Valve-regulated sealed lead acid (VRLA) batteries are widely used in various industries because of their small size, explosion-proof, stable voltage, no pollution, light weight, high discharge performance, low maintenance and low price. In the post and telecommunications, electricity, transportation, aerospace, emergency lighting, military communications and many other fields. VRLA battery has become one of the key components of the system, and its safe and reliable operation is directly related to the reliable operation of the entire equipment. However, in the course of use, due to the inability to accurately predict the remaining capacity, the light caused accidents and caused a tragedy. Therefore, an effective battery management system must be established. Accurate and reliable prediction of battery residual capacity becomes the most basic and primary task in battery management systems [1][2]. Reference [7] considers the charge and discharge rate of the battery, the temperature, the aging of the battery and the self-discharge rate, corrects the accumulated error by self-tuning, and uses the relationship between the voltage value and the capacity of the cell obtained by a large number of experiments. The inconsistency is corrected, and the correction formula is shown in equation (4). Among them: Ks is the relationship coefficient, △U is the difference between the lowest voltage of the cell in the battery pack and the average voltage of all the cells: in [8], the initial SOC is obtained by the open circuit voltage method, and then the An-time method is performed. For various compensations, the SOC estimation accuracy is less than 6%. In addition, the An-time method is often used in conjunction with the Kalman method (discussed in detail in the Kalman filter method). Among them: VBO is the open circuit voltage of the battery, Vα is the open circuit voltage when fully charged, and Vb is the open circuit voltage when fully discharged, and the corresponding relationship of the size is slightly different with different battery manufacturers. (2) For online use of VRLA batteries with a capacity of more than 80%, the SOC of the battery cannot be detected online according to the internal resistance (conductance) value; (3) According to the battery conductance value or internal resistance value, the battery performance can be determined to some extent. . Figure 1 Randles battery model In practical applications, the Kalman filter method is usually used in combination with the open circuit voltage method and the ampere-hour method. The basic process is as follows: the voltage on the capacitor in the model is taken as the state of the system. After Kalman estimates the voltage, the battery electromotive force (or open circuit voltage) is obtained by the mathematical relationship in the model. Finally, the relationship between the electromotive force and the SOC is obtained. Out of SOC. The Kalman mathematical form of the battery model is: Where SOC(k) is the SOC value at the current time; ΔSOC(k) is the change value of SOC; v(k) and i(k) are the voltage and current at the current time. Β0, Β1, Β2, Β3 are linear model coefficients obtained by least squares using reference data. In [54], for the nonlinear relationship between battery internal resistance and residual capacity, the online gray GM (1,1) model group method is used to predict the SOC of the hybrid vehicle battery unit. In [55], the SOC state equation is established based on the An-time method, and a robust filtering algorithm is proposed to predict the SOC of the battery. We are manufacturer of Marine Jib Crane in China, if you want to buy Marine Cranes For Sale,Marine Knuckle Boom Crane,Offshore Cranes For Sale please contact us. Marine Jib Crane,Marine Cranes For Sale,Marine Knuckle Boom Crane,Offshore Cranes For Sale JIANGSU OUCO HEAVY INDUSTRY AND TECHNOLOGY CO., LTD , https://www.marinelifting.com
At present, the state of charge SOC (State of Charge) is commonly used at home and abroad to indicate the remaining capacity of the battery. The SOC is an important parameter that directly reflects the sustainable power supply and health of the battery. Since VRLA batteries have different types, uses, and external environments, SOC has many influencing factors, so the methods used for prediction are various, and the battery models used are also different. The general battery modeling methods can be divided into two categories: one is the physical modeling method; the other is the system identification and parameter estimation modeling method [3].
2 physical modeling method to predict SOC
2.1 Discharge Test The discharge test method is recognized as the most reliable SOC estimation method. The battery is continuously discharged to a predetermined SOC zero point according to a current of a certain discharge rate, and the product of the discharge current and time is the remaining capacity.
The discharge test method is mainly used for laboratory calculation of battery pack charging efficiency, verification of SOC estimation accuracy, or for battery maintenance, and is applicable to all batteries. However, this method has two obvious disadvantages: (1) it requires a lot of time and manpower; (2) the ongoing work of the battery has to be interrupted and cannot be predicted online in real time. It can be used for static backup batteries, but for important occasions, this method takes a certain risk, because during the discharge, the system runs without battery backup. Once the main power supply is faulty or the mains is interrupted, the whole system will be paralyzed. Causes immeasurable losses. The literature [4] describes the discharge test method and precautions in detail, but requires a lot of manual operation; the literature [5] uses the dynamic environment monitoring system to realize the remote discharge test management of the battery pack, saving time and efficiency, but the accuracy is very low. Only the performance of the battery pack can be qualitatively judged, and the remaining capacity cannot be accurately estimated.
2.2 The ampere-hour method is the most commonly used method for SOC estimation. The calculation formula is:
(1)
Among them: SOC0 is the state of charge at the start of charge and discharge, CN is the rated capacity, η is the charge and discharge efficiency and is not constant (assuming that the direction of the charge current is positive, the direction of the discharge current is negative), and the SOC is the state of charge at the current time.
The essence of the Anshi method is to regard the battery as a black box. It is considered that the amount of electricity flowing into the battery has a certain proportional relationship with the amount of electricity flowing out of the battery, regardless of the internal structure of the battery and the external electrical characteristics. Therefore, this method is applicable to various types. battery. At the same time, we can see from the formula (1) that the problems existing in the application of the Anshi method are: (1) the calibration of the initial value of the SOC is required; (2) the accurate calculation of the charge and discharge efficiency is required; (3) the current needs to be accurately measured, and the current measurement is not Accuracy will result in SOC calculation error, and there will be accumulated error of current integration in the long run; (4) The error is large in the case of high temperature and high current fluctuation.
Therefore, when the Anshi method is used in practical applications, factors such as charge and discharge rate, temperature, battery aging, and self-discharge rate are generally compensated according to the use environment and conditions.
In [6], the SOC of static backup valve-regulated lead-acid batteries is estimated by the combination of Anshi method, Peukert equation, temperature correction and SOH. The two states are one cycle from battery capacity to capacity. During this period, the measurement battery is converted to calculate the SOH at the standard temperature at the total capacity of the standard current discharge or charge. The SOC calculation accuracy can reach 0.1% or less, and the calculation formula is:
2.3 Density method Density method is mainly used for lead-acid batteries. Since the electrolyte density gradually becomes higher during the charging process, the discharge process is gradually reduced, and the battery capacity and density have a linear relationship. Therefore, the SOC size can be predicted by measuring the density of the electrolyte [9]. Since the density method requires measurement of the electrolyte, it is mainly applied to open-type lead-acid batteries. If a density-capacity sensor with higher precision can be developed, it can be implanted into a sealed battery at the time of production in extremely important cases. The literature [10][11][12] used ultrasonic sensors, low-energy γ-rays, lead-acid battery capacity sensors to measure the electrolyte density of lead-acid batteries, and the literature [11] used fuzzy neural networks to predict the density. Preferably, none of them gives a definite relationship between the electrolyte and the SOC.
2.4 Open circuit voltage Open circuit voltage (Open Circuit Voltage) refers to the terminal voltage of the battery in the open state, in value close to the battery electromotive force. The open circuit voltage method is established based on a linear (proportional) relationship between the remaining capacity of the battery and the open circuit voltage. By measuring the open circuit voltage, the remaining capacity can be directly obtained. The advantage is that it does not depend on the size, size and discharge speed of the battery. It is relatively simple to use the open circuit voltage as the test parameter [13][14][15]. Literature [16] describes the relationship between open circuit voltage, residual capacity and electrolyte density of lead-acid batteries, and gives the formula for calculating the SOC and open circuit voltage:
When using this method, by measuring the open circuit voltage of the battery, the estimated SOC value can be obtained by general lookup. However, the open circuit voltage method also has obvious shortcomings: (1) the battery needs to be left standing for a long time to reach a stable state, and how to determine the rest time is also a problem; (2) as the battery ages and the remaining power decreases, the open circuit voltage changes. It is not obvious, so it is impossible to accurately predict the remaining power; (3) For the conventionally used series battery pack, the battery used is in a loaded state, and generally cannot measure the open circuit voltage, and online measurement cannot be realized. From the current literature, the open circuit voltage method is generally not used alone. Because the open circuit voltage method has a good SOC estimation effect at the initial stage and the end stage of charging, it is often used in combination with the Anshi method and the Kalman method.
In view of the shortcomings that the battery needs to be left for a long time, the literature [14] uses the test result that the recovery curve of the open circuit voltage of the battery is almost the same under various conditions, and obtains the prediction formula of the open circuit voltage to calculate the SOC, the predicted value and the measured value. The relative error is within 6%.
The literature [17] [18] [19] normalizes the discharge curves of VRLA batteries at different discharge rates, and finds that the discharge curves have good consistency, and the discharge mode, discharge rate, ambient temperature and discharge termination voltage, etc. The change of factors has little influence on this consistency. It is proposed to predict the SOC using only the discharge voltage. The calculation formula is as follows:
Where: tT is the length of the entire discharge time, Vend is the discharge termination voltage, and Vp is the initial discharge voltage. At any time, when the discharge voltage V(t) of the battery is known, VU(tU) can be calculated, and the normalized tU is obtained against the normalized curve, thereby obtaining a state of charge (the estimated accuracy is within 10%, Suitable for some occasions where the requirements are not high).
In [20][21], different initial discharge voltages are used to correspond to different discharge times. By periodically charging a constant current load to the battery under working conditions, a series of operating voltages are measured to establish voltage, The temperature is input, and the remaining time is the output SOC fuzzy estimation system, thereby obtaining the SOC of the single power battery, and the relative error is within 1%. This method is also called load voltage method. The method can estimate the SOC of the battery on-line, and has a good effect in constant current discharge, but is not suitable for a variable current or a highly fluctuating discharge condition.
2.5 Internal resistance (conductance) method The internal resistance of the battery has the alternating internal impedance (impedance) and direct current resistance (resistance). They are closely related to the SOC and can realize online measurement. When the battery is in different power or different service life, its internal resistance value is different. The internal resistance (conductance) method predicts the change of SOC by measuring the change of internal resistance (conductance) of the battery during discharge. [twenty two].
There is still controversy in the application of internal resistance method to predict SOC. In [23], the conductivity of the valve-regulated sealed lead-acid battery was tested and statistically analyzed by the conductivity tester. It was found that there was a linear correlation between the discharge time and the conductance value, and the correlation coefficient reached 0.825. In the IEEE1188-1996 standard, the measurement was also proposed. The necessity of internal resistance clearly stipulates that the battery internal resistance test is performed at least once a quarter [24]. However, the literature [25] [26] [27] [28] has studied and analyzed the relationship between internal resistance (conductance) and residual capacity of the battery through experimental test and theoretical analysis. The results show that: (1) valve-controlled sealed lead When the SOC of the battery is 50% or more, the internal resistance (or conductance) is basically unchanged, but when the SOC is lower than 40%, the internal resistance of the battery rises rapidly;
The occurrence of disputes is mainly related to the accuracy of the test battery itself and the internal resistance (conductance) tester, except for statistical methods. Because even the same manufacturer, the same batch, the same specification of the battery, its internal resistance (conductance) is also inconsistent, which is determined by the battery manufacturer's technical level. Moreover, when the internal resistance of the battery is extremely small and the SOC changes widely, the internal resistance does not change much. If the accuracy of the measuring instrument does not meet the requirements, it will be difficult to obtain the correspondence between the internal resistance and the remaining capacity. In [29], the impedance spectrum measurement shows that the change of ohmic internal resistance can correctly reflect the change of SOC. However, when the SOC increases from 16% to 91%, the ohmic internal resistance changes very little, about 0.6mΩ. It is proposed that when the internal impedance of the battery changes from capacitive to inductive, the corresponding excitation signal frequency has a monotonic function relationship with its SOC, and the frequency variation range is large. The resonant frequency of the VRLA battery is used as the transmission of the battery SOC. Sensing parameters, this theory is still in the research stage. At the same time, the literature [30] proposes to use the internal resistance (conductance) as an indicator of the remaining capacity of the battery and the state of health (SOH) in the case of large-scale use of the battery, and to regulate the manufacturer by selecting a battery with stable internal resistance (conductance). Production, not directly as an accurate indicator of the state of charge of the battery.
From the current literature, data and internal resistance (conductance) testing products [31] [32] [33] [34], the internal resistance (conductance) method is mainly applied to battery failure warning, which is directly applied to SOC prediction. Less (generally used as one of the factors affecting SOC in combination with voltage methods, neural networks, etc.) [36]. And the literature [30] has obtained a lot of experiments and concluded that when the conductance of a single cell is more than 80% of the reference value, the battery is normal and the capacity is above 80%; when the conductance is 60%-80% of the reference value The capacity is likely to be less than 80%. The battery is in the "normal danger" state and needs to be fully discharged. When the conductivity is less than 60% of the reference value, the battery is in a "seriously dangerous" state and needs to be replaced in time.
3 system identification and parameter estimation model method for predicting SOC
Around 2000, the system identification and parameter estimation model method began to be applied to battery SOC estimation, which is currently popular in domestic and foreign research. It mainly uses some new methods (mainly artificial intelligence algorithms) to model the battery system, integrate various factors affecting the SOC into the battery model, and systematically identify and parameterize the model through a large number of tests to obtain a battery. The relationship between these parameters and the SOC, and thus the SOC. More commonly used artificial neural network method, vector machine method, fuzzy inference method and Kalman filtering method.
3.1 Neural Network Method Because the battery is a complex nonlinear system, it is difficult to establish an accurate mathematical model for its charging and discharging process. The neural network has the characteristics of distributed parallel processing, nonlinear mapping and adaptive learning, which can better reflect the basic characteristics of nonlinearity, and can give corresponding output when there is external excitation, so it can simulate battery dynamics to some extent. Characteristics, estimated SOC [36] [37].
Most of the estimated battery SOC uses a typical 3-layer artificial neural network [38] [39]. Generally, the discharge current, the terminal voltage and the temperature of the battery are directly collected or a variable current measurement method is adopted, and the electromotive force and the internal resistance are determined as inputs of the neural network model, and the SOC is used as an output. The input and output layer neurons are generally linear functions; the number of hidden layer nodes depends on the complexity of the problem and the analysis accuracy, which can be determined according to the convergence speed of the network during the training process and the error after the training is completed. The artificial neural network method is applicable to various storage batteries, but the error of this method is greatly affected by the training data and training methods, and the actual use of noise interference affects the learning and application of the network. From the current literature, neural networks are mainly theoretical studies.
In [40] [41], another neural network, Support Vector Machine (SVM) method, is used for battery SOC estimation, which avoids the defects of traditional neural network in training time, local optimum and convergence speed. The literature [42] further proposes the use of correlation vector machine (RVM) to predict the SOC of the battery, which is more accurate than the support vector machine, and the prediction model is more sparse, but the algorithm is more complex and requires large computer resources.
3.2 Fuzzy logic method Fuzzy logic method is the fuzzy modeling of the battery, based on the input and output test data of the system, and is not limited by prior knowledge, experience and behavior. The method generally blurs the parameters (such as voltage, current, temperature, internal resistance, etc.) of the input variables of the model, and obtains the relationship between the SOC and current, voltage, temperature and the like according to a large number of battery characteristic test data. The fuzzy rules are designed and fuzzy reasoning is performed, and the battery SOC is estimated by defuzzification [43][44][45].
The main disadvantage of the fuzzy logic method is that it requires a large amount of experimental data, and the fuzzy inference rules and empirical formulas are obtained based on the experimental data. At present, this method is mainly applied to simulation and theoretical analysis, and has not been applied to practice.
3.3 Kalman Filtering The core idea of ​​Kalman filter theory is to make the best estimate of the state of the dynamic system in terms of minimum variance. It is applicable to both linear and nonlinear systems [46].
When using Kalman filter to estimate SOC, we first need to establish a battery model suitable for Kalman filter estimation, and the model must have two characteristics: (1) can better reflect the dynamic characteristics of the battery, and the order can not be too high In order to reduce the amount of computation of the processor, it is easy to implement the project; (2) The model must be able to accurately reflect the relationship between the battery electromotive force and the terminal voltage, so that the closed-loop estimation has higher accuracy. Commonly used equivalent circuit models are Randle model (see Figure 1), MassimoCeraolo model, Thevenin model, Shepherd model, etc., each of which is a pending parameter, which needs to be calculated according to experimental data [47][48].
Equation of state:
(9)
Observation equation:
(10)
An-time equation:
(11)
The input vector uk of the system usually includes variables such as battery current, temperature, residual capacity and internal resistance. The output yk of the system is usually the operating voltage of the battery. The battery SOC is included in the state quantity xk of the system. Ak and Bk are obtained by experiment. The parameters determine that ωk and vk are system noise. The core of the estimation SOC algorithm is to establish a set of recursive equations including the SOC estimation value and the covariance matrix reflecting the estimation error. The covariance matrix is ​​used to give the estimation error range. Equation (11) is the battery model state equation, which describes the SOC as the basis for the state vector.
Kalman filter can maintain good precision in the estimation process, and has a strong correction effect on the initialization error, which has a strong inhibitory effect on noise. Currently, it is mainly applied to the SOC prediction of hybrid vehicle battery with fast current change. . Based on the Kalman filter, the literature [49] [50] [51] uses the extended Kalman and colorless Kalman filtering method to estimate the SOC. The biggest disadvantage of the Kalman filter method is that its estimation accuracy depends heavily on the accuracy of the battery equivalent circuit model. Establishing an accurate battery model is the key to the algorithm. Another disadvantage is that the amount of computation is relatively large, and it is necessary to select a simple and reasonable battery model and a processor with a relatively fast operation speed.
3.4 Other methods The linear model method mentioned in [52] uses the linear model to be highly robust to the initial conditions of measurement error and error. Based on a large number of battery charge and discharge experiments, the SOC and its variation battery are established. The linear equation of terminal voltage and current is shown in equations (12) and (13). This method is suitable for small currents and slow changes in SOC, but this feature also limits the scope of its use, which has not been seen in practical applications.
In [53], the nonlinear autoregressive moving average (NARMAX) model is proposed to have high accuracy, simple structure and fast convergence. The battery operating voltage and current are used as model inputs, and other influencing factors of SOC are used as system noise. The SOC is estimated in real time with a relative error of only 1%. The applicability of this method remains to be further studied. The identification model is shown in equation (14), where y(t) is the SOC sequence, u1(t) is the current sequence, and u2(t) is the voltage sequence.
It can be seen from the various methods described above that both the physical modeling method and the system identification and parameter estimation model methods are based on the measurable parameters of the battery (mainly voltage, current, internal resistance, temperature, etc.) and the remaining The relationship between capacity, through a large number of experiments to establish a stable battery system model to predict SOC.
4 Summary In summary, the SOC prediction method is affected by many factors (discharge current, voltage, temperature, depth of discharge, internal resistance, electrolyte density, self-discharge, aging, etc.), the prediction technology of the remaining capacity of VRL A battery and Modeling is quite complex, and there is currently no accurate and universal prediction method. The above various SOC prediction methods have their own advantages and disadvantages. However, under different use environments and different prediction accuracy, a single prediction method cannot meet the actual requirements. Therefore, a high-precision data detection circuit is designed, and multiple methods are used for combined prediction. SOC, especially the combination of multiple intelligent algorithms and new theories, is a real-time, online and accurate prediction of SOC, which has become the development direction of battery residual capacity prediction.
references
[1]JinlongZhang, ChaoyingXia.ApplicationofStatisticalParameterIdentificationinBatteryManagementSystem.Proceedingsofthe8thWorldCongressonIntelligentControlandAutomationJuly6-92010
[2]Zhang Xiuling, Zhu Chunying. Application of Elman neural network based on genetic algorithm in capacity prediction of nickel-hydrogen battery[J].Industrial Instrumentation & Automation.2009.4.100-102
[3] Zhan Yiju. Application of Neural Network Method in Battery Modeling[J]. Journal of Solar Energy 1997.18(4).
[4] Liu Kaipei, Lei Ming. Safely carrying out battery discharge test [J]. Communication Power Technology. 2008.25(5).54-55
[5] Hu Yuanqing. On-line verification capacity test of remote real-time load of communication base station battery[J].Communication Power Technology.2011.28(4).83-84
[6]Wang Wei. Research on hardware and software realized by lead-acid battery management system [D]. Master's thesis of Zhejiang University. 2005.16-26
[7]Zheng Hangbo. Research and implementation of new lithium battery management system for electric vehicles [D]. Master's thesis of Tsinghua University. 2004.58-65
[8] Kong-SoonNg, Yao-FengHuang, Chin-SienMoo.AnEnhancedCoulombCountingMethodforEstimatingState-of-ChargeandState-of-HealthofLead-AcidBatteries.TelecommunicationsEnergyConference,2009.INTELEC2009.31stInternationalIssueDate:18-22Oct.2009
[9] Liu Guanglin. Introduction to the technology of lead-acid batteries [M]. Beijing: Mechanical Industry Press. 2008.265-274
[10]Luo Kai. Research on on-line detection system of residual capacity of lead-acid battery based on DSP[D]. Chongqing University of Technology Master's Thesis.2010
[11]XIONG Wenqiang. Study on Ultrasonic Measurement Method of Electrolyte Density of Lead-acid Battery[D]. Master thesis of Dalian University of Technology.2010
[12] Hou Yuexin, Li Gang, Song Changqing. Development of Battery Electrolyte Density Measuring Instrument [J]. Nuclear Electronics and Detection Technology. 2006.26(6).735-738
Current Status and Development of Battery Residual Capacity Prediction Technology
introduction