Model predictive control charging optimization method based on dynamic power state

Abstract

The invention relates to a model predictive control charging optimization method based on a dynamic power state, which initiates a'dynamic power state collaborative optimization 'mechanism, takes a real-time power upper limit as an active optimization target instead of a fixed constraint condition, and breaks through the technical bottleneck of power limitation passive response in a traditional charging strategy. The method specifically comprises the following steps: constructing an electric-thermal-aging multi-physics field coupling model of the lithium ion battery, updating electric-thermal characteristic parameters in real time through an online parameter identification algorithm, and synchronously estimating a core temperature and an aging state in combination with a double-Kalman filtering state observer; innovatively establishing a four-dimensional objective function optimization model containing a dynamic power state, and performing multi-objective collaborative optimization on a power upper limit, a charging speed, a capacity fading rate and a current fluctuation rate; and designing a dynamic rolling optimization algorithm based on a model prediction control framework, and solving the optimal charging current meeting the dynamic power distribution requirement of the power grid in real time under the hard constraint of ensuring the maximum core temperature and terminal voltage.

Description

Technical Field

[0001] This invention relates to the field of lithium battery charging technology, specifically to a model predictive control charging optimization method based on dynamic power state. Background Technology

[0002] With the global energy structure transformation, electric vehicles, as the core carrier of clean transportation, are rapidly replacing traditional fuel vehicles. However, electric vehicles have always faced two major challenges: high energy costs (mainly from battery systems) and long charging times. Although battery prices are expected to decrease, extending battery life is currently the most effective means of reducing costs. Battery aging is mainly affected by charging strategies, and there is often a trade-off between charging speed and battery life. Therefore, optimizing charging strategies is crucial to minimizing the impact on battery life while achieving fast charging.

[0003] To address this issue, breakthroughs in charging technology are crucial for the large-scale commercial application of electric vehicles. Constant current and constant voltage methods were proposed as the most basic approaches. Due to their simplicity, constant current and constant voltage charging, along with its derivatives such as multi-stage constant current charging, pulse charging, and boost charging, are widely used in many electric vehicles. However, these methods are inherently static, offering limited understanding of battery dynamics and failing to precisely control factors affecting battery safety and aging rates. In contrast, model-based charging methods demonstrate significant advantages in battery system simulation, parameter optimization, state estimation, and closed-loop control, making them more suitable for high-precision, adaptive battery charging management.

[0004] Developing model-based battery charging algorithms faces two major challenges: First, the complex nonlinear dynamics of lithium-ion batteries, involving multi-physics coupling (electrical, electrochemical, thermal, and aging), result in computationally expensive partial differential equation models with multiple time / spatial scales, making real-time applications difficult. Second, key state parameters (such as the state of charge reflecting remaining capacity and the health state characterizing overall health) rely on internal variables that cannot be directly measured (e.g., ion concentration and capacity decay), forcing research to combine model simplification and state estimation techniques to achieve feasible solutions. Existing model-based charging strategy research mostly focuses on voltage and temperature-constrained models, neglecting dynamic power and aging rate limitations. In summary, a fast-charging method for electric vehicles that can be applied online while balancing speed, safety, and lifespan expectations remains lacking. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention proposes a lithium battery fast charging method based on model predictive control. Based on the actual physical characteristic equations of the battery model, constraints for practical applications are provided to ensure efficient and rapid charging of the battery.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for fast charging a lithium battery based on dynamic power state, comprising the following steps: S1. Establish an electro-thermal-aging coupling model for lithium batteries. And the electro-thermal-aging coupled model Discretization is performed to obtain the discretized electro-thermal-aging coupled model. ; S2. Conduct pulse charge-discharge experiments on lithium batteries to obtain real-time terminal voltage and surface temperature data, and input the collected terminal voltage and surface temperature data into the discretized electro-thermal-aging coupled model. ; S3. Design a dual Kalman filter state observer for the dispersed electro-thermal-aging coupled model. Observations were conducted, and the discretized electro-thermal-aging coupling model was used. Simultaneous estimation yields the battery state of charge and core temperature-aging joint state variables; S4. Conduct constant current and constant voltage charging experiments on lithium batteries, construct a four-dimensional objective function with charging current as the control variable, considering charging time, capacity decay rate, current fluctuation rate and power deviation, and combine the physical constraints and safety constraints of the battery to form a multi-constraint optimization problem. S5. Establish a power state model and calculate the power state and power deviation. Construct a weight matrix for a four-dimensional objective function based on the power state, power deviation, state of charge, and core temperature-aging joint state variables. S6. Use a quadratic programming solver to solve the multi-constraint optimization problem of the model at the current time step and generate the optimal charging sequence.

[0007] Furthermore, the electro-thermal-aging coupling model of the lithium battery in S1 includes an electrical model, which adopts a first-order RC equivalent circuit model. The expression of the circuit model is as follows: ; ; in, This represents the derivative of the state of charge with respect to time. It refers to the battery's state of charge. It is the battery's input current. This represents the derivative of the polarization voltage with respect to time. Indicates polarization voltage. Indicates the battery's rated capacity. and These represent the polarization resistor and polarization capacitor, respectively.

[0008] Furthermore, the electro-thermal-aging coupling model of the lithium battery in S1 also includes a lumped parameter thermal model, where the rate of change of battery temperature is determined by the relationship between heat generation and heat dissipation, and its expression is as follows: ; in, This represents the derivative of temperature with respect to time. It is a heat-generating component of the battery. It is the ambient temperature. It's the battery temperature; It's thermal resistance. It is heat capacity.

[0009] Furthermore, the electro-thermal-aging coupling model of the lithium battery in S1 also includes a semi-empirical aging model, expressed as follows: ; ; in It refers to the battery charge / discharge rate. It is the capacity decay, and yes The pre-exponential factor of the multiplication function, It is activation energy. It is the basic number of aging rate. It is an index of the sensitivity of aging rate to dynamic scaling factor. It is the time step Instantaneous current, It is the universal gas constant. It is the absolute temperature of the battery.

[0010] Furthermore, the S5 power state model is a multi-constraint joint model that considers voltage, temperature, and aging, and its expression is as follows: ; in, This is the upper limit of voltage. Open circuit voltage, The upper limit of the core temperature, It is the core temperature. For thermal resistance, It is the battery current. It is the capacity decay threshold. It is the capacity decay value. It is the battery's internal resistance in ohms. and It is an aging factor.

[0011] Furthermore, the discretization process in S1 employs a recursive least squares method with a forgetting factor to identify the electrical model parameters and the lumped parameter thermal model parameters online. The specific method is as follows: The expression obtained after discretizing the circuit model is as follows: ; in, and Let be the terminal voltage and current at time k, respectively. Sampling time, The ohmic internal resistance of the battery is mainly composed of the intrinsic resistance of the electrolyte and the electrodes. and They jointly characterize the polarization effect and describe the dynamic behavior of the battery during charging and discharging.

[0012] Furthermore, the expression obtained after discretizing the lumped parameter thermal model is as follows: ; in and It is thermal resistance and heat capacity. , yes The ambient temperature at any given time yes Battery temperature at any time yes The amount of heat generated at any given moment.

[0013] Furthermore, the charged state of the S3 electrical model is used as the system's state variable, and the unscented Kalman-Busch algorithm is used to construct the state observer. The state equations and observation equations are as follows: ; ; in, The SOC prediction value at the current moment. Let be the state transition function given by the equivalent circuit model. This is the posterior estimate of SOC from the previous time step. Given the input from the previous time step, This is the actual measured value at the current moment. For the output function given by the model, map the SOC to the terminal voltage. Given the input at the current time, Is and It consists of process noise and measurement noise.

[0014] Furthermore, the physical and safety constraints of the S4 battery are as follows: ; ; ; ; in, The maximum charge and discharge current, and Upper and lower limits of battery voltage and These are the upper and lower limits of the battery's core temperature. It is the maximum capacity decay in a single step. This is the nominal capacity.

[0015] Furthermore, the four-dimensional objective function expression in S4 is as follows: ; Among them, the symbol " "This represents an estimated value." It is the target SOC value. and These represent the target input current and capacity loss term, respectively, and N is the prediction time domain. It is the weighted square 2-norm. For real-time charging power, The power state upper limit at the current moment is represented by P, Q, R, and S, which are weight matrices that need to be adjusted.

[0016] Compared with the prior art, the technical solution of this application has the following beneficial effects: 1. This model predictive control charging optimization strategy based on dynamic power state identifies parameters such as battery internal resistance and polarization capacitance in real time using a recursive least squares method with a forgetting factor, while simultaneously integrating external constraints such as grid load factor and energy storage SOC to establish a real-time correction model for the power upper limit. During peak grid periods, i.e., when the load factor is >90%, the system automatically adjusts the power upper limit to 60%-70% of the rated value, and controls charging power fluctuations within ±8% through a power deviation penalty term with a weighting coefficient δ=1.8, ensuring grid stability. During off-peak periods, the power upper limit is dynamically increased to 110%-120% of the rated value. Combined with a current fluctuation rate suppression strategy with a weighting coefficient γ=0.6, the power utilization rate is increased from 75% in the traditional strategy to over 92%. Actual testing on a 30kWh power battery pack shows that after adopting this mechanism, grid interoperability improved by 40%, and the difference in charging efficiency during peak and off-peak periods was reduced to within 15%.

[0017] 2. This model predictive control charging optimization strategy based on dynamic power state utilizes a lumped-parameter thermal model and unscented Kalman filtering for core temperature estimation, constructing a power-temperature bivariate constraint mechanism: when the core temperature exceeds 45℃, the power limit decreases linearly with increasing temperature, i.e., the power is reduced by 3% for every 1℃ increase. Simultaneously, the temperature constraint weight is adaptively adjusted, increasing the weight coefficient β from 0.5 to 1.2, suppressing heat generation by reducing the charging current. This mechanism reduces the battery core temperature by 8-10℃ under 2C charging conditions compared to traditional strategies, decreasing the probability of thermal runaway by 70%. Furthermore, the online identification of power-thermal resistance parameters has an identification error of <3%, supporting dynamic adaptation of power boundaries under different ambient temperatures, maintaining charging efficiency fluctuations of ≤12% within the temperature range of -10℃ to 50℃.

[0018] 3. The system innovatively incorporates dynamic power limit, charging time, capacity decay rate, and current fluctuation rate into a unified optimization objective, achieving multi-objective synergy through adaptive adjustment of the weight matrix. In fast charging mode, i.e., when the target SOC increases from 20% to 80%, the system automatically increases the charging speed weight (α=1.5). Simultaneously, the capacity decay penalty term β=0.8 calculated by the aging model controls the single-step decay amount within 0.01% Qnom, shortening the charging time by 25% compared to the CCCV strategy while extending cycle life by 18%-22%. When the battery state of health (SOH) is below 85%, the aging constraint weight automatically increases by 30%, suppressing irreversible capacity loss by reducing the power limit by 15%-20%. Experiments verify that this strategy can reduce the aging rate by 28%.

[0019] 4. A semi-empirical model of power change rate and SEI film growth rate is established. A power change penalty term is embedded in the objective function with a weighting coefficient η=0.4. When the power change exceeds 20% of the rated value, a current smoothing control strategy is automatically triggered. This mechanism can control the power change rate within 5W / s, effectively reducing lithium deposition side reactions and decreasing the SEI film thickness growth rate by 35%. Cycling tests of ternary lithium batteries show that after 1000 cycles, the capacity retention rate after adopting this mechanism increases from 72% to 81% compared to the traditional strategy, while the internal resistance increase is reduced by 22%.

[0020] 5. Employing model discretization and linear approximation techniques, the nonlinear problem involving the coupling of electricity, heat, and aging is transformed into a quadratic programming model. A parallel computing architecture enables control in the time domain, generating the optimal current sequence within N=10⁻¹⁵, with a single solution time of <20ms, meeting real-time control requirements. When a sudden change in grid dispatch instructions is detected, such as a power limit change of ±30%, the system adjusts the charging plan two control cycles in advance through a feedforward compensation mechanism, ensuring a power tracking error of <5%, representing a 3-fold improvement in response speed compared to traditional feedback control strategies.

[0021] 6. A dual Kalman filter algorithm is used to simultaneously estimate SOC, core temperature, aging state, and grid power margin. The state estimation error is <1.5% SOC and 2℃, providing a highly reliable input for power boundary generation. When battery parameters drift, such as a 20% increase in internal resistance due to aging, the recursive least squares method with a forgetting factor can update the parameters within 5-10 sampling periods, ensuring model prediction accuracy and avoiding power overshoot caused by model mismatch. Experiments verify that this mechanism enables the system to maintain the effectiveness of the charging strategy during battery aging, with charging efficiency degradation ≤8% after SOH ≤ 70%.

[0022] 7. Constructing a multi-layered protection threshold system for power, temperature, and voltage: When the power approaches the upper limit and the voltage reaches 4.15V, a dual derating strategy of "power-voltage" is automatically activated to limit the current to within 1.2C. If the core temperature exceeds 50℃, a temperature priority mode is activated, with a weighting coefficient β=2.0, forcibly reducing the power to below 50% of the rated value. This mechanism can maintain safe operation even under conditions such as a battery cell voltage difference >50mV and sudden changes in ambient temperature (±10℃ / min), reducing the probability of charging interruption by more than 95% compared to traditional strategies. Extreme condition tests show that the system can quickly identify and execute zoned power regulation when a battery cell overheats (local temperature 60℃), preventing fault propagation and ensuring battery safety. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating the optimized fast charging method for lithium batteries based on model predictive control, as described in this invention. Figure 2 This is a schematic diagram of the electrothermal coupling model of the cylindrical lithium battery of the present invention; Figure 3 This is the result of parameter identification using the least squares method with a forgetting factor in this invention; Figure 4 This is a comparison chart of the voltage performance of the present invention and the traditional constant current constant voltage charging algorithm; Figure 5 This is a comparison chart of the temperature performance of the present invention and the traditional constant current constant voltage charging algorithm; Figure 6 This is a comparison chart of the state of charge performance of the present invention and the traditional constant current constant voltage charging algorithm. Figure 7 This is a comparison chart of the current performance of the present invention and the traditional constant current constant voltage charging algorithm. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figure 1-7 This example provides a model predictive control charging optimization method based on dynamic power state, including: S1. Establish an electro-thermal-aging coupling model for lithium batteries. And the electro-thermal-aging coupled model Discretization is performed to obtain the discretized electro-thermal-aging coupled model. ; This invention employs, as follows Figure 1 The electrothermal coupling model shown is used to describe the characteristics of lithium-ion batteries. The electrical model expression is as follows: ; ; in, This represents the derivative of the state of charge with respect to time. It refers to the battery's state of charge. It is the battery's input current. This represents the derivative of the polarization voltage with respect to time. Indicates polarization voltage. Indicates the battery's rated capacity. and These represent the polarization resistor and polarization capacitor, respectively.

[0026] The expression for the battery's terminal voltage is as follows: ; in This is the open-circuit voltage of the battery. It is the ohmic internal resistance of the battery.

[0027] The constructed thermal model is a lumped-parameter thermal model. The rate of change of battery temperature is determined by the heat generated and the heat dissipated, and its expression is as follows: ; in, This represents the derivative of temperature with respect to time. It is a heat-generating component of the battery. It is the ambient temperature. It's the battery temperature; It's thermal resistance. It is heat capacity.

[0028] Heat-generating items The expression is as follows: ; in, This is the partial derivative of the battery open-circuit voltage with respect to the battery temperature.

[0029] The aging model uses a semi-empirical aging model, expressed as follows: ; ; in It refers to the battery charge / discharge rate. It is the capacity decay, and yes The pre-exponential factor of the multiplication factor function, the power-law factor z=0.55, It is the basic number of aging rate. It is an index of the sensitivity of aging rate to dynamic scaling factor. It is the time step Instantaneous current, It is the universal gas constant. It is the absolute temperature of the battery. The activation energy is calculated using the following formula: ; The State of Power (SOP) model is a multi-constraint joint model that considers voltage, temperature, and aging, and its expression is as follows: ; in, This is the upper limit of voltage. Open circuit voltage, The upper limit of the core temperature, It is the core temperature. For thermal resistance, It is the battery current. It is the capacity decay threshold. It is the capacity decay value. It is the battery's internal resistance in ohms. and It is an aging factor.

[0030] Furthermore, the parameters of the electrothermal model are identified using a recursive least squares method with a forgetting factor; First, based on real battery data acquisition, the parameters of the electro-thermal model are identified using Forgetting Recursive Least Squares (FFRLS). Then, the state of charge is estimated in real time using an Unscented Kalman–Bucy Filter (UKBF).

[0031] The expression obtained after discretizing the electrical model is as follows: ; in, and Let be the terminal voltage and current at time k, respectively. Sampling time, The ohmic internal resistance of the battery is mainly composed of the intrinsic resistance of the electrolyte and the electrodes. and They jointly characterize the polarization effect and describe the dynamic behavior of the battery during charging and discharging.

[0032] Regression vector and parameter vector as follows: ; ; The regression vector is the historical observation data input to the model, and the parameter vector is the unknown parameters to be identified in the model.

[0033] The electrical identification model is as follows: ; in, The current output voltage of the model. This is the model's estimate of the output. The system noise is represented by this formula. The relationship between the predicted value (model estimate) of the battery terminal voltage at the current moment and the actual measured value can be obtained from this formula, which is the output equation of the discretized electrical model.

[0034] The online parameter identification process using the FFRLS algorithm is as follows: ; in This represents the model's predicted value for the terminal voltage at the current moment. This represents the transpose of the regression vector. This represents the parameter estimate from the previous time step; ; in This represents the terminal voltage prediction error; ; in This represents the gain matrix at the current time step, which is used to determine the weight of the prediction error on the parameter update. This represents the parameter estimation covariance matrix of the previous time step. The forgetting factor controls the rate at which the weights of older data decay. It is a scalar value used to normalize the gain; ; in Estimate the covariance matrix for the updated parameters. This indicates the amount of uncertainty reduction resulting from the update; ; in, This represents the updated parameter estimation vector. This represents the parameter estimation vector from the previous time step. This represents the parameter correction amount determined by both the gain matrix and the prediction error.

[0035] The discretization expression for the lumped parameter thermal model is as follows: ; in and It is thermal resistance and heat capacity. , yes The ambient temperature at any given time yes Battery temperature at any time yes The amount of heat generated at any given moment.

[0036] Regression vector and parameter vector The expression is as follows: ; ; The heat identification model is as follows: ; in, The model outputs the temperature at the current moment. This is the model's estimate of the output. The system noise is represented by this formula. The relationship between the predicted value (model estimate) of the battery temperature at the current moment and the actual measured value can be obtained from this formula, which is the output equation of the discretized thermal model.

[0037] The online parameter identification process using the FFRLS algorithm is as follows: ; in This is the predicted temperature value at the current moment. This represents the transpose of the regression vector. This represents the vector of thermal model parameters estimated at the previous moment; ; in This indicates the temperature prediction error at the current moment. This indicates the temperature measured by the sensor at the current moment; ; in This represents the gain matrix at the current time step, which is used to determine the weight of the prediction error on the parameter update. This represents the parameter estimation covariance matrix of the previous time step. The forgetting factor controls the rate at which the weights of older data decay. It is a scalar value used to normalize the gain; ; in This represents the current thermal model parameter vector. This is the parameter estimation vector of the thermal model at the previous time step. The parameter correction amount is determined by both gain and error.

[0038] S2. Conduct pulse charge-discharge experiments on lithium batteries to obtain real-time terminal voltage and surface temperature data, and input the collected terminal voltage and surface temperature data into the discretized electro-thermal-aging coupled model. ; S3. Design a dual Kalman filter state observer for the dispersed electro-thermal-aging coupled model. Observations were conducted, and the discretized electro-thermal-aging coupling model was used. Simultaneous estimation yields the battery state of charge and the joint state variables of core temperature-aging; the specific scheme is as follows: Since the state of charge (SOC) values ​​are unmeasurable in the electrical model, a state observer is needed for real-time estimation. The state equations and observation equations are as follows: ; ; in, The SOC prediction value at the current moment. Let be the state transition function given by the equivalent circuit model. This is the posterior estimate of SOC from the previous time step. Given the input from the previous time step, This is the actual measured value at the current moment. For the output function given by the model, map the SOC to the terminal voltage. Given the input at the current time, Is and These are process noise and measurement noise. This is the standard observation equation in the observer, used to compare new measurements with predicted values ​​to generate residuals, thus updating the SOC measurement.

[0039] The unscented Kalman-Busch filter, compared to the unscented Kalman filter, features continuous-time propagation and discrete-time measurement updates. The core process is as follows: Generate Sigma points: ; This step yields 2n+1 deterministic sampling points.

[0040] The expression for continuous-time propagation based on each Sigma point is as follows: ; This step yields the trajectory of each Sigma point in the continuous time domain. .

[0041] Predicted state after propagation Covariance The expression is as follows: ; ; in and These are weighting coefficients. It is the process noise covariance matrix.

[0042] This step yields the prior state estimate. and prior covariance .

[0043] The discrete-time measurement update is as follows: ; This step yields the output prediction for each Sigma point. ; ; This step allows us to observe the predicted value. ; ; This step yields the observed-predicted covariance. ; ; This step yields the cross-covariance between the state and the observation. ; ; From this step, the Kalman gain matrix can be obtained. ; ; in For the observed variables, this step yields the updated state variables. ; ; This step yields the updated covariance. ; in It is the predicted output. It is a prediction of the observed mean. It is the measurement noise covariance matrix. It predicts the observed covariance. It is the cross-covariance between the state and the observation. express Inverse of a matrix To represent the transpose of a matrix, It is the Kalman gain matrix. It is the updated state variable. It updates the covariance.

[0044] S4. Conduct constant current and constant voltage charging experiments on lithium batteries, construct a four-dimensional objective function with charging current as the control variable, considering charging time, capacity decay rate, current fluctuation rate and power deviation, and combine the physical constraints and safety constraints of the battery to form a multi-constraint optimization problem. By determining the core temperature, capacity decay safety threshold, and terminal voltage boundary values, and using input current, capacity decay, charging speed, and dynamic power state as objective functions, a multi-objective, multi-constraint charging strategy based on model predictive control algorithm is constructed.

[0045] During battery charging, the battery's maximum current, terminal voltage, capacity decay rate, and core temperature need to be controlled below a reasonable threshold. Based on these requirements, the charging constraints are set as follows: ; ; ; ; in, The maximum charge and discharge current, and Upper and lower limits of battery voltage and These are the upper and lower limits of the battery's core temperature. It is the maximum capacity decay in a single step. This is the nominal capacity.

[0046] The four-dimensional objective function expression is as follows: ; Among them, the symbol " "This represents an estimated value." It is the target SOC value. and These represent the target input current and capacity loss term, respectively, and N is the prediction time domain. It is the weighted square 2-norm. For real-time charging power, The power state upper limit at the current moment is represented by P, Q, R, and S, which are weight matrices that need to be adjusted.

[0047] S5. Establish a power state model and calculate the power state and power deviation. Construct a weight matrix for a four-dimensional objective function based on the power state, power deviation, state of charge, and core temperature-aging joint state variables. S6. Use a quadratic programming solver to solve the multi-constraint optimization problem of the model at the current time step and generate the optimal charging sequence. The specific implementation method is as follows: use a quadratic programming solver to solve the model prediction optimization problem at the current time step, and based on the feedback from the state observer in the current closed-loop system, dynamically adjust the weight matrix according to the real-time state of charge, power state, and deviation to generate the optimal current sequence.

[0048] The standard quadratic programming form, which transforms the objective function into linear constraints, is as follows: ; in, express The transpose of the matrix, To optimize the variable vector, The matrix is ​​a Hessian matrix. express The transpose of the matrix, It is a linear coefficient vector. It is an equality constraint matrix. It is an equality constraint vector. It is an inequality constraint matrix. It is an inequality constraint vector. It is the upper and lower bound vector of the variable.

[0049] The quadratic programming (QP) approach transforms the real-time optimization problem of model predictive control (MPC)—that is, dynamically adjusting the weight matrix based on the state of charge, power, and deviation feedback from the state observer, and comprehensively considering terminal voltage, temperature, aging rate, and power constraints—into a standard mathematical framework. By efficiently solving this QP problem, it generates the optimal current prediction sequence for the future time domain, and takes only the first current value in the sequence as the optimal charging current command actually applied at the current moment. Through closed-loop feedback rolling optimization, it dynamically tracks the target while satisfying all safety constraints, and overcomes model errors and avoids the risk of overshoot due to voltage, temperature, aging, or power.

[0050] The optimized fast charging decision current is input into a closed-loop feedback system based on model predictive control to avoid battery terminal voltage, core temperature, aging rate, and power overshoot caused by model errors, and to determine the electric vehicle fast charging current at the current moment.

[0051] In this specific embodiment, specifically in this patent embodiment, the current obtained from the simulation of the Model Predictive Control (MPC) charging strategy is applied to the actual battery charging process. For example... Figure 5 As shown in the summary, while the CCCV method is simple and reliable, it has limitations in dynamic adjustment and constraint handling, and requires a relatively long charging time in the high-charge-state region because the current gradually decreases during the constant-voltage phase, leading to reduced charging efficiency. In contrast, the MPC strategy can provide a better balance between robustness and response speed in the battery system, effectively optimizing changes in voltage, temperature, current, SOP, and SOC, thereby improving charging efficiency and battery life.

[0052] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0053] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A model predictive control charging optimization method based on dynamic power state, comprising the following steps, characterized in that: S1. Establish an electro-thermal-aging coupling model for lithium batteries. And the electro-thermal-aging coupled model Discretization is performed to obtain the discretized electro-thermal-aging coupled model. ; S2. Conduct pulse charge-discharge experiments on lithium batteries to obtain real-time terminal voltage and surface temperature data, and input the collected terminal voltage and surface temperature data into the discretized electro-thermal-aging coupled model. ; S3. Design a dual Kalman filter state observer for the dispersed electro-thermal-aging coupled model. Observations were conducted, and the discretized electro-thermal-aging coupling model was used. Simultaneous estimation yields the battery state of charge and core temperature-aging joint state variables; S4. Conduct constant current and constant voltage charging experiments on lithium batteries, construct a four-dimensional objective function with charging current as the control variable, considering charging time, capacity decay rate, current fluctuation rate and power deviation, and combine the physical constraints and safety constraints of the battery to form a multi-constraint optimization problem. S5. Establish a power state model and calculate the power state and power deviation. Construct a weight matrix for a four-dimensional objective function based on the power state, power deviation, state of charge, and core temperature-aging joint state variables. S6. Use a quadratic programming solver to solve the multi-constraint optimization problem of the model at the current time step and generate the optimal charging sequence.

2. The model predictive control charging optimization method based on dynamic power state as described in claim 1, characterized in that, The electro-thermal-aging coupling model of the lithium battery in S1 includes an electrical model, which is a first-order RC equivalent circuit model. The expression of the circuit model is as follows: ； ； in, This represents the derivative of the state of charge with respect to time. It refers to the battery's state of charge. It is the battery's input current. This represents the derivative of the polarization voltage with respect to time. Indicates polarization voltage. Indicates the battery's rated capacity. and These represent the polarization resistor and polarization capacitor, respectively.

3. The model predictive control charging optimization method based on dynamic power state as described in claim 1, characterized in that, The electro-thermal-aging coupling model of the lithium battery in S1 also includes a lumped parameter thermal model, where the rate of change of battery temperature is determined by the relationship between heat generation and heat dissipation, and its expression is as follows: ； in, This represents the derivative of temperature with respect to time. It is a heat-generating component of the battery. It is the ambient temperature. It's the battery temperature; It's thermal resistance. It is heat capacity.

4. The model predictive control charging optimization method based on dynamic power state as described in claim 1, characterized in that, The electro-thermal-aging coupling model of the lithium battery in S1 also includes a semi-empirical aging model, the expression of which is as follows: ； ； in It refers to the battery charge / discharge rate. It is the capacity decay, and yes The pre-exponential factor of the multiplication function, It is activation energy. It is the basic number of aging rate. It is an index of the sensitivity of aging rate to dynamic scaling factor. It is the time step Instantaneous current, It is the universal gas constant. It is the absolute temperature of the battery.

5. The model predictive control charging optimization method based on dynamic power state as described in claim 1, characterized in that, The S5 power state model is a multi-constraint joint model that considers voltage, temperature, and aging, and its expression is as follows: ； in, This is the upper limit of voltage. Open circuit voltage, The upper limit of the core temperature, It is the core temperature. For thermal resistance, It is the battery current. It is the capacity decay threshold. It is the capacity decay value. It is the battery's internal resistance in ohms. and It is an aging factor.

6. The model predictive control charging optimization method based on dynamic power state as described in claim 2, characterized in that, The discretization of the electrical model is achieved by using a recursive least squares method with a forgetting factor to identify the electrical model parameters and thermal model parameters online. The specific method is as follows: The expression obtained after discretizing the circuit model is as follows: ； in, and Let be the terminal voltage and current at time k, respectively. Sampling time, The ohmic internal resistance of the battery is mainly composed of the intrinsic resistance of the electrolyte and the electrodes. and They jointly characterize the polarization effect and describe the dynamic behavior of the battery during charging and discharging.

7. The model predictive control charging optimization method based on dynamic power state as described in claim 3, characterized in that, The expression obtained after discretizing the lumped parameter thermal model is as follows: ； in and It is thermal resistance and heat capacity. , yes The ambient temperature at any given time yes Battery temperature at any time yes The amount of heat generated at any given moment.

8. The model predictive control charging optimization method based on dynamic power state as described in claim 1, characterized in that, The charged state of the S3 electrical model is used as the system's state variable, and the unscented Kalman-Busch algorithm is used to construct the state observer. The state equation and observation equation are as follows: ； ； in, The SOC prediction value at the current moment. Let be the state transition function given by the equivalent circuit model. This is the posterior estimate of SOC from the previous time step. Given the input from the previous time step, This is the actual measured value at the current moment. For the output function given by the model, map the SOC to the terminal voltage. Given the input at the current time, Is and It consists of process noise and measurement noise.

9. The model predictive control charging optimization method based on dynamic power state as described in claim 1, characterized in that, The physical and safety constraints of the S4 battery are as follows: ； ； ； ； in, The maximum charge and discharge current, and Upper and lower limits of battery voltage and These are the upper and lower limits of the battery's core temperature. It is the maximum capacity decay in a single step. This is the nominal capacity.

10. The model predictive control charging optimization method based on dynamic power state as described in claim 1, characterized in that, The four-dimensional objective function expression in S4 is as follows: ； Among them, the symbol " "This represents an estimated value." It is the target SOC value. and These represent the target input current and capacity loss term, respectively, and N is the prediction time domain. It is the weighted square 2-norm. For real-time charging power, The power state upper limit at the current moment is represented by P, Q, R, and S, which are weight matrices that need to be adjusted.

Images