Lithium ion battery state-of-charge estimation method based on fusion of genetic algorithm and harmony search

Abstract

The invention provides a lithium ion battery state-of-charge estimation method based on fusion of a genetic algorithm and harmony search. The method comprises the steps of firstly establishing a lithium battery equivalent circuit model and a dynamic system equation; obtaining current, voltage and SOC data through testing; then obtaining an elite population corresponding to the initial identification parameter by using a genetic algorithm; further optimizing the solution by using a harmony search algorithm to obtain a secondary parameter identification result; and finally, constructing a battery model by using a secondary parameter identification result, and carrying out SOC estimation by using Kalman filtering. According to the method, the genetic algorithm and the harmony search algorithm are combined, the stability of the genetic algorithm is utilized to search the approximate positioning potential solution space, and then the flexibility of the harmony search algorithm is utilized to carry out finer optimization, so that high-precision SOC online estimation is realized, and the problem of low SOC prediction precision of the genetic algorithm is effectively solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of lithium ion battery state evaluation, in particular to a lithium ion battery state of charge (SOC) prediction method based on genetic algorithm and harmony search fusion. BACKGROUND

[0002] Lithium ion batteries have high energy density, long cycle life, high working voltage, low self-discharge rate, and are light and small, and are widely used in new energy vehicle field and new energy storage system of smart grid. In actual use, the battery management system is generally used to manage and control the battery, which is one of the ways to obtain battery state estimation. Battery SOC estimation is one of the core technologies of battery management system, and is the basis for preventing overcharge and overdischarge of the battery and realizing balanced management. The higher the accuracy of SOC estimation, the higher the efficiency of the battery, and the longer the service life. At the same time, SOC is also an important information transmitted to the vehicle control system during vehicle driving. The vehicle control system estimates the remaining driving range and the energy feedback from the regenerative braking system to the battery by obtaining the remaining battery capacity and the change of the capacity from SOC. Therefore, the theoretical research and engineering implementation of SOC estimation have long-term significance for the development of electric vehicles. At present, the SOC prediction methods mainly include traditional method, model method, filtering algorithm, data-driven method and fusion method. The traditional method includes open circuit voltage method (OCV), alternating current impedance method, ampere-hour integral method (AHI). The models include equivalent circuit model (ECM), electrochemical model (EM) and electrochemical impedance model (EIM). The filtering methods include Kalman filter (KF), particle filter (PF) and H∞ filter (HIF). The data-driven methods include support vector machine (SVM), Gaussian process regression (GPR), neural network algorithm (NN) and random forest algorithm (RFA). According to the literature "A review of lithium-ion battery state of charge estimation and management system in electric vehicle applications: Challenges and recommendations", the traditional method is simple to implement, but has large error; the model method is affected by the complexity of the model in calculation amount and accuracy, and the model established is often not universal in different types of batteries; the filtering method needs to combine with the model for SOC prediction, and the prediction accuracy is affected by the model accuracy; the data-driven method relies on historical data to establish a network model, and has large calculation amount, and the prediction accuracy is related to the training. The fusion method combines the traditional method, model and algorithm, and combines two or more algorithms to realize the complementary advantages of the methods and obtain better calculation speed or prediction accuracy. The fusion method has better time-varying tracking ability and state prediction ability, and realizes the complementary advantages of the algorithms.

[0003] The SOC prediction of lithium battery is affected by internal and external factors, the lithium battery system shows strong nonlinearity and time-varying characteristics, resulting in difficulty in SOC prediction. In the prior art, using a genetic algorithm for SOC estimation is a main method, the genetic algorithm has adaptability and randomness in intelligent optimization, has good global optimization ability, and can perform global optimization on an optimal solution problem.

[0004] Although good results can be achieved, the genetic algorithm itself is a biological behavior of reference to the cross variation of chromosomes, which makes the relevant information of the original chromosome coding change very limited between generations, so that the effect of the later application is not ideal. The harmony search algorithm is a new heuristic global search algorithm that has recently emerged and has been successfully applied to many combinatorial optimization problems. The algorithm introduces two parameters, harmony memory consideration rate (HMCR) and pitch adjustment rate (PAR), in order to achieve a balance between the algorithm's development and exploration capabilities of the solution space. This is similar to the value law of the crossover probability and mutation probability in the genetic algorithm, but greatly widens the development space, allowing the initial solution obtained by the genetic algorithm to be further optimized. At present, using a biological heuristic-based intelligent algorithm for estimation has become a trend and can be used to improve the final application performance.

[0005] The existing SOC algorithm has the problem of insufficient accuracy, because the recognition parameters searched by the genetic algorithm are not accurate enough. In order to ensure biological rationality, the genetic algorithm exchanges the gene fragments of individuals each time. This change mode makes it impossible for the gene differences between individuals to be completely different, which limits the search space of the to-be-recognized parameters. Because different arrangements of parameters will have different effects on the final result, if more diverse arrangements and combinations can be made, better results can be obtained. This difference between parameters and the random change process of each instrument in the harmony search are very similar. Although the space difference flexibility of the harmony search is greater, this makes it unstable to directly use the harmony search in SOC prediction, which is similar to a random combination of the tones of multiple instruments. With a high probability, it will be noise rather than beautiful harmony. SUMMARY

[0006] Therefore, the application provides a lithium ion battery state of charge estimation method based on a genetic algorithm and a harmony search.

[0007] The specific technical scheme is as follows, as shown in the following formula: Figure 2 The lithium ion battery state of charge estimation method based on a genetic algorithm and a harmony search comprises the following steps. Step 1: establishing a lithium battery equivalent circuit model and a dynamic system equation; Step 2: performing OCV testing on the lithium battery, measuring the open circuit voltage of the lithium battery under different SOC conditions, and fitting the test data to obtain a relationship curve between the open circuit voltage and the SOC; Step 3: optimizing the parameter identification process in the lithium battery equivalent circuit model by using a genetic algorithm to obtain an elite population corresponding to the initial identification parameters; Step 4: mapping the elite population to a harmony search library as an initial solution group, and performing further optimal parameter identification by using a harmony search algorithm to obtain secondary parameter identification results; Step 5: constructing the lithium battery equivalent circuit model by using the secondary parameter identification results, and constructing an extended Kalman filter to realize online estimation of the SOC of the lithium battery under a cycle working condition.

[0008] Further, the lithium battery equivalent circuit model adopts a second-order RC lithium battery equivalent circuit model, and the to-be-identified parameters in the second-order RC lithium battery equivalent circuit model comprise an electrochemical polarization resistance R1, an electrochemical polarization capacitance C1, a concentration polarization resistance R2, and a concentration polarization capacitance C2.

[0009] Further, the input of the genetic algorithm comprises the charge and discharge data of the lithium battery, the relationship between the open circuit voltage U oc and the SOC, the ohmic resistance value R0 of the lithium battery, and the charge and discharge data of the lithium battery, wherein the charge and discharge data comprises a battery charge and discharge current I L and a battery terminal voltage U L ​The method for obtaining the ohmic resistance value R0 is: obtaining the residual capacity SOC value of the lithium battery in each period, and obtaining the ohmic resistance value R0 in the offline condition through pulse discharge.

[0010] Further, the genetic algorithm maps the search space of the to-be-identified parameters to a genetic space, each set of parameter values of the to-be-identified parameters is a chromosome in the solution space, and any to-be-identified parameter is referred to as a gene. The genetic algorithm exchanges a gene segment of a chromosome each time to search and roughly locate a potential solution space by using the stability of the genetic algorithm.

[0011] Further, the step 4 specifically comprises, Step 401: the optimal individual obtained by the genetic algorithm each time is encoded into the harmony memory as an initial solution, and the individual in the elite population generated in step 3 is also encoded into the harmony memory HM; Step 402: a cycle for updating the optimal solution is started, and the number of cycles N is set; Step 403: a random number between 0 and 1 is generated for all to-be-identified parameters, the random number is compared with a preset harmony memory consideration rate HMCR, if the random number is less than the HMCR, a variable is randomly selected in the HM, otherwise, a variable is randomly selected in a preset allowable range; Step 404: the variable selected in step 403 is randomly disturbed according to a preset pitch adjustment rate PAR; Step 405: the disturbed parameter obtained in step 404 is judged to be better, if the disturbed parameter is better, the better disturbed parameter is added to the HM as a harmony instrument pitch; the judgment standard that the disturbed parameter is better is that: the disturbed parameter is used as the parameter of the equivalent circuit model of the lithium battery to construct the equivalent circuit of the lithium battery; the error of the output average voltage of the constructed circuit is determined based on the dynamic system equation of the constructed circuit; if the error is less than the minimum value of the voltage error corresponding to the harmony instrument pitch in the existing harmony memory, the disturbed parameter is better; Step 406: steps 403 to 405 are repeated until the number of cycles reaches N, and the optimal solution in the HM at this time is selected as the final secondary parameter identification result.

[0012] Further, the step 5 comprises, Step 501: the equivalent circuit model of the lithium battery is constructed according to the secondary parameter identification result, and the state equation and the observation equation of the equivalent circuit; Step 502: a discrete nonlinear system equation of an extended Kalman filter is established; Step 503, using the obtained voltage and current data, under the iterative recursive calculation of the extended Kalman filter discrete nonlinear system equation, repeatedly iterate the last cycle under the condition of completing the online estimation of SOC.

[0013] Compared with the prior art, the present application has the following beneficial effects: 1. The present application uses real number coding and harmony search algorithm to optimize the equivalent circuit model parameters, compared with the traditional genetic algorithm which only uses the optimal individual, the present application breaks through the genetic structure limitation of genetic algorithm and identifies the optimized model parameters; 2. The present application combines genetic algorithm and harmony search algorithm, uses the stability of genetic algorithm to search and roughly locate the potential solution space, uses the identification parameters corresponding to all elite population individuals in the last generation of traditional genetic algorithm as different instrument tones in harmony, initializes the harmony memory bank in harmony search, and then uses the flexibility of harmony search algorithm to obtain more fine and better results. The present application not only solves the problem of poor search space diversity of genetic algorithm, but also solves the problem of poor stability of directly using harmony search for parameter identification; 3. The voltage output characteristic of the optimized model obtained by the present application is basically consistent with the actual voltage output characteristic of the battery, and has high identification accuracy; The present application substitutes the battery parameters identified by the harmony search algorithm into the extended Kalman filter for SOC prediction, which greatly improves the accuracy of the estimation model. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 The equivalent circuit model of the lithium battery; Figure 2 The overall method flowchart provided by the present application; Figure 3 The harmony search method flowchart proposed by the present application. DETAILED DESCRIPTION

[0015] In order to make the technical scheme of the present application more specific, a specific embodiment is given here, and it should be noted that the present embodiment is only one of them, as long as the scheme of combining genetic algorithm and harmony search algorithm for better battery parameter identification and then SOC prediction mentioned in the present application is within the protection scope of the present patent.

[0016] Step 1, establish the equivalent circuit model of the lithium battery (attached Figure 1 ) and the dynamic system equation, wherein the equivalent circuit model of the lithium battery adopts the second-order RC lithium battery equivalent circuit model, and the parameters in the second-order RC lithium battery equivalent circuit model include the electrochemical polarization resistance R1, the electrochemical polarization capacitance C1, the concentration polarization resistance R2 and the concentration polarization capacitance C2.

[0017] The dynamic system equation expression is as follows: SOC(t) = SOC(0) - (1) (2) (3) (4) In the formula, U OC is the open-circuit voltage of the lithium battery, U L is the battery terminal voltage, R0 is the ohmic resistance value, R1 and R2 are the electrochemical polarization resistance and concentration polarization resistance, respectively, C1 and C2 are the electrochemical polarization capacitance and concentration polarization capacitance, respectively, U1 and U2 are the terminal voltages of the battery on both ends of R1 and R2, SOC(0) is the initial value of the state of charge of the lithium battery, SOC(t) is the remaining capacity of the battery at time t, I L is the battery charging and discharging current, and t is the charging and discharging time.

[0018] Step 2, perform OCV test of lithium battery, measure open-circuit voltage value of lithium battery under different SOC, fit test data to obtain OCV-SOC relationship curve; Specifically, the OCV test is as follows, Step 2.1. Constant current discharge to cut-off voltage: discharge the lithium battery (room temperature) at 0.3A constant current to 2.8V cut-off voltage, and define the SOC at this time as 0.

[0019] Step 2.2 Constant current constant voltage charging to full charge: After standing for 2 hours, constant voltage charging is performed at 0.3A constant current and 0.03A cut-off current to 4.2V, and when the current drops to 0.03A or below, the SOC at this time is defined as 100%, and the battery is left to stand for 2 hours.

[0020] Step 2.3. OCV test: cycle the following steps ten times: after standing for 10 seconds, constant current discharge for 2 hours (record the current and terminal voltage of charging and discharging), and then stand for 2 hours (SOC decreases by about 10%), measure the open-circuit voltage OCV corresponding to the SOC at this moment, until the SOC decreases to 0.

[0021] Step 2.4. Curve fitting: use software to perform exponential fitting on the measured OCV and SOC data to obtain the OCV-SOC relationship curve.

[0022] Since the ohmic internal resistance is obviously different from the polarization internal resistance, the characteristics of the polarization internal resistance and the polarization capacitance in the battery, the measurement method and the application scene, etc., it is beneficial to more accurately analyze and evaluate different performance indicators of the battery to separately identify the ohmic internal resistance, so as to provide more targeted basis for the design, manufacture, use and management of the battery. Therefore, the HPPC (Hybrid Pulse Power Characterization) is used to obtain the ohmic internal resistance.

[0023] Step 3, the genetic algorithm is used to optimize the parameter identification process in the equivalent circuit model of the lithium battery, and the elite population corresponding to the initial identification parameters is obtained; Before using the genetic algorithm to identify the parameters, data preparation is needed. The charge and discharge data of the lithium battery are obtained, including the current (I L ) and the voltage (U L ), the SOC value of the battery at each time period is obtained, the ohmic resistance value R0 in the offline condition is obtained by pulse discharge, the relationship between the open circuit voltage OCV and the SOC, the ohmic resistance value R0, and the charge and discharge data are obtained. Then these data are input to the genetic algorithm for optimization to obtain the parameter identification result values of R1, R2, C1 and C2.

[0024] The step of optimizing by the genetic algorithm specifically includes: Step 301, mapping the search space of the to-be-identified parameters to the genetic space, the parameter value of each group is a chromosome or individual in the solution space (chromosome and individual one-to-one correspondence), the preset population size is used to generate the initial population using a random sequence; the to-be-identified parameters are the electrochemical polarization internal resistance R1, the electrochemical polarization capacitance C1, the concentration polarization resistance R2 and the concentration polarization capacitance C2; Step 302, encoding the initial population, a combination of all to-be-identified parameters is called a chromosome, and any to-be-identified parameter is called a gene; Step 303, crossing the current population, exchanging part of the gene fragments to generate new individuals; the new individuals are offspring; Step 304, simulating the mutation behavior in the biological evolution process, randomly generating mutations in the offspring genes; Step 305, decoding the offspring after crossing and mutation to obtain the numerical value of the offspring; Step 306, the decoded offspring is substituted into the fitness function, and it is determined whether the individual meets the pre-set condition, that is, the output average voltage error of the circuit composed of the identification parameters corresponding to the individual gene sequence is less than the pre-set value, here 15mV, if it meets, the individual that meets the condition the most is taken as the identification parameter, if it does not meet, the offspring is subjected to elite selection, the individual whose output average voltage error of the circuit composed of the identification parameters corresponding to the individual gene sequence is relatively small is selected to replace the poor individual in the last generation population to generate a new population, and the operation is cycled from step 303 to step 306, until there is an individual that meets the pre-set condition, or the number of generations exceeds the pre-set number of generations, the operation is stopped. At this time, the population composed of all individuals in the last generation is called an elite population.

[0025] Further, the initial solution of the chromosome in the genetic algorithm is set for the identification parameter, here the value range of R1 can be selected as 1-10mΩ, the value range of R2 can be selected as 0.1-1mΩ, the value range of C1 can be selected as 1000-5000F, the value range of C2 can be selected as 50-200F, and the size of the genetic population is set as N g Here, N g times of initialization are performed in the selected range, and an initial population containing N g individuals is obtained The identification parameter corresponding to each individual is input into formulas (2), (3), and (4), and the current is input at the same time, and the open circuit voltage is obtained according to the fitted curve, and the terminal voltage corresponding to the current individual can be obtained, and then the error of the terminal voltage is calculated. The reciprocal of the error is called the fitness of the individual, which will be used as a screening condition to select the optimal individual. In this case, the higher the fitness, the more advantageous the individual is in the genetic process.

[0026] Two individuals are randomly selected from the individuals to exchange gene fragments to obtain new individuals, and part of the individuals are selected for local gene mutation evolution, and the fitness of the new generation of individuals is calculated. If the fitness is higher than that of the last generation, the corresponding individual of the last generation is replaced to form a new population after evolution The above process is repeated until an individual that meets the pre-selected condition (the output average voltage error of the circuit composed of the identification parameters corresponding to the individual gene sequence is less than the pre-set value, here 15mV) appears or the number of generations exceeds the pre-set number of generations (here the pre-set number of generations is 1000) Unlike the traditional genetic algorithm that retains the optimal individual, the genetic algorithm is used in this method to obtain an approximate initial solution to reduce the instability of the harmony search, so the new population composed of individuals with higher fitness in the last population is called an elite population (the entire last population can also be used as an elite population, which can be flexibly set) Step 4, using the elite population obtained in step 3 as the initial solution group to map into the harmony search library, further optimal parameter identification is performed by using the harmony search algorithm to obtain the quadratic parameter identification result; Further, as shown in Figure 3 Step 4 specifically includes: Step 401, the optimal individual obtained in step 3 is encoded into the harmony memory (Harmony Memory, HM) as the initial solution, and the individuals in the elite population generated in step 3 are also included in the harmony memory; Step 402, start the loop of updating the optimal solution, and set a fixed number of cycles N; Step 403, generate a random number between 0 and 1 for the i-th to-be-identified parameter, and compare it with the harmony memory consideration rate X set in advance, if the generated number is less than HMCR, randomly select the variable in HM, otherwise, randomly select the variable within the allowed range; the harmony memory consideration rate (Harmony Memory Consideration Rate, HMCR) is a probability parameter for determining the source of the next pitch value of the variable in the memory library, representing the probability decision mechanism of selecting a value from the harmony memory library or randomly generating a value from the memory library. Step 404, randomly perturb the variable selected in step 403 according to the pitch adjustment rate (Pitch Adjustment Rate, PAR); the pitch adjustment rate is the change rate of each instrument pitch; Step 405, repeat steps 403-404 until M to-be-identified parameters are perturbed; Step 406, make a better judgment on the perturbed parameters obtained in step 405, if the perturbed parameters are better, add the harmony to HM; wherein, the better judgment means that the output average voltage error of the circuit composed of the identification parameters corresponding to the harmony instrument pitch is less than the minimum value of the voltage error corresponding to the harmony in the existing harmony library; Step 407, repeat 403 to 406 until the number of cycles reaches N, and select the optimal solution in HM at this time as the final quadratic parameter identification result.

[0027] The process of step 4 is further described below.

[0028] First, the initialization of the harmony memory library, the harmony search regards each to-be-identified parameter as a kind of instrument, and the value corresponding to the parameter as the pitch of the instrument, so during the initialization process, each individual of the elite population is mapped to each harmony in the harmony memory library, each gene of the individual is mapped to a kind of instrument in each harmony, and the value corresponding to the gene is the pitch of the instrument, so that the initial harmony memory library can be obtained.

[0029] Then, the harmonies in the harmony library are updated, and a total of N updates are set. In each update, each instrument needs to be processed individually, which is why the harmony search is more flexible. In contrast, the genetic algorithm operates on fragments. Here, it is assumed that there are a total of M parameters to be identified. For each parameter, the first thing to determine is whether the value of this parameter in this update comes from the harmony memory library or is randomly selected within a predetermined range. This is determined by the Harmony Memory Consideration Rate X (X ranges from 0 to 1). In practice, a random number between 0 and 1 is generated using a random number generator, and then compared with X. If the generated number is less than HMCR, a random tone is selected for the instrument from the existing tones in HM. Otherwise, a tone is randomly set for the instrument within the allowed range. To increase the variation of the instrument tones, a Pitch Adjustment Rate (PAR) is set to adjust the existing tones within a certain range, thereby obtaining an updated tone. After M cycles, a new harmony is obtained.

[0030] Next, the new harmony obtained is judged for optimality. The standard for harmony selection is to be harmonious and pleasant to listen to, while the selection criterion for the harmony search is whether the corresponding solution is optimal. This is similar to the genetic algorithm, which only needs to map the harmony in the harmony search to the parameters to be identified, and then calculate the corresponding terminal voltage error. If the new harmony is better, it is added to the harmony memory library. At the same time, to reduce the amount of calculation, the harmony with the worst harmony degree in the harmony memory library is removed to keep the total number of harmonies in the harmony memory library unchanged. After the above process is repeated for a total of N updates, the most beautiful harmony in the harmony memory library is selected, i.e., the identification parameters corresponding to the model prediction with the smallest terminal voltage error.

[0031] Step 5: Construct a battery model using the secondary parameter identification results obtained in step 4 to implement online estimation of the SOC of a lithium battery under cyclic conditions using an extended Kalman filter.

[0032] Further, the step 5 specifically includes: Step 501: First, construct the corresponding model according to the optimal identification parameters obtained in step 4, and establish the state equation and observation equation of the lithium battery model, which are specifically formulas (5-6): = + (5) (6) where τ1 and τ2 represent two different time constants, which can be calculated as τ1 = R1C1, τ2 = R2C2, U1 and U2 are the voltages on the polarization capacitors C1 and C2, respectively, SOC(k) is the estimation of SOC at time k; I L (k) is the current at time k, and is the input variable of the state equation; Q N is the rated capacity of the battery, T is the sampling period, U L (k) is the estimation of the total polarization voltage at time k, U1(k) and U2(k) are the estimations of the polarization voltages on R1 and R2 at time k, respectively, v(k) and ω(k) are Gaussian noises; Step 502, establish the discrete nonlinear system equation of the extended Kalman filter, expressions (7-8) are: = (7) = (8) where x k is the state vector, f(x k-1 , θ k-1 , u k-1 ) is the process equation, θ k-1 is the model parameter vector, u k-1 is the input vector, ω k-1 is the process excitation noise, z k is the observation vector, h(x k-1 , θ k-1 , u k-1 ) is the observation equation, and v k-1 is the observation noise; Step 503, according to the actual situation, using voltage and current data, under the iterative recursion calculation of the extended Kalman filter, repeatedly iterate to finally complete the online estimation of SOC under the cyclic working condition. Among them, the voltage and current data sources include direct collection in the test process, calculation by existing formulas / equations, public data sets (such as NASA database), etc.

[0033] The performance of the identification parameters obtained by combining the genetic algorithm and the harmony search method in the SOC estimation process is better than that of the genetic algorithm or the harmony search method alone.

Claims

1. A lithium-ion battery state of charge estimation method based on genetic algorithm and harmony search, characterized in that: Step 1, establishing a lithium battery equivalent circuit model and a dynamic system equation; Step 2, performing OCV test of the lithium battery, measuring the open circuit voltage value of the lithium battery under different SOC conditions, and fitting the test data to obtain the relationship curve between the open circuit voltage and the SOC; Step 3, using genetic algorithm to optimize the parameter identification process of the lithium battery equivalent circuit model, and obtaining an elite population corresponding to the initial identification parameters; Step 4, mapping the elite population to a harmony search library as an initial solution group, and using harmony search algorithm to further identify the optimal parameters to obtain a secondary parameter identification result; Step 5, using the secondary parameter identification result to construct the lithium battery equivalent circuit model and to construct an extended Kalman filter to realize online estimation of the lithium battery SOC under cyclic conditions. 2.The lithium-ion battery state of charge estimation method based on genetic algorithm and harmony search according to claim 1, characterized in that: the lithium battery equivalent circuit model adopts a second-order RC lithium battery equivalent circuit model, and the to-be-identified parameters in the second-order RC lithium battery equivalent circuit model include an electrochemical polarization resistance R1, an electrochemical polarization capacitance C1, a concentration polarization resistance R2, and a concentration polarization capacitance C2. 3.The lithium-ion battery state of charge estimation method based on genetic algorithm and harmony search according to claim 1, characterized in that: 4.The lithium-ion battery state of charge estimation method based on genetic algorithm and harmony search according to claim 1, characterized in that: The genetic algorithm maps the search space of the to-be-identified parameters to a genetic space, each set of parameter values of the to-be-identified parameters is a chromosome in the solution space, and any to-be-identified parameter is referred to as a gene. The genetic algorithm exchanges a gene segment of a chromosome each time to search and roughly locate a potential solution space by using the stability of the genetic algorithm. 5.The lithium-ion battery state of charge estimation method based on genetic algorithm and harmony search according to claim 1, characterized in that: Step 401, encoding the optimal individual obtained by the genetic algorithm each time into the harmony memory as an initial solution, and also encoding the individuals in the elite population generated in step 3 into the harmony memory HM; Step 402, starting a loop for updating the optimal solution, and setting the number of loops N; Step 403, generating a random number between 0 and 1 for each to-be-identified parameter, comparing the random number with a preset harmony memory consideration rate HMCR, and if the random number is less than the HMCR, randomly selecting a variable in the HM, otherwise, randomly selecting a variable within a preset allowable range; Step 404, randomly perturbing the variable selected in step 403 according to a preset pitch adjustment rate PAR. ​ ​ The input of the genetic algorithm includes: the charge-discharge data of the lithium battery, the open circuit voltage U oc The relationship with the SOC, the ohmic resistance value R0 of the lithium battery, and the charge-discharge data of the lithium battery, wherein the charge-discharge data includes the battery charge-discharge current I L And the battery terminal voltage U L The ohmic resistance value R0 is obtained by obtaining the residual capacity SOC value of the lithium battery at each period, and performing pulse discharge to obtain the ohmic resistance value R0 under offline condition. ​ ​ ​ ​ ​ ​ ​ ​ Step 405, the disturbed parameters obtained in step 404 are judged to be better, if the disturbed parameters are better, the better disturbed parameters are added to the HM as a harmonic instrument tone; the judgment standard of the better disturbed parameters is that the disturbed parameters are used as the parameters of the lithium battery equivalent circuit model to construct a lithium battery equivalent circuit; based on the constructed circuit, a dynamic system equation is used to determine the error of the output average voltage of the constructed circuit; if the error is smaller than the minimum value of the voltage error corresponding to the harmonic instrument tone in the existing harmony library, the disturbed parameters are better; Step 406, repeat steps 403 to 405 until the number of cycles reaches N, and select the optimal solution in the HM at this time as the final secondary parameter identification result.

6. The lithium ion battery state of charge estimation method fused with the genetic algorithm and the harmony search according to any one of claims 1-5, characterized in that: Step 5 comprises, Step 501, constructing the lithium battery equivalent circuit model according to the secondary parameter identification result, and the state equation and the observation equation of the equivalent circuit; Step 502, establishing the discrete nonlinear system equation of the extended Kalman filter; Step 503, using the obtained voltage and current data, under the iterative recursive calculation of the discrete nonlinear system equation of the extended Kalman filter, iteratively iterating to finally complete the online estimation of the SOC under the cycle working condition.

Images