An Active Equalization Control Strategy and Method for Power Lithium Batteries Based on PSO-GA-FCM Clustering

Abstract

The present invention relates to an active equalization control strategy and method for power lithium batteries based on PSO-GA-FCM clustering, which is characterized in that it includes the following steps: First, the real-time state of charge SOC of each lithium battery cell is found according to the SOC-OCV curve; after obtaining the SOC value, the FCM clustering algorithm optimized by PSO and GA is used to perform data clustering and optimization analysis on each single lithium battery to judge the state of each single lithium battery; finally, the clustering result is transmitted through a single-chip microcomputer, the equalization state is judged, an equalization command is sent, and the conduction and cutoff of the MOSFET tube are controlled to charge and discharge and equalize each lithium battery. The present invention has applicability, can solve the problem of inconsistent power in a discrete battery pack, reduce the energy difference between each single battery in the battery pack, and improve the SOC consistency of the battery pack.

Description

Technical Field

[0001] The present invention belongs to the technical field of battery balancing, and particularly relates to an active balancing control strategy and method for power lithium batteries based on PSO-GA-FCM clustering. Background Art

[0002] The active balancing management system for lithium batteries: It consists of a lithium battery pack, a main control unit, a data acquisition unit, RS485, a CAN bus communication module, an upper computer and a display unit, and a balancing control unit. The single-chip microcomputer calculates parameters such as the voltage, current, and temperature of the lithium battery collected by the balancing module, and performs battery balancing control through the collected battery parameters, so as to realize the balancing management during the charging and discharging process of the power lithium battery.

[0003] Ternary lithium-ion batteries have the advantages of high energy density, good temperature characteristics, and high safety, and are widely used in various energy storage systems. However, the voltage level and capacity of a single battery cell are limited. In order to meet the requirements of high voltage and high power of the energy storage system, it is necessary to connect battery cells in series and parallel to form a battery pack. However, due to the uneven temperature field of each single battery cell and the inconsistency of parameters such as the internal resistance and Coulomb efficiency of the single battery cell, the remaining power of the single battery cells in the battery pack will be inconsistent during repeated charging and discharging processes, which limits the overall available capacity of the battery pack. In order to solve the problem of inconsistent remaining power of the batteries, it is necessary to use a battery management system to control the charging and discharging of the batteries.

[0004] The battery management system is a link between the battery and the user. Its main target is lithium iron phosphate batteries or ternary lithium batteries. It is mainly to improve the utilization rate of the battery, increase the service life of the battery, and prevent the phenomena of overcharging and over-discharging of the battery. It can be used in electric vehicles, underwater vehicles, robots, drones, etc.

[0005] In order to address the inconsistency problem in the lithium battery pack, an equalization technology is usually integrated into the battery management system BMS. Equalization technology generally refers to a special technical means introduced to avoid or reduce the adverse effects on aspects such as the capacity utilization rate, output power, and service life of the battery pack caused by the inconsistency problem of the battery pack. The current equalization technology has become a key technology in the battery management system BMS. An efficient equalization measure can improve the effective use capacity of the entire battery and extend the service life of the battery. The current research on equalization technology mainly focuses on two aspects: equalization control strategy and equalization circuit topology design. The research on the equalization strategy of the battery pack focuses on establishing evaluation indicators for the inconsistency of each single battery cell in the battery pack and proposing effective equalization control methods based on this; while the equalization circuit topology design focuses on the design and improvement of an equalization circuit structure with high efficiency, simple control structure, and relatively low cost.

[0006] The problems to be studied in the balancing control strategy are the selection of the balancing variable and the control of the start and end times of balancing. Most of the existing technologies study based on voltage consistency, capacity consistency, and state of charge (SOC) of the battery as the balancing variable. Using SOC as the balancing variable can achieve better balancing effects, and the system is also easy to control, which can better reflect the true state of the lithium battery pack. The balancing strategy based on SOC consistency is to set the SOC threshold, discharge the battery with a SOC greater than the threshold, and perform charge management on the battery with a SOC less than the threshold. Since the SOC threshold is based on the OCV-SOC curve characteristics of the lithium battery itself, there are problems of inaccurate balancing accuracy and estimation. When the battery ages or the voltage measurement is inaccurate, it will directly affect the estimation, resulting in unnecessary balancing and switching losses. The traditional FCM algorithm requires artificial determination of the initial clustering center and the number of clusters, which has great subjectivity and uncertainty, and the algorithm is prone to falling into local extrema.

[0007] Therefore, the present invention proposes an active balancing control strategy and method for power lithium batteries based on PSO-GA-FCM clustering to achieve the SOC consistency of each single battery in the battery pack and the balancing control of the battery pack. Summary of the Invention

[0008] Aiming at the deficiencies in the existing technologies in the above background content, an active balancing control strategy and method for power lithium batteries based on PSO-GA-FCM clustering of the present invention can solve the problem of inconsistent power of discrete battery packs, and can reduce the energy difference between battery monomers in the battery pack, thereby improving the consistency of the battery pack.

[0009] To achieve the above objectives, the technical solution adopted by the present invention is characterized by including the following steps:

[0010] Use the improved PSO-GA-FCM algorithm to cluster the sample data, and the algorithm steps are as follows.

[0011] Step S101: Given the number of categories c, the fuzzy index m, the population size N, the learning factors C1, C2, the crossover probability P c , the mutation probability P m , the inertia weight ω, the threshold ε, and the maximum number of iterations T.

[0012] Step S102: The battery SOC reflects the usage of the remaining capacity of the current battery, which is defined as the ratio of the current remaining capacity to the storage capacity when the battery is fully charged. As shown in formula (1), according to the SOC-OCV curve of the lithium battery pack characteristics, the SOC value of the battery under real-time working conditions is accurately calculated using the look-up table method. According to formula (2), n lithium batteries are sorted according to the size of SOC. The data with the maximum SOC value is defined as the center μ1 of class 1, the data of the arithmetic mean of the maximum SOC value and the minimum SOC value is defined as the center μ2 of class 2, and the value of the minimum SOC value is defined as the center μ3 of class 3.

[0013]

[0014] Where: C n is the nominal capacity of the battery, i(τ) represents the current value of the battery at time τ, represents the SOC value at time t0, η represents the Coulomb efficiency, and SOC t represents the SOC value at time t;

[0015] Where: SOC max is the maximum SOC value, SOC min is the minimum SOC value, is the arithmetic mean of SOC, and μ i is the clustering center;

[0016] Step S103: Three initial clustering centers μ1, μ2, μ3 are initially set to form three first-generation examples. The current position of each particle is its P besti , and the best position among all particles in the current population is g best . The fitness value is calculated using formula (3). The most fit individual in the population is the particle with the largest fitness value among all particles.

[0017]

[0018] Where: k is a constant, j(U, V) is the objective function of FCM. The smaller j(U, V) is, the better the clustering effect is, and the higher the particle fitness f(x i ) is.

[0019] Step S104: Update the velocity and position of each particle using formula (4) and formula (5). According to the crossover probability P c , and the mutation probability P m , perform selection, crossover, and mutation operations to generate the next generation of particle swarm.

[0020] V (i+1)d = ωV id + c1rand1(P id-x id ) + c2rand2(P gd -x gd ) (4)

[0021] X (i+1)d = X id + V (i+1)d (5)

[0022] In the formula, ω is the inertia weight, usually linearly decreasing from 0.9 to 0.2, c1 and c2 are acceleration constants, V (i+1)d corresponds to the velocity of the particle, X (i+1)d corresponds to the position of the particle, and rand1 and rand2 are two random functions varying within the range [0, 1].

[0023] Step S105: Calculate the fitness value of each individual in the new particle swarm using formula (3). First, compare it with the individual of the previous generation. If it is greater than the fitness value of the previous generation, then replace the individual of the previous generation to become P besti , and the fitness value is the best fitness value of the individual. Otherwise, keep it unchanged.

[0024] Step S106: Then compare it with the fitness of the optimal individual in the population. If it is greater than the fitness value of the optimal individual in the population, then replace the global optimal individual g best with this individual, and its fitness value is the best fitness value of the population at this time. Otherwise, the global optimal and the best fitness value of the population remain unchanged.

[0025] Step S107: The iteration number T = T + 1.

[0026] Step S108: When the algorithm reaches the maximum number of iterations of evolution or the set threshold ε, that is, when the fitness of the population has not improved, the algorithm stops. Otherwise, jump to step S104.

[0027] Step S109: Generate the population with the best fitness and the best initial clustering center.

[0028] Step S110: Input the generated initial clustering center in FCM for data classification.

[0029] Step S111: Output the clustering result.

[0030] The steps of using the FCM algorithm are as follows.

[0031] Step S201: FCM fuzzifies the degree to which each data in the data criterion set belongs to each clustering center into a value from 0 to 1, mainly by dividing the samples according to the membership degrees of the data to be clustered to the C clustering centers. The clustering model of FCM is shown in formula (6):

[0032]

[0033] Step S202: In the formula, x i is the i-th data to be clustered; v k is the k-th clustering center; u ki is the fuzzy membership degree that the i-th data to be clustered x i belongs to the k-th clustering center; m is the fuzzy exponent, m ∈ [1, ∞], and its value can not only affect the clustering performance of FCM, but also measure the fuzzy degree of the FCM algorithm. U is the fuzzy membership degree matrix composed of u ki , also known as the partition matrix; V is the matrix composed of C clustering centers v k .

[0034] Step S203: Transform the above formula into an optimization problem with constraints, and change the model into the form shown in Equation (7):

[0035]

[0036] Step S204: Therefore, the FCM problem is transformed into a conditional extreme value problem with Lagrange multipliers λ. Respectively, take partial derivatives of the clustering center v k and the fuzzy membership degree u ki to obtain the update iteration rules for both as shown in Equation (8):

[0037]

[0038] Step S204: By continuously iteratively updating v k and u ki , when FCM stops running, and outputs the fuzzy membership degree matrix U and the center matrix V.

[0039] The steps for judging the lithium battery balancing operation are as follows:

[0040] Step S301: Use formula (9) to calculate the range ΔSOC of the battery pack.

[0041] ΔSOC = max(SOC[i]) - min(SOC[j]) (9)

[0042] In the formula, i and j are the battery numbers, and SOC[i] represents the SOC value of the battery numbered i.

[0043] Step S302: When the result of ΔSOC is less than the set threshold, end and wait for the next estimated value of the single battery SOC. Otherwise, perform the next balancing operation;

[0044] Step S303: Make a judgment based on the clustering result output in Step S111. Charge and balance the single lithium battery whose SOC is lower than the first clustering center, and discharge and balance the single lithium battery whose SOC is higher than the second clustering center.

[0045] The balancing module is used to balance the batteries that need to be balanced according to the control of the secondary MCU in the small battery pack.

[0046] Preferably, the balancing module adopts a bi-directional flyback transformer.

[0047] Preferably, the bi-directional flyback transformer adopts the LTC3300-1 chip.

[0048] The present invention has the following beneficial effects:

[0049] (1) It has strong parameter adaptability, is easy to adjust parameters, and is convenient for operation;

[0050] (2) The active balancing control strategy and method for power lithium batteries based on PSO-GA-FCM clustering in the present invention has a faster balancing speed and better improvement in battery capacity compared with the traditional battery SOC balancing. Description of the Drawings

[0051] Figure 1 is the flow chart of the optimized FCM algorithm of the present invention

[0052] Figure 2 is the flow chart of the FCM clustering analysis of the present invention

[0053] Figure 3 is the flow chart of the clustering balance of the present invention

[0054] Figure 4 is the schematic circuit diagram of the active balancing unit of the present invention

[0055] Figure 5 is the balance control diagram of the present invention

[0056] Figure 6 is the schematic diagram of the SOC balance effect of the present invention Detailed Embodiment

[0057] The technical solution of the present invention will be specifically described below with reference to the drawings. There is provided an active balancing control strategy and method for power lithium batteries based on PSO-GA-FCM clustering, including the following steps:

[0058] S1. Sort the n lithium batteries according to the SOC value using formula (1). Define the data with the maximum SOC value as the center μ1 of class 1, the arithmetic mean of the maximum SOC value and the minimum SOC value as the center μ2 of class 2, and the minimum SOC value as the center μ3 of class 3. The SOC is calculated using the ampere-hour integration method, as shown in equation (2). First, look up the real-time state of charge SOC of the n lithium battery cells according to the SOC-OCV curve of the lithium battery pack characteristics. n ;

[0059]

[0060]

[0061] S2. Initially set three initial clustering centers μ1, μ2, μ3 to form three first-generation examples. The current position of each particle is its P besti , and the best position among all particles in the current population is g best . Calculate the fitness value using equation (3). The most fit individual in the population is the particle with the largest fitness value among all particles.

[0062]

[0063] In the formula: k is a constant, and j(U,V) is the objective function of FCM. The smaller j(U,V) is, the better the clustering effect, and the higher the particle fitness f(x i ).

[0064] S3. Update the velocity and position of each particle using formula (4) and formula (5). Based on the crossover probability P c , and the mutation probability P m , perform selection, crossover, and mutation operations to generate the next generation of particle swarm.

[0065] V (i+1)d = ωV id + c1rand1(P id - x id )+ c2rand2(P gd - x gd ) (4)

[0066] X (i+1)d = X id + V (i+1)d (5)

[0067] In the formula, ω is the inertia weight, usually linearly decreasing from 0.9 to 0.2, c1 and c2 are acceleration constants, and rand1 and rand2 are two random functions that vary within the range [0,1].

[0068] S4. Calculate the fitness value of each individual in the new particle swarm using formula (3). First, compare it with the individual in the previous generation. If it is greater than the fitness value of the previous generation, then replace the individual in the previous generation to become P besti , and the fitness value is the best fitness value of the individual. Otherwise, keep it unchanged. Then compare it with the fitness value of the optimal individual in the group. If it is greater than the fitness value of the optimal individual in the group, then replace the global optimal individual g with this individual best , and its fitness value is the best fitness value of the group at this time. Otherwise, the optimal individual and the best fitness value of the group remain unchanged (the optimized FCM partial process is as shown in Figure 1 ). When the algorithm reaches the maximum number of iterations of evolution or the set threshold ε, that is, when the fitness of the population has not improved, the algorithm stops. Otherwise, jump to step S3

[0069] S5. Generate the population with the best fitness, and the best initial clustering centers

[0070] S6. Use the FCM algorithm to perform clustering analysis on SOC according to the generated optimal initial clustering centers (the FCM clustering analysis flowchart is as shown in Figure 2 ).

[0071] S7. FCM fuzzifies the degree to which each data in the data criterion set belongs to each clustering center into a value between 0 and 1. It mainly divides the samples according to the membership degrees of the data to be clustered to the C clustering centers. The clustering model of FCM is shown in formula (6):

[0072]

[0073] S8. In the formula, x i is the i-th data to be clustered: v k is the k-th clustering center; u ki is the fuzzy membership degree of the i-th data to be clustered x i belonging to the k-th clustering center; m is the fuzzy index, m ∈ [1, ∞]. Its value can not only affect the clustering performance of FCM, but also measure the fuzzy degree of the FCM algorithm. U is the fuzzy membership degree matrix composed of u ki , also known as the partition matrix; V is the matrix composed of C clustering centers v k .

[0074] S9. Transform the above formula into an optimization problem with constraints, and change the model into the form shown in formula (7):

[0075]

[0076] S10. Therefore, the FCM problem is transformed into a conditional extreme value problem with Lagrange multipliers λ. Respectively, for the clustering center v k and the fuzzy membership degree u kiTaking the partial derivative, the update iteration rules for both are shown in Equation (8):

[0077]

[0078] S10. By continuously iteratively updating v k and u ki until FCM stops running, and the fuzzy membership matrix U and the center matrix V are output.

[0079] S7. As shown in Figure 3 is the clustering equilibrium flowchart. According to the clustering results output in step S6, the single lithium batteries with SOC lower than the first clustering center are charged for balancing, and the single lithium batteries with SOC higher than the second clustering center are discharged for balancing.

[0080] Principle of the active unit balancing circuit: The active unit balancing circuit is as shown in Figure 4 where C n 、C n -1、C m 、C m -1 represent the nth, (n - 1)th, mth, and (m - 1)th battery cells respectively. BOOT represents the top of the 6 series-connected battery cells (the positive electrode of the 6th battery cell), V- represents the end of the 6 series-connected battery cells (the negative electrode of the 1st battery cell), T n 、T m represent flyback transformers. G n P、C m P、I n P、I m P、G n S、C m S、I n S、I m S are respectively connected to the G n P、C m P、I n P、I m P、G n S、C m S、I n S、I m S pins of LTC3300. S n 1、S n 2、S m 1、S m 2 are all MOSFET switches.

[0081] If the voltage of the C n th battery cell is too high and needs to be balanced, the main control module unit sends a balancing control command to LTC3300-1, and the G n P pin of LTC3300-1 becomes high level, and Sn 1 conducts, and current flows through transformer T n The primary side discharges to battery C n I n The current flowing through is detected by the GP pin. When the current reaches the peak value of 10 A, the equalization is completed, and the main control module unit sends a stop equalization command to LTC3300-1 n The GP pin becomes low level, and n1 turns off. The main control module unit sends an equalization control command to LTC3300-1, and the GS pin of LTC3300-1 outputs high level m T m of S m 2 conducts, and S m 2 and S n 2 form a loop, and current flows through the secondary side of transformer T m I m The current flowing through is detected by the GS pin. When the current reaches the peak value of 10 A, the main control module unit sends a stop equalization command to LTC3300-1 m The GS pin becomes low level, and S m 2 turns off. The main control module unit sends an equalization control command to LTC3300-1, and the GP pin of LTC3300-1 becomes high level m Current flows through the primary side of T m to charge battery C m I m The current flowing through is detected by the GP pin. When the current reaches the peak value of 10 A, the main control module unit sends a stop equalization command to LTC3300-1 m The GP pin becomes low level, and S m 2 turns off, and the equalization is completed. By repeating this process multiple times, active equalization of 6 series-connected batteries can be achieved

[0082] Overall equalization control principle: The overall equalization control process is as follows Figure 5As shown, the equalization control mainly consists of genetic algorithm, particle swarm optimization, and FCM clustering. The single-chip microcomputer first obtains the data sampled by the AD and the battery data collected by the battery acquisition chip, calculates the initial state of charge of each battery unit through these data, and then divides the N battery units into three categories through the FCM clustering algorithm optimized by particle swarm optimization and genetic algorithm inside the single-chip microcomputer, and saves and outputs the category relationship. The equalization chip LTC3300-2 controls the switch states of each battery pack according to the equalization command issued by the single-chip microcomputer, so as to achieve equalization control. At the same time, during the switch control process, it is possible that the voltage of a single battery cell exceeds the threshold value (4.17V). The single-chip microcomputer judges whether to activate the active equalization through the data sampled by the AD. When the voltage exceeds the threshold value, the single-chip microcomputer gives a signal to the active equalization drive unit to drive the corresponding battery switch state. When the voltage of the single battery cell exceeding the threshold value is less than or equal to the average battery voltage of the corresponding battery unit, the equalization stops.

[0083] The simulation results are as Figure 6 shown Figure 6 in the figure is the schematic diagram of the active equalization effect of 6 series-connected batteries in different SOC states. It can be seen from the figure that the battery with the minimum SOC value is preferentially charged, and the other batteries are discharged. When the SOC values of all batteries tend to be the same state, the individual batteries are charged as a whole. It can be seen from the figure that under this method, the charging time of each battery is short and the charging speed is fast.

[0084] The above embodiments are only illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Any person familiar with this technology can modify or change the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or changes made by those with ordinary knowledge in the technical field without departing from the spirit and technical idea disclosed by the present invention should still be covered by the claims of the present invention.

Claims

1. A power lithium battery active equalization control strategy and method based on PSO-GA-FCM clustering, characterized in that: It includes the following steps: S1. The battery SOC reflects the usage of the remaining capacity of the current battery, which is defined as the ratio of the current remaining capacity to the storage capacity when the battery is fully charged, as shown in formula (1): Where: C n is the nominal capacity of the battery, i(τ) represents the battery current value at time τ, represents the SOC value at time t0, η represents the Coulomb efficiency, and SOC t represents the SOC value at time t; Accurately calculate the SOC value of the battery under real-time working conditions by using the look-up table method based on the SOC-OCV curve of the lithium battery pack characteristics; S2. Sort the SOC values of each battery calculated by the look-up table method according to their magnitudes, and set the maximum value, minimum value, and arithmetic mean of SOC as the clustering centers of the initial clustering algorithm according to formula (2); Where: SOC max is the maximum value of SOC, SOC min is the minimum value of SOC, is the arithmetic mean of SOC; S3. Use GA and PSO to optimize the set particle centers, calculate the fitness value using formula (3), and the most fit individual in the population is the particle with the largest fitness among all particles; where: k is a constant; j(U, V) is the objective function of FCM. The smaller j(U, V) is, the better the clustering effect is, and the higher the particle fitness f(x i ) will be; S4. The solution of the PSO algorithm corresponds to a particle in the solution space of the problem to be solved. Each particle has a corresponding velocity, position, and fitness value determined by the objective function. Each iteration of the particle in the solution space is based on the currently found better solution to find the next solution. The \(i\)-th particle is represented as \(x\) i =(x i1 ,x i2 ,…,x id ), and the best position it has experienced is denoted as \(P\) i =(P i1 ,P i2 ,…,P id ), also known as \(P\) besti . The velocity of particle \(i\) is represented by \(V\) i =(V i1 ,V i2 ,…,V id ). For each generation of particles, its velocity and position in the \(d\)-th dimension change according to the following equations: V (i+1)d = ωV id + c1rand1(P id - x id ) + c2rand2(P gd - x gd ) (4) X (i+1)d = X id + V (i+1)d (5) In the formula: ω is the inertia weight, usually linearly decreasing from 0.9 to 0.2, c1 and c2 are acceleration constants, and rand1 and rand2 are two random functions varying within the range of [0, 1]; Update the velocity and position of each particle using equations (4) and (5), and perform selection, crossover, and mutation operations according to the crossover probability P c , and the mutation probability P m to generate the next generation of particle swarm; S6. Obtain the population with the best fitness and the initial clustering center with the best fitness, input the generated output clustering center in FCM for data classification. FCM fuzzifies the degree to which each data in the data standard set belongs to each clustering center into a value between 0 and 1. It mainly divides the samples according to the membership degrees of the data to be clustered to the C clustering centers. The clustering model of FCM is shown in formula (6): where: x i is the i-th data to be clustered: v k is the k-th clustering center; u ki is the fuzzy membership degree that the i-th data x to be clustered i belongs to the k-th clustering center; m is the fuzzy exponent, m ∈ [1, ∞], and its value can affect the clustering performance of FCM and measure the fuzziness of the FCM algorithm. U is the fuzzy membership matrix composed of u ki and is also called the partition matrix; V is the matrix composed of C clustering centers v k ; Transform the above formula into a constrained optimization problem, and change the model into the form shown in formula (7); Therefore, the FCM problem is transformed into a conditional extreme value problem with Lagrange multiplier λ. Respectively, taking partial derivatives with respect to the cluster center v k and the fuzzy membership degree u ki yields the update iteration rules for both as shown in Equation (8): By continuously iterating and updating v k and u ki , when FCM stops running, and the fuzzy membership matrix U and the center matrix V are output; S7. Transmit the output clustering result to the single-chip microcomputer control chip, and the single-chip microcomputer sends an equalization control command to the active equalization control chip LTC3300-1; S8. The single-chip microcomputer controls the conduction and cut-off of each MOSFET tube in the active equalization circuit according to the equalization control command to charge and discharge the single lithium battery.

Images