基于ADAM优化卷积神经网络的火电厂大数据平台预警模型测试与应用

为了提高火电厂大数据平台的生产维护安全化、监控管理精细化、经济效益持续化，提出了大数据平台内开发故障诊断预警系统，采用基于自适应力矩估计(adaptive moment estimation, Adam)算法优化二维卷积神经网络方法建模技术融入于大数据平台中，并结合大数据平台和专家故障预警诊断功能进行测试与应用。首先对故障预警模型进行数理建模及模型训练优化，直至满足模型功能要求，实施模型算法代码与大数据平台的合库部署上线、满足提前发现系统故障的功能，并结合机理分析对故障系统进行细致化分类，最终发现根本的故障原因。实现了火电厂生产过程中各系统运行特性的全周期监控，在系统或设备发生故障前进行预警并推送异常信息，规范化了模型部署在大数据平台后的测试与实施工作，进一步发现模型缺陷，提高模型准确率。

Test and Application of Early Warning Model of Thermal Power Plant Big Data Platform based on ADAM Optimization Convolutional Neural Network

In order to improve the safety of production and maintenance, refinement of monitoring and management, and sustainable economic benefits of the big data platform of thermal power plants, a fault diagnosis and early warning system is developed within the big data platform, and the modeling technology of optimizing the two-dimensional convolutional neural network method based on ADAM algorithm is integrated into the big data platform, and the big data platform and expert fault warning diagnosis function are combined for testing and application. Firstly, the fault early warning model is mathematically modeled and model training and optimized until the model function requirements are met, the combined library deployment of the model algorithm code and the big data platform is implemented to meet the function of discovering system faults in advance, and the fault system is classified in detail based on mechanism analysis, and finally the root cause of the fault is found. It realizes the full-cycle monitoring of the operating characteristics of each system in the production process of thermal power plants, warns and pushes abnormal information before system or equipment failure, standardizes the test and implementation of the model after deployment on the big data platform, further discovers the model defects, and improves the accuracy of the model.