To solve the problem of too many variable numbers which makes the model complex, a kind of auxiliary variables selection method is established. After that, soft sensing of lead-acid battery capacity is put forward. First, the RReliefF method is adopted to define quantitatively the influence of auxiliary variables. Then, the soft sensing model is built up with all the combination of auxiliary variables with BP neural network. Simulation results show that the soft sensing of battery capacity is established ideal. It provides theoretical feasibility to omit the battery discharge capacity in the process of production inspection process.
Soft Sensing Model Visualization: Fine-tuning Neural Network from What Model Learned