Study On The Optimal Lead-acid Battery Charging With Neural Network Prediction And Variable Structure Fuzzy Control

The charging of lead-acid battery, the most widely accepted secondary battery, constrained to technological limitation, mainly adopted traditional charging methods such as constant voltage, constant voltage with current limitation, constant current and etc. Those methods may not follow the internal physical and chemical laws within the battery, leading to serious overcharge and gas-generation, and resulting in low-efficiency, time-consumming and prone-to-damage of the charging operation. Therefore, new technique that elevates the efficiency, shortens the time, and prolongs the battery life, now becomes the focus of the fast developing power electronic and control engineering. The author's project "New-type Intelligent Charging Device", funded by former Ministry of Coal Industry, is a symbol of this focus. With a study in depth on the charging process the author reached the following understanding:The ideal charging current at a specific moment is dependent on the amount of active material onthe battery plates and the reactive environment, that is to say, the charging current should followa J.A.Mas proposed acceptable curve or optimal curve;The polarization is unavoidable but controllable in charging and discharging. The elimination ofpolarization by short discharge is the key to optimal charging. Because of the technical complexity such as parameter discreteness and nonlinearity, the confirmationof a theoretically existing acceptable charging current curve is quite difficult and the research and implementation of optimal charging technique are still in their first steps. The author's attempt in this dissertation is by utilizing the technical advances and newest fruits in power electronics, pattern recgnition and parameter identification, artificial intelligence, and DSP application to study the problem of multi-aim optimization, striving for a high-efficiency, fast and damage-free approach of lead-acid battery charging. In writing the dissertation, the author has done his work on following aspects:An in-depth study on the electrochemical mechanism of lead-acid battery. Based on a number of charging and discharging experiments, a thought is put forward that the polarization may be detected according to the terminal voltages at charging, terminal voltage at the short pause, and the SOC (state of charge);The depolarization by momentary discharge according to the reference of optimal curve. A realtime depolarization strategy is put forward that takes polarization voltage and SOC as inputs, and the width of depolarization pulse, revised by solution temperature, as output. The adoption of a fuzzy neural network control strategy that features satisfactory ability of self-learning and nonlinear approaching. The charging current traces dynamically the battery-dependent acceptable curves to maintain the charging process under optimal status. A new-type Buck-Boost topology that uses a two-unit IPM as main switching devices to perform charger and depolarizer;The software/hardware implementation of a DSP(TMS320F240) based system controller that performs high-speed data-acquisition, event management, control algorithm and output control. The experiment results indicated that by application of new control strategy, the charging efficiency was raised to about 90%, the charging period was reduced to within 2 hours, and there was no apparent electrolyte temperature-rise, which means high efficiency, fast and damage-free charge is realized. Other problems such as the charging in series battery groups were also studied in the dissertation...