DC System Insulation Online Monitoring And Battery Failure Prediction

As an important part of the power system, DC system provides power supply forkey loads such as monitoring and protection devices. So the quality of power supplyplays an important role in the security and stability of the power system. Consideringthat DC grounding and battery failure are common faults of DC system, it is meaningfulto online monitor the insulation and predict the battery failure.For frequently occurred DC grounding fault, the online insulation monitoringdevice is designed to monitor the DC system. Based on the analysis of existinginsulation monitoring devices and principles, an improved insulation online monitoringscheme is proposed, which is a combination of bus and branch grounding fault detection.The proposed scheme incorporates the advantages of balanced bridge and switchingbridge methods and takes AC detection into account. The hardware and software of thedevice are designed in detail in this project. A decentralized, unitized multi-controllerdetecting structure is established, which uses the voltage detection unit and feeder busacquisition unit detecting the bus voltage and leakage current. The detected informationis processed and analyzed by host controller to calculate the insulation resistance on busand feeder. Then, grounding fault point can be located, alarmed and displayed. Thesoftware is designed to complete different functions of device based on the STM32series controller platform. In addition, the interference immunity of hardware andsoftware is also studied.The performance of prototype is tested and analyzed by simulated DC groundingin laboratory. It is demonstrated that the accuracy of detecting the insulation resistanceon the bus and feeder is limited in2%. Additionally, functions like fault location,alarming and displaying, fault information storage are tested and proved to meetcorresponding technical specifications. It is shown that the device performance isgreatly improved compared with others.As a backup power of DC system, battery failure seriously endangers theuninterruptible power supply. The gray prediction model GM(1,1) is put forward topredict battery failure. Furthermore, an improved GM(1,1) based on genetic algorithmis constructed by using battery capacity and internal resistance as feature vectors. It isverified that the improved GM(1,1) has higher prediction accuracy and can predict thebattery failure parameters in advance. In this way, the drawbacks of experimental testsare compensated, and battery failure can be accordingly judged.