A Neural Network Based Method for Real-Time Measurement of Capacity and SOC of Vanadium Redox Flow Battery

Abstract

The state of charge (SOC) of Vanadium Redox Flow Battery (VRFB) plays an important role in the operation and control of the Battery system. The value of SOC can be defined as the ratio of the remaining capacity to the rated capacity of the battery. Current measurement of SOC of VRFB is limited to one certain charge-discharge circulation so the rated capacity is known and can be regarded as a constant. However, during long time cycling, the capacity of VRFB will reduce gradually to a relatively low level so that the capacity of the battery cannot be seen as the constant value of rated capacity, which makes it difficult to measure the SOC accurately in real-time operation. This work presents a neural network based method of measuring the capacity and SOC for VRFB in real time. The capacity is firstly classified into three levels in terms of the loss degree by a Probabilistic Neural Network (PNN) using the values of the voltage per second and the average power of the cell stack in any period of the circulation. The values of capacity which fall within different levels are then given by different Back Propagation Neural Networks (BPNN) trained by the battery operation values in corresponding level. Finally, the SOC can be obtained by the calculated capacity. All the networks are validated by experimental data and the results indicate that the method is suitable for the measurement of VRFB capacity and SOC in the practical application.