Research On IGBT Based Solar Inverter Reliability And Power Loss Forecast Due To Inverter Failure

The reliability of a solar PV plant is very important for its stable energy transform. As a major part of the PV system, the solar inverter is considered as the weakest link in the system. This dissertation conducted a fundamental research into solar inverter reliability considerations on system level as well as semiconductor device level with electrical, thermal and statistical theories, especially focusing on its function in PV system, failure mechanisms, lifetime assessment and power loss prediction.(1) Taking inverter efficiency characteristics into consideration, we analyzed the match principles and requirements of the solar inverter with a comparable example according to electrical parameters such as voltage, current and the rated power of the solar inverter and PV panels(2) We built the thermal network of the IGBT module based on its power consumption theory and failure mechanisms. We achieved the curve of its transient thermal impedances through finite element analysis. The value of steady state thermal resistance were ensured by referring to the data sheet of the product with relative error of 2.86%.(3) We built a thermal-electrical model of the IGBT module with its thermal impedances and achieved the junction temperature distribution of the chip under power cycling dissipation. The simulation result was validated by comparison with theoretical calculation with a relative low deviation, namely less than 1.36%. With the IGBT lifetime assessment model, we predicted the lifetime of the SKM75GB063D(600V/75A)product.(4) Finally, we presented a new method to predict the power loss due to inverter faults. This method was based on correlation of the historical solar inverters' data. For a day and a week prediction, the accuracy is applicable to the industry if the correlation coefficient between two inverters is higher than 99.9% since the average relative error for a day and a week are below 10% and 3%, respectively. As for an hour's prediction, the prediction accuracy is unreliable for industry.