Automatic control method and device for thermal management system of electric vehicle

Abstract

The invention discloses an automatic control method and device for a thermal management system of an electric vehicle, relates to the field of intelligent control, and realizes multi-step prediction of battery and motor heat production and thermal load change in the future by fusing external information such as a navigation path and weather forecast, plans a thermal management strategy in advance, and improves the response perspectiveness of the system. A multi-source disturbance dynamic structured correlation equalization technology is adopted, the coupling relation between heat sources is modeled, cross disturbance is compensated through a joint state function, and multi-heat-loop cooperative control is achieved. On the basis of a model predictive control (MPC) framework, temperature tracking errors and energy consumption penalty are optimized, the energy consumption of actuators such as a compressor and a water pump is reduced while a safe temperature interval is met, and the endurance is prolonged. And in combination with a state estimator fusion model and observation data, the control precision and robustness under a complex working condition are enhanced, and finally global optimal thermal management considering the service life of parts, the comfort of passengers and the energy efficiency is realized.

Description

Technical Field

[0001] This application relates to the field of intelligent control, and more specifically, to an automatic control method and device for an electric vehicle thermal management system. Background Art

[0002] As an important direction for future transportation development, the performance, driving range, and service life of electric vehicles largely depend on an efficient and reliable thermal management system. The thermal management system of an electric vehicle is responsible for precisely controlling the temperatures of the battery pack, motor, power electronics, and passenger cabin. The battery pack needs to operate within a specific temperature range to ensure optimal charge-discharge performance and cycle life; the motor and power electronics need to dissipate heat to avoid overheating damage and efficiency degradation; the temperature of the passenger cabin directly affects passenger comfort. These thermal loads and demands change dynamically with the vehicle's driving conditions, environmental conditions, and user settings. Therefore, it is crucial to build a thermal management system that can respond intelligently and work efficiently in coordination.

[0003] Most existing electric vehicle thermal management control solutions adopt rule-based control, PID control, or simple state feedback control. These methods are usually reactive, i.e., they adjust according to current sensor signals such as temperature and pressure. For example, when the battery temperature rises to a certain threshold, the cooling system is activated, or the air conditioner is turned on according to the set passenger cabin temperature. Although these methods can achieve basic thermal management functions, they lack the ability to predict future operating conditions and cannot foresee upcoming thermal load changes (such as motor heating caused by climbing, increased battery heating due to long-term high-speed driving, or sudden increase in external environmental temperature). This reactive control often lags behind actual needs, resulting in untimely system response when a large amount of heat dissipation is required, or overworking after the thermal load decreases, thus causing unnecessary energy consumption, shortening the driving range, and making it difficult to achieve global optimal energy management while ensuring comfort and component health. In addition, existing methods are difficult to effectively coordinate the control of multiple interrelated thermal circuits and actuators, handle the coupling relationship between different heat sources (battery, motor, cabin), and are also difficult to dynamically balance between strict temperature constraints and energy consumption targets.

[0004] In view of the above problems, there is an urgent need for a more advanced and intelligent automatic control method for electric vehicle thermal management. Summary of the Invention

[0005] To solve the above technical problems, this application is proposed.

[0006] According to one aspect of this application, there is provided an automatic control method for an electric vehicle thermal management system, which includes: Obtain real-time vehicle state data and external prediction information, where the external prediction information includes navigation path information and weather forecast information for the next N time steps; Construct a current system state vector based on the real-time vehicle state data; Convert the external prediction information into a predictable perturbation sequence; Construct an MPC optimization problem instance based on the current system state vector and the predictable perturbation sequence; Input the MPC optimization problem instance into an online optimization solver module to obtain a current optimal control instruction; Send the current optimal control instruction to the drive actuator of the electric vehicle thermal management system to obtain a physical drive signal and updated real-time vehicle state data.

[0007] According to another aspect of the present application, an automatic control device for an electric vehicle thermal management system is provided, which includes: A data acquisition module for obtaining real-time vehicle state data and external prediction information, where the external prediction information includes navigation path information and weather forecast information for the next N time steps; A current system state vector construction module for constructing a current system state vector based on the real-time vehicle state data; An external prediction information conversion module for converting the external prediction information into a predictable perturbation sequence; An MPC optimization problem instance construction module for constructing an MPC optimization problem instance based on the current system state vector and the predictable perturbation sequence; An optimal control instruction generation module for inputting the MPC optimization problem instance into an online optimization solver module to obtain a current optimal control instruction; A drive execution module for sending the current optimal control instruction to the drive actuator of the electric vehicle thermal management system to obtain a physical drive signal and updated real-time vehicle state data.

[0008] Compared with the prior art, an automatic control method and device for an electric vehicle thermal management system provided by the present application fuse external information such as navigation paths and weather forecasts to achieve multi-step prediction of future heat generation and heat load changes of the battery and motor, plan the thermal management strategy in advance, and improve the forward-lookingness of the system response. The multi-source perturbation dynamic structured correlation equilibrium technology is adopted to model the coupling relationship between heat sources, and the cross-perturbation is compensated through the joint state function to achieve coordinated control of multiple heat circuits. Based on the model predictive control (MPC) framework, the temperature tracking error and energy consumption penalty are optimized to reduce the energy consumption of actuators such as compressors and water pumps while meeting the safe temperature range, and the battery life is extended. By combining the state estimator to fuse the model and observation data, the control accuracy and robustness under complex working conditions are enhanced, and finally the global optimal thermal management that takes into account component life, passenger comfort and energy efficiency is realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The above and other objects, features and advantages of the present application will become more obvious by describing the embodiments of the present application in more detail with reference to the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification, and are used to explain the present application together with the embodiments of the present application, and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 It is a flowchart of an automatic control method for an electric vehicle thermal management system according to an embodiment of the present application.

[0011] Figure 2 It is a flowchart of step S3 in the automatic control method for an electric vehicle thermal management system according to an embodiment of the present application.

[0012] Figure 3 It is a flowchart of step S5 in the automatic control method for an electric vehicle thermal management system according to an embodiment of the present application.

[0013] Figure 4 It is a block diagram of an automatic control device for an electric vehicle thermal management system according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0014] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although some embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. On the contrary, these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not used to limit the protection scope of the present disclosure.

[0015] It should be understood that the various steps described in the method embodiments of the present disclosure may be executed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this regard.

[0016] In view of the problems in the above background art, in the technical solution of the present application, an automatic control method for an electric vehicle thermal management system is proposed. Figure 1 It is a flowchart of an automatic control method for an electric vehicle thermal management system according to an embodiment of the present application. As Figure 1 shown, the automatic control method for an electric vehicle thermal management system according to an embodiment of the present application includes: S1, obtaining real-time vehicle state data and external prediction information, where the external prediction information includes navigation path information and weather forecast information for the next N time steps; S2, constructing a current system state vector based on the real-time vehicle state data; S3, converting the external prediction information into a predictable disturbance sequence; S4, constructing an MPC optimization problem instance based on the current system state vector and the predictable disturbance sequence; S5, inputting the MPC optimization problem instance into an online optimization solving module to obtain a current optimal control instruction; S6, sending the current optimal control instruction to the drive actuator of the electric vehicle thermal management system to obtain a physical drive signal and updated real-time vehicle state data.

[0017] In step S1, real-time vehicle state data and external prediction information are obtained, where the external prediction information includes navigation path information and weather forecast information for the next N time steps. In particular, the navigation path information includes altitude, speed limit, and congestion prediction; the weather forecast information includes ambient temperature, ambient humidity, and solar radiation intensity. It should be understood that in order to achieve forward-looking and optimized control of the electric vehicle thermal management system, it is not only necessary to accurately grasp the current operating state, but more importantly, to predict the factors that will affect the heat load and heat dissipation conditions in the future. Traditional reactive control strategies only make adjustments based on the current state, often with a lag, and are unable to cope with upcoming drastic operating condition changes or environmental changes, resulting in low energy efficiency or poor temperature control effects. By introducing external prediction information, especially the navigation path and weather forecast in the next period of time, the control system can foresee the future driving resistance, heat generation of the motor and battery, and the impact of the environment on the heat load of the cockpit and the heat exchange of the battery compartment. Therefore, the real-time vehicle state data provides the starting point and feedback for control, while the external prediction information provides feedforward disturbances for advanced control algorithms such as MPC, enabling the system to make more intelligent, energy-saving, and predictive control decisions.

[0018] In particular, in a specific example of this application, the specific implementation of step S1 is as follows: The acquisition of real-time vehicle state data means that at the beginning of each control cycle, through various sensors integrated inside the vehicle, feedback from actuators, and network communication of the electronic control unit (ECU), the operating parameters of each subsystem of the vehicle at the current moment are collected in real time. These parameters constitute the core elements of the current system state vector. These data are rich and diverse, including but not limited to: the battery management system (BMS) provides data such as the total voltage, total current, temperature of each single cell, overall temperature, state of charge (SOC), and state of health (SOH) of the battery pack; the motor controller (MCU) provides temperature, speed, torque of the drive motor, and temperature information of power electronic devices (such as inverters); the vehicle control unit (VCU) or a dedicated thermal management controller (HTC) collects feedback signals such as the working state, flow rate, pressure, and temperature from the occupant compartment temperature and humidity sensors, external environmental temperature sensors, and key components in the thermal management system, such as compressors, water pumps, fans, valves, etc. These real-time data are transmitted to the computing platform that executes the thermal management control algorithm through a high-speed communication network inside the vehicle, such as CAN FD or in-vehicle Ethernet. The control platform performs necessary filtering, calibration, and integration on these massive real-time data to form a multi-dimensional vector that can comprehensively reflect the current operating state of the system, that is, the current system state vector.

[0019] Meanwhile, the external prediction information is obtained in parallel. This part of the information provides the control system with predictions of the environment and operating conditions over a period of time in the future, enabling the controller to anticipate and proactively respond to possible changes in heat load. The external prediction information covers the navigation path information and weather forecast information for the next N time steps. The navigation path information is provided by the vehicle's navigation system or connected cloud service, and its acquisition is based on the destination set by the user or the currently planned route. The navigation system not only provides real-time position and direction, but also predicts the road conditions along the way. The navigation path information specifically includes the altitude, speed limit, and congestion prediction for the next N time steps. The altitude data can be obtained through a digital elevation map to predict future slope changes; the speed limit information reflects the maximum allowable speed on different sections of the road; the congestion prediction is based on historical traffic data and real-time road conditions to predict future traffic flow and reduced driving speed. This information is provided in a sequential form, that is, for each of the consecutive N time steps in the future, the predicted altitude, speed limit, and congestion status are given respectively. The weather forecast information is obtained through the vehicle's communication module, such as a cellular network or Wi-Fi, from an external meteorological service platform and is associated with the vehicle's current geographical location and the expected driving path. The weather forecast information specifically includes the ambient temperature, ambient humidity, and solar radiation intensity for the next N time steps. These data services can provide weather predictions for the locations where the vehicle is expected to be in the future. The ambient temperature and humidity directly affect system heat dissipation and cabin heat load; the solar radiation intensity mainly affects the solar heat gain of the cabin. Similarly, this information is also provided in a sequential form, predicting the ambient temperature, ambient humidity, and solar radiation intensity at each time point within the next N time steps.

[0020] Specifically, N in the future N time steps is preset as a key parameter in the overall design of the MPC-based control system. N represents the prediction horizon length of the MPC controller, that is, how far forward the controller looks when performing optimization calculations. The selection of N is a trade-off process: a larger N can provide a longer-term prediction ability, enabling the controller to better plan long-term strategies, but will significantly increase the computational amount and complexity of model prediction; a smaller N has a lighter computational burden, but may lead to short-sighted control decisions and cannot effectively cope with heat load changes on a long time scale. Therefore, the value of N is determined through offline simulation and performance evaluation during the system development stage, in combination with the processing power of the in-vehicle computing platform and the requirements for control performance to obtain an optimal or acceptable fixed value. The time span of the external prediction information obtained in this step, that is, it must cover the time range defined by the preset N time steps to meet the input requirements of the subsequent MPC optimization problem. For example, N can be set to 90, indicating that the system predicts information within the next 90 seconds. If each time step is set to 1 second, it contains 90 predicted data points for future moments. These preset N values and time steps are determined during system design and calibration and are used fixedly during vehicle operation or adjusted dynamically according to specific conditions.

[0021] In step S2, based on the real-time vehicle state data, a current system state vector is constructed. Correspondingly, since MPC is an advanced control strategy, its core is to predict the future behavior of the system within the prediction horizon based on the dynamic model of the system and obtain the optimal control input at the current moment through optimization calculations. To accurately predict future behavior, the control algorithm must know the exact state of the system at the current moment. Although the real-time vehicle state data provides rich sensor information and other operating parameters, these raw data are scattered, diverse, contain redundancy or noise, and cannot be directly used as variables describing the internal physical states of the system, such as energy, mass, temperature distribution, etc., and these internal physical states are the key factors determining the future dynamic response of the system. Therefore, it is necessary to screen, process, and integrate these real-time acquired, multi-source sensor data and information into a refined and standardized mathematical representation form, that is, the current system state vector.

[0022] Specifically, in a specific example of the present application, the specific implementation process of step S2 is as follows: The real-time vehicle state data are the original operating parameters collected in real time from each subsystem, sensor, and actuator inside the vehicle at each control cycle or sampling moment. These real-time data are transmitted through the in-vehicle communication network, such as CAN bus, in-vehicle Ethernet, etc., to the computing platform responsible for executing the thermal management control algorithm.

[0023] After receiving this real-time data from different sampling frequencies and different buses, a series of processes are required to construct the current system state vector. First, the received data is time-synchronized to ensure that all collected data points can reflect the state of the system at the same moment as much as possible. Then, necessary preprocessing is performed on the original sensor data, including but not limited to: filtering to remove measurement noise, calibration to correct sensor biases, and validity checks to eliminate abnormal or incorrect data. For some state variables, they need to be obtained through simple calculations or look-up tables. For example, the average temperature or the highest temperature of the battery pack is calculated based on the data of multiple temperature sensors.

[0024] Next, from the processed real-time vehicle state data, key state variables that can represent the dynamic characteristics of the thermal management system are selected and extracted. The selection of these state variables is based on the physical model and control requirements of the thermal management system, and they are the minimum set necessary to describe the current thermodynamic and electrochemical state of the system. The state variables include: the key temperatures of the battery pack (such as average temperature, highest temperature), the drive motor temperature, the power electronics temperature, the internal air temperature of the passenger compartment, and even the SOC of the battery. Importantly, these selected variables should be the dynamic states of the system, that is, their future values depend not only on the current control inputs and external disturbances but also on their current own values.

[0025] Finally, these selected, processed, and synchronized key state variables are arranged in a predefined order and assembled into a vector with a fixed dimension, that is, the current system state vector. For example, this vector is defined as [battery average temperature, motor temperature, inverter temperature, passenger compartment temperature, battery SOC]. Each component of the vector corresponds to the current value of a specific state variable. This constructed current system state vector concisely and comprehensively quantifies the operating state of the thermal management system at the current moment and serves as the starting point and key input for state estimation and model predictive control algorithms in subsequent steps, for predicting the future behavior of the system and solving the optimal control instructions.

[0026] In step S3, the external prediction information is converted into a predictable disturbance sequence. It should be understood that the core mechanism of MPC is to predict the system behavior within a certain period in the future through the system dynamic model and make optimized decisions based on this. The future behavior of the system not only depends on the current initial state and the applied control input, but is also affected by various uncontrollable factors from the external environment and driving conditions. For the thermal management system, these external factors are manifested as "disturbances" that change dynamically over time and are known or predictable, such as the heat generation power changes of different components, the heat load changes of the external environment on the cockpit and the battery, etc. The external prediction information itself is not a direct heat or energy value, but the root cause or influencing factor that leads to heat generation, transfer, or exchange. For example, climbing (altitude change) or accelerating / decelerating (speed limit, congestion impact) will cause changes in the loads of the motor and the battery, thereby affecting their heat generation power; while the external environmental temperature, humidity, and solar radiation directly affect the heat exchange between the cockpit and the battery compartment and the outside world and the heat load of the cockpit by solar radiation. In order to enable MPC to utilize this information in its prediction model to predict the future thermodynamic state evolution, the external prediction information reflecting future working conditions and the environment needs to be converted into a predictable disturbance sequence that can be directly input into the system dynamic model and affect the system state.

[0027] Figure 2 It is a flowchart of step S3 in the automatic control method of the electric vehicle thermal management system according to the embodiment of the present application. Specifically, in the embodiment of the present application, as Figure 2 shown, step S3 of converting the external prediction information into a predictable disturbance sequence includes: S31, predicting the battery heat generation power and the motor / electronic control unit heat generation power for the next P steps based on the navigation path information for the next N time steps to obtain a battery heat generation power sequence and a motor / electronic control unit heat generation power sequence; S32, predicting the heat load of the environment on the cockpit and the heat exchange amount with the battery compartment for the next P steps based on the weather forecast information for the next N time steps to obtain a heat load sequence and a heat exchange amount sequence; S33, respectively performing multi-source disturbance dynamic structured correlation equalization on the battery heat generation power sequence, the motor / electronic control unit heat generation power sequence, the heat load sequence, and the heat exchange amount sequence to obtain a battery heat generation power equalization sequence, a motor / electronic control unit heat generation power equalization sequence, a heat load equalization sequence, and a heat exchange amount equalization sequence; S34, integrating the battery heat generation power equalization sequence, the motor / electronic control unit heat generation power equalization sequence, the heat load equalization sequence, and the heat exchange amount equalization sequence to obtain the predictable disturbance sequence.

[0028] Specifically, in step S31, based on the navigation path information for the future N time steps, predict the battery heat generation power and the motor / electronic control unit (ECU) heat generation power for the next P steps to obtain a battery heat generation power sequence and a motor / ECU heat generation power sequence. Correspondingly, the main heat sources that need to be managed by the thermal management system of an electric vehicle are the battery, the drive motor, and its power electronic controller (ECU). The heat generation power of these components is not constant but highly dependent on the current operating conditions of the vehicle, especially the demand for traction and braking power. For example, when driving at high speed, accelerating rapidly, climbing a slope, or performing energy recovery, the battery and the motor / ECU generate a large amount of heat; while when driving at a low constant speed, coasting, or parking, the heat generation is relatively less. The traditional reactive thermal management system only determines the cooling intensity based on the current component temperature and is difficult to cope with the upcoming drastic changes in operating conditions. Therefore, by obtaining future navigation path information, including altitude, speed limits, and congestion predictions, the system can anticipate the driving cycle and power demand that the vehicle will experience in the next period of time and convert them into a specific and quantifiable prediction sequence of the battery and motor / ECU heat generation power, providing an early warning about future heat load changes for the MPC controller.

[0029] Specifically, in a specific example of the present application, the prediction process in step S31 is implemented using a machine learning model, and the specific implementation process is as follows: First, a prediction model needs to be established, which can learn the mapping relationship from the future driving conditions of the vehicle (reflected by the navigation path information) to the heat generation power of the battery and the motor / ECU. To train this model, a large amount of historical data needs to be collected. This data includes the driving records of the vehicle under various driving conditions, which contain data corresponding to the future navigation path information (such as altitude changes, speed, and whether it is in a congested state of the actual driving path) and the battery heat generation power and the motor / ECU heat generation power recorded at the same time. The collected data needs to cover different environmental conditions, driving styles, and operating conditions to ensure the generalization ability of the model.

[0030] Preprocess the collected historical data to form a data set for training the machine learning model. For each segment of historical driving data, divide it into an input sequence and an output sequence. The input sequence is the navigation path information corresponding to the future N time steps (for example, altitude sequence, speed sequence, congestion status sequence), and the output sequence is the battery heat generation power sequence and the motor / ECU heat generation power sequence corresponding to the future P time steps. P is the time step of the heat generation power predicted by the model and is consistent with the prediction of the subsequent MPC control. In practical applications, P is less than or equal to N, indicating that the navigation information for the next N steps is used to predict the heat generation power for the nearest P steps. For example, if N is set to 90 and P is set to 60, the model will use the navigation data for the next 90 seconds to predict the battery and motor / ECU heat power for the next 60 seconds.

[0031] Select a suitable machine learning model architecture for training. In particular, the prediction model adopts an encoder-decoder architecture. Specifically, the encoder (which can use LSTM, GRU, or Transformer) processes the input sequence of navigation path information for the next N time steps, compressing it into a fixed-length context vector or a more advanced sequence representation. The decoder (also an LSTM, GRU, or Transformer) then receives the output of the encoder and gradually generates the sequence of battery heat generation power and motor / electronic control heat generation power for the next P time steps. When generating each output step, the decoder can utilize the attention mechanism, enabling it to focus on the most relevant parts of the encoder output, thereby improving the prediction accuracy. That is to say, considering that both the input and output are sequence data and the prediction task has temporal dependence, sequence models such as RNN, LSTM, or GRU are suitable choices. More advanced model architectures such as Transformer can also be adopted. The input layer of the model receives the navigation path information for the next N time steps (for example, altitude, speed limit, congestion prediction, etc. can be input as feature vectors at each time step), and the output layer outputs the battery heat generation power and motor / electronic control heat generation power for the next P time steps. The weights and biases of the model are trained by minimizing the error (such as mean squared error) between the predicted output and the actual measured data through an optimization algorithm (such as Adam).

[0032] After the model training is completed, it is deployed to the vehicle's thermal management control unit for online prediction. When the vehicle is running, the system obtains the sequence of navigation path information for the next N time steps. These sequences serve as the input to the trained machine learning model. After receiving these inputs, the model, through internal calculations, predicts the sequence of battery heat generation power and the sequence of motor / electronic control heat generation power corresponding to the next P time steps. For example, if the input is the sequence of altitude, speed limit, and congestion prediction for the next 90 seconds, the model will output the predicted sequence of battery heat generation power for the next 60 seconds (a sequence containing 60 power values) and the predicted sequence of motor / electronic control heat generation power (another sequence containing 60 power values). These predicted sequences are the output of this step and provide key information for subsequent heat load prediction and balancing processing. The preset values of N and P are determined according to factors such as the availability of prediction information, prediction accuracy requirements, and system computing resources during the system design phase and are fixedly used during system operation.

[0033] Specifically, P represents the prediction horizon length of the model predictive controller. Different from N (navigation / weather prediction horizon), P is a core parameter determined during the design of the MPC controller, which determines the time length for the controller to predict the system dynamics forward. The selection of P depends on the thermal dynamics characteristics of the controlled object (electric vehicle thermal management system), the response time of the system, and the computing power of the in-vehicle computing platform. A larger P can better capture and utilize the thermal inertia of the system for optimization. For example, the cooling system can be turned on in advance to cope with the upcoming high heat load, thereby reducing the peak temperature and energy consumption, but the computational burden is also greater. Among them, P is less than or equal to N because the navigation path information for predicting heat generation has only N time steps. The preset values of N and P are determined according to factors such as the availability of prediction information, the requirements for prediction accuracy, and the system computing resources during the system design stage and are fixedly used during system operation.

[0034] Specifically, in step S32, based on the weather forecast information for the next N time steps, predict the heat load on the cabin and the heat exchange amount with the battery compartment for the next P steps to obtain a heat load sequence and a heat exchange amount sequence. It should be understood that the thermal management system of an electric vehicle not only needs to cope with the heat loads of its own heat-generating components (battery, motor, electronic control), but also needs to manage the heat input or output from the external environment, especially the environmental impacts on the passenger compartment and the battery compartment. These environmental factors, such as environmental temperature, humidity, and solar radiation intensity, will significantly change the cooling / heating requirements of the cabin and the heat exchange rate between the battery compartment and the external environment. Traditional control methods often can only passively adjust the air conditioner and cooling system according to the current environmental conditions or the internal temperature of the cabin. This reaction mode cannot foresee the heat load fluctuations caused by future environmental changes, resulting in the system's inability to respond in time when the environmental conditions are about to deteriorate, affecting the riding comfort or component temperature; on the contrary, when the environmental conditions are about to improve, the system overworks, causing energy waste. Therefore, by using the obtained weather forecast information for the next N time steps, the system can understand the future change trend of the external environment in advance and convert these future environmental impacts based on the weather forecast into specific and quantified cabin heat load sequences and battery compartment heat exchange amount sequences, providing key feedforward information about future environmental disturbances for the MPC controller.

[0035] Specifically, in a specific example of the present application, the specific implementation process of step S32 This prediction process can also be implemented using a machine learning model, especially the encoder-decoder architecture suitable for processing sequence input and output: First, a prediction model needs to be established to learn the complex mapping relationship from future weather information to the cockpit heat load and the heat exchange amount in the battery compartment caused by the environment. The cockpit heat load is mainly affected by factors such as environmental temperature, humidity, solar radiation, window heat transfer, and occupant heat generation. Among them, environmental weather factors are the main external disturbances. The heat exchange amount between the battery compartment and the environment is affected by environmental temperature, battery compartment structure, heat dissipation duct design, and possible external cooling or heating devices. To train this model, a historical dataset is required, which includes weather forecast information (such as sequences of environmental temperature, environmental humidity, and solar radiation intensity) for the next N time steps as input, and the cockpit heat load and the heat exchange amount in the battery compartment obtained through calculation or measurement as output.

[0036] After data collection, preprocessing is carried out to form input-output sequence pairs required for training the machine learning model. The input sequence is the weather forecast information for the next N time steps, and the output sequence is the sequence of the cockpit heat load for the next P time steps and the sequence of the heat exchange amount in the battery compartment. P is the time step predicted by the model, which is consistent with P of the heat power prediction model and the prediction of the subsequent MPC, and is less than or equal to N. For example, if N is set to 90 and P is set to 60, the model uses the weather information for the next 90 seconds to predict the cockpit heat load and the heat exchange amount in the battery compartment for the next 60 seconds.

[0037] An encoder-decoder architecture is adopted to implement this prediction model. The encoder is responsible for processing the input sequence of weather forecast information for the next N time steps and encoding it into a fixed-length vector or a more advanced sequence representation. The encoder can use multi-layer LSTM or GRU networks, or a Transformer encoder. Taking the GRU encoder as an example, it receives feature vectors such as environmental temperature, environmental humidity, and solar radiation intensity at each time step and updates the hidden state internally. Finally, the hidden state at the last time step of the encoder or the context vector formed by integrating the hidden states of multiple time steps captures the information of the entire input weather sequence.

[0038] The decoder receives the output of the encoder and is responsible for gradually generating the sequence of cockpit heat loads and the sequence of battery compartment heat exchange amounts for the next P time steps. The decoder is also an LSTM or GRU network, or a Transformer decoder. Taking the LSTM decoder as an example, it receives the context vector output by the encoder at the initial moment as its initial hidden state. Then, at each time step, the decoder takes the output of the previous time step (or forces the true value of the previous time step during training) and the current hidden state as inputs, predicts the cockpit heat load and the battery compartment heat exchange amount at the current time step through a linear layer, and updates its own hidden state for prediction at the next time step. To improve the prediction accuracy, an attention mechanism can be introduced into the decoder, enabling the decoder to focus on the most relevant parts of the encoder input sequence when generating each output.

[0039] The entire encoder-decoder model is trained using an optimization algorithm (such as Adam) by minimizing the error between the predicted output and the actual heat load and heat exchange amount. After training, the model is deployed to the vehicle thermal management control unit. During vehicle operation, a sequence of weather forecasts for the next N time steps is obtained and input into the trained encoder-decoder model. The encoder processes the input sequence, and the decoder generates a sequence of predicted cockpit heat loads and a sequence of predicted battery compartment heat exchange amounts for the next P time steps based on the output of the encoder. For example, when inputting a sequence of ambient temperature, humidity, and solar radiation for the next 90 seconds, the model outputs a sequence of predicted cockpit heat loads for the next 60 seconds (a sequence containing 60 heat load values) and a sequence of predicted battery compartment heat exchange amounts (another sequence containing 60 heat exchange amount values). Specifically, the determination principle of the preset values N and P is the same as that of the thermal power prediction model, and the prediction range, accuracy, and computing resources need to be balanced.

[0040] Specifically, in step S33, multi-source perturbation dynamic structured correlation equalization is performed on the battery heat generation power sequence, the motor / electronic control heat generation power sequence, the heat load sequence, and the heat exchange amount sequence respectively to obtain a battery heat generation power equalization sequence, a motor / electronic control heat generation power equalization sequence, a heat load equalization sequence, and a heat exchange amount equalization sequence. Correspondingly, when integrating the battery heat generation power sequence, the motor / electronic control heat generation power sequence, the heat load sequence, and the heat exchange amount sequence, since the obtained predictable perturbations essentially belong to multi-source perturbations, which represent the complex non-linear correlations of the thermal management state based on each perturbation source, it is easy to cause unbalanced integration of each perturbation source during integration. Therefore, if the battery heat generation power sequence, the motor / electronic control heat generation power sequence, the heat load sequence, and the heat exchange amount sequence can first be mapped to an associated space with a stable state structure, the balanced expression of the predictable perturbation sequence with respect to the above variables can be achieved.

[0041] Specifically, in the embodiments of the present application, multi-source disturbance dynamic structured correlation equilibrium is respectively performed on the battery heat generation power sequence, the motor / electronic control heat generation power sequence, the heat load sequence, and the heat exchange quantity sequence to obtain a battery heat generation power equilibrium sequence, a motor / electronic control heat generation power equilibrium sequence, a heat load equilibrium sequence, and a heat exchange quantity equilibrium sequence, including: Represent the battery heat generation power sequence, the motor / electronic control heat generation power sequence, the heat load sequence, and the heat exchange quantity sequence as row vectors respectively to obtain a set of state vectors, that is , , and ; Calculate the mean vector between the sets of state vectors to obtain the state mean vector, and this process is expressed as: ; Wherein, is vector addition by position point, is the state mean vector; Based on the state mean vector, calculate the covariance matrix of each state vector in the set of state vectors respectively to obtain a set of state characteristic covariance matrices, and this process is expressed as: ; Wherein, is the th state vector in the set of state vectors, is vector subtraction by position point, is matrix multiplication, is the transpose operation, is the th state characteristic covariance matrix in the set of state characteristic covariance matrices; Perform sensitive correlation on each group of corresponding state vectors and state characteristic covariance matrices in the set of state vectors and the set of state characteristic covariance matrices to obtain a set of joint state function values, and this process is expressed as: ; Wherein, is the exponential function value with the natural constant e as the base, is the th joint state function value in the set of joint state function values, to describe the correlation coupling relationship between each state vector and the joint state; Based on each joint state function value in the set of joint state function values, construct the stable state prediction value of the corresponding state vector to obtain a set of stable state prediction values, and this process is expressed as: ; Wherein, is the calculation of the two-norm, is the A stable state prediction value, that is, by constructing an embedding mapping from the state vector to the joint state function, to obtain the stable state based on the non-degenerate norm representation; Using each stable state prediction value in the set of stable state prediction values as a compensation coefficient, performing state equilibrium association on each state vector in the set of state vectors to obtain a set of state equilibrium vectors; wherein, each state equilibrium vector in the set of state equilibrium vectors is the battery heat generation power equilibrium sequence, the motor / electronic control heat generation power equilibrium sequence, the heat load equilibrium sequence, and the heat exchange amount equilibrium sequence, and this process is expressed as: ; wherein, is the observation coefficient, is the th state equilibrium vector in the set of state equilibrium vectors, that is, taking this stable state prediction value as a dynamic structured constraint compensation and superimposing it on each state vector.

[0042] Specifically, the observation coefficient is dynamically adjusted through experimental calibration or an online optimization algorithm, and its value depends on the sensitivity requirements of the system to state vector deviations (such as ) and the noise suppression requirements, and can be specifically determined by historical data fitting or Kalman filter gain calculation.

[0043] In this way, the state association can be quantified through the calculation of the joint state function value, and the non-equilibrium association state caused by joint state degradation can be eliminated based on the calculation of the stable state prediction value, so as to finally dynamically constrain each state to the structured association space and realize the balanced expression of the predictable perturbation sequence with respect to the above variables.

[0044] Specifically, in step S34, integrating the battery heat generation power equilibrium sequence, the motor / electronic control heat generation power equilibrium sequence, the heat load equilibrium sequence, and the heat exchange amount equilibrium sequence to obtain the predictable perturbation sequence. It should be understood that although the foregoing steps have respectively predicted and processed and balanced the predictable thermodynamic perturbations from different sources (battery heat generation, motor / electronic control heat generation, cockpit heat load, environmental heat exchange), these perturbations ultimately act on the system and jointly affect the temperature states of key components such as the battery, motor, and cockpit. The core of the MPC algorithm is to predict the future state evolution based on a unified system dynamic model, and this model treats all external predictable influences as one or a group of structured perturbation input terms. Therefore, it is necessary to logically integrate these perturbation sequences from different sources, but after being balanced and aligned to the same prediction time domain, P time steps, according to their action modes in the system model, to form a single predictable perturbation sequence that contains all the predictable thermodynamic input information.

[0045] Specifically, in a specific example of the present application, the specific implementation process of step S34 is as follows: For each of the future P time steps k (where k ranges from 1 to P), the k-th value of the battery heat generation power equilibrium sequence, the k-th value of the motor / electronic control heat generation power equilibrium sequence, the k-th value of the heat load equilibrium sequence, and the k-th value of the heat exchange amount equilibrium sequence are respectively extracted. Then, these four values are arranged in a predefined order to form a perturbation vector at time step k. For example, this perturbation vector can be defined as [predicted battery heat generation power value_k, predicted motor / electronic control heat generation power value_k, predicted cockpit heat load value_k, predicted battery compartment heat exchange amount value_k]. Repeat the above combination process to construct corresponding perturbation vectors for the future P time steps (from time step 1 to time step P) in sequence. In this way, a sequence containing P perturbation vectors is obtained. This sequence formed by concatenating P perturbation vectors in chronological order is the predictable perturbation sequence.

[0046] In step S4, based on the current system state vector and the predictable perturbation sequence, an MPC optimization problem instance is constructed. Correspondingly, the working mechanism of MPC requires that in each control cycle, an optimization problem is solved online for the current system state and foreseeable future external influences to determine the optimal control input at the current moment. The current system state vector precisely describes the physical state of the system at the current moment and is the starting point for future state prediction. The predictable perturbation sequence provides information on the external perturbations that the system will face within the future prediction time domain, including known or predicted influences from driving conditions (battery, motor heat generation) and the environment (cockpit heat load, battery compartment heat exchange). To construct a mathematical optimization problem that can guide optimal control decisions, these two need to be integrated.

[0047] Specifically, in the embodiment of the present application, step S4, based on the current system state vector and the predictable perturbation sequence, constructs an MPC optimization problem instance, including: S41, performing state estimation optimization on the current system state vector based on a state estimator to obtain an optimized current system state vector; S42, constructing the MPC optimization problem instance based on the optimized current system state vector and the predictable perturbation sequence.

[0048] More specifically, in a specific embodiment of the present application, step S41, based on a state estimator, optimizes the state estimation of the current system state vector to obtain an optimized current system state vector, including: S411, inputting the state estimation at the previous moment and the control input at the previous moment into the system dynamic model to predict the state at the current moment to obtain a predicted current system state vector and the uncertainty of the predicted current system state vector; S412, inputting the predicted current system state vector, the uncertainty of the predicted current system state vector, and the current system state vector into the observation model to obtain the optimized current system state vector. The state estimator mentioned here is an intelligent algorithm module, whose core role is to fuse information from different sources, including the mathematical model prediction of the system, the system state estimation at the previous moment, and the actual sensor measurement data at the current moment to provide the best estimate of the actual internal state of the system. In particular, the state estimator includes an input system dynamic model and an observation model.

[0049] Correspondingly, the system dynamic model contains the physical laws governing the evolution of the system state over time. This prediction process is based on the internal mechanism of the system, providing a smooth and physically consistent current state estimate, avoiding the noise impact caused by directly using the current raw measurements. At the same time, since the model is not perfect, there are process noise and modeling errors, and this model-based prediction has uncertainty. Quantifying this uncertainty (e.g., through a covariance matrix) is crucial because it reflects the reliability of the model prediction and provides weight information for subsequent correction in combination with actual measurements. For this purpose, the state estimation at the previous moment and the control input at the previous moment are input into the system dynamic model to predict the state at the current moment to obtain a predicted current system state vector and the uncertainty of the predicted current system state vector.

[0050] Specifically, in a specific example of the present application, the specific implementation process of step S411 is as follows: Here, the state estimate at the previous moment refers to the optimal estimated value of the internal state variables of the system obtained by the state estimator after comprehensive data fusion (for example, filtering and optimizing by combining the predicted value at the previous moment and the measured value at the current moment) at the end of the immediately previous control cycle. It is a vector that contains the most credible numerical representations of the key physical quantities in the system (such as the temperatures of various regions of the battery pack, the temperatures of the motor and the electronic control unit, the temperature and flow rate of the coolant, the cabin temperature, etc.) at the previous moment. This vector is the output of the state estimation process in the previous step (or the previous control cycle). The control input at the previous moment refers to the control quantity actually applied by the controller of the thermal management system to each actuator or the actual response of these actuators within the immediately previous control cycle. For example, the actual rotational speed or power of the coolant pump, the actual rotational speed of the fan, the actual torque or power of the electric compressor, the heating power of the heater, etc. These values are the actual actions taken by the system at the previous moment to achieve the control objective, and they directly affect the change of the system state within the current control cycle. This control input is also a vector, which is composed of the input values of all controlled actuators in the system.

[0051] Taking these two vectors, the state estimate at the previous moment and the control input at the previous moment, as inputs, they are fed into the system dynamic model. This system dynamic model is the basis for state estimation and MPC prediction. It captures the physical laws or empirical relationships of the evolution of the internal state of the thermal management system over time. For the physics-based modeling method, the content of the system dynamic model is a set of mathematical equations, such as differential equations or difference equations describing energy conservation, heat transfer, fluid dynamics, electrical energy conversion to heat energy, etc. The parameters in these equations (such as the heat capacity of the material, the thermal conductivity, the convective heat transfer coefficient, the flow characteristic curve of the pump, the efficiency curve of the fan, the relationship between the resistance heating power and the current, etc.) constitute the core content of the model, and they characterize the physical properties of each component in the system and their interactions. For the data-driven or hybrid modeling method, the content of the model is the state transition function or parametric relationship learned from a large amount of historical data, which describes in a mathematical form (not necessarily explicit physical equations) how the input (previous state, control input, disturbance) leads to state changes.

[0052] Specifically, the state estimate at the previous moment is used as the starting state of the system, and the control input at the previous moment is used as the external excitation applied within this time interval. Then, the system dynamic model is utilized to deduce how the system state will change after one control cycle (i.e., the time interval from the previous moment to the current moment). If the system dynamic model is in the form of a differential equation, this is solved through numerical integration methods (such as Euler's method, Runge-Kutta method, etc.); if the model is in the form of a difference equation, a one-step recursive calculation is directly performed. This calculation process simulates physical processes such as how heat flows within the system and how temperature responds to control actions.

[0053] In addition to predicting the state vector itself, this step also requires obtaining the uncertainty of the predicted current system state vector. This is because the system dynamic model may not be completely accurate (model error), and there may be unmodeled random perturbations (process noise) in the system itself. In probabilistic state estimation algorithms (such as the Kalman filter and its variants), the system dynamic model not only includes the state transition equation but also contains the statistical characteristics (covariance matrix) describing the process noise. When making a prediction, not only the mean of the predicted state (i.e., the predicted current system state vector) is predicted, but also based on the linearized approximation of the state transition equation (or using methods such as unscented transformation to handle non-linearity) and the known process noise covariance, how the current state estimate covariance matrix propagates over time is predicted, thereby obtaining the uncertainty of the predicted current system state vector, represented as a covariance matrix. The diagonal elements of this covariance matrix represent the variances of the predicted values of each state variable, and the off-diagonal elements represent the covariances between them. This part of the system dynamic model (the process noise covariance matrix and its propagation method) is also an important component of the model. The setting of model parameters (such as heat transfer coefficients, heat capacities, etc.) and the process noise covariance matrix is carried out through system identification experiments or by referring to component specifications and physical principles.

[0054] In a specific example: The state estimate vector at the previous moment includes the estimated battery point temperature of 42°C and the estimated coolant temperature of 38°C. The control input vector at the previous moment includes the cooling pump power command of 50%. The content of the system dynamic model is a set of difference equations describing battery heat conduction and convective heat transfer, and its parameters (such as the equivalent heat capacity of the battery, the thermal resistance from the battery to the cooling plate, the relationship between the convective heat transfer coefficient from the cooling plate to the coolant and the coolant flow rate (determined by the pump power), etc.) are determined based on the battery pack design and experimental data. The model also includes a process noise covariance describing the uncertainty of internal heat generation in the battery and other unmodeled perturbations.

[0055] The model receives the estimated battery temperature of 42°C and the estimated coolant temperature of 38°C at the previous moment, as well as the cooling pump power command of 50% at the previous moment. The difference equation inside the model, based on these inputs and using the thermophysical parameters it contains, calculates how the battery temperature and the coolant temperature will change during the current control cycle (e.g., 100 milliseconds), considering processes such as the battery's own heat generation (which may be related to the battery current, internal resistance, etc., and the current may also be an input or a predicted perturbation to the model), heat transfer from the battery to the cooling plate, and heat transfer from the cooling plate to the coolant. This calculated result is the predicted current system state vector. For example, the predicted battery temperature is 42.1°C and the predicted coolant temperature is 38.05°C. At the same time, the model, according to the content describing the process noise in it and considering the uncertainty of the previous moment's state estimate (if available), calculates and outputs the uncertainty of the predicted current system state vector. For example, the variance of the predicted battery temperature is 0.02 and the variance of the predicted coolant temperature is 0.01. These values represent the possible fluctuation ranges of the predicted values. These predicted values and their uncertainties will be used in the subsequent state update steps.

[0056] It should be understood that the predicted current system state vector is the result of forward deduction based on the system dynamic model, reflecting the physical law of the system state evolution over time. However, it does not utilize the real-time sensor data at the current moment, so it cannot correct model errors or cope with unmodeled perturbations. The current system state vector directly comes from sensor measurements and contains the real information of the system at the current moment, although there is noise or incompleteness. For this reason, the predicted current system state vector, the uncertainty of the predicted current system state vector, and the current system state vector are input into the observation model.

[0057] Specifically, in a specific example of the present application, the specific implementation process of step S412 is as follows: The described observation model can be understood as a machine learning model specifically trained for data fusion. It learns how to calculate the best estimate of the system state at the current moment based on the state predicted by the model, the predicted uncertainty, and the actual sensor measurements. This model can be a feedforward neural network, also known as a multi-layer perceptron. Its architecture includes an input layer, several hidden layers, and an output layer.

[0058] The input layer of the model receives a combination of information from three parts: First, the predicted current system state vector, which is the speculation of the current state by the dynamic model based on the information from the previous moment; second, the uncertainty of the predicted current system state vector, which quantifies the credibility of the prediction by the dynamic model; third, the current system state vector, which is the original sensor measurement obtained by the system at the current moment, containing real information but also accompanied by noise. These three vectors are concatenated or combined in other ways into a single input vector and fed into the input layer of the feedforward neural network. The hidden layer processes the input signal through a non-linear activation function (such as ReLU), learning the complex relationships between the predicted value, uncertainty, and measurement value, and how they jointly affect the final state estimation. The output layer then generates a vector, namely the optimized current system state vector. This output vector is the optimal estimate of all key state variables of the system at the current moment. It combines prediction and measurement information, filters out noise, and may include inferences about states that cannot be directly measured.

[0059] Under the feedforward neural network architecture, this machine learning observation model needs to determine its internal parameters (weights and biases) through offline training. The training data includes a large number of historical records of: the predicted current system state vector corresponding to the same time point, the uncertainty of the predicted current system state vector (from the output of the trained dynamic model), the current system state vector (original sensor data), and the true current system state vector for supervised learning (obtained through high-precision simulation or experimental calibration). The model is trained by minimizing the error (e.g., mean squared error) between the output optimized current system state vector and the corresponding true current system state vector. During the training process, the model learns how to intelligently weight and combine the predicted value and the measurement value according to the credibility of the prediction (the magnitude of uncertainty) and the characteristics of the measurement value (the noise distribution in the training data) to obtain the optimal estimate.

[0060] More specifically, in a specific embodiment of the present application, in step S42, based on the optimized current system state vector and the predictable disturbance sequence, constructing the MPC optimization problem instance includes: S421, predicting the system state evolution of the electric vehicle thermal management system within the next P steps if a series of control inputs are applied based on the optimized current system state vector and the predictable disturbance sequence to obtain a predicted state trajectory and a predicted control trajectory; S422, defining a cost function for the predicted state trajectory and the predicted control trajectory, the cost function including temperature tracking error, energy consumption, control quantity change penalty, and weight coefficients; S423, imposing constraints on the cost function, the constraints including state constraints and input constraints to obtain the MPC optimization problem instance. Correspondingly, MPC is a control strategy based on online optimization, and its working mechanism requires constructing a mathematical optimization problem and solving it for the current system state and future expected external influences at each control cycle to obtain the optimal control instruction at the current moment. Therefore, in the present application, based on the optimized current system state vector and the predictable disturbance sequence, constructing the MPC optimization problem instance is the core link for integrating the thermal management requirements, system capabilities, and future prediction information into a solvable mathematical problem.

[0061] It should be understood that MPC has predictive ability because it can predict the state change of the system within a period of time in the future at the current moment. To achieve this prediction, MPC requires a mathematical model that can describe the dynamic behavior of the system and perform simulation and deduction based on this. Using the optimized current system state vector as the starting point of the simulation because it is the most reliable estimate of the current state of the system. At the same time, incorporating the future known external influences (reflected by the predictable disturbance sequence) into the model input to ensure that the prediction is closer to the actual situation.

[0062] Specifically, in a specific example of the present application, the specific implementation process of step S421 is as follows: First, the optimized current system state vector is the most accurate estimate of the current internal state of the system output by the state estimator at the current time step (referred to as time t). It is a vector containing key state information that has been fused and filtered, such as the current optimal estimated temperatures of various parts of the battery, motor temperature, cabin temperature, coolant state, etc. This vector will be used as the starting point for future predictions.

[0063] Second, the predictable disturbance sequence refers to the prediction of the external environment and operating conditions of the system within the next P time steps (from time t to t + P - 1). This can be a sequence of vectors, with each vector representing the predicted disturbance at a future moment.

[0064] It is borne by the trained machine learning system model. This model learns the law of how the system state evolves over time given the current state, control input, and external disturbance. This model can be a neural network with a sequence-to-sequence architecture or a recurrent neural network (RNN) capable of multi-step prediction, such as a long short-term memory network (LSTM) or a gated recurrent unit (GRU).

[0065] The input structure of the model can be designed to receive the predicted state of the previous step, the control input of the current step, and the predicted disturbance of the current step at each prediction step (from step k = 0 to P - 1). Specifically, for the first step of prediction (k = 0), the model receives the optimized current system state vector, the first element of the future control input sequence, and the first element of the predictable disturbance sequence. The model predicts the state at time t + 1. Then, for the second step of prediction (k = 1), the model receives the predicted state, the second element of the future control input sequence, and the second element of the predicted disturbance sequence. The model predicts the state at time t + 2, and so on, for P steps of prediction.

[0066] The internal architecture of this machine learning system model can be multiple layers of LSTM or GRU units. These recurrent units are able to capture the internal dynamic connections and nonlinear behaviors of the system and predict future state changes by learning the relationships between states, controls, and disturbances in historical data. At each prediction step, the recurrent unit processes the input of the current step (predicted state of the previous step, current control, current disturbance), updates its internal state, and then outputs the predicted value of the system state at the end of the current step through a fully connected layer.

[0067] Training this machine learning system model requires a large amount of historical data, including records of the state evolution of the system under different operating conditions and control inputs and the corresponding disturbance data. The model learns to predict the true state at the next moment by inputting the true state, control input, and disturbance at a certain moment, and extends to multi-step prediction, and optimizes the model parameters by minimizing the error between the entire predicted trajectory and the true trajectory.

[0068] During the prediction process of MPC, this is not a single prediction, but the above P-step prediction process is repeatedly executed for a series of alternative future control input sequences. The MPC optimizer generates multiple sets of different series of control input sequences (i.e., different predicted control trajectories). For each set of control sequences, they are input into the machine learning system model together with the optimized current system state vector and the predictable disturbance sequence, and the model then outputs the corresponding predicted state trajectory.

[0069] Give a specific example: The optimized current system state vector represents the battery maximum temperature of 45°C, the motor temperature of 70°C, and the cabin temperature of 25°C. Predict the future for P = 5 time steps (each step is 10 seconds). The predictable disturbance sequence is the environmental temperature prediction within the next 5 10 - second intervals, such as [30°C, 31°C, 32°C, 31°C, 30°C]. The MPC optimizer generates a future control input sequence (i.e., a predicted control trajectory), such as a cooling pump power command sequence [80%, 80%, 70%, 60%, 50%] and a fan speed command sequence [70%, 70%, 60%, 50%, 40%]. Input the starting state [45, 70, 25], the disturbance sequence [30, 31, 32, 31, 30], and this set of control sequences [80, 80, 70, 60, 50] and [70, 70, 60, 50, 40] into the trained machine - learning system model. The model first receives [45, 70, 25] (the current state) and [80, 70] (the first - step control and disturbance combined input) to predict the next - moment state, for example, the prediction is [45.2, 70.5, 25.3]. Then use this predicted state [45.2, 70.5, 25.3] and the next set of control [80, 70] and disturbance

[31] to predict the state at the next moment, and so on for 5 steps. Finally, a predicted state trajectory is obtained. For example, the battery temperature sequence is [45.2, 45.3, 45.2, 45.1, 45.0]. The MPC optimizer evaluates multiple such predicted state trajectories (generated by different control sequences), selects the control sequence that best meets the control objective, and uses its first element as the control instruction actually applied in the current cycle.

[0070] Correspondingly, optimization requires a clear cost function to measure the advantages and disadvantages of different control strategies. The goal of MPC is to find a control input sequence that can optimize the comprehensive performance within the future prediction time domain. To achieve the multi - objective optimization of the thermal management system, multiple performance indicators such as the accuracy of temperature control (e.g., making the battery temperature and cabin temperature as close as possible to the target values), the energy consumption of the system, and the smoothness of control actions need to be quantified and integrated into a single numerical index, which is the cost function. By taking the predicted state trajectory and the predicted control trajectory as inputs, the cost function can calculate the total cost generated if the system operates according to these trajectories.

[0071] Specifically, in a specific example of this application, the specific implementation process of step S422 is as follows: The cost function is constructed as the accumulation of instantaneous costs over the next P time steps, multiplied by corresponding weight coefficients. First is the temperature tracking error term. For each key temperature variable in the predicted state trajectory, the difference between its value at each of the next P time steps and the set target temperature is calculated. These differences are squared to penalize large deviations and then summed over the next P time steps. Second is the energy consumption term. The total energy consumed by the system actuators (such as compressors, water pumps, fans) over the next P time steps is calculated based on the predicted control trajectory. This involves mapping the control input at each time step to the actuator power and then accumulating. Finally is the control variable change penalty term. The difference between control inputs at adjacent time steps in the predicted control trajectory is calculated, squared, and then summed over all control inputs and all future time steps to smooth the control action. The total temperature tracking error, total energy consumption, and total control variable change penalty obtained from the above accumulation are respectively multiplied by the preset weight coefficients. These weight coefficients are positive values used to balance the importance of each performance metric. The final cost function value is the sum of these three weighted terms. In particular, during the system design and calibration phase, engineers set the initial weights according to the specific control objectives and performance requirements of the electric vehicle thermal management system. Then, a large number of simulations are performed by running the system model under various typical driving cycles and environmental conditions. By observing performance metrics such as the temperature control effect, energy consumption performance, and smoothness of control actions under different weight combinations, the weight coefficients are iteratively adjusted until the desired performance balance is achieved.

[0072] Specifically, in a specific example of the present application, the specific implementation process of step S423 is as follows: A series of mathematical constraint conditions are defined according to the physical characteristics of the thermal management system, component safety specifications, and actuator specifications. These constraints are applied to each time step of the predicted state trajectory and the predicted control trajectory generated by the first two sub-steps. For example, the state constraints include: for each time step k within the prediction horizon, the predicted temperature of a certain key component must be greater than or equal to its safety lower limit and less than or equal to its safety upper limit. The input constraints include: for each time step k within the prediction horizon, the predicted control input value of a certain actuator must be greater than or equal to its minimum output capacity and less than or equal to its maximum output capacity. Taking the cost function as the objective function of the optimization problem and the above state constraints and input constraints as the limiting conditions of the optimization problem forms the MPC optimization problem instance.

[0073] In step S5, the MPC optimization problem instance is input into the online optimization solver module to obtain the current optimal control instruction. Figure 3 FIG. is a flowchart of step S5 in the automatic control method for an electric vehicle thermal management system according to an embodiment of the present application. Specifically, in the embodiment of the present application, as Figure 3As shown, in step S5, inputting the MPC optimization problem instance into the online optimization solver to obtain the current optimal control instruction includes: S51, using a numerical optimization algorithm to solve the MPC optimization problem instance to obtain an optimal control input sequence; S52, extracting the control input at the current moment from the optimal control input sequence as the current optimal control instruction. The online optimization solver is a dedicated algorithm or software library for receiving the definition of the above MPC optimization problem instance and solving the problem under real-time requirements to find a series of future control input sequences that minimize (or maximize) the cost function while satisfying all constraint conditions.

[0074] It should be understood that the MPC optimization problem instance constructed in the previous step is a mathematical problem that describes how to select a control input sequence for the next P steps to minimize a cost function that measures the future system performance while complying with all physical constraints. To find this optimal control input sequence, powerful mathematical tools - numerical optimization algorithms - are needed. These algorithms can iteratively search for combinations of control inputs, evaluate the future state trajectories they predict through the system model, calculate the corresponding cost function values, and check whether all state and input constraints are satisfied.

[0075] Specifically, in a specific example of the present application, the specific implementation process of step S51 is as follows: The MPC optimization problem instance is a formal mathematical description of the control decision problem to be solved at the current time step. It contains all the necessary information for the numerical optimization algorithm to solve: First, it contains the current system state vector to be optimized, which serves as the starting point for prediction and the initial state during the optimization process.

[0076] Secondly, it includes a predictable disturbance sequence, where these external influences are predicted within the next P time steps and are considered when predicting the evolution of the system state. Thirdly, it includes a description of the system dynamic model, which can be a set of physical equations or a simplified form thereof, used to predict the state at the next moment based on the current state, control input, and disturbance. The core part is the definition of the cost function, which is a mathematical expression quantifying the control performance and cost, and the goal of the algorithm is to minimize this function. For example, the cost function may be defined as a weighted sum (such as the sum of squares) of the deviations between the predicted system states (such as temperature) and the desired setpoints within the next P time steps, plus a weighted sum (such as the sum of squares of the control input or the sum of squares of the change in the control input) of the "costs" caused by the magnitude or rate of change of the applied control input. The setting of the weights reflects the priorities of different optimization objectives, such as whether energy conservation or precise temperature control is prioritized. Finally, it includes the constraint conditions, which are a set of mathematical inequalities or equalities to be satisfied. They describe the physical limitations and operating requirements of the system. For example, the state variables (such as battery temperature, motor temperature) need to be maintained within a safe or comfortable range, and the control inputs (such as cooling pump power, fan speed) need to be within the upper and lower limits physically allowed by the actuator.

[0077] The numerical optimization algorithm is an iterative process used to find the minimum (or maximum) value of the cost function while satisfying the above constraint conditions. For the MPC problem, the goal is to find a sequence of control inputs for the next P time steps such that starting from the current state and evolving according to the system dynamic model, the state trajectory in the next P steps can minimize the cost function while all state and control constraints are satisfied.

[0078] Specifically, the numerical optimization algorithm includes: an evaluation mechanism for the cost function and constraint functions. In each iteration of the algorithm, it is necessary to be able to, based on a future P-step control input sequence to be evaluated, use the system dynamic model to simulate and predict the corresponding future P-step state trajectories, and calculate the current value of the cost function and check whether all constraint conditions are satisfied (or the degree of violation) based on these trajectories and the control sequence. Sensitivity information calculation method. Most efficient numerical optimization algorithms need to know the "sensitivity" of the cost function and constraint functions to the control input, that is, the rate at which they change with the control input. This involves calculating the partial derivatives (i.e., gradients or Jacobian matrices) of the cost function and constraint functions with respect to the future P-step control input sequence, and sometimes even the second-order partial derivatives (Hessian matrix). The calculation of these partial derivatives can be achieved through analytical differentiation, automatic differentiation, or numerical methods (such as finite differences). Search direction determination strategy. Based on the current cost function value, constraint violation situation, and the calculated sensitivity information, the algorithm needs to decide in which "direction" to adjust the control input based on the current control input sequence in order to more effectively reduce the cost function value or satisfy the constraints. Different algorithms have different direction determination strategies. For example, the gradient descent method moves along the negative gradient direction, the Newton method uses second-order information, and methods such as sequential quadratic programming (SQP) determine the direction by solving a quadratic programming subproblem. Step size selection method. Determine how far to move along the calculated search direction. The appropriate step size should not only ensure that the cost function decreases (for unconstrained optimization) or the constraints are improved, but also guarantee the stability and convergence of the algorithm. This involves techniques such as line search or trust region. Iterative update rule. Update the current control input sequence according to the determined search direction and step size to generate a new and better estimated value of the control input sequence. Stopping criterion. Define when it is considered that a satisfactory solution has been found and the iteration is terminated. Common stopping criteria include: the change in the cost function value is less than a certain threshold, the degree of constraint violation is less than a certain tolerance, the norm of the gradient is small enough, or the preset maximum number of iterations is reached.

[0079] Specifically, in a specific example of this application, the MPC optimization problem instance includes: the current estimated cockpit temperature of 26°C, the target cockpit temperature of 22°C, the predicted sequence of ambient temperatures for the next 3 steps [30°C, 31°C, 32°C], and the upper and lower limits of the PTC heater power [0W, 1000W], and the upper and lower limits of the air-conditioning compressor power [0W, 2000W]. The system dynamic model describes the thermodynamic relationship of how the cockpit temperature changes with the heater power, compressor power, ambient temperature, and occupant heat generation (which may be treated as a known disturbance). The cost function is set to minimize the sum of the squared deviations of the predicted cockpit temperature for the next 3 steps from the target value of 22°C, plus the sum of the squares of the heater power and compressor power for the next 3 steps (as the energy consumption cost). The constraints are that the predicted cockpit temperature needs to be within the range of [21°C, 24°C], and the heater and compressor powers are within their respective upper and lower limits. An SQP solver in the online optimization solution module receives this information. The solver initializes a control sequence for the next 3 steps, such as heater power [0W, 0W, 0W], compressor power [500W, 500W, 500W]. Then it starts iterations: simulating the system dynamic model to predict the cockpit temperature trajectory; calculating the cost function value and constraint violation; calculating the gradient with respect to the power sequence; solving the QP sub-problem to determine the search direction and step size; updating the power sequence; repeating until an optimal sequence is found, such as heater power sequence [0W, 0W, 0W], compressor power sequence [800W, 750W, 700W], so that the predicted cockpit temperature trajectory can quickly and comfortably reach and maintain near the target, while satisfying all constraints and having low energy consumption. The solver finally returns this optimal control input sequence [[0W, 800W], [0W, 750W], [0W, 700W]].

[0080] Accordingly, the MPC algorithm adopts a predictive control strategy, that is, in each control cycle, the optimal control sequence for a period of time in the future is calculated, but only the first control instruction in this sequence is actually applied to the system actuator. This is because MPC is executed cyclically: when the next control cycle arrives, the system state has changed, and new predicted disturbance information is also available. At this time, a new MPC optimization problem will be reconstructed and solved. Therefore, among the future control sequences calculated in the previous cycle, except for the first instruction, the rest of the instructions will no longer be optimal due to the new information. This "cyclic optimization and rolling execution" characteristic of MPC enables the controller to continuously use the latest system state and prediction information to correct its control strategy, thereby improving the adaptability and robustness of the control. Therefore, from the entire calculated optimal control input sequence, only the control instruction corresponding to the current moment (the first time step of the prediction horizon) is extracted and applied, which is the current optimal control instruction.

[0081] Specifically, in a specific example of the present application, the specific implementation process of step S52 is as follows: Obtain the optimal control input sequence corresponding to the optimal control instructions within the next P time steps. Extract the control vector at the first position (corresponding to the first time step starting from the current moment) from this sequence. This extracted control vector is the current optimal control instruction. This control instruction is a multi-dimensional vector that contains the set values or action instructions to be sent to each actuator (such as water pumps, compressors, electronic expansion valves, fans, etc.) of the thermal management system during the current control cycle. This instruction will be sent to the drive actuator of the thermal management system, converted into a physical drive signal, and act on the system.

[0082] In step S6, send the current optimal control instruction to the drive actuator of the electric vehicle thermal management system to obtain a physical drive signal and updated real-time vehicle state data. It should be understood that the output of the control algorithm must act on the actual physical system to achieve the regulation of the system. The ultimate goal of the foregoing complex calculation and optimization process is to generate this current optimal control instruction. This instruction is a digital-form signal that represents how to operate various actuators of the thermal management system at the current moment to achieve optimal performance. However, actuators (such as water pumps, fans, compressors, electronic expansion valves, etc.) cannot directly understand digital instructions, and they need to receive specific physical signals (such as voltage, current, PWM signals, CAN bus commands, etc.) to drive their actions. Therefore, it is necessary to convert the digital control instruction into a physical drive signal that the actuator can recognize and respond to. When these physical drive signals act on the actuator, the actuator will adjust its working state according to the instruction (such as changing the water pump speed, compressor power, valve opening, etc.), thereby changing the coolant flow rate, refrigerant flow rate, air volume, etc., and further affecting the temperature, pressure, etc. of each component in the thermal management system. The change in the system state will be measured by sensors to generate updated real-time vehicle state data reflecting the current actual situation of the system. These updated data will be used as the perception input for the next control cycle to construct a new current system state vector, drive the state estimation, disturbance prediction, and MPC optimization processes of the next cycle, and form a closed-loop control.

[0083] Specifically, in a specific example of the present application, the specific implementation process of step S6 is as follows: Send the current optimal control instruction of the previous step to the controller of the thermal management system of the electric vehicle or the relevant drive interface module. This current optimal control instruction is a vector that contains the target values to be set by each actuator in the thermal management system during the current control cycle. For example, the instruction includes the target water pump speed, the target compressor torque or speed, the target electronic expansion valve opening, the target fan speed, etc. After receiving these digital instructions, the controller of the thermal management system or the drive interface module will convert these digital values into corresponding physical drive signals according to the preset drive strategy or interface specification. For example, convert the target water pump speed into a specific PWM signal or voltage signal and send it to the water pump driver; convert the target compressor torque or speed into the corresponding current or CAN bus command and send it to the compressor controller; convert the target valve opening into the number of stepping motor pulses or voltage signal and send it to the electronic expansion valve actuator. These physical drive signals directly act on each actuator of the thermal management system, prompting it to perform the actions specified by the instruction.

[0084] When the actuator changes its working state according to the physical drive signal, the physical state of the thermal management system (such as the temperature, pressure, and flow rate of each component) will change. These changes are measured in real time by sensors (temperature sensors, pressure sensors, flow sensors, etc.) arranged everywhere in the system. The original analog or digital signals generated by these sensors are collected by the vehicle's bus system (such as the CAN bus) and transmitted to the main controller or the thermal management domain controller. After the collected original sensor data undergoes the necessary signal processing, filtering, and format conversion, it constitutes the updated real-time vehicle state data that reflects the current actual operating condition of the system. These data are updated periodically and are used at the beginning of the next control cycle to construct a new current system state vector, thereby driving the iterative execution of the entire MPC control cycle.

[0085] In summary, the automatic control method of the electric vehicle thermal management system based on the embodiments of the present application is elucidated. By integrating external information such as navigation paths and weather forecasts, it realizes multi-step prediction of future battery and motor heat generation and heat load changes, plans the thermal management strategy in advance, and improves the forward-lookingness of the system response. It adopts the multi-source perturbation dynamic structured correlation equalization technology to model the coupling relationship between each heat source, and compensates for cross-perturbations through the joint state function to achieve coordinated control of multiple heat circuits. Based on the model predictive control (MPC) framework, it optimizes the temperature tracking error and energy consumption penalty, reduces the energy consumption of actuators such as compressors and water pumps while meeting the safe temperature range, and extends the battery life. By combining the state estimator to fuse the model and observation data, it enhances the control accuracy and robustness under complex working conditions, and finally realizes the global optimal thermal management that takes into account component life, passenger comfort, and energy efficiency.

[0086] Figure 4 This is a block diagram of an automatic control device for an electric vehicle thermal management system according to an embodiment of the present application. As Figure 4 shown, the automatic control device 100 for an electric vehicle thermal management system according to an embodiment of the present application includes: a data acquisition module 110, configured to acquire real-time vehicle state data and external prediction information, where the external prediction information includes navigation path information and weather forecast information for the next N time steps; a current system state vector construction module 120, configured to construct a current system state vector based on the real-time vehicle state data; an external prediction information conversion module 130, configured to convert the external prediction information into a predictable disturbance sequence; an MPC optimization problem instance construction module 140, configured to construct an MPC optimization problem instance based on the current system state vector and the predictable disturbance sequence; an optimal control instruction generation module 150, configured to input the MPC optimization problem instance into an online optimization solver to obtain a current optimal control instruction; and a drive execution module 160, configured to send the current optimal control instruction to a drive actuator of the electric vehicle thermal management system to obtain a physical drive signal and updated real-time vehicle state data.

[0087] Here, those skilled in the art can understand that the specific operations of each step in the above automatic control device for the electric vehicle thermal management system have been described in detail above with reference to Figures 1 to 3 the description of the automatic control method for the electric vehicle thermal management system, and therefore, the repeated description thereof will be omitted.

[0088] In summary, it is intended that the above detailed description be regarded as illustrative rather than restrictive, and it should be understood that the above embodiments should be understood as only for illustrating the present invention and not for limiting the protection scope of the present invention. After reading the content recorded in the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. An automatic control method for a thermal management system of an electric vehicle, characterized in that, including: obtaining real-time vehicle state data and external prediction information, where the external prediction information includes navigation path information and weather forecast information for the next N time steps; constructing a current system state vector based on the real-time vehicle state data; converting the external prediction information into a predictable disturbance sequence; constructing an MPC optimization problem instance based on the current system state vector and the predictable disturbance sequence; inputting the MPC optimization problem instance into an online optimization solving module to obtain a current optimal control instruction; sending the current optimal control instruction to a drive actuator of an electric vehicle thermal management system to obtain a physical drive signal and updated real-time vehicle state data.

2. The automatic control method of the electric vehicle thermal management system according to claim 1, characterized in that, The navigation path information includes altitude, speed limit, and congestion prediction; the weather forecast information includes ambient temperature, ambient humidity, and solar radiation intensity.

3. The automatic control method of the electric vehicle thermal management system according to claim 2, characterized in that, Converting the external prediction information into a predictable disturbance sequence includes: predicting the battery heat generation power and motor / electronic control heat generation power for the next P steps based on the navigation path information for the next N time steps to obtain a battery heat generation power sequence and a motor / electronic control heat generation power sequence; predicting the heat load on the cockpit and the heat exchange amount with the battery compartment for the next P steps based on the weather forecast information for the next N time steps to obtain a heat load sequence and a heat exchange amount sequence; performing multi-source disturbance dynamic structured correlation equalization on the battery heat generation power sequence, the motor / electronic control heat generation power sequence, the heat load sequence, and the heat exchange amount sequence respectively to obtain a battery heat generation power equalization sequence, a motor / electronic control heat generation power equalization sequence, a heat load equalization sequence, and a heat exchange amount equalization sequence; integrating the battery heat generation power equalization sequence, the motor / electronic control heat generation power equalization sequence, the heat load equalization sequence, and the heat exchange amount equalization sequence to obtain the predictable disturbance sequence.

4. The automatic control method of the electric vehicle thermal management system according to claim 3, wherein Performing multi-source disturbance dynamic structured correlation equalization on the battery heat generation power sequence, the motor / electronic control heat generation power sequence, the heat load sequence, and the heat exchange amount sequence respectively to obtain a battery heat generation power equalization sequence, a motor / electronic control heat generation power equalization sequence, a heat load equalization sequence, and a heat exchange amount equalization sequence includes: representing the battery heat generation power sequence, the motor / electronic control heat generation power sequence, the heat load sequence, and the heat exchange amount sequence as row vectors respectively to obtain a set of state vectors; calculating a mean vector between the sets of state vectors to obtain a state mean vector; calculating a covariance matrix of each state vector in the set of state vectors based on the state mean vector to obtain a set of state characteristic covariance matrices; performing sensitive correlation on each pair of corresponding state vectors and state characteristic covariance matrices in the set of state vectors and the set of state characteristic covariance matrices to obtain a set of joint state function values; constructing a stable state prediction value of the corresponding state vector based on each joint state function value in the set of joint state function values to obtain a set of stable state prediction values; Using each of the steady state prediction values in the set of steady state prediction values as a compensation coefficient, perform state equilibrium association on each state vector in the set of state vectors to obtain a set of state equilibrium vectors; Among them, each state equilibrium vector in the set of state equilibrium vectors is the battery heat generation power equilibrium sequence, the motor / electronic control heat generation power equilibrium sequence, the heat load equilibrium sequence, and the heat exchange amount equilibrium sequence.

5. The automatic control method of the electric vehicle thermal management system according to claim 1, characterized in that Based on the current system state vector and the predictable disturbance sequence, construct an MPC optimization problem instance, including: Based on a state estimator, perform state estimation optimization on the current system state vector to obtain an optimized current system state vector; Based on the optimized current system state vector and the predictable disturbance sequence, construct the MPC optimization problem instance.

6. The automatic control method of the electric vehicle thermal management system according to claim 5, characterized in that, Based on a state estimator, perform state estimation optimization on the current system state vector to obtain an optimized current system state vector, including: Input the state estimation at the previous moment and the control input at the previous moment into the system dynamic model to predict the state at the current moment to obtain a predicted current system state vector and the uncertainty of the predicted current system state vector; Input the predicted current system state vector, the uncertainty of the predicted current system state vector, and the current system state vector into the observation model to obtain the optimized current system state vector.

7. The automatic control method of the electric vehicle thermal management system according to claim 6, characterized in that, Based on the optimized current system state vector and the predictable disturbance sequence, construct the MPC optimization problem instance, including: Input the optimized current system state vector and the predictable disturbance sequence, and predict the system state evolution of the electric vehicle thermal management system within the next P steps if a series of control inputs are applied to obtain a predicted state trajectory and a predicted control trajectory; Define a cost function for the predicted state trajectory and the predicted control trajectory, and the cost function includes temperature tracking error, energy consumption, control quantity change penalty, and weight coefficients; Apply constraints to the cost function, and the constraints include state constraints and input constraints to obtain the MPC optimization problem instance.

8. The automatic control method of the electric vehicle thermal management system according to claim 1, wherein, Input the MPC optimization problem instance into an online optimization solving module to obtain the current optimal control instruction, including: Use a numerical optimization algorithm to solve the MPC optimization problem instance to obtain an optimal control input sequence; Extract the control input at the current moment from the optimal control input sequence as the current optimal control instruction.

9. An automatic control device for an electric vehicle thermal management system, characterized in that, Including: A data acquisition module, configured to acquire real-time vehicle state data and external prediction information, where the external prediction information includes navigation path information and weather forecast information for the next N time steps; A current system state vector construction module, configured to construct a current system state vector based on the real-time vehicle state data; An external prediction information conversion module, configured to convert the external prediction information into a predictable disturbance sequence; An MPC optimization problem instance construction module, configured to construct an MPC optimization problem instance based on the current system state vector and the predictable disturbance sequence; The optimal control instruction generation module is used to input the MPC optimization problem instance into the online optimization solver to obtain the current optimal control instruction; The driving execution module is used to send the current optimal control instruction to the driving actuator of the electric vehicle thermal management system to obtain the physical driving signal and the updated real-time vehicle state data.

Images