基于深度卷积神经网络的工业循环冷却水系统运行状态预测

Operation state prediction for industrial circulating cooling water system based on deep convolutional neural network

工业循环冷却水系统运行状态预测是维护工业设备正常运行的重要保障。为此,提出一种基于深度卷积神经网络的智能工业循环冷却水系统运行状态预测方法。该方法根据工业循环冷却水水质特征,使用工业循环水在线仪表采集pH值、碱度、硬度、氯离子等实时数据,设计了深度卷积神经网络框架,在每层网络中加入Dropout层,以避免神经网络训练过拟合。利用某电厂的实际水质测量数据对该方法进行有效性验证,结果表明,该方法的预测结果准确率可达94%,且泛化能力良好,优于现有其他方法。

The operation state prediction of industrial circulating cooling water system is an important guarantee to maintain normal operation of industrial equipment.Therefore,an intelligent operation state prediction method for industrial circulating cooling water system based on deep convolutional neural network is proposed.According to the characteristics of water quality of industrial circulating cooling water,a deep convolutional neural network framework is designed using online instrument sampling data of industrial circulating water(including real-time data of pH value,alkalinity,hardness,chloride ion,etc.).A Dropout layer is added to each layer of the network to avoid over-fitting by neural network training.Moreover,the validity of this method is verified by the actual water quality measurement data of a power plant.The results show that,this methodhas a prediction accuracy of 94%,and has good generalization ability,which is better than other existing methods.