灰色模型的改进及其在电厂实时测量数据预处理中的应用

Improvement of grey model and application to preprocessing of real-time measurement data in power plants

针对火电厂性能在线监控平台测点数据随机影响因素较多、信息量少、实时性要求高的特点提出了一种基于改进的灰色模型动态数据预测方法。首先使用弱化算子对原始测点数据进行预处理,然后通过对灰色模型背景值重构和初值选择建立改进的灰色模型,用拟合度来检验其预测结果的相对误差精度。对火电厂流量和温度实时动态测量数据集进行仿真计算表明,与回归分析算法、基于LM算法的BP神经网络算法和已有灰色模型算法相比,对于流量和温度实测数据预测,改进的灰色模型算法都有效地提高了预测结果的精度和数据预测效率。

Since the measured points data of performance monitoring platform on line in power plants are influenced by many random factors, and have less information and the real-time requirement, an improved Grey Model （GM） was proposed. Firstly, the proposed method used weakening buffer operator was used to weaken the original measuring point data. Secondly, the improved model was established by modifying the initial value of the background values and choosing the initial data. Finally, predictions relative error accuracy was verified by degree of fitting. The actual data sets of flow and temperature were simulated. The simulation results show that, compared with regression analysis algorithm, BP neural network based on Levenberg-Marquardt （LM） algorithm and grey model algorithm, the improved GM algorithms increase the accuracy of prediction results and data prediction efficiency in real-time data forecasting of flow and temperature.