Research On Fiber State Detection Method Based On Multi-source Information Fusion

Power fiber,as an important part of power grid,undertakes the important task of power communication.However,the route laying is complex and changeable,which is vulnerable to natural disasters.Among them,the icing phenomenon formed by condensation of water vapor or water droplets and the dancing phenomenon formed under the influence of wind are the most common and harmful.There are many potential problems due to the aging of the electric fiber with long running time.Therefore,it is necessary to master the running status of power optical fibers in real time and obtain the line running status information in time,which provides important reference for line maintenance personnel.Existing state of power optical fiber line recognition methods usually use a single optical fiber state parameter identification,optical fiber state recognition based on single parameter easy failure problem such as false positives,fault type wrongly,it is difficult to accurately identify line state,aiming at this problem,multi-source information fusion method is adopted to analyze state of electric power optical fiber.The main contents of the paper include the following two points:(1)Research on optical fiber ice-coated dance state identification method based on fiber temperature,vibration and strain parameters.The temperature,vibration and strain parameters of power fiber are measured by φ-OTDR(phase-sensitive optical time domain reflectometer)and BOTDR(Brilouin Optical Time domain reflectometer).Firstly,the data are normalized and denoised,and the feature vectors after processing are extracted.The fusion vector is obtained by fusion of feature vectors through correlation fusion.By fusion vector training ELM(extreme learning machine)algorithm model,training after the ELM algorithm for optical fiber ice and wave state recognition rate at 94.44%,given the initial weights and bias of ELM is random,adopts the LSO(algorithm)of the optimize the ELM model,improve the identification accuracy of the dynamic state electric power optical fiber cladding ice dance,The recognition accuracy is improved to 98.89%,and the recognition effect is good.(2)Research on power fiber health state evaluation model based on fuzzy comprehensive evaluation.Selection of electric power optical fiber temperature,strain,vibration,wind speed,humidity and light power parameter for assessment of the health status of electric power optical fiber properties,but considering the several properties may exist between the redundant attributes,using rough set theory to the attribute reduction,select after the reduction of attributes: temperature,strain and optical power data of electric power optical fiber properties.The evaluation matrix of attribute fusion is established.AHP(analytic hierarchy process)to obtain the subjective weight,using CRITIC(inter-layer relevance importance criteria)method to obtain the objective weight,and then use the addition method to combine the subjective weight and objective weight,improve the reliability of weight;The weighted average operator is used to calculate the health status score of the power fiber.According to the grade score table,the current health status of the power fiber is obtained,and the real-time operation status of the power fiber line is evaluated and analyzed.