Physically Constrained Gray-Box Model Reduction Construction Method and System

By constructing a physical constraint gray box reduced-order model, the global optimization problem of the electric vehicle winter thermal management system in complex environments is solved, realizing fast and stable temperature prediction and control, which is suitable for MPC prediction models.

CN123116804APending Publication Date: 2026-11-17FAW JIEFANG AUTOMOTIVE CO
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611279292.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-21
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing electric vehicle winter thermal management systems struggle to achieve global optimization under complex road gradients, load variations, and low-temperature environments. Furthermore, the black-box model lacks parameter symbol constraints and physically interpretable features, leading to false heat sources and unstable predictions.

Method used

A physical constraint gray box model for order reduction was constructed. By collecting winter thermal management data of electric vehicles, the training set and the validation set were divided, the state variables were determined and the physical heat flow feature matrix was constructed. Symbolic constraints were applied, and regularized least squares solution and multi-step rolling simulation were used for verification. The parameters of the order reduction model were then derived.

Benefits of technology

It achieves rapid and stable multi-state temperature prediction in low-temperature environments, avoids anti-physical responses, improves the interpretability and engineering credibility of the model, and is suitable for predictive control of MPC.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116804A_ABST
    Figure CN123116804A_ABST
Patent Text Reader

Abstract

This invention discloses a method and system for constructing a physically constrained gray-box model for thermal management reduction, relating to the field of thermal management control and model reduction technology for new energy vehicles. The method includes: A1. Collecting thermal management runtime sequence data of electric vehicles under winter operating conditions; cleaning, smoothing, outlier removal, and missing value filling of the data; and dividing the data into training and validation sets in chronological order. The runtime sequence data includes at least ambient temperature, vehicle speed, battery electrical parameters, four core temperature states, thermal management actuator states, and drive loss data. A2. Determining the set of state variables for the gray-box model for thermal management reduction, including at least battery temperature, battery coolant inlet temperature, passenger compartment temperature, and electric drive temperature. This invention replaces the high-dimensional thermal management model with a four-state physical thermal network, reducing the computational load and lowering the computational efficiency from minutes to milliseconds, enabling the model to be used for rapid prediction at the controller level.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of thermal management control and model reduction technology for new energy vehicles, and particularly to a method and system for constructing physical constraint gray box reduction models. Background Technology

[0002] In winter operating conditions, electric vehicles face challenges such as low-temperature performance degradation of the power battery, high energy consumption for passenger compartment heating, short window for utilizing waste heat from the electric drive, and complex coupling of heat pumps and PTC heating actuators. These factors cause the thermal management system to simultaneously face requirements for temperature safety, energy consumption control, and comfort control. Traditional control methods based on empirical rules or single PID controllers rely heavily on calibration, making it difficult to achieve global optimization under complex road gradients, load variations, and low-temperature environments.

[0003] High-fidelity thermal management simulation models can describe the coupling relationships between components such as coolant circuits, valves, pumps, fans, compressors, PTCs, and electric drive losses. However, they involve large computational loads and numerous state variables, making them unsuitable for direct embedding into the vehicle controller for real-time predictive optimization. While conventional data-driven models offer fast computation speeds, they are prone to anti-physics responses if they lack constraints on heat flow direction, actuator physical effects, and energy conservation. For example, a heating actuator might cause a predicted temperature drop, a cooling actuator might cause a predicted temperature rise, or spurious heat sources might be generated when the temperature difference condition is not met.

[0004] Existing battery thermal degradation models primarily focus on reducing the temperature field of the battery pack itself, making it difficult to cover the system-level dynamics of battery coolant, electric drive waste heat, passenger compartment, environmental heat dissipation, and actuator linkage in winter vehicle thermal management. Existing black-box neural networks or general machine learning models are highly dependent on the distribution of training data, lack parameter sign constraints and physically interpretable features, making them difficult to use as long-term stable prediction models for MPC.

[0005] Therefore, a gray-box ROM modeling method is needed that has low computational cost while maintaining the main physical channels, heat flow direction constraints, and control availability of the thermal management system, so that it can achieve fast, stable, and interpretable multi-state temperature prediction in the winter thermal management control of electric vehicles. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for constructing a physical constraint gray box reduction model, thereby solving at least one of a number of technical problems.

[0007] Core technical issues: It is difficult to achieve global optimization under complex road slopes, load changes, and low temperature environments; false heat sources are generated when the temperature difference conditions are not met; and the lack of parameter symbol constraints and physically interpretable features makes it difficult to use as a prediction model for MPC in the long term.

[0008] This invention provides the following solution: According to a first aspect of the present invention, a method for constructing a physically constrained gray-box reduced-order model is provided, comprising: A1. Collect the timing data of electric vehicle thermal management operation in winter, and divide it into training set and validation set according to the timing sequence. A2. The state variables of the gray box reduced-order model are determined to include battery temperature, battery coolant inlet temperature, passenger compartment temperature, and electric drive temperature. A3. Construct a corresponding physical heat flux characteristic matrix for each state variable, such that the rate of change of state temperature satisfies: ; This is the physical heat flux characteristic matrix. For the channel coefficients to be identified, Equivalent heat capacity; Each heat flow characteristic corresponds to a heat transfer channel in active heating, driven waste heat coupling, ambient heat dissipation, cooling circuit heat exchange, and battery internal resistance heating. A4. Apply physical sign constraints to the channel coefficients corresponding to each heat flow characteristic: the channel coefficient constraint for active heating is non-negative, the channel coefficient constraint for active cooling is non-positive, and the heat transfer direction for temperature difference driven heat exchange channels is determined by the sign of the temperature difference, with the coefficient constraint being non-negative. A5. Traverse the candidate equivalent heat capacities and solve the corresponding channel coefficients for each candidate heat capacity by combining signed constraint regularized least squares. The minimum heat capacity of RMSE is predicted by multi-step rolling simulation temperature, and combined with the channel coefficient as the parameters of the reduced-order model for this state. A6. Use the validation set to conduct closed-loop simulation verification, and export the reduced-order model parameters after successful verification.

[0009] Furthermore, in step A3, the heat flow characteristics corresponding to the battery temperature include at least the bidirectional linear heat transfer characteristics between the coolant inlet and the battery. The bidirectional linear heat transfer characteristics are constructed by multiplying the actual temperature difference with the pump speed and the loop flow distribution coefficient. The heat transfer direction is determined autonomously by the sign of the temperature difference.

[0010] Furthermore, including: in step A3, the heat flow characteristics corresponding to the battery coolant inlet temperature include the characteristics of the drive waste heat coupling group; the characteristics of the drive waste heat coupling group are constructed by multiplying the drive loss square root linear compression with the softplus soft gating function, the water pump speed and the loop allocation coefficient. The softplus soft-gating function only takes effect when the electric drive temperature is higher than the battery coolant inlet temperature and exceeds a set margin.

[0011] Furthermore, in step A3, the heat flow characteristics corresponding to the electric drive temperature include the heat generation characteristics of electric drive losses after square root linear compression, and the bidirectional linear temperature difference heat transfer characteristics between the electric drive and the coolant.

[0012] Furthermore, in step A5, the multi-step rolling simulation uses the midpoint second-order Runge-Kutta integral method for state recursion.

[0013] Furthermore, in step A6, the derived reduced-order model parameters include at least the state order, sampling period, equivalent heat capacity, feature name, channel coefficient, symbolic constraint, and state boundary.

[0014] Furthermore, the state variables can be expanded to include multi-point temperatures of the battery module, motor temperature, electronic control temperature, compressor outlet temperature, or condenser pressure.

[0015] Furthermore, in step A5, the channel coefficient identification method can be replaced by quadratic programming with physical constraints, Bayesian regression, robust regression, or recursive least squares.

[0016] According to a second aspect of the present invention, a system for constructing a physically constrained gray-box reduced-order model is provided, comprising: The data acquisition module is used to collect thermal management operation sequence data of electric vehicles under winter conditions. After cleaning, smoothing, outlier removal and missing value filling, the data is divided into training set and validation set in chronological order. The runtime sequence data includes at least ambient temperature, vehicle speed, battery electrical parameters, four types of core temperature states, thermal management actuator status, and drive loss data; the four types of core temperature states include: battery temperature, battery coolant inlet temperature, passenger compartment temperature, and electric drive temperature; The set determination module is used to determine the set of state variables of the gray box reduced-order model. The state variables include at least the battery temperature, battery coolant inlet temperature, passenger compartment temperature, and electric drive temperature. The matrix construction module is used to construct the corresponding physical heat flux characteristic matrix for each state variable, such that the rate of change of state temperature satisfies the following relationship: ; in, This is the physical heat flux characteristic matrix for this state. For the channel coefficients to be identified, This is the equivalent heat capacity in this state; Each heat flow characteristic corresponds to at least one type of heat transfer channel with a clear physical meaning, including active heating, driven waste heat coupling, ambient heat dissipation, cooling circuit heat exchange, and battery internal resistance heating. The constraint application module is used to apply physical sign constraints to the channel coefficients corresponding to each heat flow characteristic: the constraint for active heating channel coefficients is non-negative, the constraint for active cooling channel coefficients is non-positive, and the heat transfer direction for temperature difference driven heat exchange channels is determined by the sign of the temperature difference, with the coefficient constraint being non-negative. The model parameter module is used to traverse candidate equivalent heat capacities within a preset heat capacity search interval, and solve the corresponding channel coefficients for each candidate heat capacity by combining regularized least squares with sign constraints. Using the temperature prediction RMSE of single-state multi-step rolling simulation as the evaluation index, the combination of heat capacity and channel coefficient with the minimum RMSE is selected as the reduced-order model parameter for this state. The data embedding module is used to perform closed-loop simulation verification of the multi-state coupled reduced-order model using a validation set. After successful verification, the parameters of the reduced-order model can be exported and directly embedded into the vehicle controller.

[0017] According to a third aspect of the present invention, an electronic device is provided, comprising: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus; The memory stores a computer program, which, when executed by the processor, causes the processor to perform steps such as constructing a physical constraint gray-box reduction model.

[0018] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, comprising: storing a computer program executable by an electronic device, wherein when the computer program is run on the electronic device, the electronic device performs steps such as constructing a physically constrained gray-box reduction model.

[0019] The above solution achieves the following beneficial technical effects: This invention replaces the high-dimensional thermal management model with a four-state physical thermal network, reducing the amount of prediction computation and lowering the computational efficiency from minutes to milliseconds, making the model usable for rapid prediction at the controller end.

[0020] This invention uses physical heat flow characteristics and symbolic constraints to identify and avoid anti-physical results such as cooling channels cooling down and cooling channels heating up in black-box models, thereby improving the interpretability and engineering credibility of the model.

[0021] This invention enables the model to express the impact of electric drive waste heat on the battery coolant inlet temperature under winter mountainous conditions or high load conditions by driving waste heat soft gating and driving loss square root compression, while avoiding oversensitivity caused by high load samples.

[0022] This invention reduces false heat sources and improves the consistency of heat exchange direction between the battery and coolant, and between the electric drive and coolant by removing the fixed softplus bias and adopting bidirectional heat exchange based on real temperature difference.

[0023] This invention uses thermal capacity search and multi-step simulation RMSE evaluation to not only fit the derivative but also directly optimize the rolling prediction capability of the model, making ROM more suitable for MPC. Attached Figure Description

[0024] Figure 1 This is a schematic diagram comparing the accuracy of battery temperature prediction between the reduced-order model and the original data provided in a specific embodiment of the present invention.

[0025] Figure 2 This is a schematic diagram comparing the accuracy of the reduced-order model and the original data for predicting the battery inlet temperature, provided in a specific embodiment of the present invention.

[0026] Figure 3 This is a schematic diagram comparing the accuracy of the reduced-order model and the original data for electric drive temperature prediction, provided in a specific embodiment of the present invention.

[0027] Figure 4 This is a schematic diagram of the physical constraint gray box reduction model construction process provided in a specific embodiment of the present invention.

[0028] Figure 5 This is a flowchart of a method for constructing a physical constraint gray box reduction model provided by one or more embodiments of the present invention.

[0029] Figure 6 This is a structural diagram of a physical constraint gray box reduction model construction system provided by one or more embodiments of the present invention.

[0030] Figure 7 This is a block diagram of an electronic device structure based on a physical constraint gray box reduction model construction method provided by one or more embodiments of the present invention. Detailed Implementation

[0031] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0032] Figure 5 This is a flowchart of a method for constructing a physical constraint gray box reduction model provided by one or more embodiments of the present invention.

[0033] Example 1, as Figure 5 The physical constraint gray box reduction model construction method shown is aimed at winter thermal management of electric vehicles, including: A1. Collect thermal management operation sequence data of electric vehicles under winter conditions. After cleaning, smoothing, outlier removal and missing value filling, the data is divided into training set and validation set according to time sequence. The runtime sequence data includes at least ambient temperature, vehicle speed, battery electrical parameters, four types of core temperature states, thermal management actuator status, and drive loss data. A2. Determine the set of state variables for the gray box reduced-order model. The state variables should include at least the battery temperature, battery coolant inlet temperature, passenger compartment temperature, and electric drive temperature. A3. For each state variable, construct the corresponding physical heat flux characteristic matrix such that the rate of change of state temperature satisfies the following relationship: ; in, This is the physical heat flux characteristic matrix for this state. For the channel coefficients to be identified, This is the equivalent heat capacity in this state; Each heat flow characteristic corresponds to at least one type of heat transfer channel with a clear physical meaning, including active heating, driven waste heat coupling, ambient heat dissipation, cooling circuit heat exchange, and battery internal resistance heating. A4. Apply physical sign constraints to the channel coefficients corresponding to each heat flow characteristic: the channel coefficient constraint for active heating is non-negative, the channel coefficient constraint for active cooling is non-positive, and the heat transfer direction for temperature difference driven heat exchange channels is determined by the sign of the temperature difference, with the coefficient constraint being non-negative. A5. Traverse the candidate equivalent heat capacities within the preset heat capacity search interval, and solve the corresponding channel coefficient for each candidate heat capacity by combining regularized least squares with signed constraints. Using the temperature prediction RMSE of single-state multi-step rolling simulation as the evaluation index, the combination of heat capacity and channel coefficient with the minimum RMSE is selected as the reduced-order model parameter for this state. A6. Use a validation set to perform closed-loop simulation verification on the multi-state coupled reduced-order model. After successful verification, export the parameters of the reduced-order model that can be directly embedded into the vehicle controller.

[0034] Closed-loop verification includes at least the following: starting a multi-state rolling simulation with the initial state of the verification set, recursively calculating each state according to the same sampling period as the controller, calculating the RMSE, MAE and maximum absolute error of each state, and simultaneously checking the channel coefficient sign, state boundary and heat flow direction constraints; and exporting a model parameter file containing the state sequence, sampling period, equivalent heat capacity, feature name, channel coefficient, sign constraint and state boundary only when the error index and physical constraints meet the preset conditions.

[0035] In this embodiment, step A3 includes: the heat flow characteristics corresponding to the battery temperature include at least the bidirectional linear heat transfer characteristics between the coolant inlet and the battery. The bidirectional linear heat transfer characteristics are constructed by multiplying the actual temperature difference with the pump speed and the loop flow distribution coefficient. The heat transfer direction is determined autonomously by the sign of the temperature difference.

[0036] In this embodiment, step A3 includes the following: the heat flow characteristics corresponding to the battery coolant inlet temperature include the characteristics of the drive waste heat coupling group; the characteristics of the drive waste heat coupling group are constructed by multiplying the drive loss square root linear compression with the softplus soft gating function, the water pump speed and the loop allocation coefficient. The softplus soft-gating function only takes effect when the electric drive temperature is higher than the battery coolant inlet temperature and exceeds a set margin.

[0037] In this embodiment, step A3 includes the following: the heat flow characteristics corresponding to the electric drive temperature include the electric drive loss heat generation characteristics after square root linear compression, and the bidirectional linear temperature difference heat transfer characteristics between the electric drive and the coolant.

[0038] In this embodiment, step A5 includes the following: the multi-step rolling simulation uses the midpoint second-order Runge-Kutta integral method for state recursion.

[0039] In this embodiment, step A6 includes the following: the derived reduced-order model parameters include at least the state order, sampling period, equivalent heat capacity, feature name, channel coefficient, symbolic constraint, and state boundary.

[0040] In this embodiment, the state variables can be expanded to include multi-point temperatures of the battery module, motor temperature, electronic control temperature, compressor outlet temperature, or condenser pressure.

[0041] In this embodiment, step A5 includes the following: the parameter identification method can be replaced by quadratic programming with physical constraints, Bayesian regression, robust regression, or recursive least squares.

[0042] Specifically, in Example 2, a predictive control method for a new energy vehicle's anticipatory thermal management model is disclosed. This method is used for forward-looking coordinated control of a multi-loop coupled thermal management system that includes a battery temperature control loop, an electric drive waste heat loop, and a passenger compartment air conditioning loop. The method includes the following steps: T1. Construct a physical constraint gray box reduced-order thermal model (gray box reduced-order thermal model, nonlinear reduced-order thermal model) for winter operating conditions. The model should include at least four types of state variables: battery temperature, battery coolant inlet temperature, passenger compartment temperature, and electric drive temperature, which will serve as the prediction engine for model predictive control. T2. Obtain the predicted disturbance sequence (predicted sequence) of road slope, driving loss, vehicle speed and ambient temperature in the future prediction time domain. Generate multi-dimensional scene semantic score (four types of continuous scene score) through hierarchical look-ahead planning and map it to obtain the actuator bias trajectory (actuator control trajectory). T3. Construct a multi-layered composite stage cost function. The cost function should integrate at least the time-varying reference trajectory tracking term, the asymmetric under-temperature thermal debt penalty term (asymmetric thermal safety penalty term), the physical path and scene semantic reward and penalty term (physical path topology reward and penalty term), and the actuator action constraint term as the optimization objective. T4. The optimal actuator control sequence is obtained by using a convex solution link with linear time-varying approximation along the trajectory and state condensation elimination, combined with a real nonlinear cost line search acceptance mechanism and a rolling hot start mechanism. T5. Extract the first step control quantity of the optimal control sequence and output it to the thermal management actuator. In the next control cycle, update the current thermal state and the predicted disturbance sequence. Repeat steps T2 to T4 to form a rolling time-domain closed-loop control (rolling closed-loop control).

[0043] In this embodiment, the method for constructing the physically constrained gray-box reduced-order thermal model in step T1 specifically includes: constructing a corresponding physical heat flux characteristic matrix for each state variable, such that the rate of change of state temperature satisfies... ;in This is the physical heat flux characteristic matrix. For the channel coefficients to be identified, The equivalent heat capacity is used; each heat flow characteristic corresponds to at least one type of heat transfer channel with a clear physical meaning, including active heating, driven waste heat coupling, environmental heat dissipation, cooling circuit heat exchange, and battery internal resistance heating; physical sign constraints are applied to the coefficients of each channel: the coefficients of active heating channels are constrained to be non-negative, the coefficients of active cooling channels are constrained to be non-positive, and the coefficients of temperature difference driven heat exchange channels are constrained to be non-negative, with the heat exchange direction determined autonomously by the sign of the temperature difference; the candidate equivalent heat capacities are traversed within the preset heat capacity search interval, and the channel coefficients are solved by regularized least squares with sign constraints. The final model parameters are determined based on the criterion of minimizing the temperature prediction RMSE of single-state multi-step rolling simulation.

[0044] In this embodiment, the heat flow characteristics corresponding to the battery coolant inlet temperature include the drive waste heat coupling characteristics. The characteristics are constructed by multiplying the drive loss square root linear compression with a softplus soft gating function, water pump speed, and loop allocation coefficient. The softplus soft gating function only takes effect when the electric drive temperature is higher than the battery coolant inlet temperature and exceeds a set margin.

[0045] In this embodiment, the heat exchange characteristics between the battery and the coolant, and between the electric drive and the coolant, are all constructed using a bidirectional linear heat exchange form based on the actual temperature difference, with the heat exchange direction adaptively switching according to the sign of the temperature difference.

[0046] In this embodiment, the hierarchical look-ahead planning process in step T2 specifically includes: applying exponential decay weights to the predicted disturbance sequence for sliding window weighted calculation to generate semantic scores for four scenarios: waste heat recovery, preheating, precooling, and maintenance; based on the semantic scores of the four scenarios and the current thermal state, generating a five-dimensional semantic heat flow target vector containing the battery active heating heat power, the heat power of driving waste heat to the battery, the heat power of driving waste heat to the passenger compartment, the heat power of battery active cooling, and the heat power of front-end active heat dissipation; combining the capacity constraints of each heat flow path, mapping the heat flow target vector to the actuator bias trajectory, which serves as both the actuator reference term in the cost function in step T3 and the initial value for hot start optimization in step T4.

[0047] In this embodiment, the preheating score has a built-in waste heat recovery suppression term. When the waste heat recovery score increases, the preheating score automatically decreases, forming a mutually exclusive control logic that prioritizes waste heat. The uphill score is calculated using a dual mechanism that takes the maximum value of the continuous weak slope weighting term and the short-term strong slope extreme value term. An event proximity score is also introduced to amplify the gain of the near-end uphill scenario.

[0048] In this embodiment, the five-dimensional semantic heat flow target vector is smoothed by a first-order low-pass filter and then mapped to the actuator bias trajectory. Furthermore, the actuator trajectory generation incorporates a heating and cooling path conflict (heating-cooling conflict) suppression mechanism.

[0049] In this embodiment, the multi-layer composite stage cost function in step T3 specifically includes a three-layer progressive structure: The first layer is the basic tracking and actuator cost: generating a time-varying exponential approximation reference trajectory starting from the current state, constructing state tracking cost and terminal cost; constructing actuator bias cost by combining actuator bias trajectory, and constructing actuator rate of change cost by combining control quantity difference between adjacent steps; The second layer is the thermal safety differential penalty cost: extracting the unilateral under-temperature deviation of battery temperature and battery coolant inlet temperature below the reference trajectory, applying additional high penalty, forming an asymmetric under-temperature thermal debt penalty term; setting state soft constraint boundaries for all thermal states, and independently decoupling the electric drive high temperature risk; The third layer is the physical path and scene semantic reward and punishment cost: including at least waste heat reuse reward term, heating and cooling conflict penalty term, cooling actuator matching penalty term, water pump invalid action penalty term, and state maintenance action penalty term, all reward and punishment terms are continuously differentiable algebraic expressions.

[0050] In this embodiment, the asymmetric under-temperature thermal debt penalty term only applies to the battery temperature and the battery coolant inlet temperature. When the state is higher than the reference trajectory, the penalty term is 0. The electric drive high temperature risk is handled separately through an independent soft constraint upper limit and a dedicated amplification factor, and is not intertwined with the battery under-temperature penalty.

[0051] In this embodiment, the waste heat reuse reward is a negative cost, and it only takes effect when the electric drive temperature is higher than the battery coolant inlet temperature and the margin is met, and the water pump and the corresponding valve passage are connected; the heating and cooling conflict penalty is constructed by multiplying the battery-side PTC heating power and the compressor power; the hold-state action penalty dynamically amplifies the actuator change rate penalty weight according to the hold-state score.

[0052] In this embodiment, the convex solution link and acceptance mechanism in step T4 specifically includes: calling the inlined physical constraint gray box reduced-order thermal model (inlined nonlinear reduced-order thermal model) to perform forward rolling prediction in the full prediction time domain of the current candidate control sequence to obtain the state trajectory, and calculating the true nonlinear cost based on the composite stage cost function; calculating the state Jacobian matrix and control Jacobian matrix step by step along the current state trajectory to establish a linear time-varying approximate model in the control correction increment space; eliminating all state variables through cascaded recursion to compress the original nonlinear optimization problem in the state-control hybrid space into a standard convex quadratic programming subproblem with only the control correction as the sole decision variable; solving the convex quadratic programming subproblem using a hot-start method to obtain the control correction direction, using the control correction direction as the search direction, and accepting the updated control sequence by combining backtracking search with the true nonlinear cost.

[0053] In this embodiment, a near-optimal convergence guard is set: when the infinite norm of the control correction is lower than a set threshold and the actual cost change satisfies the near-optimal margin, the iteration is terminated directly; a multi-level safety backoff mechanism is set: when the line search fails, damping factor correction, trust region shrinking, and output shift of the control sequence of the previous cycle are used in sequence.

[0054] In this embodiment, the inlined physical constraint gray-box reduced-order thermal model is implemented with inline code that has a fixed physical connection order and a single function entry point, eliminating runtime dynamic parsing overhead; the line search uses the Armijo descent criterion as the acceptance criterion.

[0055] In this embodiment, the rolling hot start mechanism in step T4 is as follows: the optimal control sequence of the previous control cycle is shifted to the left by one step, and the tail is filled by zero-order maintenance as the initial control sequence of the current control cycle; during cold start, the calibration reference control quantity is used for filling.

[0056] Figure 6 This is a structural diagram of a physical constraint gray box reduction model construction system provided by one or more embodiments of the present invention.

[0057] Example 3, as Figure 6 The physical constraint gray box reduction model construction system shown includes: The data acquisition module is used to collect thermal management operation sequence data of electric vehicles under winter conditions. After cleaning, smoothing, outlier removal and missing value filling, the data is divided into training set and validation set in chronological order. The runtime sequence data includes at least ambient temperature, vehicle speed, battery electrical parameters, four types of core temperature states, thermal management actuator status, and drive loss data. The set determination module is used to determine the set of state variables of the gray box reduced-order model. The state variables include at least the battery temperature, battery coolant inlet temperature, passenger compartment temperature, and electric drive temperature. The matrix construction module is used to construct the corresponding physical heat flux characteristic matrix for each state variable, such that the rate of change of state temperature satisfies the following relationship: ; in, This is the physical heat flux characteristic matrix for this state. For the channel coefficients to be identified, This is the equivalent heat capacity in this state; Each heat flow characteristic corresponds to at least one type of heat transfer channel with a clear physical meaning, including active heating, driven waste heat coupling, ambient heat dissipation, cooling circuit heat exchange, and battery internal resistance heating. The constraint application module is used to apply physical sign constraints to the channel coefficients corresponding to each heat flow characteristic: the constraint for active heating channel coefficients is non-negative, the constraint for active cooling channel coefficients is non-positive, and the heat transfer direction for temperature difference driven heat exchange channels is determined by the sign of the temperature difference, with the coefficient constraint being non-negative. The model parameter module is used to traverse candidate equivalent heat capacities within a preset heat capacity search interval, and solve the corresponding channel coefficients for each candidate heat capacity by combining regularized least squares with sign constraints. Using the temperature prediction RMSE of single-state multi-step rolling simulation as the evaluation index, the combination of heat capacity and channel coefficient with the minimum RMSE is selected as the reduced-order model parameter for this state. The data embedding module is used to perform closed-loop simulation verification of the multi-state coupled reduced-order model using a validation set. After successful verification, the parameters of the reduced-order model can be exported and directly embedded into the vehicle controller.

[0058] It is worth noting that although this system / device only discloses the above-mentioned modules / units, it does not mean that this system / device is limited to the above-mentioned basic functional modules. On the contrary, what this invention intends to express is that, based on the above-mentioned basic functional modules, those skilled in the art can add one or more functional modules in combination with the prior art to form an infinite number of embodiments or technical solutions. That is to say, this system is open rather than closed. It cannot be assumed that the scope of protection of the claims of this invention is limited to the above-disclosed basic functional modules just because this embodiment only discloses a few basic functional modules.

[0059] Example 4 discloses a physical constraint gray-box ROM modeling method for winter thermal management of electric vehicles. This method uses real vehicle driving data as input, and after time series cleaning, state smoothing, outlier removal, and training / validation partitioning, constructs a four-state gray-box thermal network model including battery temperature, battery coolant inlet temperature, passenger compartment temperature, and electric drive temperature.

[0060] The model represents the temperature change rate of each thermal state as a combination of several heat flow features with clear physical meanings and coefficients to be identified, and normalizes them using heat capacity parameters. Each heat flow feature corresponds to physical channels such as active heating, drive waste heat coupling, environmental heat dissipation, cooling circuit heat exchange, compressor refrigeration, and battery internal resistance heating. Each channel is given positive, negative, or free sign constraints to ensure that the parameter identification results meet the thermodynamic directionality requirements. Furthermore, for winter electric vehicle operating conditions, a soft-gated feature for drive waste heat to the battery coolant inlet, a sublinear compression feature based on the square root of drive loss, a bidirectional linear heat transfer feature to remove spurious softplus temperature difference bias, and a numerically stable softplus function are introduced. This allows the model to express the nonlinear thermal response at low winter temperatures while maintaining the differentiability and numerical stability required for MPC solutions. Through steps such as heat capacity search, regularized least squares identification with sign constraints, single-state multi-step simulation RMSE evaluation, and validation set closed-loop simulation evaluation, a ROM model that can be exported to a controller or MPC controller is obtained.

[0061] This embodiment specifically includes the following steps (such as...). Figure 4 (as shown) S1. Collect winter thermal management operation data of electric vehicles: The data includes vehicle time series, ambient temperature, vehicle speed, slope, battery current, battery voltage, SOC, battery temperature, battery coolant inlet temperature, passenger compartment temperature, electric drive temperature, compressor power, PTC power, water pump speed, fan duty cycle, valve opening, expansion valve opening, electric drive loss or drive loss that can be estimated from motor / electric control temperature and power, etc.

[0062] S2. Preprocess the raw data: This includes removing non-monotonic time points, calculating the sampling period, pruning outliers by quantile, smoothing state variables and control variables respectively, linearly filling missing data, and dividing the training set and validation set according to time order.

[0063] S3. Determine the state variables of the gray box ROM: The model state includes: Battery temperature 'T_bat'; Battery coolant inlet temperature 'T_bat_in'; Passenger compartment temperature 'T_cab'; Electric drive temperature 'T_drive'.

[0064] S4. Construct physical heat flux characteristics. For each state, construct a characteristic matrix Phi such that the rate of change of state satisfies: Where s represents any state, This is the physical heat flux characteristic matrix for this state. For the channel coefficients to be identified, This is the equivalent heat capacity in this state.

[0065] S5. Battery Temperature Characteristics: Battery temperature characteristics include bidirectional heat exchange between the coolant inlet and the battery, battery Joule heating, ambient natural convection, series heat transfer in the electric drive circuit, compressor cooling, and indirect cooling from the radiator. ; in The heat transfer coefficient is obtained from the normalized speed of the water pump. This is the flow distribution coefficient for the battery circuit. The coefficient of this feature is constrained to be non-negative, and the heat transfer direction is determined by the sign of the temperature difference, thus avoiding fixed positive or spurious heat sources (such as...). Figure 1 , 2 (As shown in Figure 3).

[0066] S6. Battery Coolant Inlet Temperature Characteristics: The battery coolant inlet characteristics are divided into an active heating group, a drive waste heat coupling group, a heat dissipation / environment group, and a mixed flow group. The active heating group includes direct PTC2 heating and low-temperature heating gating; the drive waste heat coupling group includes direct heat source from drive losses and temperature difference heat transfer from electric drive to coolant; the heat dissipation / environment group includes radiator cooling, chiller cooling, and ambient convection; and the mixed flow group includes battery regeneration.

[0067] The characteristics of the direct heat source driving waste heat are as follows: ; ; ; Here, softplus is a smoothed 'max(x,0)' substitution function. This feature is only effective when the electric drive temperature is higher than the battery coolant inlet temperature and exceeds a set margin, avoiding the generation of spurious drive waste heat when heat transfer conditions are not met, while maintaining the model's differentiability for MPC optimization.

[0068] S7. Crew compartment temperature and electric drive temperature characteristics: Crew compartment temperature characteristics include heater core heat exchange, PTC air heating, evaporator cooling, ambient convection, and passive heat sources. Electric drive temperature characteristics include electric drive losses, electrical power generation, ambient convection, bidirectional coolant heat exchange, forced air cooling, radiator loop cooling, and chiillery auxiliary cooling.

[0069] The electric drive loss heat generation adopts sqrt(loss_drive_kw) linear compression to reduce the oversensitivity problem caused by excessive characteristic amplitude under high load and large slope conditions; the heat exchange between the electric drive and the coolant adopts bidirectional linear temperature difference characteristics, and non-negative constraints are applied to the coefficients so that when the electric drive temperature is higher than the coolant temperature, it behaves as cooling the electric drive, and vice versa, it behaves as the coolant transfers heat to the electric drive.

[0070] S8. Set physical symbol constraints for each feature: For heating channels, set the lower bound of the coefficient to 0; for cooling channels, set the upper bound of the coefficient to 0; for channels where the direction changes with the state, such as natural convection, set it to a free symbol or let the temperature difference term bear the direction. Symbol constraints prevent the identification of inverse physical parameters.

[0071] S9. Search for the equivalent heat capacity Cs of each state: Preset the heat capacity search range for each state, for example, set different search ranges for battery temperature, battery coolant inlet temperature, passenger compartment temperature, and electric drive temperature. For each candidate heat capacity, multiply the target derivative by the heat capacity and then perform a regularized least squares solution with signed constraints to obtain the channel coefficient.

[0072] S10. Select the optimal heat capacity using single-state multi-step simulation error: For each candidate heat capacity and corresponding coefficient, starting from the measured initial value of that state, perform multi-step rolling simulation using the midpoint RK2 method or the equivalent second-order integral method, and calculate the RMSE between the simulated temperature and the measured temperature. Select the heat capacity and coefficient combination with the smallest RMSE as the ROM parameter for that state.

[0073] S11. Verify the trained ROM: Use the validation set to perform multi-state simulation, calculate the RMSE, MAE and maximum error for each state, and output the characteristic coefficients, optimal heat capacity, search curve, validation graph and model file.

[0074] S12. Export the ROM to the thermal management controller or MPC controller: The exported content includes the state sequence, sampling period, heat capacity, feature name, channel coefficient, symbol constraints and state boundaries, which are used for subsequent rapid temperature prediction and control optimization.

[0075] This embodiment achieves the following: 1. The winter thermal management system for electric vehicles is reduced to a four-state gray box ROM (Rotational Memory Unit) for battery temperature, battery coolant temperature, cab temperature, and electric drive temperature. 2. A heat flow characteristic matrix is ​​constructed based on physical channels such as active heating, drive waste heat, ambient heat dissipation, cooling circuit heat exchange, and mixed flow. 3. Positive, negative, or free sign constraints are set for the heat flow channel coefficients, and parameters are identified using constrained regularized least squares. 4. The characteristic construction of drive loss heat sources is performed using the square root method with sublinear compression. 5. The gray box ROM parameters are determined jointly through equivalent heat capacity search and multi-step rolling simulation RMSE (Real-Time Sequence of Existing Memory).

[0076] This embodiment also includes: 1. The state variables can be expanded to include multi-point temperatures of the battery module, motor temperature, electronic control temperature, compressor outlet temperature, or condenser pressure, etc., depending on the vehicle model and sensor configuration. 2. The heat capacity can be determined using a fixed physical prior, search-based determination, or joint optimization with a prior penalty term. 3. The parameter identification method can be replaced by constrained least squares with physically constrained quadratic programming, Bayesian regression, robust regression, or recursive least squares. 4. In addition to the square root method, the drive loss characteristics can also be compressed using power functions, saturation functions, or piecewise linear functions.

[0077] Figure 7 This is a block diagram of an electronic device structure based on a physical constraint gray box reduction model construction method provided by one or more embodiments of the present invention.

[0078] Example 6, as Figure 7 As shown, this application provides an electronic device, including: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; The memory stores a computer program that, when executed by a processor, causes the processor to perform steps of a physical constraint gray box reduction model construction method.

[0079] Example 7: This application also provides a computer-readable storage medium storing a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform the steps of a physical constraint gray box reduction model construction method.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a physically constrained gray box reduced-order model, characterized in that, include: A1. Collect time-series data of electric vehicle winter thermal management operation, and divide the time series into training set and validation set; A2. The state variables of the gray box reduced-order model are determined to include battery temperature, battery coolant inlet temperature, crew compartment temperature, and electric drive temperature. A3. Construct a corresponding physical heat flux characteristic matrix for each state variable, such that the rate of change of state temperature satisfies: ; This is the physical heat flux characteristic matrix. For the channel coefficients to be identified, Equivalent heat capacity; The heat flow characteristics correspond to the heat transfer channels in active heating, driven waste heat coupling, environmental heat dissipation, cooling circuit heat exchange, and battery internal resistance heating. A4. Apply physical sign constraints: The coefficient constraint for active heating channels is non-negative, the coefficient constraint for active cooling channels is non-positive, and the heat transfer direction for temperature difference driven heat exchange channels is determined by the sign of the temperature difference, with the coefficient constraint being non-negative. A5. Traverse the candidate equivalent heat capacities and solve the corresponding channel coefficients for each candidate heat capacity by combining signed constraint regularized least squares. The minimum heat capacity of RMSE is predicted and combined with the channel coefficient as the parameters of the reduced-order model for this state. A6. Use the validation set to conduct closed-loop simulation verification, and export the reduced-order model parameters after successful verification.

2. The method for constructing a physically constrained gray box reduced-order model according to claim 1, characterized in that, include: In step A3, the heat flow characteristics corresponding to the battery temperature include at least the bidirectional linear heat transfer characteristics between the coolant inlet and the battery. The bidirectional linear heat transfer characteristics are constructed by multiplying the actual temperature difference with the water pump speed and the loop flow distribution coefficient. The heat transfer direction is determined autonomously by the sign of the temperature difference.

3. The method for constructing a physically constrained gray box reduced-order model according to claim 1, characterized in that, include: In step A3, the heat flow characteristics corresponding to the battery coolant inlet temperature include the characteristics of the drive waste heat coupling group; The characteristics of the drive waste heat coupling group are constructed by multiplying the square root linear compression of the drive loss with a softplus soft gating function, the pump speed and the loop allocation coefficient. The softplus soft-gating function only takes effect when the electric drive temperature is higher than the battery coolant inlet temperature and exceeds a set margin.

4. The method for constructing a physically constrained gray box reduced-order model according to claim 1, characterized in that, include: In step A3, the heat flow characteristics corresponding to the electric drive temperature include the heat generation characteristics of electric drive loss after square root linear compression, and the bidirectional linear temperature difference heat transfer characteristics between the electric drive and the coolant.

5. The method for constructing a physically constrained gray box reduced-order model according to claim 1, characterized in that, include: In step A5, the multi-step rolling simulation uses the midpoint second-order Runge-Kutta integral method for state recursion.

6. The method for constructing a physically constrained gray box reduced-order model according to claim 1, characterized in that, include: In step A6, the exported reduced-order model parameters include at least the state order, sampling period, equivalent heat capacity, feature name, channel coefficient, symbolic constraint, and state boundary.

7. The method for constructing a physically constrained gray box reduced-order model according to claim 1, characterized in that, include: The state variables can be expanded to include multi-point temperatures of the battery module, motor temperature, electronic control temperature, compressor outlet temperature, or condenser pressure.

8. A system for constructing a physically constrained gray box reduced-order model, characterized in that, include: The data acquisition module is used to collect the timing data of electric vehicle winter thermal management operation, which is divided into training set and validation set according to the time sequence. The set determination module is used to determine the state variables of the gray box reduced-order model, including battery temperature, battery coolant inlet temperature, passenger compartment temperature, and electric drive temperature. The matrix construction module is used to construct the corresponding physical heat flux characteristic matrix for each state variable, such that the rate of change of state temperature satisfies: ; This is the physical heat flux characteristic matrix. For the channel coefficients to be identified, Equivalent heat capacity; Each heat flow characteristic corresponds to a heat transfer channel in active heating, driven waste heat coupling, ambient heat dissipation, cooling circuit heat exchange, and battery internal resistance heating. The constraint application module is used to apply physical sign constraints to the channel coefficients corresponding to each heat flow characteristic: the constraint for active heating channel coefficients is non-negative, the constraint for active cooling channel coefficients is non-positive, and the heat transfer direction for temperature difference driven heat exchange channels is determined by the sign of the temperature difference, with the coefficient constraint being non-negative. The model parameter module is used to traverse the candidate equivalent heat capacity and solve the corresponding channel coefficient for each candidate heat capacity by combining signed constraint regularized least squares. The minimum heat capacity of RMSE is predicted by multi-step rolling simulation temperature, and combined with the channel coefficient as the parameters of the reduced-order model for this state. The data embedding module is used to conduct closed-loop simulation verification using a validation set, and after successful verification, the parameters of the reduced-order model are exported.

9. An electronic device, characterized in that, include: The processor, communication interface, memory, and communication bus are connected, with the processor, communication interface, and memory communicating with each other via the communication bus. The memory stores a computer program that, when executed by a processor, causes the processor to perform the steps of constructing a physical constraint gray-box reduction model as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, include: The device stores a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform the steps of constructing a physical constraint gray box reduction model as described in any one of claims 1 to 7.