Research On Modeling And Forecasting HVDC Valve Cooling System Based On Temperature Data

Valve cooling system is one of the complex hydromechatronics equipment systems, the research involves the knowledge of many disciplines, so it’s difficult to getting a reasonable mechanism model, nor can ensure the accuracy of failure forecasting. Traditional methods for failure forecasting of valve cooling system mainly depend on the experience of the staff, it’s low reliability for lack of theoretical guidance. At present, the theory of modeling and method for failure forecasting of valve cooling system are lack of research, has not been an ideal mathematical model. To achieve high accuracy fault forecasting has become a serious problem of the power system.The goal of this paper is to model a high accuracy forecasting model for temperature failure of valve cooling system, for Providing theoretical guidance to modeling valve cooling system failure prediction model through the research on modeling the temperature data of the valve cooling system.First of all, the paper introduces the composition and problems of valve cooling system, and analyzes the characteristics of the temperature variable.meanwhile introduces the research status of domestic and international modeling forecasting theory and method. Secondly, due to the existence of outliers and missing values in the data, using the boxplot and the robust Markov distance method for univariate and multivariate outliers detect; then, taking the average close to the value and similar days at the same time for single missing values (outliers) and continuous missing values (outliers), thirdly, in order to establish the high precision model, using the logarithmic transformation to mitigate the volatility of the data. Modeling valve inlet temperature model by regression analysis. ARMAX-GARCH and VAR approachs, using SARMA-EGARCH.SARMAX-EGARCH and VAR approachs to modeling valve outlet temperature variable, using SARMA-EGARCH. FT-ARMA-EGARCH and WNN methods to modeling valve hall temperature variable, each model has a satisfactory goodness of fit. Finally, the breakpoint test methods to verify the rationality of the model, and through short term, medium short term assessment of the evaluation criteria to ensure the high precision of theforecasting model forecast.By modeling the temperature data of a Domestic converter station in2011, and showing that the forecasting accuracy of the proposed forecasting method is satisfied by enterprise, thereby proves the effectiveness and practicality of modeling theory and methods of this paper.