Application Of Statistical Model In Modeling Of Solar Cell Quality Data

Pico satellite power supply system consists of solar cells and lithium batteries,the Ga In P/Ga As/Ge structure of the solar cell is the most efficient.Pico satellite energy supply system is a key component of satellite.and its performance study is one of the important topics at present,for example,whether the performance of a lithium battery has an impact on the performance of a solar cell,how to analyze the performance change of Ga In P/Ga As/Ge solar cells,for the monitoring data obtained under complex conditions how to find the appropriate performance evaluation index,and how to deal with the model of non-adaptive problems.In order to solve these problems,the first is the data preprocessing.By analyzing the change characteristics of output power and temperature of the delayed telemetry,we select the peak value of the solar cells output power during each operation period of the satellite as the performance evaluation amount.Secondly,observed by the data shows that the output power of solar cells there has been a leap decline,and the reason of this jump-drop is verified by hypothesis testing.The significance of this difference before and after the decline was measured by U statistic.Lithium battery completely invalid point in time as a mutation point to divide the data in different situations and carried on the staged modeling analysis to it.A method for modeling the performance degradation of Ga In P/Ga As/Ge solar cells in a complex and ever-changing space environment was proposed.The ARIMA model was constructed to characterize the performance of solar cells.Aiming at the long-term prediction performance drop of ARIMA model,we established support vector regression(SVR)model,finally,concept drift is introduced into the ARIMA model,and the model is not adapted to the problem has been improved.The Spearman rank correlation coefficient was introduced to test whether the output power of solar cells was affected by the lithium battery.The concept drift detection mechanism is introduced in terms of model adaptability.By analyzing the in-orbit monitoring data,a reasonable method for calculating the solar cell performance degradation factor was obtained,and the impact of lithium batteries were compensated analysis.The results show that the output power of solar cells and lithium batteries have a strong working relationship.The predictive value of the performance parameter of the segmented ARIMA model is basically the same as the actual value,and the relative error of the prediction results of the piecewise model reached 3.3% and 0.71%,respectively.The predictive effect of SVR model is similar to that of the segmented ARIMA model,the relative error of the prediction results also reaching 0.89%,however,the ability to predict backward is superior to the ARIMA model.The concept drift detection mechanism can significantly improve the model of non-adaptive problems.