基于大型水电机组实测数据的LSTM-SVM抬机预测OA
LSTM-SVM-based lifting prediction for large hydropower units using measured data
对于大型成熟水力发电厂,依托机组过渡过程中抬机运行的历史数据开展抬机量预测,能够有效规避数学建模难题,为抬机量预测提供可行思路.文中选取金沙江流域某大型水力发电厂机组过渡过程的大量现场实测数据,构建 LSTM-SVM混合算法模型(长短时记忆网络与支持向量机相结合),通过对数据进行分类、训练、验证与测试,得到优化后的抬机量预测模型,并设计针对性的模型评价指标.将模型预测结果与现场实测数据对比,同时选取 BiLSTM算法、单一LSTM算法开展对比试验,分析所提算法的预测优势.结果表明:LSTM-SVM混合算法预测精度可达 98%以上,具备优异的抬机量预测性能.
For large mature hydropower plants,predicting the lift amount based on massive historical data of unit lifting during transient processes can effectively circumvent mathematical modeling difficul-ties and provide a feasible approach for lift amount prediction.Extensive on-site measurement data from unit transient processes in a large hydropower plant on the Jinsha River Basin was selected,and an LSTM-SVM hybrid algorithm model(combining long short-term memory network and support vector machine)was constructed.Through data classification,training,validation,and testing,an optimized lift amount prediction model was obtained,and targeted model evaluation indicators were designed.The prediction results of the model were compared with on-site measured data,while comparative expe-riments using the BiLSTMalgorithm and the single LSTMalgorithm were conducted to analyze the pre-dictive advantages of the proposed algorithm.Experimental results show that the LSTM-SVMhybrid al-gorithm achieves a prediction accuracy of over 98%,demonstrating excellent lift amount prediction per-formance.
蒋树;钱晶;李佳;曾云;鲍友洪
昆明理工大学冶金与能源工程学院,云南 昆明 650032||白鹤滩水力发电厂,四川 凉山 615499昆明理工大学冶金与能源工程学院,云南 昆明 650032昆明理工大学冶金与能源工程学院,云南 昆明 650032||西安许继电力电子技术有限公司,陕西 西安 710000昆明理工大学冶金与能源工程学院,云南 昆明 650032白鹤滩水力发电厂,四川 凉山 615499
农业科技
大型水电机组抬机LSTM-SVM预测现场数据人工智能
large-scale hydropower unitliftingLSTM-SVMpredictionon-site measured dataartificial intelligence
《排灌机械工程学报》 2026 (5)
479-487,9
国家自然科学基金资助项目(52269020,52079059)
评论