Research On Measurement Anomaly Identification Method Based On Stream Pattern

Energy measurement is widely used in industries and livelihood fields, i.e., hydropower, natural gas, petrochemical and so on, which presents a multi-source, multi-dimensional, multi-modal complex information field characteristics, involved in the fluid quantity, the electricity quantity, mechanical quantity, optical quantity and other parameters. Energy measurement has a great important implication for saving energy, enhancing the competitiveness of enterprises, and improving energy efficiency. Compared with the traditional measurement model, intelligent energy measurement has a higher demand in the efficiency of the measurement, real-time nature of data, automation degree and so on. To study detection and analysis technology of anomaly in the process of energy measurements is an important safeguard to ensure measurement safe, accurate and efficient.This paper, with the background of energy measurement, focuses on a measurement anomaly detection methods based on stream model. With this method, detect measurement stream data anomaly model in real-time and analyze the location of anomaly source of measurement equipment. The main contents of this paper include the following several aspects:First, study processing model for multi-source multi-dimensional measurement stream data. According to the characteristics of measurement stream data, combined with existing sliding window model, we analyzed data processing mode of multi-sensor measurement equipment cluster data and proposed sliding window model for multi-source multi-dimensional measurement stream.Second, study the method of anomaly identification based on stream data. This paper combined measurement detection techniques with the device anomaly identification technology, projects the measurement stream data to subspace, and according to the energy model of the subspace, identifies measurement anomaly. This method can not only test measurement device errors, but also can identify whether measurement device itself has anomaly which leads to errors. Provide a new idea in the accuracy and the efficiency of measurement stream data detection.Then, study the anomaly location method based stream pattern. By establishing anomaly location model to solve the subspace matrix, analyze the subspace matrix to obtain every anomaly sources’ state, which resolves the most likely source of anomaly. Anomaly location technology widely is applied in the field of measurement stream detection, and achieves the abnormal localization of measuring equipment.Last, research the measurement anomaly analysis platform. Based on the measurement anomaly identification and location algorithm, develop measurement detection anomalies identification platform which was applied in the real measurement detection and by the experiment achieved good effect. The overall measurement process is more intelligent and more accurate in measurement results, effectively improves the measurement overall efficiency of detection.