基于GSABO-BP和Bootstrap的电力负荷区间预测

Power load interval prediction based on GSABO-BP and Bootstrap

针对电力负荷序列波动性强、预测精度低的问题,提出一种基于GSABO-BP模型和Bootstrap的电力负荷区间预测方法。首先提出一种改进的减法优化算法(GSABO),在保留减法优化算法(SABO)良好的收敛性基础上,融合黄金正弦算法(Gold-SA)来提升其搜索能力;然后,利用所提方法对BP神经网络的权值和阈值进行寻优,构建GSABO-BP预测模型,对电力负荷进行点预测;最后,采用Bootstrap方法分析电力负荷功率预测误差,结合点预测结果确定输出结果的波动区间。经仿真测试,所提方法寻优能力强、鲁棒性好;且相比于其他算法,该方法的预测精度、区间可靠性、区间宽度等均有显著提升。综合点预测和区间预测效果可知,二者结合有助于准确评估预测误差,具有较高的实际应用价值。

In allusion to the issue of high volatility in electricity load sequences and low predictive accuracy,an electricity load interval prediction method based on the GSABO-BP(golden sine algorithm-subtraction average based optimizer-back propagation)model and Bootstrap is proposed.An improved GSABO algorithm is proposed,which can integrate the golden sine algorithm(Gold-SA)to enhance its search ability while preserving the good convergence of the subtraction average based optimizer(SABO).The proposed method is used to optimize the weights and thresholds of the BP neural network,constructing a GSABO-BP prediction model for point prediction of power load.The Bootstrap method is used to analyze the error of power load prediction,and determine the fluctuation range of the output results by combining with the point prediction results.The simulation testing shows that the proposed method has strong optimization ability and good robustness.In comparison with other algorithms,this method has significantly improved in prediction accuracy,interval reliability,interval width,etc.The combination of point prediction and interval prediction results can help accurately evaluate prediction errors and has high practical application value.