基于混合模型驱动的电力物联系统数据质量优化方法

The stability and reliability of industrial internet of things data play an important guiding role in the healthy and scientific operation of the new industrial chain. Currently, the potential physical model and the data-driven model used to address the problem are imperfect. Each of them encounters its limits in industrial application. The optimization method of the power industrial internet of things data based on a hybrid model is proposed in this project. Such as LSSVM, Brent’s method, Fuzzy Logic, Adaptive Kalman Filtering, RNN, and Broad Learning are considered to solve the problem. The main content is aiming to complete the quantitative performance of characterization theory and the physical model, improve the discrimination and recovery methods for the timing-exception in communication, and establish a neural network model merged with broad learning reducing the residuals. The effectiveness of this method will be confirmed by the comparison of the data derived from several different IOT systems, in the power industry scenario. Eventually, the results could help establish the data-quality standard of the IOT system, make the platform data displayed continuously and accurately, more importantly, make a contribution to the smart-grid and the reformation of electric power system.

保障物联系统数据质量的稳定可靠性，对新型产业链的健康、科学化运行具有重要指导作用。目前潜在解决该问题的物理模型与数据驱动模型皆有不完善之处，且单独采用信息系统物理模型或神经网络模型的方法都存在局限性。本项目提出并研究了基于混合模型驱动的电力物联系统数据质量优化方法。以最小二乘支持向量机、Brent’s整体优化算法、模糊逻辑、动态自适应Kalman滤波、循环神经网络、宽度学习等为解决问题的技术手段，进一步拓展完善物联系统性能定量表征理论与物理模型、提升通讯数据异常尤其是时序异常的判别及恢复方法，建立基于宽度学习改进的神经网络模型。并在电力工业场景下，通过多个不同来源物联系统展示数据的对比，论证该方法的有效性。最终建立物联系统数据质量评价标准，使平台数据展示连续、准确。研究成果对于推动国家智能电网信息化建设以及中央电力改革深入能源服务市场具有重要意义。

保障物联系统数据质量的稳定可靠性，对新型产业链的健康、科学化运行具有重要指导作用。本项目研究前潜在解决该问题的物理模型与数据驱动模型皆有不完善之处，且单独采用信息系统物理模型或神经网络模型的方法都存在局限性。本项目完成了基于混合模型驱动的电力物联系统数据质量优化方法。以无监督学习、支持向量机、模糊神经网络、自适应Kalman滤波、宽度学习等为解决问题的技术手段，进一步拓展并完善了物联系统性能定量表征理论与物理模型、提升通讯数据异常尤其是时序异常的判别及恢复方法；并在电力工业场景下，通过多个不同来源物联系统展示数据的对比，论证该方法的有效性。最终建立物联系统数据质量评价标准，使平台数据展示连续、准确。研究已经应用于有关电气综合系统平台，取得了发明专利等成果，并得到了政府的科技创业支持。

内容获取失败，请点击重试