机组启动态温度预测模型自适应测试技术

基于机器学习的温度预测模型的预测准确度依赖于故障样本,而发电厂的故障样本较少,不同电厂的故障样本往往不能通用.提出故障样本的自适应生成技术和基于灵敏度的测试技术,即通过自适应生成故障样本,并使用故障样本对已部署的温度预测算法应用效果进行灵敏度测试的方法.首先统计温度测点开关量信号测值的历史数据,然后综合历史统计、温度测...

Adaptive Test Technology of Temperature Prediction of Units in Start-up State

The prediction accuracy of temperature prediction model based on machine learning depends on the number of fault samples.The number of unit fault samples is limited.And fault samples from different pumped store power station are often not interchangeable.The generation technology and sensitivity test technology of fault samples are proposed in this paper.It provides fault samples and test technology for intelligent trend judgment algorithm.The test can completely cover the trend characteristics under fault conditions.The limitation of relying on the fault samples is overcome.The quantitative evaluation for the implementation effect of intelligent technology is achieved.Intelligent technology parameter adjustment,selection of appropriate algorithms and implementation means are provided with index support.The verification of the implementation effect of the intelligent technology is realized in advance to avoid losses and uncertain influences caused by the verification through actual projects.