1、 翻译部分： 英文原文 Research upon the High-capacity Lead-Acid BatteryCharge Model Kun Yang and GuangyaoOuYang Department ofPower Engineering Naval University ofEngineering Wuhan, Hubei Province, China Ping Zhang and Jianpingzhang Department of Power Engineering Naval University of Engineering Wuhan, Hubei
2、Province, China Abstract - The simulation of the submarine battery charge process is an important part of the submarine battery modeling, which is especially a firm foundation for the further optimization usage of the modules of the whole submarine power system. In this paper, the modeling strategy
3、 of the submarine battery charge process is discussed in detail firstly, in succession an integrated algorithm in which both the improved back propagation neural network method and the linear simulation technology are used is brought forward for the building of the battery charge model, finally the
4、outputs of the simulation experimentation show us a comparatively high precision. Index Terms-： Neural network； Linear simulation； Battery； Charge model. I. INTRODUCTION Battery is the only power source for the conventional submarine underwater propulsion, so the submarine has to move up to the snor
5、kel state to charge for the battery when the energy is exhausted, however the adoption of the snorkel sailing state throw the submarine into a fatal danger, because the snorkel state make the submarine easier spied by its enemy for its diversified increscent physical fields, such as the infrared rad
6、iation navigational wake etc., thus the research upon the submarine battery Charge process seems to be very important. Not much the same as the civilian battery, the submarine battery works under a very complex and execrable situations, this leads to that the charge initial states of the battery dif
7、fer in thousands ways, so the battery charge model should be built based on diversified initial state. On account of that the voltage of the battery can be affected by many factors, such as the SOC (SOC: StateOf Charge) battery polarization delamination former discharge rate etc., moreover the facto
8、rs above are mutual influential； it is hard to decouple for them. The conventional electrochemistry mechanism modelelectrochemistry experiential formula model and the equivalent circuit model considers that there must be some formula relations among the main factors of the battery 123, actually the
9、inner reaction of the battery is a very complex and highly nonlinear process, so the hypothesis seems not conform to the fact, so by farther is not an effective mathematic model for the battery. Further battery model aiming at engineering application calls for that the model not only can reflect the
10、 battery dynamic process well, but also be simple and convenient for use. In view of the electrochemistry mechanism complexity of the battery, an integrated algorithm in which both the improvedBP(BP: Back-Propagation) neural network method and the linear simulation technology are used is introduced
11、in the paper, basing on the deep analysis on the battery charge test curve data. Finally it is proved that the algorithm has avery famous practicality in the battery charge process simulation. II. THE BUILDING OF THE BATTERY MODEL A. The Analysis on the Battery Charge Process Characteristic Consider
12、ing the submarine batterys life limit and the high expenditure, it is unlikely to do large numbers of tests to obtainplatitudinous data for the simulation, so we have to think of ways to simulate for it only through the limited several battery charge test curves. Fig. 1 shows us several typical prac
13、tical charge process test curves under different former discharge rate in a certain charge period (for convenience, some of the data is normalized.) From Fig. 1, it is easy to know that the battery charge process is a highly nonlinear process and the curve trend of the several middle stages of the c
14、harge process differ a lot with the first stage and the last stage, which can be concretely described as follows: The charge times of the middle stages are comparatively short, the data points are comparatively centralized and the curves trend changes acuter. The neural network algorithm is once use
15、d to try to simulate the whole five stages charge process directly, however the net is so hard to be convergent, even along with the introduction of the improved algorithm making the net performance threshold met, the output curve cannot reflect the charge process correctly all the same, even worse
16、the data is intercrossed. This may be because that the object performance error of the neural network algorithm is global, so even the global error is met, it may not indicate that each local part of the curve has been expressed correctly. In a word, even the neural network algorithm provides us a n
17、ew method for the battery dynamic process modeling, considering the characteristic of the battery charge curve, some other technologies should be supplement into the modeling course. Fig. 1 Battery charge process test curve B. Modelingfor the Battery Five Stages Charge Processes Considering the high
18、ly nonlinear characteristic of the five stages charge process test curve that is showed in Fig. I, it is conceived to simulate each stage of the five separately. I) The Simulation of the First Stage Charge Process It is well-known that both the nonlinear system and the uncertainty system can be desc
19、ribed commendably by the ANN (ANN : Artificial Neural Network) algorithm because of its excellent performance in parallel processing and self-learning, therefore the ANN algorithm provides us a feasibleresolvent for the modeling of the first stage of the battery dynamic charge process4. Besides it h
20、as been proved that the three-layer feed forward BP neural network can be trained to approximate any multi-input and multi-output real function (with a finite number of discontinuities) arbitrarily well. It is obvious that the first stage charge process can be easily simulated through the ANN techno
21、logy. In this paper, a BP neural network model is to be constructed to simulate the first stage of the battery charge process under either former discharge rate, firstly the Neural network principium sketch map is showed as Fig. 2 below: Fig. 2 The neural network principium sketch map When the Sigmo
22、id function is chosed as the neuron transfer function, namely = 11+，considering the weights updating formula: Updating Increment= (Learning Rate) x (Teacher signal-NeuronOutput) x (Sigmoid Function Differential Value) x (Neuron Output) (1) Actually from the differential of the Sigmoid function as fo
23、llows: = 1 + 2 = 1 + 2 1 + 1 = 1 (2) It is obvious that the output of the neuron always ranges from 0 to 1. when the approach 0 or I, the Updating increment becomes smaller, thus the net stability boosts up, however virtually the huge frequency happening of this situation often leads to an even slow
24、er learning speed. Aiming at suchlike problems, large numbers of improved algorithms are put forward by scholars recent years, and the improved algorithm by adding the item of the momentum is adopted in this paper, which is widely used for the moment. Namely when the weights Updating increment is to
25、 be calculated at the time of day t equals n, the weights Updating increment corresponding to the time of day n-I is also considered. The idiographic formula please consults the (3) below: = 11+ 1(3) There into1is the error square sum of the output layer until the time of day n-I, 1 is the learning rate, andthe is the momentum constant. If the current correction direction (the first
