基于改进PSO优化RNN的短期电力负荷预测模型  

Short-term electric load forecasting model based on improved PSO optimized RNN

作　　者：程换新[1] 黄震 Cheng Huanxin;Huang Zhen(College of Automation and Electrical Engineering,Qingdao University of Science and Technology,Qingdao 266061,China)

机构地区：[1]青岛科技大学自动化与电子工程学院

出　　处：《电子测量技术》2019年第20期94-98,共5页Electronic Measurement Technology

摘　　要：随着热电厂电力负荷的数据量越来越大,传统的热电厂电力负荷预测方法难以应付巨大的数据量和数据的随机性。为了合理的进行电力系统的规划和优化运行,提出了基于改进粒子群算法优化循环神经网络的短期电力负荷预测研究。基于深度学习中循环神经网络善于处理时间序列型数据的特点,对未来短期的电力负荷数据进行预测。然后针对循环神经网络在电力负荷预测中易陷入局部极小和全局搜索能力较弱的缺点,提出采用在粒子群算法搜索过程中引入模拟退火算法概率突变的思想,最后利用改进后的粒子群算法优化循环神经网络的结点权值参数,并利用MATLAB进行仿真。结果证明该模型相较于循环神经网络和未改进粒子群算法优化的循环神经网络预测方法具有更高的预测准确度,可以精确预测负荷的变化,克服了数据量大且随机的难点,有较高的工业应用价值。With the increasing data volume of thermal power plant electrical load,traditional thermal power plant power load forecasting methods are difficult to cope with the huge amount of data and the randomness of data.In order to rationally plan and optimize the operation of power system,a short-term power load forecasting based on improved particle swarm optimization algorithm for cyclic neural network is proposed.Based on the characteristics of cyclic neural network in deep learning,which is good at processing time series data,it predicts short-term power load data in the future.Then,in view of the shortcomings of cyclic neural network in power load forecasting,which is easy to fall into local minimum and weak global search ability,the idea of introducing simulated annealing algorithm probability mutation in particle swarm optimization is proposed.Finally,the improved particle swarm optimization algorithm is used.The node weight parameters of the cyclic neural network are optimized and simulated using MATLAB.The results show that the model has higher prediction accuracy than the cyclic neural network and the improved neural network prediction method optimized by the improved particle swarm optimization algorithm.It can accurately predict the load change and overcome the difficulty of large data volume and randomness.High industrial application value.

关 键 词：循环神经网络 深度学习 短期负荷预测 粒子群算法 模拟退火算法 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]