粒子群-神经网络混合算法在短期电价预测中的应用

Application of particle group and neural network hybrid algorithm for short term electric power price forecast

为了提高电力市场环境下的电价预测精度,在研究短期电价预测中采用了粒子群和反向传播神经网络相结合的混合算法,先利用粒子群算法确定初值,再采用神经网络完成给定精度的学习。对我国四川电网电价进行预测的结果表明,粒子群优化的神经网络算法收敛速度快于神经网络算法,预报精度显著提高,平均百分比误差可控制在2%以内,平均绝对误差最大值为1.87$/MWh。该算法可有效用于电力系统的短期电价预测。

In order to improve the forecast precision of the power market, the mixed algorithm of particle swarm and back propagation network is used for the short term price forecast. Firstly the particle swarm algorithm is used to determine the initial values, the network is used for the given accuracy, then the electric power price in Si Chuan power market is forecasted. The results indicate that the mixed algorithm has a quicker convergence rate than back propagation algorithm, and it has high predicting precision, the average percentage error is not more tan 2 %, the largest average absolute error is 1.87 ￥MWh. This algorithm can be used efficaciously for short term electric price forecast of power market.