Electric vehicle battery thermal management method and device based on predictive control

Abstract

The invention discloses an electric vehicle battery thermal management method and device based on prediction control, relates to the field of battery thermal management, and generates a fusion working condition prediction profile by fusing a long-term information source, a short-term information source and current vehicle state data. Secondly, based on the weighted evaluation of the weather forecast accuracy, the traffic prediction deviation and the short-term correction uncertainty level, the prediction uncertainty level is quantified, and the prediction uncertainty level is embedded into the MPC optimization objective function weight dynamic adjustment and constraint margin design, so that the excessive dependence of a controller on ideal prediction is avoided. In addition, through a real-time deviation index monitoring and fuzzy logic fine tuning mechanism, a control instruction is corrected on line when the deviation between a predicted state and an actual state exceeds a threshold value. According to the scheme, the perspectiveness, the robustness and the real-time performance of the battery thermal management system are remarkably improved, and the problems that a traditional method is inaccurate in prediction and poor in interference immunity are solved.

Description

Technical Field

[0001] This application relates to the field of battery thermal management, and more specifically, to a method and device for electric vehicle battery thermal management based on predictive control. Background Art

[0002] The performance of the power battery pack directly affects the endurance, power performance, service life and safety of electric vehicles, among which temperature management is particularly crucial. Lithium-ion batteries are sensitive to temperature changes. Excessive or too low temperature and uneven temperature distribution will reduce efficiency, accelerate aging, and even cause thermal runaway. Therefore, an efficient battery thermal management (BTM) system is essential for ensuring the stable operation of vehicles. Traditional control strategies mostly adopt rule-based or PID methods, relying on temperature threshold deviations for passive adjustment, lacking foresight and unable to achieve active intervention. To overcome the above problems, researchers have proposed a thermal management method based on model predictive control (MPC). Compared with traditional methods, this method combines the dynamic models of battery heat generation and heat dissipation, and uses the prediction of future driving conditions to optimize the control input in real time under various constraint conditions, so as to achieve energy conservation and consumption reduction and precise temperature control.

[0003] Although the battery thermal management scheme based on model predictive control has good application prospects, it still faces many challenges in actual operation. At present, some predictive control methods have problems of insufficient accuracy and robustness in working condition prediction. They often rely on a single or limited information source and are difficult to effectively integrate long-term strategic information (such as navigation routes, weather forecasts) and short-term tactical information (such as traffic conditions, driving behaviors), resulting in inaccurate prediction results and unable to accurately reflect the future heat load change trend. In addition, there is uncertainty in prediction, which comes from factors such as traffic flow randomness and weather forecast errors. Many existing MPC-BTM strategies have not fully quantified these uncertainties and have not reasonably utilized this type of information in control optimization, resulting in the controller being overly dependent on inaccurate prediction values and showing vulnerability when facing actual disturbances, and the control effect deteriorates. When the deviation between the prediction and the actual state is large in some systems, there is also a lack of an effective online correction mechanism, making it difficult to adjust the control strategy in time, affecting the real-time response ability and adaptability of the system, and restricting its wide application in complex real scenarios.

[0004] Therefore, how to improve the accuracy of working condition prediction, effectively evaluate and integrate the uncertainty of prediction, and perform robust MPC optimization and real-time control correction accordingly is the key issue to improve the actual performance of the electric vehicle battery thermal management system based on predictive control. Summary of the Invention

[0005] In order to solve the above technical problems, this application is proposed.

[0006] According to one aspect of the present application, a method for thermal management of an electric vehicle battery based on predictive control is provided, which includes: Obtain long-term information sources, short-term information sources, and current vehicle state data; Based on the long-term information source, the short-term information source, and the current vehicle state data, perform long-term and short-term operating condition fusion prediction to obtain a fusion operating condition prediction profile and a prediction uncertainty level; Input the fusion operating condition prediction profile and the current battery measurement state into a battery model to perform battery state prediction to obtain a battery state prediction trajectory; Based on the battery state prediction trajectory and the prediction uncertainty level, perform MPC optimization with uncertainty awareness to obtain an optimal BTM control sequence; Extract the current BTM control instruction from the optimal BTM control sequence and send it to the underlying actuator of the BTM system to obtain a real-time deviation index; Based on the comparison between the real-time deviation index and a preset threshold, determine whether to correct the current BTM control instruction.

[0007] According to another aspect of the present application, a device for thermal management of an electric vehicle battery based on predictive control is provided, which includes: An information data acquisition module for obtaining long-term information sources, short-term information sources, and current vehicle state data; A fusion operating condition prediction module for performing long-term and short-term operating condition fusion prediction based on the long-term information source, the short-term information source, and the current vehicle state data to obtain a fusion operating condition prediction profile and a prediction uncertainty level; A battery state prediction trajectory generation module for inputting the fusion operating condition prediction profile and the current battery measurement state into a battery model to perform battery state prediction to obtain a battery state prediction trajectory; An information data acquisition module for performing MPC optimization with uncertainty awareness based on the battery state prediction trajectory and the prediction uncertainty level to obtain an optimal BTM control sequence; A real-time deviation index calculation module for extracting the current BTM control instruction from the optimal BTM control sequence and sending it to the underlying actuator of the BTM system to obtain a real-time deviation index; An instruction correction module for determining whether to correct the current BTM control instruction based on the comparison between the real-time deviation index and a preset threshold.

[0008] Compared with the prior art, a method and device for electric vehicle battery thermal management based on predictive control provided by the present application generate a fused operating condition prediction profile that takes into account both global and local features by integrating long-term information sources, short-term information sources, and current vehicle state data, and introduce deviation dynamic compensation and conformal mapping optimization to improve the prediction accuracy and overcome the limitations of traditional single information source prediction. Secondly, based on the weighted evaluation of weather forecast accuracy, traffic prediction deviation, and short-term correction uncertainty level, the prediction uncertainty level is quantified and embedded in the dynamic adjustment of the weight of the MPC optimization objective function and the design of the constraint margin to achieve active perception and robust optimization of uncertainty and avoid the over-reliance of the controller on ideal predictions. In addition, through real-time deviation index monitoring and fuzzy logic fine-tuning mechanism, the control instruction is corrected online when the deviation between the prediction and the actual state exceeds the threshold, enhancing the adaptive ability of the system to disturbances. This solution significantly improves the forward-looking, robustness, and real-time performance of the battery thermal management system through multi-source information fusion prediction, uncertainty-aware optimization, and closed-loop feedback correction, and solves the problems of inaccurate prediction and poor disturbance resistance of traditional methods. Description of the Drawings

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

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

[0011] Figure 2 It is a data flow diagram of a method for electric vehicle battery thermal management based on predictive control according to an embodiment of the present application.

[0012] Figure 3 It is a flowchart of step S2 in a method for electric vehicle battery thermal management based on predictive control according to an embodiment of the present application.

[0013] Figure 4 It is a block diagram of a device for electric vehicle battery thermal management based on predictive control according to an embodiment of the present application. Detailed 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. Instead, 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] In view of the problems in the above background art, the present application proposes a method for thermal management of an electric vehicle battery based on predictive control. Figure 1 It is a flowchart of a method for thermal management of an electric vehicle battery based on predictive control according to an embodiment of the present application. Figure 2 It is a data flow diagram of a method for thermal management of an electric vehicle battery based on predictive control according to an embodiment of the present application. As Figure 1 and Figure 2 shown, the method for thermal management of an electric vehicle battery based on predictive control according to an embodiment of the present application includes: S1, obtaining a long-term information source, a short-term information source, and current vehicle state data; S2, performing long-term and short-term operating condition fusion prediction based on the long-term information source, the short-term information source, and the current vehicle state data to obtain a fusion operating condition prediction profile and a prediction uncertainty level; S3, inputting the fusion operating condition prediction profile and the current battery measurement state into a battery model to perform battery state prediction to obtain a battery state prediction trajectory; S4, performing MPC optimization with uncertainty perception based on the battery state prediction trajectory and the prediction uncertainty level to obtain an optimal BTM control sequence; S5, extracting a current BTM control instruction from the optimal BTM control sequence and sending it to a lower-level actuator of the BTM system to obtain a real-time deviation index; S6, determining whether to correct the current BTM control instruction based on a comparison between the real-time deviation index and a preset threshold.

[0016] In step S1, long-term information sources, short-term information sources, and current vehicle state data are obtained. In particular, the long-term information sources described here include navigation route data, long-term weather forecasts, user charging plans, and vehicle historical energy consumption data; the short-term information sources include GPS positioning data, real-time vehicle speed / acceleration sensor data, V2X traffic information, short-term weather forecasts / measured ambient temperatures, and analysis results of the driver's recent operating habits. It should be understood that long-term information sources such as navigation routes and long-term weather forecasts provide a macroscopic and trend-based prediction of the operating scenario for the system, which helps in formulating strategic thermal management plans; short-term information sources such as real-time GPS, V2X information, and driver behavior capture immediate and local dynamic changes, enabling effective correction of long-term predictions and enhancing the immediate accuracy of predictions; current vehicle state data such as battery SoC and internal temperature provide accurate initial conditions and feedback benchmarks for all predictions and controls. The comprehensive utilization of such multi-dimensional information aims to overcome the problems of large prediction errors and poor adaptability caused by relying solely on a single information source or failing to fully integrate long- and short-term information, and to accurately predict the battery state and formulate a more reliable model predictive control optimization strategy with uncertainty awareness for subsequent use.

[0017] Specifically, step S1 obtains the following: The acquisition of long-term information sources is first initiated through the in-vehicle human-machine interface or preset. For example, when the user enters a destination, such as the city center library, in the navigation system, the system will obtain detailed navigation route data from the navigation module. This data not only includes the total mileage and estimated total travel time but is also refined into a series of road segment characteristics, such as the next 5 kilometers being an urban expressway with a speed limit of 80 km / h, followed by a 3-kilometer urban congestion section with an estimated average speed of 20 km / h, and may include altitude profile information of key road segments. At the same time, the vehicle connects to the Internet weather service through its communication module and downloads the long-term weather forecast for the next few hours, for example, the next 6 hours, to obtain the environmental temperature profile along the route or in the target area, such as a temperature prediction value per hour and the solar radiation intensity level. If the user sets a charging plan through the vehicle settings or the mobile phone App, such as starting to charge at the home charging pile at 22:30 tonight with a target SOC of 90%, this user charging plan information is recorded by the system. The vehicle will also retrieve vehicle historical energy consumption data related to the current driver or the current route from its internal storage or the cloud database. For example, based on the historical driving style of the current driver and the average energy consumption of similar historical trips on the planned route, it is 14.2 kWh / 100km.

[0018] In parallel, short-term information sources and current vehicle state data are continuously collected at high frequency. The in-vehicle GPS positioning module provides the precise latitude and longitude coordinates of the vehicle at a preset frequency, such as 1 Hz. The vehicle's CAN bus broadcasts in real time the real-time vehicle speed from the wheel speed sensor and acceleration sensor, such as the instantaneous vehicle speed of 45 km / h, and the vehicle acceleration, such as the current acceleration of 0.5 m / s². If the vehicle is equipped with a V2X communication unit, it will receive real-time traffic information from the surrounding infrastructure or vehicles, such as the remaining 20 seconds of green light for the traffic signal 800 meters ahead or a traffic jam 1 kilometer ahead, with an average vehicle speed of 10 km / h. The short-term weather forecast, such as the predicted environmental temperature within the next 30 minutes, can be obtained by fusing the measured values of the in-vehicle environmental temperature sensor and the network meteorological data with a shorter time scale. The analysis results of the driver's recent operating habits, such as judging the current driving style as steady or aggressive by analyzing the throttle opening, braking frequency, and intensity of the driver in the past few minutes. The current vehicle state data is directly obtained from the battery management system (BMS), mainly including the current state of charge (SOC) of the battery, such as 68%, and the real-time temperature distribution data measured by multiple temperature sensors inside the battery pack, such as the current average battery temperature of 29.5°C. All these long-term information sources, short-term information sources, and current vehicle state data obtained in real time are integrated and processed by the system as the input for the subsequent modules.

[0019] In step S2, based on the long-term information source, the short-term information source, and the current vehicle state data, long-term and short-term operating condition fusion prediction is performed to obtain a fusion operating condition prediction profile and a prediction uncertainty level. Accordingly, considering that traditional prediction methods often rely on information of a single time scale and are difficult to balance strategic long-term planning and tactical immediate adjustment, resulting in low prediction accuracy and poor adaptability especially in complex and changeable real driving environments. By fusing the macro trends provided by the long-term information source and the immediate corrections brought by the short-term information source, and combining the current vehicle state, a fusion operating condition prediction profile closer to the actual future situation can be generated. More importantly, by analyzing the inherent uncertainties of each information source and the deviations between them, the uncertainty level of the overall prediction can be quantified. This not only improves the robustness of the prediction but also enables the subsequent model predictive control to actively consider this uncertainty during the optimization process, thereby formulating a battery thermal management strategy that performs better and is safer under actual disturbances, effectively solving the problems of inaccurate prediction and lack of uncertainty perception mentioned in the background art.

[0020] Specifically, in a possible embodiment, Figure 3 is a flowchart of step S2 in the method for battery thermal management of an electric vehicle based on predictive control according to an embodiment of the present application. As Figure 3As shown, in step S2, based on the long-term information source, the short-term information source, and the current vehicle state data, long-term and short-term driving condition fusion prediction is performed to obtain a fusion driving condition prediction profile and a prediction uncertainty level, including: S21, inputting the long-term information source into a long-term prediction module to obtain a long-term power demand profile, a long-term ambient temperature profile, a solar radiation profile, and a long-term charging event sequence; S22, inputting the short-term information source and the current vehicle state data into a short-term prediction module to obtain a short-term power demand profile and a short-term ambient temperature profile; S23, updating the long-term power demand profile based on the short-term power demand profile to obtain an updated long-term power demand profile; S24, updating the long-term ambient temperature profile based on the short-term ambient temperature profile to obtain an updated long-term ambient temperature profile; S25, integrating the updated long-term power demand profile, the updated long-term ambient temperature profile, the solar radiation profile, and the long-term charging event sequence to obtain the fusion driving condition prediction profile.

[0021] It should be understood that in order to convert the original and macroscopic long-term planning information into a quantitative future load and environmental condition prediction that has direct guiding significance for the battery thermal management system. In this application, by inputting the long-term information source into the long-term prediction module, that is, the long-term power demand profile indicates the main energy consumption and corresponding heat generation during future driving; the long-term ambient temperature profile and the solar radiation profile define the influence of the external environment on the battery heat exchange conditions during vehicle operation; the long-term charging event sequence indicates the concentrated heat load that may be brought by the charging behavior. These structured prediction profiles are the basis for forward-looking energy optimization and heat management strategy formulation, enabling the system to pre-judge the possible risks of battery overheating or overcooling in the future, thereby effectively addressing the problem of the lack of foresight in the traditional control strategy mentioned in the background art.

[0022] Specifically, step S21 is processed as follows: The generation of the long-term power demand profile first depends on the detailed analysis of the navigation route data, from which the characteristics of each section in the future journey are extracted, such as the section type (e.g., highway, urban area, mountain road), the average speed of each section (e.g., the urban expressway section is expected to be 80 km / h, and the congested urban area section is expected to be 20 km / h), the slope information (e.g., a 2% uphill or a -1% downhill), and the expected driving duration of each section. These characteristics are then input into a pre-trained power prediction model. This model is a deep neural network that includes an embedding layer to process categorical features such as section type, followed by two layers of LSTM (Long Short-Term Memory) units, each layer having, for example, 64 hidden units to capture the temporal dependence of the driving conditions, and finally outputs the predicted power value through a fully connected layer. The input feature vector of this neural network includes the normalized average speed, slope value, one-hot encoded section type, and the estimated fixed power consumption of auxiliary devices such as air conditioners (e.g., preset to 1.2 kilowatts when the summer cooling is turned on). The model is offline trained with a large amount of vehicle historical energy consumption data that includes various driving conditions and their corresponding actual power consumption. After training is completed, the feature sequence parsed from the current planned route is input into the model, and a long-term power demand profile covering the future, for example, 60 to 120 minutes, with a time resolution of, for example, one data point per minute can be obtained.

[0023] The generation of the long-term environmental temperature profile and solar radiation profile is mainly based on the obtained long-term weather forecast. The module extracts the hourly predicted environmental temperature values and solar radiation intensity predicted values (unit: watts per square meter) along the predetermined route or target area within the next few hours, for example, within the next 6 hours, from the weather forecast. To obtain a continuous profile that matches the time resolution of the power demand profile (e.g., one data point every 15 minutes), the module will use the cubic spline interpolation algorithm to smooth the original hourly predicted data points, respectively forming a time-series long-term environmental temperature profile and a long-term solar radiation profile.

[0024] The generation of the long-term charging event sequence precisely analyzes the user's charging plan. For example, the user sets to start charging at 22:30 tonight at the home charging pile, with a target SOC of 90%. First, obtain the preset rated charging power of the home charging pile, such as 7 kilowatts. Then, combine the current state of charge of the battery, such as obtained from the vehicle status data as 35%, and the rated total capacity of the battery, such as preset as 60 kWh, calculate the charging power required to reach the target SOC of 90%, and consider the preset average charging efficiency, such as 90%. Based on this, estimate the required charging duration, (90% - 35%) * 60 kWh / (7 kW * 90%) is approximately equal to 5.2 hours. Therefore, a time series will be generated, in which starting from 22:30, within a time period of approximately 5.2 hours, the corresponding charging power value is set to 7 kW; at all other time points outside this time period, the charging power value is 0. The time resolution of this sequence is consistent with the power demand profile, such as one data point per minute. These precisely generated profiles together provide key long-term prediction inputs for subsequent battery thermal management.

[0025] Accordingly, although the long-term prediction gives the macroscopic trend, it is difficult to cope with sudden traffic conditions, immediate changes in driver behavior, or rapid changes in the microenvironment. Therefore, in order to precisely capture the recent dynamic changes of the vehicle through immediate and high-frequency data updates, so as to generate a short-term prediction result that is more accurate and responsive than the long-term prediction, the short-term prediction module uses information such as GPS positioning, real-time vehicle speed / acceleration, V2X information, short-term weather, and driver habit analysis to estimate the power demand and ambient temperature in the next few seconds to minutes, in order to obtain the short-term prediction profile.

[0026] Specifically, step S22 is processed as follows: To generate the short-term power demand profile, the short-term prediction module receives short-term information sources as inputs. These data are fed into a pre-trained neural network model as input features in their original acquisition frequency, such as GPS at 1 Hz, vehicle speed acceleration at 0.1 seconds, or in a properly processed form. The model can be a recurrent neural network (RNN) that includes an input layer, an LSTM (long short-term memory) network or a GRU (gated recurrent unit) layer (for example, each layer is configured with 32 to 64 neurons to effectively capture the dynamic dependencies between time series), and a fully connected output layer. The input layer receives the serialized data of the above short-term information sources within a relatively short past time window, such as the last 5 to 10 seconds. The RNN model is trained through offline supervised learning. The training data set contains a large amount of high-frequency sensor data in real driving scenarios and their corresponding actual vehicle power consumption within a short future period (for example, the next 30 seconds to 90 seconds, with a time resolution of, for example, 1 second). During the training process, the root mean square error (RMSE) is used as the loss function, and the network weights are adjusted through the backpropagation algorithm. After training, the model can predict the power demand profile within a short future time based on real-time inputs.

[0027] For the generation of the short-term environmental temperature profile, the short-term prediction module mainly relies on the data that fuses the measured values of in-vehicle environmental temperature sensors and short-term weather forecasts of the network. For example, the current in-vehicle sensor measures the external environmental temperature as 29.5 °C, and the obtained short-term weather service forecasts that the temperature will remain stable at this level within the next 30 minutes. At this time, the module can adopt a simple persistence model, that is, predict that the environmental temperature will basically remain at 29.5 °C within the next 15 to 30 minutes. Or, if the short-term forecast shows a clear temperature change trend, for example, the forecast is that the temperature will drop by 0.5 °C after 15 minutes, a linear interpolation or a simple autoregressive model can be used. For example, the parameters p, d, q can be preset as (1, 1, 1), and the measured temperature sequence in recent minutes and the forecast trend are used to generate a short-term environmental temperature profile with a relatively high time resolution, such as one data point per minute, within a short future time, such as 15 to 30 minutes. The preset parameters, such as the order of the AR model, can be determined based on historical data analysis. The finally output short-term power demand profile and short-term environmental temperature profile will be used for subsequent fusion and refinement of the long-term prediction profile.

[0028] It should be understood that although the long-term power demand profile is based on macroscopic planning (such as a navigation route), its prediction time span is relatively large, making it difficult to fully capture all the dynamic changes during actual driving, such as sudden traffic congestion, immediate adjustments in driver behavior, or unexpected road condition changes. Therefore, updating the long-term power demand profile can effectively correct the possible biases in the initial stage of long-term prediction, making the entire prediction profile closer to the actual upcoming working conditions, providing more accurate inputs for subsequent model predictive control, and thus enhancing the effectiveness of the battery thermal management strategy and system energy efficiency.

[0029] Specifically, in a possible embodiment, step S23 of updating the long-term power demand profile based on the short-term power demand profile to obtain an updated long-term power demand profile includes: replacing the corresponding part of the long-term power demand profile with the short-term power demand profile to obtain the updated long-term power demand profile. Correspondingly, although the long-term power demand profile provides the trend for a relatively long time in the future, its response to immediate changes is relatively lagged. The short-term power demand profile is generated based on high-frequency real-time data and can more accurately reflect the vehicle's dynamic demands in an extremely short time in the future. By directly replacing the initial segment of the long-term prediction with this part of the high-precision and high-resolution short-term prediction, it can be ensured that when the MPC controller makes an optimization decision, the prediction data it first relies on is closest to the real situation, thereby improving the effectiveness and robustness of the control actions and better coping with unexpected situations in actual driving, which is a key step in optimizing the battery thermal management efficiency.

[0030] Specifically, step S23 is processed as follows: First, obtain the previously generated long-term power demand profile and short-term power demand profile. The short-term power demand profile covers a shorter future time window, such as the next 30 - 90 seconds, and has a high time resolution, such as one data point per second. The long-term power demand profile covers a longer time span, such as the next 15 minutes (MPC prediction horizon Np) or even longer, and its time resolution may be lower, such as one data point per minute. The update process first determines the time length covered by the short-term power demand profile, such as Ns minutes. Then, in the long-term power demand profile, starting from the current moment, intercept the initial part corresponding to Ns minutes. This part will be completely replaced by the short-term power demand profile. In particular, before replacement, it may be necessary to process the short-term power demand profile to match the time resolution of the subsequent part of the long-term power demand profile, or perform smoothing at the splicing point. For example, if the short-term power demand profile is one data point per second and the long-term power demand profile is one data point per minute, the short-term power demand profile can be averaged or sampled at each minute time point to obtain a minute-level resolution before replacement. The finally formed updated long-term power demand profile has a high-precision short-term prediction for the initial Ns minutes, and the subsequent part is the remaining content of the original long-term prediction.

[0031] Furthermore, based on the deviation between the corresponding parts of the short-term power demand profile and the long-term power demand profile, the updated long-term power demand profile can be corrected based on this deviation. That is, instead of directly replacing the corresponding part of the long-term power demand profile with the short-term power demand profile, the updated long-term power demand profile is obtained by combining the short-term power demand profile and the corresponding part of the long-term power demand profile. This can improve the accuracy of the updated long-term power demand profile, thereby improving the accuracy of the fused operating condition prediction profile. And since the fused operating condition prediction profile also takes into account the deviation between the corresponding parts of the short-term power demand profile and the long-term power demand profile, this can further improve the consistency based on long- and short-term information between the fused operating condition prediction profile and the prediction uncertainty level. That is, here, the deviation between the corresponding parts of the short-term power demand profile and the long-term power demand profile essentially stems from the non-stationary characteristics of the power demand time series data of the electric vehicle's electrothermal system, that is, the statistical characteristic differences related to the power time scale distribution in the time series direction.

[0032] Based on this, in another possible embodiment, step S23, updating the long-term power demand profile based on the short-term power demand profile to obtain an updated long-term power demand profile, includes: Obtain the deviation values of the corresponding parts in the short-term power demand profile and the long-term power demand profile in the time series to obtain the time series of power demand deviation, that is: First, it is necessary to obtain the power demand deviation in the time series of the corresponding parts in the short-term power demand profile and the long-term power demand profile on the time series .

[0033] Construct the deviation delay phase space of each power demand deviation value in the time series of the power demand deviation to obtain the time series of power demand delay phase space deviation values, that is: ; where and respectively represent the th and the th time points, represents the time difference, and are respectively the power demand deviations corresponding to and in the time series of power demand deviation, is the th power demand delay phase space deviation value in the time series of power demand delay phase space deviation values, that is, in the case of evenly dividing the time series, the deviation delay phase calculation for time series scale decomposition can be performed, so as to construct the deviation delay phase space in the case of scale decomposition under the time series.

[0034] Perform conformal mapping calculation on each power demand delay phase space deviation value in the time series of the power demand delay phase space deviation values to obtain the time series of power demand deviation conformal mapping values, that is: ; where is the th power demand delay phase space deviation value in the time series of power demand delay phase space deviation values, is the logarithmic function value with the natural constant e as the base, is the th power demand deviation conformal mapping value in the time series of power demand deviation conformal mapping values, that is, through phase space conformal mapping, the distortion elimination between different time series phases is performed, so as to ensure the unified mapping of deviations under scale decomposition, that is, the decomposition scale coupling mapping of the time series fractal characteristics of deviations in the generalized phase space is realized.

[0035] Based on the time series of the power demand deviation conformal mapping values and the time series of the power demand delay phase space deviation values, perform dynamic compensation optimization on the time series of the power demand deviation to obtain the time series of power demand compensation deviation, that is: ; where is the coefficient of the correction term weight. For example, , of course, this is only an example here, and it can be determined according to prior knowledge, calibrated through experiments or adjusted by an optimization algorithm, is the th power demand compensation deviation in the time series of the power demand compensation deviation, that is, the compensated power demand compensation deviation, that is, the initial deviation value is introduced into the conformal mapping phase space to perform the decomposition scale time series distribution redundancy in the time dimension, and the time series dependence of the cross-deviation delay phase space is constructed.

[0036] Based on the time series of the power demand compensation deviation, the long-term power demand profile is updated to obtain the updated long-term power demand profile. In this way, according to the optimized power demand compensation deviation to obtain the updated long-term power demand profile. For example, taking the half deviation position between the corresponding parts of the short-term power demand profile and the long-term power demand profile as the updated long-term power demand profile can improve the accuracy of the fusion working condition prediction profile and the consistency based on long-term and short-term information between the fusion working condition prediction profile and the prediction uncertainty level.

[0037] Similarly, the inherent uncertainty and low time and space resolution of long-term weather forecasts make it difficult to accurately capture the local and instantaneous environmental temperature changes actually encountered during vehicle driving. For example, long-term forecasts may not be able to reflect the micro-environment temperature fluctuations caused by the vehicle entering a tunnel, under-bridge shadow area or urban heat island effect in a short time. Therefore, updating the long-term profile can effectively correct the deviation in the initial stage of long-term prediction, make the updated environmental temperature prediction closer to the real situation that the vehicle is about to experience, provide a more reliable environmental input for the subsequent accurate battery state simulation and optimization of the thermal management control strategy, and improve the overall control performance and energy efficiency.

[0038] Specifically, in a possible embodiment, step S24 of updating the long-term ambient temperature profile based on the short-term ambient temperature profile to obtain an updated long-term ambient temperature profile includes: replacing the corresponding part of the long-term ambient temperature profile with the short-term ambient temperature profile to obtain the updated long-term ambient temperature profile. It should be understood that although the long-term ambient temperature profile can provide the ambient temperature trend for a relatively long future time period, its time resolution is low and it is difficult to capture the local and rapid ambient temperature changes caused by microclimate, occlusion, etc. on the actual driving path of the vehicle. The short-term ambient temperature profile is based on real-time data from in-vehicle sensors and short-term meteorological information updated at high frequency, and can more accurately reflect the ambient temperature that the vehicle will actually experience in the next few minutes to dozens of minutes. By directly replacing the initial corresponding part of the long-term prediction with this high-precision short-term prediction, it is possible to ensure the closest-to-actual ambient temperature input in the near-term prediction period, thereby avoiding control deviation or energy consumption waste caused by inaccurate ambient temperature prediction. In particular, the processing process of this step is the same as that of the first embodiment of updating the long-term power demand profile described above.

[0039] Specifically, in another possible embodiment, step S24 can also be processed in the same way to obtain the updated long-term ambient temperature profile by the method of obtaining the updated long-term power demand profile by compensating the deviation between the short-term power demand profile and the long-term power demand profile described above.

[0040] Correspondingly, the temperature change of the battery is the result of the combined action of multiple factors, including power demand during driving (heat generation), ambient temperature (convective heat transfer), solar radiation (radiative heat absorption), and power input during charging (charging heat generation). Considering any one factor alone cannot accurately predict the temperature rise of the battery. Therefore, these key external and internal disturbances affecting the thermal state of the battery need to be integrated to form a comprehensive integrated operating condition prediction profile.

[0041] Specifically, step S25 is processed as follows: First, ensure that all input profiles, namely the updated long-term power demand profile, the updated long-term ambient temperature profile, the solar radiation profile, and the long-term charging event sequence, have a unified prediction time range and time resolution. For example, the predicted future time range is set to the next Np time steps. For instance, Np can be set to 30 steps, and each step Δt is 1 minute, so the total prediction duration is 30 minutes. If the time resolutions of the respective profiles are different when they are generated, resampling processing needs to be performed first. For example, if the solar radiation profile is given with one data point every 15 minutes, while the other profiles are given with one data point every 1 minute, then interpolation (such as linear interpolation or constant-hold interpolation) needs to be performed on the solar radiation profile to match the 1-minute resolution. Subsequently, for each discrete time point k within the prediction time range, from k = 0 to k = Np - 1, the predicted values of the respective profiles at this time point are combined. Specifically, the entries of the fused operating condition prediction profile at time point will include: the power demand value from the updated long-term power demand profile at time point , the ambient temperature value from the updated long-term ambient temperature profile, the solar radiation intensity value from the solar radiation profile, and the charging power value from the long-term charging event sequence. In this way, a multi-dimensional time series is formed, where each time point is associated with a complete set of operating condition parameters. For example, at the 5th minute (k = 4), the fused operating condition prediction profile may contain: {power demand: 10 kW, ambient temperature: 25 °C, solar radiation: 400 W / m², charging power: 0 kW}. This structured fused operating condition prediction profile will then serve as the key input to the battery thermal model and the MPC controller.

[0042] Specifically, in a possible embodiment, based on the long-term information source, the short-term information source, and the current vehicle state data, long-term and short-term operating condition fusion prediction is performed to obtain a fused operating condition prediction profile and a prediction uncertainty level, including: obtaining the weather forecast accuracy rate and the navigation traffic prediction accuracy; calculating the deviation between the corresponding parts of the short-term power demand profile and the long-term power demand profile, and generating a short-term correction uncertainty level based on the deviation; calculating the weighted sum of the weather forecast accuracy rate, the navigation traffic prediction accuracy, and the short-term correction uncertainty level to obtain the prediction uncertainty level.

[0043] The specific processing is as follows: First, obtain the accuracy rate of weather forecasts and the accuracy of navigation traffic predictions. The accuracy rate of weather forecasts, for example, is set as a value between 0 and 1, such as 0.85. This value can be obtained based on a long-term statistical evaluation of the historical forecast data of the meteorological service provider used, such as the comparison of temperature, radiation with actual observed values. For example, if the historical data shows that the probability that the average absolute error of its 24-hour temperature forecast is less than a preset threshold such as 1.5 degrees Celsius is 85%, then the accuracy rate can be set to 0.85. The accuracy of navigation traffic predictions, similarly, is set to 0.75, for example, and can be quantified by analyzing the deviation between the average vehicle speed or travel time prediction provided by the navigation module historically and the actual driving records of the vehicle. For example, if the historical data shows that in 75% of the cases, the error between the predicted travel time and the actual travel time is within the preset 10%.

[0044] Subsequently, calculate the deviation between the corresponding parts of the short-term power demand profile and the long-term power demand profile, and generate a short-term correction uncertainty level based on this deviation. Here, for example, the short-term power demand profile covers the next Ns minutes, such as Ns = 3 minutes. Extract the part of the long-term power demand profile corresponding to these Ns minutes. Compare the power values of the two profiles point by point at each time point within this time period, for example, one point every 10 seconds, and calculate the average value of the absolute value of their differences to obtain an average power deviation value, such as 3 kW. The short-term correction uncertainty level is then generated based on this average power deviation value. For example, it can be normalized. If a significant upper limit of the preset power deviation is 10 kW, then the short-term correction uncertainty level can be calculated as: the average power deviation value divided by the preset power deviation upper limit, that is, 3 kW divided by 10 kW to get 0.3. This value reflects the correction amplitude of short-term real-time information to long-term planning. The greater the correction, the greater the contribution of uncertainty in this aspect.

[0045] Finally, calculate the weighted sum of the weather forecast accuracy rate, the navigation traffic prediction accuracy, and the short-term correction uncertainty level to obtain the prediction uncertainty level. First, convert the accuracy rate index into an uncertainty index. For example, the weather forecast uncertainty is 1 minus the weather forecast accuracy rate (i.e., 1 - 0.85 = 0.15), and the navigation traffic prediction uncertainty is 1 minus the navigation traffic prediction accuracy (i.e., 1 - 0.75 = 0.25). Then, assign preset weight factors to these three uncertainty sources: weather forecast uncertainty, navigation traffic prediction uncertainty, and short-term correction uncertainty level. For example, the weights can be set as w1 = 0.2, w2 = 0.3, w3 = 0.5 respectively. The sum of these weight factors is 1, and its specific value can be determined in advance according to the sensitivity analysis of the influence of different factors on battery thermal management or expert experience. The prediction uncertainty level is finally calculated by the sum of the products of these three uncertainty components and their corresponding weights, that is, (0.15 * 0.2) + (0.25 * 0.3) + (0.3 * 0.5), and finally the prediction uncertainty level is obtained as 0.255.

[0046] In step S3, input the fused operating condition prediction profile and the current battery measurement state into the battery model to perform battery state prediction to obtain the battery state prediction trajectory. It should be understood that the performance, life, and safety of an electric vehicle battery highly depend on its operating state, especially temperature. The goal of thermal management is to actively maintain the battery within the optimal operating range. To achieve this active control, it is necessary to know in advance how the battery state will change under expected future driving and environmental conditions. The fused operating condition prediction profile provides detailed information on future external excitations, while the current battery measurement state provides accurate initial conditions for the prediction. The battery model can then, based on these inputs, follow physical and chemical laws, calculate, and output the sequence of battery states changing over time within the future prediction time domain, that is, the battery state prediction trajectory. This trajectory is the basis for formulating the optimal thermal management strategy in subsequent model predictive control.

[0047] Specifically, in a possible embodiment, inputting the fused operating condition prediction profile and the current battery measurement state into the battery model to perform battery state prediction to obtain the battery state prediction trajectory includes: using the current battery measurement state as the initial condition, and using the battery model to forward simulate the evolution of the future battery state under the fused operating condition prediction profile to obtain the battery state prediction trajectory.

[0048] The specific treatment is as follows: First, a calibrated electro-thermal coupling battery model is adopted. This model mainly includes an electrical sub-model and a thermal sub-model. The electrical sub-model, such as a second-order RC equivalent circuit model, whose core parameters include ohmic internal resistance, polarization internal resistance, and polarization capacitance. These parameters and the open-circuit voltage are all functions of the state of charge (SOC) and temperature of the battery, and are pre-stored in the form of a look-up table, and the data is sourced from laboratory tests of battery cells or modules under different working conditions. The thermal sub-model is established using the lumped parameter method to describe the average temperature change of the entire battery pack, or a more refined distributed parameter model is used to describe the internal temperature distribution. Its key parameters include the total heat capacity of the battery pack, for example, preset to 400 kilojoules per kelvin, this value is estimated and determined according to the specific heat capacity and mass of the battery material, the convective heat transfer coefficient between the battery surface and the environment, for example, looked up according to the wind speed or calculated using an empirical formula, which can be set to 5 watts per square meter per kelvin when stationary and up to 20 watts per square meter per kelvin when driving, the surface area of the battery, for example, 1.5 square meters, and the solar radiation absorption coefficient, for example, 0.7.

[0049] In the simulation practice, at the starting moment of prediction, the time step is recorded as 0. The current battery measurement status obtained by the controller from the sensor, for example, the average temperature of the battery pack is 29.5 degrees Celsius, the highest single cell temperature monitored inside the pack is 30.8 degrees Celsius, and the current state of charge of the battery is 68%, is used as the initial calculation conditions for the battery model. Subsequently, the simulation calculation will advance step by step forward according to the time step defined by the integrated operating condition prediction profile. For example, this time step is set to 1 minute, and the total number of prediction steps is advanced, for example, the total number of prediction steps is 30, corresponding to a total prediction time domain of 30 minutes. At each discrete time step from 0 to the total number of prediction steps minus 1: First, the external input data corresponding to this time step is extracted from the integrated operating condition prediction profile, that is, the power demand value at this moment, for example, discharging 10 kW, the ambient temperature value, for example, 25 degrees Celsius, the solar radiation intensity value, for example, 400 W / m², and the charging power value, for example, 0 kW. Second, based on the power demand or charging power at the current time step, combined with the state of charge and average temperature of the battery at this moment, the actual working current value of the battery is calculated through the electrical sub-model. For example, the discharge current can be obtained by dividing the power demand value by the currently estimated battery terminal voltage, and the terminal voltage itself is the open circuit voltage minus the product of the current value and the total internal resistance of the battery; this calculation may involve iterative solutions or rely on pre-stored detailed look-up tables. Then, according to the calculated current value and the total internal resistance of the battery in the current state, the Joule heat generated inside the battery is calculated. At the same time, the reversible entropy change heat (related to the current, temperature, and the rate of change of the open circuit voltage with temperature) may also be taken into account in the total heat generation. The total heat generation in this time step is obtained. Next, based on the ambient temperature value and solar radiation intensity value predicted by the integrated operating condition, combined with the current average temperature of the battery surface, the heat exchange between the battery and the external environment is calculated using the thermal sub-model. This includes the convective heat dissipation calculated through the convective heat transfer coefficient, surface area, and temperature difference, and the solar radiation absorption heat calculated through the solar radiation absorption coefficient, effective irradiated area, and solar radiation intensity. Subsequently, considering the total heat generation inside the battery and the net heat exchange with the environment (heat absorption minus heat dissipation) comprehensively, the net increase in heat inside the battery is obtained. Then, based on the total heat capacity of the battery pack, the change in the average temperature of the battery in this time step is calculated. For example, if the net heat input in one time step is 120 kJ and the total heat capacity is 400 kJ / K, the temperature will rise by 0.3 degrees Celsius. Thus, the average temperature of the battery at the start of the next time step is updated. If a distributed parameter thermal model is used, the temperatures of each node inside the battery pack will be updated similarly, and the predicted highest single cell temperature will be recorded. Finally, according to the calculated working current value and time step, the state of charge of the battery is updated. The new state of charge is equal to the current state of charge minus (or plus, for charging) the product of the current and the time step divided by the rated capacity of the battery (in ampere-hours).The updated average battery temperature, maximum cell temperature, and state of charge calculated at this time step are stored as a data point of the battery state prediction trajectory at this future time point. At the same time, these newly calculated state values also become the initial conditions for the simulation calculation of the next time step. This iterative calculation process is repeated for the total number of prediction steps, and finally a sequence covering the entire prediction time domain and containing the evolution of the key battery state parameters over time is generated, that is, the battery state prediction trajectory.

[0050] In step S4, based on the battery state prediction trajectory and the prediction uncertainty level, uncertainty-aware MPC optimization is performed to obtain an optimal BTM control sequence. Correspondingly, since traditional MPC optimization relying only on nominal predictions may lead to poor control effects or even violate constraints such as excessive battery temperature in practice due to prediction deviations. By introducing the prediction uncertainty level, uncertainty-aware MPC can explicitly consider the impact of this uncertainty during the optimization process, thereby improving energy efficiency as much as possible while ensuring battery safety and life.

[0051] Specifically, in a possible embodiment, performing uncertainty-aware MPC optimization based on the battery state prediction trajectory and the prediction uncertainty level to obtain an optimal BTM control sequence includes: constructing an objective function that includes an energy consumption weight, a temperature deviation weight, and a maximum temperature difference weight, and the energy consumption weight, the temperature deviation weight, and the maximum temperature difference weight are dynamically adjusted based on the prediction uncertainty level; obtaining constraint conditions, where the constraint conditions include state constraints, input constraints, and uncertainty-aware constraint margins; and combining the constraint conditions to solve and optimize the objective function based on an iterative algorithm to obtain the optimal BTM control sequence.

[0052] The specific processing is as follows: First, construct the objective function. In this solution, the objective function is designed to minimize a weighted sum over the entire prediction horizon, for example, Np time steps with a total duration of 30 minutes. That is, the objective function = (energy consumption term * energy consumption weight) + (sum of squared temperature deviation terms * temperature deviation weight) + (sum of squared maximum temperature difference terms * maximum temperature difference weight). Specifically, this weighted sum includes several key parts: The first part is the estimated total energy consumption of the thermal management actuator, such as the cumulative power consumption of the cooling fan, water pump, and refrigeration compressor at each time step, multiplied by the time step length, and then accumulated over the entire prediction horizon. Its weight is the energy consumption weight. The second part is the sum of the squares of the deviations between the battery temperature and its ideal target temperature, which is preset to 28 degrees Celsius, for example. This value is determined according to the battery type and performance requirements, and its weight is the temperature deviation weight. The third part is the square of the maximum temperature difference inside the battery pack, for example, the difference between the predicted highest single cell temperature and the average temperature, to ensure temperature uniformity. Its weight is the maximum temperature difference weight. The key is that the energy consumption weight, temperature deviation weight, and maximum temperature difference weight are not fixed but are dynamically adjusted according to the input prediction uncertainty level. If the prediction uncertainty level is high, it is inclined to increase the temperature deviation weight and the maximum temperature difference weight to adopt a more conservative temperature control strategy to ensure safety. At the same time, the energy consumption weight may be appropriately reduced because pursuing extreme energy conservation under high uncertainty may lead to the risk of temperature exceeding the limit. The specific adjustment rules can be preset, such as weight = base weight * (1 + product of adjustment factor and prediction uncertainty level), and the adjustment factor is pre-calibrated through simulation or experiment.

[0053] Secondly, obtain and set the constraint conditions. These constraints ensure that model predictive control always respects physical and safety boundaries during the optimization process. The constraint conditions mainly include: state constraints. For example, the average battery temperature in the predicted battery state trajectory must be maintained between a preset lower limit such as 15 degrees Celsius and an upper limit such as 35 degrees Celsius at each prediction time step k. The maximum temperature of a single cell must not exceed the absolute safety upper limit such as 45 degrees Celsius, and the state of charge of the battery should also be within the allowable range. Input constraints, that is, the control quantities of the battery thermal management actuator such as fan speed and cooling power cannot exceed their physical operable range at each time step k. For example, the fan speed is from 0 to 5000 revolutions per minute, and the cooling power is from 0 to 5 kilowatts. In particular, to reflect the characteristics of uncertainty perception, an uncertainty perception-constraint margin is introduced. This means that when setting state constraints, a certain tightening will be performed according to the predicted uncertainty level. For example, if the nominal maximum temperature constraint is 45 degrees Celsius and the predicted uncertainty level is 0.255, the actual constraint upper limit used for optimization can be set to 45 degrees Celsius minus a margin value determined by the uncertainty level (for example, the margin is equal to a sensitivity coefficient multiplied by the predicted uncertainty level and then multiplied by a temperature range reference value, such as 0.1 multiplied by 0.255 and then multiplied by 10 degrees Celsius, resulting in a margin of approximately 0.255 degrees Celsius, so the actual constraint becomes 44.745 degrees Celsius), so as to leave more safety margins when there is a large uncertainty in the prediction.

[0054] Finally, combine the objective function and all constraint conditions, and solve this optimization problem based on an iterative algorithm. Since the model predictive control problem is a non-linear and constrained multi-variable optimization problem, mature numerical optimization algorithms such as sequential quadratic programming (SQP) and interior point method are often used. At the beginning of each control cycle, the optimizer uses the latest predicted battery state trajectory and predicted uncertainty level, and through iterative calculations, finds a sequence of battery thermal management control inputs for Np time steps in the future prediction time domain, such as the fan gear and cooling system power at each time step, so that this sequence can minimize the value of the weighted objective function constructed earlier while satisfying all constraint conditions. This series of control actions obtained by solving is the optimal BTM control sequence.

[0055] In step S5, the current BTM control instruction is extracted from the optimal BTM control sequence and sent to the underlying actuator of the BTM system to obtain the real-time deviation index. It should be understood that model predictive control calculates an optimal control sequence for a future time domain (e.g., Np time steps) in each control cycle, but only executes the first control instruction in the sequence. Therefore, it is necessary to extract the control instruction corresponding to the current moment from the entire optimization sequence and send it down. This is a key step to convert the obtained abstract control quantity into actual physical actions by sending it to the underlying actuator. The real-time deviation index reflects the idea of closed-loop control. By monitoring the deviation between the actual execution effect and the expectation, this deviation information can be used to update and correct the battery model parameters, evaluate the control performance, or be used as the input for the next round of prediction uncertainty evaluation, so as to continuously improve the accuracy and robustness of thermal management.

[0056] Specifically, in a possible embodiment, extracting the current BTM control instruction from the optimal BTM control sequence and sending it to the underlying actuator of the BTM system to obtain the real-time deviation index includes: extracting the current BTM control instruction from the optimal BTM control sequence; inputting the current BTM control instruction into the underlying actuator of the BTM system and obtaining the current real-time measurement value; calculating the difference between the current real-time measurement value and the previously predicted current state to obtain the real-time deviation index.

[0057] The specific processing is as follows: First, the current BTM control instruction is extracted from the optimal BTM control sequence. After the optimization solution of the model predictive control MPC in the previous step, an optimal BTM control sequence covering the entire future prediction time domain, for example, Np time steps, is obtained. If each time step Δt is 1 minute, the total duration is Np minutes. This sequence contains the recommended control actions for each future time step. For example, a sequence containing Np elements, and each element may include the target speed of the cooling fan, the target power of the refrigeration compressor, the target power of the heater, etc. According to the rolling time domain characteristic of model predictive control, only the first element in the sequence is adopted as the control instruction to be executed at the current moment, that is, the current control cycle. For example, if the optimized optimal fan speed sequence is [2500 revolutions per minute, 2300 revolutions per minute,..., 1800 revolutions per minute], then the fan speed part in the extracted current BTM control instruction is 2500 revolutions per minute. Similarly, if the optimal refrigeration power sequence is [1.2 kilowatts, 1.0 kilowatts,..., 0.5 kilowatts], then the current refrigeration power instruction is 1.2 kilowatts.

[0058] Secondly, execute the input of the current BTM control instruction into the underlying actuator of the BTM and obtain the current real-time measurement values. The extracted current BTM control instruction, such as a fan speed of 2500 revolutions per minute and a refrigeration power of 1.2 kilowatts, will be converted into signals that the underlying actuator can recognize and execute (such as voltage signals, pulse width modulation PWM signals, or messages sent through a vehicle network such as the Controller Area Network CAN bus), and sent to the corresponding hardware units, such as the fan drive module and the compressor controller. These underlying actuators will implement the received instructions. After the current control instruction is sent and takes effect for a period of time, that is, before the end of the current control cycle and the start of the next control cycle, the current real-time measurement values are obtained through sensors of the battery management unit, such as temperature sensors, voltage sensors, current sensors, and other relevant sensors on the vehicle. These measurement values reflect the actual state of the battery after the control instruction has been executed. For example, the measured current average temperature of the battery pack is 30.5 degrees Celsius, the highest single-cell temperature is 31.8 degrees Celsius, and the current state of charge (SOC) of the battery is 67.2%.

[0059] Finally, calculate the difference between the current real-time measurement values and the previously predicted current state to obtain the real-time deviation index. The previously predicted current state here refers to the predicted value of the battery state at the current moment by the battery model during the MPC optimization in the previous control cycle. For example, in the previous control cycle, that is, 1 minute ago, MPC predicted that the average temperature of the battery pack at the current moment should be 30.2 degrees Celsius. Now, compare the currently measured average temperature of 30.5 degrees Celsius with this previous predicted value of 30.2 degrees Celsius. Calculate the difference: real-time deviation index for average temperature = (30.5 degrees Celsius - 30.2 degrees Celsius) = 0.3 degrees Celsius. Similarly, the deviation index for SOC, the deviation index for the highest temperature, etc. can be calculated. This real-time deviation index quantifies the accuracy of the model prediction and the actual effect of the control execution, and is an important basis for subsequent model correction, uncertainty assessment update, or control strategy adjustment.

[0060] In step S6, based on the comparison between the real-time deviation index and the preset threshold, it is determined whether to correct the current BTM control instruction. Accordingly, although model predictive control will be optimized according to the prediction, the complexity of the actual working conditions and the imperfection of the model will still result in a deviation between the prediction and the actual state, that is, the real-time deviation index. If this deviation exceeds the preset acceptable range (preset threshold), it indicates that the current control instruction may not be sufficient to effectively maintain the battery state within the desired range, or may lead to excessive energy consumption. By comparing and triggering the correction mechanism, this large deviation can be quickly responded to, and the currently issued control instruction can be fine-tuned, rather than relying entirely on the next MPC optimization cycle to correct. This helps to correct the control deviation more timely, prevent small deviations from accumulating into big problems, thereby improving the accuracy, safety of battery temperature control and the efficiency of overall thermal management, while avoiding unnecessary adjustments to small, normal fluctuations and ensuring the smoothness of control.

[0061] Specifically, in one possible embodiment, determining whether to correct the current BTM control instruction based on the comparison between the real-time deviation index and the preset threshold includes: in response to the real-time deviation index being greater than or equal to the preset threshold, fine-tuning the current BTM control instruction based on fuzzy logic.

[0062] The specific processing is as follows: First, the trigger condition for this process is that the real-time deviation index is greater than or equal to the preset threshold. For example, if the battery average temperature deviation in the real-time deviation index is +0.8 degrees Celsius (indicating that the actual temperature is 0.8 degrees Celsius higher than the previous prediction value), and the preset deviation threshold for this is +0.5 degrees Celsius, then since 0.8 degrees Celsius is greater than 0.5 degrees Celsius, the fine-tuning mechanism is activated. This preset threshold, such as 0.5 degrees Celsius, is determined based on the understanding of battery performance and safety requirements and a large amount of experimental data analysis, aiming to distinguish normal, acceptable fluctuations from significant deviations that require intervention.

[0063] Once triggered, the fine-tuning based on fuzzy logic begins. The first step is the fuzzification of the input variables. The real-time deviation index as the input (for example, the aforementioned average temperature deviation of positive 0.8 degrees Celsius) is converted into a fuzzy linguistic variable. This requires the pre-definition of fuzzy sets describing the degree of deviation and their membership functions. For example, the average temperature deviation can be divided into several fuzzy subsets such as large positive deviation (DBP), medium positive deviation (DMP), small positive deviation (DSP), zero deviation (DZ), small negative deviation (DSN), etc. For the input value of positive 0.8 degrees Celsius, it may belong to the large positive deviation with a membership degree of 0.7 and to the medium positive deviation with a membership degree of 0.3. The form of the membership function (such as triangular, trapezoidal, Gaussian) and its parameters are set according to experience or data-driven methods. The second step is fuzzy rule inference. The core is a pre-set "IF-THEN" fuzzy rule base. These rules map the fuzzified input deviation states to the adjustment amounts of the fuzzified output control instructions. For example, one rule may be: IF the average temperature deviation is large positive deviation THEN the adjustment amount of the fan speed is a large increase AND the adjustment amount of the cooling power is a large increase. Another rule may be: IF the average temperature deviation is medium positive deviation THEN the adjustment amount of the fan speed is a medium increase AND the adjustment amount of the cooling power is a medium increase. According to the membership degrees of the input deviation of positive 0.8 degrees Celsius in the large positive deviation and the medium positive deviation, these rules are activated to different degrees. The third step is the aggregation and defuzzification of the fuzzy output. Multiple rules may be activated simultaneously, and their outputs (for example, the adjustment amount of the fan speed is a large increase and the adjustment amount of the fan speed is a medium increase) need to be aggregated into a total fuzzy output. Subsequently, through defuzzification methods (such as the centroid method, the maximum membership degree method), this total fuzzy output is converted into an accurate, numerical control instruction adjustment amount. For example, after defuzzification, the obtained adjustment amount of the fan speed may be an increase of 250 revolutions per minute, and the adjustment amount of the cooling power is an increase of 0.2 kilowatts.

[0064] Finally, the calculated accurate adjustment amount is applied to the current BTM control instruction. If the original current fan speed instruction is 2500 revolutions per minute and the cooling power instruction is 1.2 kilowatts, then the fine-tuned instruction becomes a fan speed of 2750 revolutions per minute (2500 revolutions per minute plus 250 revolutions per minute) and a cooling power of 1.4 kilowatts (1.2 kilowatts plus 0.2 kilowatts). This fine-tuned control instruction will replace the original instruction and be sent to the underlying actuator.

[0065] In summary, the predictive control-based electric vehicle battery thermal management method according to the embodiments of the present application is elucidated. By integrating long-term information sources, short-term information sources, and current vehicle state data, it generates a fusion operating condition prediction profile that takes into account both global and local features, and introduces deviation dynamic compensation and conformal mapping optimization to improve the prediction accuracy and overcome the limitations of traditional single information source prediction. Secondly, based on the weighted evaluation of weather forecast accuracy, traffic prediction deviation, and short-term correction uncertainty level, the prediction uncertainty level is quantified and embedded in the dynamic adjustment of the MPC optimization objective function weight and the design of the constraint margin to achieve active perception and robust optimization of uncertainty, and avoid the over-reliance of the controller on ideal predictions. In addition, through real-time deviation index monitoring and fuzzy logic fine-tuning mechanism, when the deviation between the prediction and the actual state exceeds the threshold, the control instruction is corrected online to enhance the adaptive ability of the system to disturbances. This solution significantly improves the forward-looking, robustness, and real-time performance of the battery thermal management system through multi-source information fusion prediction, uncertainty-aware optimization, and closed-loop feedback correction, and solves the problems of inaccurate prediction and poor disturbance resistance of traditional methods.

[0066] Figure 4 FIG. is a block diagram of a predictive control-based electric vehicle battery thermal management device according to an embodiment of the present application. As Figure 4 shown, the predictive control-based electric vehicle battery thermal management device 100 according to an embodiment of the present application includes: an information data acquisition module 110 for acquiring long-term information sources, short-term information sources, and current vehicle state data; a fusion operating condition prediction module 120 for performing long-term and short-term operating condition fusion prediction based on the long-term information source, the short-term information source, and the current vehicle state data to obtain a fusion operating condition prediction profile and a prediction uncertainty level; a battery state prediction trajectory generation module 130 for inputting the fusion operating condition prediction profile and the current battery measurement state into a battery model to perform battery state prediction to obtain a battery state prediction trajectory; an information data acquisition module 140 for performing MPC optimization with uncertainty perception based on the battery state prediction trajectory and the prediction uncertainty level to obtain an optimal BTM control sequence; a real-time deviation index calculation module 150 for extracting the current BTM control instruction from the optimal BTM control sequence and sending it to the underlying actuator of the BTM system to obtain a real-time deviation index; and an instruction correction module 160 for determining whether to correct the current BTM control instruction based on the comparison between the real-time deviation index and a preset threshold.

[0067] As described above, the electric vehicle battery thermal management device 100 based on predictive control according to an embodiment of the present application can be implemented in various wireless terminals, such as a server having an electric vehicle battery thermal management algorithm based on predictive control. In a possible implementation manner, the electric vehicle battery thermal management device 100 based on predictive control according to an embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the electric vehicle battery thermal management device 100 based on predictive control can be a software module in the operating system of the wireless terminal, or can be an application program developed for the wireless terminal; of course, the electric vehicle battery thermal management device 100 based on predictive control can also be one of the many hardware modules of the wireless terminal.

[0068] Alternatively, in another example, the electric vehicle battery thermal management device 100 based on predictive control and the wireless terminal can also be separate devices, and the electric vehicle battery thermal management device 100 based on predictive control can be connected to the wireless terminal through a wired and / or wireless network and transmit interaction information in accordance with a predefined data format. Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned electric vehicle battery thermal management device based on predictive control have been described in detail in the description of the electric vehicle battery thermal management method based on predictive control in Figures 1 to 3 and thus, the repeated description thereof will be omitted.

Claims

1. A method for thermal management of an electric vehicle battery based on predictive control, characterized in that, Including: Obtain long-term information sources, short-term information sources, and current vehicle state data; Based on the long-term information source, the short-term information source, and the current vehicle state data, perform long-term and short-term driving condition fusion prediction to obtain a fusion driving condition prediction profile and a prediction uncertainty level; Input the fusion driving condition prediction profile and the current battery measurement state into a battery model to perform battery state prediction to obtain a battery state prediction trajectory; Based on the battery state prediction trajectory and the prediction uncertainty level, perform MPC optimization with uncertainty awareness to obtain an optimal BTM control sequence; Extract the current BTM control instruction from the optimal BTM control sequence and send it to the underlying actuator of the BTM system to obtain a real-time deviation index; Based on the comparison between the real-time deviation index and a preset threshold, determine whether to correct the current BTM control instruction.

2. The method for thermal management of an electric vehicle battery based on predictive control according to claim 1, characterized in that, The long-term information sources include navigation route data, long-term weather forecasts, user charging plans, and vehicle historical energy consumption data; the short-term information sources include GPS positioning data, real-time vehicle speed / acceleration sensor data, V2X traffic information, short-term weather forecasts / measured ambient temperatures, and analysis results of the driver's recent operation habits.

3. The method for thermal management of an electric vehicle battery based on predictive control according to claim 2, wherein Based on the long-term information source, the short-term information source, and the current vehicle state data, perform long-term and short-term driving condition fusion prediction to obtain a fusion driving condition prediction profile and a prediction uncertainty level, including: Input the long-term information source into a long-term prediction module to obtain a long-term power demand profile, a long-term ambient temperature profile, a solar radiation profile, and a long-term charging event sequence; Input the short-term information source and the current vehicle state data into a short-term prediction module to obtain a short-term power demand profile and a short-term ambient temperature profile; Update the long-term power demand profile based on the short-term power demand profile to obtain an updated long-term power demand profile; Update the long-term ambient temperature profile based on the short-term ambient temperature profile to obtain an updated long-term ambient temperature profile; Integrate the updated long-term power demand profile, the updated long-term ambient temperature profile, the solar radiation profile, and the long-term charging event sequence to obtain the fusion driving condition prediction profile.

4. The method for thermal management of an electric vehicle battery based on predictive control according to claim 3, wherein Based on the long-term information source, the short-term information source, and the current vehicle state data, perform long-term and short-term driving condition fusion prediction to obtain a fusion driving condition prediction profile and a prediction uncertainty level, including: Obtain the weather forecast accuracy and the navigation traffic prediction accuracy; Calculate the deviation between the corresponding parts of the short-term power demand profile and the long-term power demand profile, and generate a short-term correction uncertainty level based on the deviation; Calculate the weighted sum of the weather forecast accuracy, the navigation traffic prediction accuracy, and the short-term correction uncertainty level to obtain the prediction uncertainty level.

5. The method for thermal management of an electric vehicle battery based on predictive control according to claim 3, wherein Updating the long-term power demand profile based on the short-term power demand profile to obtain an updated long-term power demand profile, including: Obtain the deviation values of the corresponding parts of the short-term power demand profile and the long-term power demand profile in the time series to obtain the time series of the power demand deviation; Construct the deviation delay phase space of each power demand deviation value in the time series of the power demand deviation to obtain the time series of the power demand delay phase space deviation value; Perform conformal mapping calculation on each power demand delay phase space deviation value in the time series of the power demand delay phase space deviation value to obtain the time series of the power demand deviation conformal mapping value; Based on the time series of the power demand deviation conformal mapping value and the time series of the power demand delay phase space deviation value, perform dynamic compensation optimization on the time series of the power demand deviation to obtain the time series of the power demand compensation deviation; Based on the time series of the power demand compensation deviation, update the long-term power demand profile to obtain the updated long-term power demand profile.

6. The method for thermal management of an electric vehicle battery based on predictive control according to claim 1, characterized in that, Input the fusion operating condition prediction profile and the current battery measurement state into the battery model for battery state prediction to obtain the battery state prediction trajectory, including: using the current battery measurement state as the initial condition, and using the battery model to perform forward simulation calculation of the evolution of the future battery state under the fusion operating condition prediction profile to obtain the battery state prediction trajectory.

7. The method for thermal management of an electric vehicle battery based on predictive control according to claim 1, characterized in that, Based on the battery state prediction trajectory and the prediction uncertainty level, perform MPC optimization with uncertainty awareness to obtain the optimal BTM control sequence, including: Construct an objective function, which includes an energy consumption weight, a temperature deviation weight, and a maximum temperature difference weight, and the energy consumption weight, the temperature deviation weight, and the maximum temperature difference weight are dynamically adjusted based on the prediction uncertainty level; Obtain constraint conditions, which include state constraints, input constraints, and uncertainty awareness - constraint margin; Combine the constraint conditions and solve and optimize the objective function based on an iterative algorithm to obtain the optimal BTM control sequence.

8. The method for thermal management of an electric vehicle battery based on predictive control according to claim 1, wherein Extract the current BTM control instruction from the optimal BTM control sequence and send it to the underlying actuator of the BTM system to obtain the real-time deviation index, including: Extract the current BTM control instruction from the optimal BTM control sequence; Input the current BTM control instruction into the underlying actuator of the BTM system and obtain the current real-time measurement value; Calculate the difference between the current real-time measurement value and the previously predicted current state to obtain the real-time deviation index.

9. The method for thermal management of an electric vehicle battery based on predictive control according to claim 8, characterized in that Based on the comparison between the real-time deviation index and a preset threshold, determine whether to correct the current BTM control instruction, including: in response to the real-time deviation index being greater than or equal to the preset threshold, perform fine-tuning on the current BTM control instruction based on fuzzy logic.

10. An electric vehicle battery thermal management device based on predictive control, characterized in that, Including: An information data acquisition module for acquiring long-term information sources, short-term information sources, and current vehicle state data; A fusion operating condition prediction module for performing long-term and short-term operating condition fusion prediction based on the long-term information source, the short-term information source, and the current vehicle state data to obtain a fusion operating condition prediction profile and a prediction uncertainty level; A battery state prediction trajectory generation module for inputting the fusion operating condition prediction profile and the current battery measurement state into the battery model for battery state prediction to obtain the battery state prediction trajectory; An information data acquisition module, configured to perform uncertainty-aware MPC optimization based on the battery state prediction trajectory and the prediction uncertainty level to obtain an optimal BTM control sequence; A real-time deviation index calculation module, configured to extract a current BTM control instruction from the optimal BTM control sequence and send it to a low-level actuator of the BTM system to obtain a real-time deviation index; An instruction correction module, configured to determine whether to correct the current BTM control instruction based on a comparison between the real-time deviation index and a preset threshold.

Images