科学合理的中长期电力需求预测方法,是电力产业科学规划建设的前提。构建了基于高斯过程(GPR)和粒子群(PSO)的混合电力需求预测模型。采用PSO算法对协方差函数中的参数进行优化,将修正后的参数作为初始值在GPR模型中进行电力需求方面的培训。在贝叶斯框架下,对协方差函数中的参数再次进行优化。用训练好的GPR模型进行电力需求预测,并将结果与自回归积分移动平均模型和指数平滑模型进行比较。验证结果表明,基于高斯过程(GPR)和粒子群(PSO)的混合电力需求预测模型具有很好的稳定性和更高的预测精度。
Establishing a scientific and reasonable method for predicting medium and long-term power demand is a prerequisite for the scientific planning and construction of the power industry. This paper constructs a hybrid power demand forecasting model based on Gaussian process( GPR) and particle swarm optimization( PSO). This paper adopts the PSO algorithm to optimize the parameters in the covariance function,and uses the modified parameters as initial values to train the power requirements in the GPR model. Under the Bayesian framework,the parameters in the covariance function are optimized once again. Finally,the trained GPR model is used to predict the power demand,and the results are compared with the autoregressive integral moving average model and the exponential smoothing model. The verification results show that the hybrid power demand forecasting model based on Gaussian process( GPR) and particle swarm optimization(PSO) has good stability and higher prediction accuracy.
作者
黄元生
胡建军
蔡雅倩
Huang Yuansheng;Hu Jianjun;Cai Yaqian(North China Electric Power University,Baoding 071003,Hebei,China)
出处
《电测与仪表》
北大核心
2020年第2期74-80,共7页
Electrical Measurement & Instrumentation
基金
国家自然科学基金资助项目(570729264)
江苏省自然科学基金资助项目(sk20177054)
广东省科技厅科技项目(2014bf70048)
关键词
高斯过程回归
粒子群算法
电力需求预测
神经网络训练
Gaussian process regression
particle swarm optimization algorithm
power demand forecasting
neural network training
