CAJ | 학술논문

为提高抽水蓄能调节系统仿真中水泵水轮机模型精度，提出了一种集成PSO＿BP神经网络模型来描述水泵水轮机全特性。首先利用改进Suter变换对水泵水轮机全特性进行处理得到样本数据，然后采用PSO算法优化BP神经网络的初始权值和阈值，反复训练出若干个PSO＿BP神经网络，最后将单个PSO＿BP网络作为自适应Boosting集成算法的弱学习器，最终构建出水泵水轮机的集成神经网络模型。计算结果表明，与单个BP网络模型相比，该模型具有更好的拟合精度及泛化性能，为进一步研究抽水蓄能调节系统性能奠定了基础。
위제고추수축능조절계통방진중수빙수륜궤모형정도，제출료일충집성PSO＿BP신경망락모형래묘술수빙수륜궤전특성。수선이용개진Suter변환대수빙수륜궤전특성진행처리득도양본수거，연후채용PSO산법우화BP신경망락적초시권치화역치，반복훈련출약간개PSO＿BP신경망락，최후장단개PSO＿BP망락작위자괄응Boosting집성산법적약학습기，최종구건출수빙수륜궤적집성신경망락모형。계산결과표명，여단개BP망락모형상비，해모형구유경호적의합정도급범화성능，위진일보연구추수축능조절계통성능전정료기출。
In order to improve the accuracy of pump-turbine model applied in the pumped storage regulation system simulation ,this paper designs a PSO_BP ensemble neural network model to describe pump-turbine characteristics .Firstly ,the pump-turbine charac‐teristics data are processed by improved Suter-transformation to obtain the sample data .Then ,the PSO algorithm is used to optimize the initial weights and thresholds of BP neural networks and train out several PSO_BP neural networks .Finally ,a single PSO_BP network is treated as a weak learner of adaptive boosting ensemble learning algorithm ,and then the ensemble neural network model for pump-turbine characteristics is completed .The calculation results show that the model has better fitting accuracy and generaliza‐tion performance than a single BP neural network model ,and it provides a good foundation for a further study of the pumped storage regulation system .