Battery SOC prediction method and battery management system based on FOMIAUKF algorithm

Abstract

This invention discloses a battery SOC prediction method and battery management system based on the FOMIAUKF algorithm. The method includes: establishing a fractional-order lithium battery model to obtain the state-space discrete system equations and measurement equations; then, collecting battery discharge data through hybrid power pulse characteristic testing and fitting the relationship between open-circuit voltage and SOC; using a WSO-LM hybrid algorithm for parameter identification to obtain the corresponding model parameters; and introducing a fractional-order multi-innovation adaptive unscented Kalman filter FOMIAUKF algorithm to estimate the SOC value of the lithium battery pack and perform error analysis. The SOC accuracy error of the battery management system of this invention is within 0.2%, which is far superior to the requirements of the national standard QC / T897-2011.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium batteries, and in particular to a battery SOC prediction method and a battery management system based on a FOMIAUKF algorithm. Background Art

[0002] With the rapid increase in the number of traditional cars, problems such as excessive energy consumption and environmental pollution have become increasingly prominent. Therefore, in recent years, safe, clean and efficient new energy vehicles are rapidly emerging. At the same time, related technologies have also undergone rapid evolution. Lithium batteries, as an energy medium with high energy density, excellent charge and discharge performance, light weight and portability, long life and environmental protection, have been widely used in the field of new energy vehicles.

[0003] The importance of battery management system (BMS) in new energy vehicles is becoming increasingly prominent. Its responsibilities include monitoring the voltage, current and temperature of the battery pack at all times to avoid overcharging, over-discharging and over-current of the battery pack, thereby ensuring the safety of the lithium battery pack, enabling it to function more effectively and extend its service life. In the early days of the rise of new energy, due to limited resources, BMS faced a series of challenges, such as incomplete measurement information parameters, long data collection intervals, and lack of real-time status checks, making it difficult to conduct a comprehensive analysis and study of the actual battery usage status.

[0004] Although the existing lithium-ion batteries on the market have made certain progress in the detection accuracy, durability and reliability of battery management systems, there is still a lack of systematic design and research in the estimation, balance control and thermal management of battery state of charge. The estimation accuracy of battery state of charge (SOC) has an important impact on the safe use of vehicles, the extension of battery life and the decision of charging time. The SOC estimation methods commonly used at home and abroad currently include the ampere-hour integration method, the open circuit voltage method and the Kalman filter method, and each method has its advantages and disadvantages.

[0005] Among them, when the unscented Kalman filter (UKF) is used in the battery management system to estimate the battery SOC, it accurately captures the nonlinear system characteristics by selecting a set of sigma points to perform state prediction and measurement updates. Its advantage is that it is suitable for nonlinear systems. The unscented sampling points avoid the Taylor expansion of nonlinear functions and improve the accuracy of the estimation. However, the computational complexity of UKF is relatively high, which may have a certain impact on the real-time performance of the system. In addition, its performance is sensitive to the selection of sigmapoints and the processing of system noise, and the parameters need to be carefully adjusted. Summary of the invention

[0006] Technical purpose: In view of the defects in the prior art, the present invention discloses a battery SOC prediction method and a battery management system based on the FOMIAUKF algorithm, continuing the idea of ​​estimating the battery SOC by the unscented Kalman filter method, and on this basis integrating adaptive and multi-innovation theories, a fractional-order multi-innovation adaptive unscented Kalman filter algorithm (FOMIAUKF) is proposed to improve the estimation accuracy of SOC.

[0007] Technical solution: In order to achieve the above technical objectives, the present invention adopts the following technical solution.

[0008] A battery SOC prediction method based on FOMIAUKF algorithm comprises the following steps:

[0009] Step 1: Establish a fractional-order lithium battery model based on the second-order RC equivalent model to obtain the state space discrete system equation and measurement equation of the fractional-order lithium battery model;

[0010] Step 2: Conduct hybrid power pulse characteristic test, collect battery discharge data, and fit the relationship between open circuit voltage and SOC;

[0011] Step 3: Use the WSO-LM hybrid algorithm to identify parameters and obtain the parameters of the fractional-order lithium battery model, namely, the battery internal resistance R 0 , Polarization internal resistance R 1 and R 2 , polarization capacitance C 1 and C 2 ; Obtain the fractional-order lithium battery model after parameter identification;

[0012] Step 4: Take the relationship expression between open circuit voltage and SOC as the true value of SOC, and use the fractional-order multi-innovation adaptive unscented Kalman filter algorithm, i.e., FOMIAUKF algorithm, to calculate the error of the fractional-order lithium battery model after parameter identification to obtain the predicted SOC value of the lithium battery pack.

[0013] The present invention also discloses a battery management system based on the FOMIAUKF algorithm, including a lithium battery pack, a battery management chip, a central control unit, a voltage detection module, a temperature detection module, a current detection module, a battery protection and balancing circuit;

[0014] The lithium battery pack includes a plurality of lithium batteries, the lithium battery pack is connected to a battery management chip, the battery management chip is connected to a voltage detection module, a temperature detection module, and a current detection module, and the battery management chip and the voltage detection module, the temperature detection module, and the current detection module realize voltage detection, temperature detection, and current detection of the lithium battery pack;

[0015] The voltage detection module, temperature detection module, and current detection module are connected through I 2C communication mode is connected to the central control unit, which is used to feed back the voltage detection, temperature detection and current detection results of the lithium battery pack to the central control unit; the central control unit is used to implement any of the above-mentioned battery SOC prediction methods based on the FOMIAUKF algorithm, and measure the battery SOC according to the voltage detection, temperature detection and current detection results of the lithium battery pack; the central control unit is connected to the lithium battery pack through a battery protection and balancing circuit to ensure that the capacity and performance of each single battery remain consistent.

[0016] Beneficial effects:

[0017] 1. The present invention continues the idea of ​​estimating battery SOC by using the unscented Kalman filter method, and on this basis integrates the adaptive and multi-innovation theories, and proposes a fractional-order multi-innovation adaptive unscented Kalman filter algorithm (FOMIAUKF) to improve the estimation accuracy of SOC;

[0018] 2. The present invention uses the WSO-LM hybrid algorithm for parameter identification, which has higher model accuracy than the traditional least squares method;

[0019] 3. The present invention designs a corresponding battery management system based on the proposed FOMIAUKF algorithm and verifies the accuracy and feasibility of information collection of the BMS system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a system block diagram of the present invention;

[0021] Figure 2 This is a schematic diagram of an equivalent circuit model of a lithium battery of the present invention;

[0022] Figure 3 It is a schematic diagram of the WSO-LM hybrid algorithm flow of the present invention;

[0023] Figure 4 This is a comparison diagram of the model terminal voltage of the present invention;

[0024] Figure 5 is a model terminal voltage error diagram of the present invention;

[0025] Figure 6 It is a schematic diagram of a partial circuit of a central control unit of the present invention;

[0026] Figure 7 It is a schematic diagram of a voltage acquisition circuit of the present invention;

[0027] Figure 8 It is a schematic diagram of a step-down circuit of the present invention;

[0028] Fig. 9 It is a schematic diagram of the current collection circuit of the present invention;

[0029] Fig.10 This is a comparison diagram of SOC estimation accuracy of the present invention;

[0030] Fig.11 This is a comparison diagram of the SOC estimation error of the present invention. DETAILED DESCRIPTION

[0031] The following is a further description and explanation of a battery SOC prediction method and a battery management system based on the FOMIAUKF algorithm of the present invention in conjunction with the accompanying drawings.

[0032] As attached Figure 1 As shown, a battery SOC prediction method based on the FOMIAUKF algorithm includes the following steps:

[0033] Step 1: Establish a fractional-order lithium battery model based on the second-order RC equivalent model to obtain the state space discrete system equation and measurement equation of the fractional-order lithium battery model;

[0034] In the fractional order model, the present invention selects the second-order RC equivalent model as the lithium battery equivalent circuit model, including the open circuit voltage U oc , terminal voltage V o 、Battery internal resistance R 0 , Polarization internal resistance R 1 and R 2 , polarization capacitance C 1 and C 2 . Polarization internal resistance R 1 The polarization capacitor C 1 Parallel connection, polarization internal resistance R 2 With polarization capacitance C 2 In parallel, the open circuit voltage is U oc and the battery internal resistance R 0 , polarization internal resistance R after parallel connection 1 With polarization capacitance C 1 , polarization internal resistance R after parallel connection 2 With polarization capacitance C 2 Connection. Figure 2 As shown;

[0035] The fractional calculus formula for the second-order RC equivalent model is as follows:

[0036]

[0037]

[0038] Wherein, t is the upper limit of the integral, and p is the lower limit of the integral. Since R is greater than 0 in the present invention, t and p can be omitted. sis the step size of the operation. In some embodiments of the present invention, the step size of the operation is taken as 0.1 s. R is the order. In the present invention, R is taken as a decimal to represent fractional-order operation (0 < R < 1). i is the i-th sampling point, and f(t) is the given function. is the fractional-order operator.

[0039] Figure 2 The two polarization capacitors C 1 , C 2 can be expressed as:

[0040]

[0041] where α and β are both fractional-order orders of the polarization capacitors, and their values are taken as decimals. j is the imaginary number, w is the angular frequency, Z c1 and Z c2 are the impedances of the polarization capacitors C 1 and C 2 .

[0042] The state-space equation of the fractional-order lithium battery model is:

[0043]

[0044]

[0045] D = [-R] (6)

[0046] where the order of the polarization capacitor C1 is α, the order of the polarization capacitor C2 is β, U 1 is the voltage across the polarization internal resistance R 1 , U 2 is the voltage across the polarization internal resistance R 2 , V 0 is the terminal voltage, U oc is the open-circuit voltage, and D is the operation parameter.

[0047] The definition formula of SOC is as follows:

[0048]

[0049] In the formula, Q new represents the rated capacity of the battery, T s represents the sampling time, SOC k+1 is the predicted value at the next moment, i.e., the k + 1 moment, SOC k is the estimated value at the k moment, I k is the current value flowing through the battery internal resistance R 0 at the k moment.

[0050] The state-space discrete system equation and measurement equation of the fractional-order lithium battery model are as follows:

[0051]

[0052]

[0053]

[0054]

[0055]

[0056]

[0057] Among them, x k+1 is the state space discrete system equation, k is the variable at the current moment, U oc，k is the open circuit voltage value at time k, y k is the measurement equation; A, B, K j , C are parameter matrices respectively; x 0 is the initial value of the model's state variable, whose value is x 0 =SOC 0 ;x k is the state variable at the current moment;

[0058] Step 2: Conduct hybrid power pulse characteristic test, collect battery discharge data, and fit the relationship between open circuit voltage and SOC;

[0059] The battery discharge data used was obtained by connecting 10 INR18650-30Q batteries in parallel through a hybrid pulse characteristic test. The room temperature was 25°C, the nominal capacity of a single battery was 3Ah, the nominal voltage was 3.6V, and the discharge method was to discharge for 180s each time and then stand for 3h. The discharge current was 30A. The obtained experimental data was fitted with a polynomial to obtain the OCV-SOC fitting curve. Through the analysis of the fitting results of different orders, the 8th order fitting has higher accuracy, and the fitting curve is closer to the real result. The function expression of the 8th order fitting is as follows:

[0060]

[0061] Among them, U oc is the open circuit voltage, p 1 、p 2 、p 3 、p 4 、p 5 、p 6 、p 7 、p 8 、p 9 are the undetermined coefficients, SOC, SOC 2 , SOC 3, SOC 4 , SOC 5 , SOC 6 , SOC 7 , SOC 8 are the corresponding powers of the SOC values ​​respectively;

[0062] Step 3: Use the war strategy optimization-Levenberg Marquardt hybrid algorithm, that is, the WSO-LM hybrid algorithm, to perform parameter identification and obtain the parameters of the fractional-order lithium battery model, that is, the battery internal resistance R 0 , Polarization internal resistance R 1 and R 2 , polarization capacitance C 1 and C 2 ; Get the fractional-order lithium battery model after parameter identification; Based on the fractional-order lithium battery model and the known open circuit voltage U OC and SOC function, the WSO-LM hybrid algorithm is used for parameter identification to obtain the battery internal resistance R 0 , Polarization internal resistance R 1 and R 2 , polarization capacitance C 1 and C 2 The SOC in step 2 is a value obtained by OCV-SOC fitting based on the value obtained from the battery pack charge and discharge experiment, and is used to calculate the model parameters.

[0063] Based on the fractional-order lithium battery model and the known open-circuit voltage U OC The WSO-LM hybrid algorithm is used for parameter identification. The war strategy optimization algorithm WSO generates an initial population in a random manner, in which each individual consists of seven genomes (including R 0 , R 1 , R 2 , C 1 , C 2 , α, β), these seven genomes are the battery model parameters, which are represented by binary coding. For each individual, a fitness check is performed, and individuals with high fitness are selected for crossover and mutation to generate a new population. This cycle will continue until the optimal solution for the required fitness is reached. The battery internal resistance R0, polarization internal resistance R1 and R 2 , polarization capacitance C 1 and C 2 The flowchart of the WSO-LM hybrid algorithm is shown in the attached figure. Figure 3 As shown; Figure 4 and Figure 5 A model terminal voltage comparison diagram and a model terminal voltage error diagram of the present invention are respectively given.

[0064] The Levenberg-Marquardt operator is added to the war strategy optimization algorithm WSO. The LM algorithm has a strong local optimization ability, which effectively overcomes the shortcomings of the local optimization ability of the war strategy optimization algorithm and avoids the premature convergence of the war strategy optimization algorithm. In each iteration, the optimal parameter vector output by the war strategy optimization algorithm provides a priori estimates for the LM algorithm. The LM algorithm completes high-precision nonlinear fitting based on the prior estimates, and continuously adjusts the search direction by adjusting the adaptive parameters, which greatly reduces the probability of the objective function falling into the local optimum. The model parameter correction expression of the LM algorithm is:

[0065] h(l+1)=h(l)-[j T J+λI] -1 J T f(h)λ>0 (15)

[0066] Where I is the unit matrix, J is the Jacobian matrix, λ is the damping factor, h(l+1) is the battery model parameter vector of the l+1th iteration, h(l) is the battery model parameter vector of the lth iteration, and f(h) is the battery model error function.

[0067]

[0068] Where: h is the battery model parameter vector, is the actual terminal voltage of the qth sample, is the terminal voltage estimate of the battery model parameters corresponding to the qth sample, 1≤q≤N, N is the number of samples, and the samples here are obtained by performing a hybrid power pulse characteristic test HPPC on the battery at different charge states at 25°C to obtain data samples of the rebound voltage after the battery discharge pulse.

[0069] In the WSO-LM hybrid algorithm, according to the law of survival of the fittest, the crossover probability and mutation probability of high-quality individuals and low-ranking individuals in each generation of the population are different. High-quality individuals should be retained as much as possible, while low-ranking individuals should undergo as much crossover and mutation as possible. Therefore, the crossover probability and mutation probability can be expressed as follows (17):

[0070]

[0071] Among them, Cross(j) is the crossover probability of the jth individual, Mutation(j) is the mutation probability of the jth individual, J(j) is the individual fitness of the jth individual, and J max is the maximum value of the sum of squares of the differences between the output measured in the population and the output predicted by the model, J ave is the average of the sum of squares of the differences between the output measured within the group and the output predicted by the model, K1 and K 2 are the initial values ​​of crossover probability and mutation probability respectively.

[0072] Step 4: Take the relationship expression between open circuit voltage and SOC as the true value of SOC, and use the fractional-order multi-innovation adaptive unscented Kalman filter algorithm, i.e., FOMIAUKF algorithm, to calculate the error of the fractional-order lithium battery model after parameter identification to obtain the predicted SOC value of the lithium battery pack.

[0073] The implementation of the traditional unscented Kalman (UKF) filter algorithm is based on an ideal assumption of noise. Usually, the noise of the system model is assumed to be Gaussian white noise that obeys a normal distribution. However, the system model noise will change over time and is not completely known. Therefore, based on FOM, the present invention combines the adaptive covariance algorithm with the multi-information method to propose the FOMIAUKF algorithm, i.e., the fractional-order multi-information adaptive unscented Kalman filter algorithm. The adaptive covariance algorithm corrects W in real time during the SOC estimation process. k and R k , reduce the estimation error of SOC, the specific formula is as follows:

[0074]

[0075]

[0076]

[0077] d k =(1-b)(1-b k+1 ) -1 (twenty one)

[0078] Among them, Q k is the process noise covariance matrix at time k, Q k-1 is the process noise covariance matrix at time k-1, d k-1 is the scaling parameter at time k-1, d k is the scaling parameter at time k, K k is the Kalman gain, ε k is the residual, which is the difference between the real-time voltage measurement value and the model voltage prediction value. is k The transposed matrix of K k The transposed matrix, P k is the variance at time k, is the predicted value of the variance at time k, i.e., the covariance, R k is the observation noise covariance matrix at time k, R k-1 is the observation noise covariance matrix at time k-1, P y，kis the variance matrix of the observation at time k, y k is the real-time voltage measurement value at time k, is the voltage prediction value at time k, that is, the voltage prediction value of the lithium battery pack SOC prediction using the fractional-order lithium battery model after parameter identification. b represents the forgetting factor, which is generally 0.95 to 0.99.

[0079] The multi-information method is an effective means to predict the error information generated during the iterative algorithm to achieve a more accurate a posteriori correction. When only one new information is used to predict the error, it may lead to the loss of a posteriori correction information. Therefore, the combination of multiple new information and Kalman filtering helps to improve the accuracy of the algorithm. Although the use of the multi-information method will increase the amount of calculation, considering the importance of improving the estimation accuracy, a slight increase in the computational cost is acceptable. The multi-information matrix is ​​shown in formula (22). By comprehensively utilizing multiple new information, the error information can be considered more comprehensively, thereby improving the robustness and accuracy of the algorithm.

[0080]

[0081] Among them, E p，k is a multi-innovation matrix, p represents the length of the innovation, y k is the real-time voltage measurement value at time k, is the mean of the observed variable at time k.

[0082] In summary, the main steps of the FOMIAUKF algorithm in the present invention to estimate the battery SOC are as follows:

[0083] (1) Algorithm initialization

[0084]

[0085] Among them, x 0 is the initial value of the model's state variable, whose value is x 0 =SOC 0 , w 0 , v 0 is the initial value of process noise and measurement noise, the default value is 0, P xx，0 is the initial value of the covariance of the state estimate, Q 0 , R 0 is the variance of process noise and measurement noise, SOC 0 is the initial voltage value of the lithium battery pack determined by the open circuit voltage method, and E[1] represents the calculated expected value.

[0086] (2) Generate Sigma sample point X i

[0087] Get the state vector 2n+1 Sigma sample points Xi , calculate the sampling points and the corresponding weight vectors, and then according to the UT transformation (Unscented transform, that is, untraceable transformation), use the nonlinear state function to transmit the sampling points to obtain the Sigma point at time k-1:

[0088]

[0089] In the formula, γ is the proportional factor, x k-1 is the model state variable at time k-1, X i is the Sigma point, P xx，k / k-1 is the covariance of the state estimate.

[0090] (3) Calculate the weighting coefficient

[0091] In the nonlinear equation X i，k / k-1 =f(X k-1 ,i k ) into the Sigma point to obtain the next prediction

[0092] X i，k / k-1 =f(X k-1 , I k )+w k-1 i=1,...,2n (25)

[0093] Among them, X i，k / k-1 is the predicted value of Sigma point, X k-1 is the Sigma point at time k-1, i k is the current value at time k, w k-1 is the process noise at time k, f(X k-1 ,i k ) is a nonlinear transfer function, and n is n sampling points. Then the predicted value of the state variable is:

[0094]

[0095] Where i is the i-th sampling point, m is the superscript of the weight coefficient, W i m is the weight coefficient, and its value is:

[0096]

[0097] Among them, c is the superscript of different weight coefficients.

[0098] From this, the nonlinear transfer formula of the Sigma sampling point set can be obtained as follows:

[0099] λ i，k / k-1 =f(X i，k / k-1 , I k ) (28)

[0100] Among them, λ i，k / k-1 is the Sigma point after nonlinear transmission;

[0101] (4) Time update

[0102] By calculating the weights and covariance, we can get the mean and variance of the observed variables

[0103]

[0104]

[0105] in, is the mean value of the observed variable, P xx，k / k-1 is the variance of the state variable, Q k-1 is the k-1 moment process noise covariance, W i m and W i c is the weight coefficient;

[0106] (5) Measurement update

[0107] According to the measurement function ζ i，k / k-1 =G(λ i，k / k-1 ,i k ) Update the observations. By calculating the weights and covariances, we can get the mean and variance of the observed variables as

[0108]

[0109] Among them, P yy，k / k-1 is the variance of the observed variable, ζ i，k / k-1 is the updated observation value, W i m and W i c is the weight coefficient, R k-1 is the covariance of the measurement noise;

[0110] The joint covariance matrix of state variables and observation variables is:

[0111]

[0112] The Kalman gain is

[0113] K x，k =P xy，k / k-1 (P yy，k / k-1 ) -1 (33)

[0114] Among them, (P yy，k / k-1 ) -1 Pyy，k / k-1 The inverse matrix of

[0115] Define the Kalman gain matrix:

[0116] K p，k =[K x，k K x，k-1 …K X，K-P+1 ] (34)

[0117] (6) Update the posterior state value

[0118] The innovation sequence is:

[0119] e k =y k -y k / k-1 (35)

[0120] Among them, e k is the new information at time k, y k is the real-time voltage measurement value at time k, y k / k-1 is the predicted value of the observed variable at time k;

[0121] Define the multi-innovation matrix as:

[0122]

[0123] Among them, E p，k is the innovation matrix, p is the innovation length, is the mean of the observed variable at time k;

[0124] The state variable estimate is updated as:

[0125]

[0126] Among them, K p，k is the Kalman gain, is the predicted value of the state variable at time k, x k is the updated state variable at time k;

[0127] (7) Update the posterior state error covariance:

[0128]

[0129] Among them, K x，k is the Kalman gain, P xx，k / k-1 is the predicted value of the state error covariance at time k, P y，k / k-1 is the state error covariance at time k, P xx，k is the updated posterior state error covariance; K x，k The transposed matrix of

[0130] According to formulas (1) to (38), the FOMIAUKF method can be used to estimate the SOC of lithium batteries and obtain the SOC value. SOC is an element in the state space vector x, which is described in the above formula. The state variable update is to update the SOC value.

[0131] like Figure 1 As shown, a battery management system based on the FOMIAUKF algorithm includes a lithium battery pack, a battery management chip, a central control unit, a voltage detection module, a temperature detection module, a current detection module, a battery protection and balancing circuit;

[0132] The lithium battery pack includes a plurality of lithium batteries, the lithium battery pack is connected to a battery management chip, the battery management chip is connected to a voltage detection module, a temperature detection module, and a current detection module, and the battery management chip and the voltage detection module, the temperature detection module, and the current detection module realize voltage detection, temperature detection, and current detection of the lithium battery pack;

[0133] The voltage detection module, temperature detection module, and current detection module are connected through 1 2 C communication mode is connected with the central control unit, and is used to feed back the voltage detection, temperature detection and current detection results of the lithium battery pack to the central control unit;

[0134] The central control unit is used to implement the above-mentioned battery SOC measurement method based on the FOMIAUKF algorithm, and the battery SOC measurement is performed according to the voltage detection, temperature detection and current detection results of the lithium battery pack; the central control unit is connected to the lithium battery pack through the battery protection and balancing circuit to ensure that the capacity and performance of each single battery remain consistent; the central control unit is also connected to the upper computer management system interface through CAN communication;

[0135] Specifically, the overall framework of the battery management system in the present invention is divided into two modules: hardware design and software design. The hardware design module mainly designs the schematic diagram, and on this basis completes the production of the BMS, i.e., the development board of the battery management system; the software design module mainly uses the fractional-order multi-information adaptive unscented Kalman filter method to improve the SOC estimation accuracy, and designs a host computer to monitor the data in real time. The present invention describes in detail the design principles and implemented functions of each module of the peripheral circuit, wherein the main control chip of the central control unit is selected as STM32F103C, which has up to 80 IO ports, 20kB of SRAM storage space and multiple communication modes such as SPI, USART, CAN, etc., and is widely used in embedded designs, such as Figure 6 shown.

[0136] Its hardware design mainly consists of the following modules:

[0137] (1) Central Control Unit:

[0138] The core of the battery management system is the central controller, such as Figure 7 As shown, its main work is to receive data such as battery current, voltage, temperature, etc., and then use the battery management method of the present invention to estimate the battery SOC, and issue instructions for control and data communication with the PC. The main functions implemented include the following parts:

[0139] 1) Charge and discharge balance control.

[0140] 2) Estimate the battery pack SOC.

[0141] 3) Provide timely warnings for battery over-discharge, over-current, high temperature, etc.

[0142] 4) Detect the single cell and overall voltage and current of the battery pack.

[0143] 5) Communicate with the host computer through UART to initialize the protection parameters such as overcharge, over discharge, over current and short circuit.

[0144] (2) Voltage detection module:

[0145] The current mainstream voltage collection methods include: resistance voltage division method, flying capacitor method, operational amplifier differential amplification method, etc. Since traditional voltage and temperature detection circuits often require each single cell to be equipped with an independent conditioning circuit to transmit the voltage and temperature data of the single cell to the microcontroller through an isolation amplifier, a filter circuit and an analog-to-digital converter (ADC), thereby obtaining accurate voltage and temperature readings, this method has the problems of high cost, large space and multiple IO requirements, especially in large-scale battery monitoring applications. Therefore, the present invention uses the battery management chip BQ76930 to collect the voltage of the single cell, which has the advantages of high precision and high computing speed compared to the traditional peripheral circuit to measure the battery voltage. The lithium battery pack consists of 10 18650 lithium batteries, such as Figure 7 The P6 port in the circuit is the voltage input port of the lithium battery pack. To prevent the battery pack voltage from being too high, the MP9486A buck chip is used to perform DC-DC buck processing on the battery pack. Figure 7 and 8As shown, the VBAT2 signal output by pin 11 of the P6 port is connected to pin 3 of the MP9486A buck chip, and pin 5 of the MP9486A buck chip outputs 13V-1. In order to improve the accuracy of voltage acquisition, the acquisition is carried out from the P6 port to the battery management chip BQ76930, and RC is used for filtering, that is, a low-pass filter is used for filtering on the acquisition channel, where the resistance value is 1kΩ and the capacitance is 1uf, and the noise with a frequency greater than 159Hz is removed. The main control chip initializes the front-end chip through the IC2 port, and the value of the single cell voltage is obtained by reading the value of the ADC register.

[0146] (3) Temperature detection module:

[0147] In order to avoid potential safety risks caused by long-term high-temperature operation of the battery pack, the design of a real-time temperature acquisition module is essential, which can effectively avoid safety hazards caused by long-term overheating of the battery pack. This system uses thermistors to detect battery temperature. NTC thermistors are sensitive to temperature, and their resistance changes with temperature. Temperature acquisition is achieved by converting the temperature signal into a voltage signal through a 10K 3950NTC thermistor chip resistor and inputting it into the BQ76930. 2 C interface to the management control unit for processing, and the temperature value can be obtained after TEMP_SEL is enabled. The register addresses for reading temperature are TS1, TS2, and TS3. The present invention uses the TS2 pin, that is, Figure 7 Part P7.

[0148] (4) Current detection module:

[0149] The monitoring subsystem of the battery management chip is carried out through the two pins of the positive terminal SRN and the negative terminal SRP of the sensitive resistor. The 16-bit CC calculates the accumulated charge on the current sensitive resistor by integrating the battery current over time, and samples the measured current value with a sampling period of 250ms, thereby realizing the monitoring of the battery pack current. Therefore, in order to test whether the current monitoring function of the chip can be used correctly, the current acquisition circuit module in the test platform designs a current sensitive resistor RSNS between the two pins SRN and SRP of the chip, and designs and adds RC filtering circuits between the current sensitive resistor RSNS and the two pins SRN and SRP to improve the quality of the current acquisition signal. The specific circuit design of the current acquisition circuit module is as follows: Fig. 9 shown.

[0150] (5) Single cell battery balancing circuit:

[0151] Lithium-ion battery packs are generally composed of several cells, and the condition of the cells is directly related to the overall performance of the battery pack. The purpose of balancing management is to ensure that the capacity and performance of each cell remain consistent, so as to avoid interruption of charging or discharging due to abnormal voltage of a certain cell. The balancing function is divided into two modes: active and passive. Active balancing transfers charge from high-voltage cells to low-voltage cells to achieve overall balance. This method has high energy utilization and excellent balancing effect, but the control circuit is complex and the cost is high. Passive balancing releases the energy of high-voltage cells through resistors to reduce the capacity difference between batteries. The circuit is simple and the technology is mature. This system adopts a passive balancing management strategy, which is implemented inside the battery management chip BQ76930. Since there is a MOS inside the chip, the high level triggers conduction, and the excess power will be consumed and released through the resistor.

[0152] (6) Communication module:

[0153] The present invention uses the CAN bus to realize information transmission between the BMS, the vehicle master controller and the host computer. The CAN receiving module waits for the arrival of the data frame and determines the validity of the message based on the predetermined frame header, frame ID, sender address and number of times. After receiving, the data is verified, analyzed and sorted to extract various parameters of the battery.

[0154] In the present invention, the data measured by the digital multimeter and the thermometer are used as the reference true value of the system, and the data detected by the BMS is compared with the reference true value to verify the accuracy of the system.

[0155] The SOC accuracy error of the battery management system of the present invention is within 0.6%, which is far better than the provisions of the national standard QC / T897-2011.

[0156] The following is an embodiment of an experiment using the inventive method:

[0157] A hybrid power pulse test was performed on the 18650-30Q lithium battery pack at room temperature of 25 degrees Celsius. The nominal capacity of the battery cell is 3.0Ah and the nominal voltage is 3.6V. The discharge method is to discharge for 180 seconds each time and then stand for 3 hours, and the discharge current is 30A. The data measured by the digital multimeter and thermometer are used as the reference true value of the system. The data detected by the BMS is compared with the reference true value to obtain the test results of 10 groups of random test points, which are the voltage, current, and temperature collection comparison results, as shown in Tables 1 to 3.

[0158] Table 1 Multimeter voltage data table

[0159] Number of comparisons Host computer / V Multimeter / V Relative error / % 1 34.88 34.93 0.14 2 33.95 33.99 0.12 3 32.66 32.62 0.12 4 31.30 31.25 0.16 5 28.86 28.79 0.24 6 25.56 25.49 0.27 7 23.39 23.45 0.25 8 22.00 22.06 0.27 9 20.89 20.86 0.14 10 20.20 20.16 0.20

[0160] Table 2 Multimeter current data table

[0161]

[0162]

[0163] Table 3 Temperature data table

[0164] Number of comparisons Host computer / ℃ Temperature meter / ℃ Relative error / % 1 25.5 25.6 0.39 2 26.6 26.3 1.14 3 26.9 27.4 1.82 4 28.3 28.6 1.05 5 30.0 29.6 1.35 6 30.6 30.3 1.00 7 31.2 31.6 1.27 8 34.1 33.9 0.59 9 35.2 35.0 0.57 10 35.5 35.7 0.56

[0165] It can be observed from Tables 1 to 3 that during the continuous discharge of the battery pack, the voltage gradually decreases, and as the discharge time increases, the temperature gradually increases, while the discharge current remains relatively uniform. The accuracy error of the terminal voltage measurement is between 0.12% and 0.27, the relative error of the current measurement is less than 0.90%, and the error of the temperature measurement is within 1.82%. This shows that the battery management system has high accuracy in the measurement of voltage, current and temperature, and its performance is stable and reliable.

[0166] The ultimate goal of the present invention is to estimate the battery SOC. After verifying the accuracy of BMS data acquisition, it is necessary to further verify the accuracy of the SOC estimation method under the WSO-LM model parameters. First, the battery pack is charged with a constant current of 0.5C to an upper limit voltage of 4.2V, and then charged with a constant voltage until the charging rate is reduced to 0.05C, which is regarded as the battery being fully charged, that is, SOC = 1. The discharge point data is still obtained by hybrid pulse testing of the 18650-30Q lithium battery pack. Under the same initial conditions, MATLAB is used to apply unscented Kalman filtering (UKF), fractional-order unscented Kalman filtering (FOUKF) and FOMIAUKF to estimate the SOC of the lithium battery pack under the discharge condition, and the results are shown in the figure. Fig.10 As shown in Figure 2. Under this model parameter, the three filtering algorithms can effectively track the SOC reference value, but the FOMIAUKF algorithm performs best.

[0167] Fig.11 is the error fitting curve between the FOMIAUIKF algorithm and the true value. In the initial stage of simulation, the SOC estimation result fluctuates greatly, which is caused by the fact that the internal chemical reaction of the battery has not reached a stable state. However, as the discharge proceeds, the SOC estimation error gradually decreases and approaches 0. The overall estimation error of the FOMIAUIKF method is stable within 0.2%. The reason is that the FOMIAUIKF algorithm does not need to perform local linearization processing on the nonlinear function, and can be directly substituted into the nonlinear battery model for estimation. In addition, the process noise Qk and measurement noise Rk are corrected online through the new information and residual sequence, which improves the accuracy of the algorithm.

[0168] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A battery SOC prediction method based on FOMIAUKF algorithm, characterized in that: The following steps are involved: Step 1: Establish a fractional-order lithium battery model based on the second-order RC equivalent model to obtain the state space discrete system equation and measurement equation of the fractional-order lithium battery model; Step 2: Conduct hybrid power pulse characteristic test, collect battery discharge data, and fit the relationship between open circuit voltage and SOC; Step 3: Use the WSO-LM hybrid algorithm to perform parameter identification to obtain the parameters of the fractional-order lithium battery model, namely, the battery internal resistance R0, polarization internal resistances R1 and R2, and polarization capacitors C1 and C2; Get the fractional-order lithium battery model after parameter identification; Step 4: Taking the relationship expression between the open circuit voltage and the SOC as the true value of the SOC, the fractional-order multi-innovation adaptive unscented Kalman filter algorithm, i.e., the FOMIAUKF algorithm, is used for the fractional-order lithium battery model after parameter identification to calculate the error and obtain the predicted value of the SOC of the lithium battery pack; Step 4 includes the following steps: (1) Algorithm initialization: Among them, x0 is the initial value of the state variable of the model, and its value is x0=SOC0, w0, v0 are the initial values ​​of process noise and measurement noise, and the default value is 0, P xx,0 is the initial value of the covariance of the state estimation, Q0, R0 are the variances of the process noise and the measurement noise, SOC0 is the initial value of the voltage of the lithium battery pack determined by the open circuit voltage method; E[·] represents the expected value of the calculation; (2) Generate Sigma sample point X i : Get the state vector 2n+1 Sigma sample points X i , calculate the sampling points and the corresponding weight vectors, and then according to the UT transformation, use the nonlinear state function to transmit the sampling points to obtain the Sigma point at time k-1: In the formula, γ is the proportional factor, x k-1 is the model state variable at time k-1, X i is the Sigma point, P xx，k / k-1 is the variance of the state variable; (3) Calculate the weighting coefficient: In the nonlinear transfer function X i，k / k-1 =f(X k-1 ,i k ) and substitute the Sigma point to get the next prediction: X i,k / k-1 =f(X k-1 ,i k )+w k-1 i=1,...,2n Among them, X i，k / k-1 is the predicted value of Sigma point, X k-1 is the Sigma point at time k-1, i k is the current value at time k, w k-1 is the process noise at time k, f(X k-1 ,i k ) is a nonlinear transfer function, and n is n sampling points; then the predicted value of the state variable is: Where i is the i-th sampling point, m and c are superscripts of different weight coefficients, W i m is the weight coefficient, and its value is: From this, the nonlinear transfer formula of the Sigma sampling point set can be obtained as follows: λ i,k / k-1 =f(X i,k / k-1 ,I k ) Among them, λ i，k / k-1 is the Sigma point after nonlinear transmission; (4) Time update: By calculating the weights and covariance, we can find the mean and variance of the observed variables: in, is the mean value of the observed variable, P xx，k / k-1 is the variance of the state variable, Q k-1 is the k-1 moment process noise covariance, W i m and W i c is the weight coefficient; (5) Measurement update: According to the measurement function ζ i，k / k-1 =G(λ i，k / k-1 ,i k ) update the observed values, the mean and variance of the observed variables are: Among them, P yy，k / k-1 is the variance of the observed variable, ζ i，k / k-1 is the updated observation value, W i m and W i c is the weight coefficient, R k-1 is the covariance of the measurement noise; The joint covariance matrix of state variables and observation variables is: The Kalman gain is: K x，k =P xy，k / k-1 (P yy，k / k-1 ) -1 Among them, (P yy，k+1 / k ) -1 P xy，k+1 / k The inverse matrix of Define the Kalman gain matrix: K p，k =[K x，k K x，k-1 … K X，K-P+1 ] (6) Update the posterior state value: The innovation sequence is: and k =and k -and k / k-1 Among them, e k is the new information at time k, y k is the real-time voltage measurement value at time k, y k / k-1 is the predicted value of the observed variable at time k; Define the multi-innovation matrix as: Among them, E p，k is the innovation matrix, p is the innovation length, is the mean of the observed variable at time k; The state variable estimate is updated as: Among them, K p，k is the Kalman gain, is the predicted value of the state variable at time k, x k is the updated state variable at time k; (7) Update the posterior state error covariance: Among them, K x，k is the Kalman gain, P xx，k / k-1 is the predicted value of the state error covariance at time k, P y，k / k-1 is the state error covariance at time k, P xx，k is the updated posterior state error covariance; K x，k The transposed matrix of .

2. The battery SOC prediction method based on the FOMIAUKF algorithm according to claim 1, characterized in that: In step 1, a second-order RC equivalent model is selected as the lithium battery equivalent circuit model in terms of the fractional-order model, including the open-circuit voltage U oc , terminal voltage V o , battery internal resistance R0, polarization internal resistance R1 and R2, polarization capacitors C1 and C2; polarization internal resistance R1 is connected in parallel with polarization capacitor C1, polarization internal resistance R2 is connected in parallel with polarization capacitor C2, and the open circuit voltage is U oc Connected in sequence with the battery internal resistance R0, the polarization internal resistance R1 and the polarization capacitor C1 after parallel connection, and the polarization internal resistance R2 and the polarization capacitor C2 after parallel connection; the state space discrete system equation and measurement equation of the fractional-order lithium battery model are: Among them, x k+1 is the state space discrete system equation, k is the variable at the current moment, U oc，k is the open circuit voltage value at time k, y k is the measurement equation; A, B, K j , C are parameter matrices, I k is the current value flowing through the battery internal resistance R0 at time k, Q new Indicates the rated capacity of the battery, SOC k is the estimated value at time k, SOC k+1 is the predicted value at the next moment, that is, the k+1 moment. The order of the polarized capacitor C1 is α, and the order of the polarized capacitor C2 is β.

3. The battery SOC prediction method based on the FOMIAUKF algorithm according to claim 2, characterized in that: The step 3 comprises: based on the fractional-order lithium battery model and the known open circuit voltage U OC The relationship between the SOC function and the WSO-LM hybrid algorithm is used for parameter identification. The war strategy optimization algorithm WSO generates an initial population in a random manner, in which each individual includes seven genomes, namely R0, R1, R2, C1, C2, α, and β. These seven genomes are the battery model parameters. For each individual, a fitness check is performed, and individuals with high fitness are selected for crossover and mutation to generate a new population. The Levenberg-Marquardt operator is added to the war strategy optimization algorithm WSO, and the model parameter correction expression of the LM algorithm is: h(l+1)=h(l)-[J T J+λI] -1 J T f(h) λ>0 Where I is the unit matrix, J is the Jacobian matrix, λ is the damping factor, h(l+1) is the battery model parameter vector of the l+1th iteration, h(l) is the battery model parameter vector of the,th iteration, and f(h) is the battery model error function; Where: h is the battery model parameter vector, is the actual terminal voltage of the qth sample, is the terminal voltage estimate of the battery model parameters corresponding to the qth sample, 1≤q≤N, N is the number of samples; In the WSO-LM hybrid algorithm, the calculation formulas for crossover probability and mutation probability are: Among them, Cross(j) is the crossover probability of the jth individual, Mutation(j) is the mutation probability of the jth individual, J(j) is the individual fitness of the jth individual, and J max is the maximum value of the sum of squares of the differences between the output measured in the population and the output predicted by the model, J ave is the average of the sum of squares of the differences between the output measured within the group and the output predicted by the model, K1 and K2 are the initial values ​​of the crossover probability and mutation probability, respectively; Finally, the battery internal resistance R0, polarization internal resistances R1 and R2, and polarization capacitors C1 and C2 are obtained.

4. A battery management system based on the FOMIAUKF algorithm, characterized in that: It includes a lithium battery pack, a battery management chip, a central control unit, a voltage detection module, a temperature detection module, a current detection module, a battery protection and balancing circuit; The lithium battery pack includes a plurality of lithium batteries, the lithium battery pack is connected to a battery management chip, the battery management chip is connected to a voltage detection module, a temperature detection module, and a current detection module, and the battery management chip and the voltage detection module, the temperature detection module, and the current detection module realize voltage detection, temperature detection, and current detection of the lithium battery pack; The voltage detection module, temperature detection module, and current detection module are connected through I 2 C communication mode is connected to the central control unit, which is used to feed back the voltage detection, temperature detection and current detection results of the lithium battery pack to the central control unit; the central control unit is used to implement a battery SOC prediction method based on the FOMIAUKF algorithm as described in any one of claims 1-3, and the battery SOC is measured according to the voltage detection, temperature detection and current detection results of the lithium battery pack; the central control unit is connected to the lithium battery pack through a battery protection and balancing circuit to ensure that the capacity and performance of each single battery remain consistent.

5. A battery management system based on FOMIAUKF algorithm according to claim 4, characterized in that: The central control unit uses the STM32F103C main control chip, the battery management chip model is BQ76930, and in the voltage detection module, the MP9486A buck chip is used to perform DC-DC buck processing on the battery pack.

6. A battery management system based on FOMIAUKF algorithm according to claim 5, characterized in that: In the current detection module, a current sensitive resistor RSNS is designed between the SRN and SRP pins of the battery management chip, and an RC filter circuit is added between the current sensitive resistor RSNS and the SRN and SRP pins for sequential current detection.

Images