The Diagnose Method Of State Of Healthfor Satellite Battery

As the primary power supplier of the satellites, rechargeable battery plays an important role for the safety of the satellites. Study on the algorithm of the battery State-of-Health(SOH) diagnostic is capable to provide approaches to estimate the current battery health condition, predict the battery performance degradation, provide useful data for the battery related maintenance and prevent the battery related failure, which causes catastrophic loss.This dissertation studies deeply on the battery SOH diagnose methods, major work can be summarized as:A Support Vector Regression(SVR-PF) based Lithium-ion Battery(LIB) SOH diagnose method is proposed. First, the battery SOH parameters are defined according to the satellite LIB impedance and capacity degradation mechanism, the battery SOH parameters identification models are established, and the parameters identification method is proposed. The proposed method improves the degeneracy phenomenon of the standard PF, and the accuracy of the estimation is improved. Second, to improve the shortcoming of the existing studies that the Remaining Useful Life(RUL) density is not updated during the prediction process, the RUL prediction model based on SVR-PF is established by using the identified SOH parameters, and the predicted impedance is treated as the measurement of the model, and the SOH estimation results are fully applied in the RUL prediction, the RUL probability density can be also updated. Therefore, the prediction accuracy is improved. The proposed method provides a novel SOH diagnose method for the next generation satellite.To provide an alternate diagnose method when the satellite battery impedance is hard to measure, a sub-optimized satellite LIB SOH diagnostic method is proposed. The capacity degradation model is established according to the capacity degradation mechanism, and the capacity degradation parameters are identified based on the SVR-PF, the RUL prediction is also based on the SVR-PF. The proposed method is also capable to update the RUL probability density. The performance of the proposed method is tested by the measured satellite LIB data, which shows the method has accurate prediction results.A combined satellite LIB RUL prediction method based on Dempster-Shafer theory(DS theory) and SVR-PF is proposed to provide a solution of the limited sample based RUL prediction. First, the application of DS theory in LIB RUL prediction is introduced, then the prediction approaches and the combined prediction equations are established, finally, the RUL prediction model based on DS theory and SVR-PF is established to guarantee the accurate prediction when lacking useful data.A satellite battery SOH diagnostic method based on improved Artificial Neural Network(ANN) and SVR is proposed to provide a solution for the Nickel-Hydrogen battery, which model is hard to establish. First, for the measured pressure data, which is time series, an imporved Wavelet Transform NN(WTNN) is proposed, the GA is improved by the Akaike Information Criterion(AIC), the WTNN parameters are optimized by the improved GA, therefore both the prediction accuracy and the model complexity can be optimized, and the prediction accuracy is improved. Second, for the quantity of battery measured voltage data is large, an improved SVR is proposed, the single kernel SVR is improved by multi-kernel SVR, therefore the advantage of every kernel can be involved and the estimation accuracy is increased, and the parameters of the multi-kernel SVR are optimized by the Quantum Particle Swarm Algorithm(QPSO). The diversity of the particles is reduced after several iterations, therefore the diversity factor is introduced to enhance the particles diversity, and the optimization results are also improved. The Nickel-Hydrogen battery voltage data is used to test the performance of the proposed method.