摘要
将支持向量机(SVM)理论与网格搜索及交叉验证相结合,应用于梯级水电系统隐随机优化调度中,实现径流不确定条件下的梯级实际优化运行。以系统结构风险最小为SVM训练目标,结合参数分布规律,采用指数划分的网格搜索对模型参数进行优选;将K-fold交叉验证技术引入到SVM训练性能评价中,降低了训练样本随机性对训练模型性能的干扰,提高了模型的泛化能力。建立VC_与MATLAB混合编程平台,对梯级水电系统隐随机优化调度运行进行仿真,结果表明基于采用最优参数SVM的隐随机优化调度在梯级系统发电量和发电过程方面取得了良好成果。
The combination of SVM （Support Vector Machine） theory combined with grid search and cross validation is applied to the implicit stochastic optimal dispatch of cascaded hydropower stations under the condition of uncertain inflow.With the minimum system structure risk taken as the training objective of SVM and the parameter distribution law combined,the exponentially divided grid is searched for determining the optimal model parameters.K-fold cross validation technique is adopted in the evaluation of SVM training performance,which decreases the influence of training sample randomness on the performance of training model and improves its generalization ability.The implicit stochastic optimal dispatch of cascaded hydropower stations is simulated on the hybrid programming platform of VC_ and MATLAB,which shows that,the implicit stochastic optimal dispatch based on SVM with optimized parameters enhances both power generation capacity and process.
出处
《电力自动化设备》
EI
CSCD
北大核心
2014年第3期125-131,共7页
Electric Power Automation Equipment
基金
国家自然科学基金资助项目(51279062
41340022)
中央高校基本科研业务费专项资金资助项目(11QX52)~~
关键词
支持向量机
网格搜索
交叉验证
混合编程
梯级水电系统
优化
水电
support vector machines
grid search
cross validation
hybrid programming
cascaded hydropower stations
optimization
hydroelectric power
