摘要
为有效利用电力资源,改进电力供需结构,建立面向电力负载的短期预测模型。利用层次分析法,对负载预测的影响因素做权重筛选,优化输入参数。通过主成分分析法对样本数据进行线性组合,压缩数据,提高网络泛化能力。引入L-M算法完善反向传播(BP)算法,加快收敛速度。同时结合改进的遗传算法,自适应调整交叉变异概率,对BP神经网络的初始权重进行动态赋值。在真实数据集上的实验结果表明,相较于传统神经网络模型,提出的模型能够加快神经网络的收敛速度,同时提高预测精度,电力负载的实际值与预测值的相对误差小于3%。
For the effective usage of power resources and improving the structure of power supply and demand,a short term forecasting model for electric load is established. The analytic hierarchy process is used to screen the weight of the factor that affects the load forecasting, thus the input parameters are optimized. Principal component analysis method is used to make linear combination of sample data, compress data, and improve the network generalization ability. L-M algorithm is introduced to improve the Back Propagation（BP） algorithm, and speed up the convergence rate. At the same time, combined with the improved genetic algorithm, the crossover mutation probability is adjusted adaptively, and the initial weights of BP neural network are dynamically assigned. Experimental results on real datasets show that compared with the traditional neural network model, the proposed model can speed up the convergence of the neural network and improve the accuracy of prediction. The relative error of power load forecasting between actual value and predicted one is less than 3%.
出处
《计算机工程》
CAS
CSCD
北大核心
2017年第10期277-282,288,共7页
Computer Engineering
基金
北京电力医院一体化运维监控与管理项目
关键词
神经网络
电力系统
负载预测
反向传播算法
自适应遗传算法
neural network
electric power system
load forecasting
Back Propagation （BP） algorithm
adaptive geneticalgorithm
