Self-adaptive heat management system of energy storage container

Abstract

The invention discloses a self-adaptive thermal management system for an energy storage container, and particularly relates to the technical field of thermal management of the energy storage container, which is characterized in that a joint probability model fusing multi-source data is constructed, Gaussian process regression and multi-precision CFD simulation fusion are introduced, and a chaotic feature extraction and anomaly recognition mechanism is combined, so that the self-adaptive thermal management of the energy storage container is realized. High-confidence dynamic modeling of complex airflow and temperature fields in the energy storage container and accurate recognition of abnormal areas are achieved, and the self-adaptive regulation and control capacity of a heat management system is improved; by collecting temperature and humidity data, constructing a condensation early warning mechanism and a micro-airflow intervention strategy and combining edge calculation and reinforcement learning, condensation risk real-time identification and control strategy optimization are achieved, the defects that in a traditional scheme, response to the problems of thermal runaway and condensation water accumulation is slow, and control lags are effectively overcome, and the method is suitable for large-scale popularization and application. And the safety and the reliability of the self-adaptive thermal management system of the energy storage container under the dynamic working condition are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy storage container thermal management, and more particularly, to an adaptive thermal management system for an energy storage container. BACKGROUND

[0002] In the closed operation and maintenance space of the energy storage container, in order to meet the heat dissipation needs of high-power battery packs or electronic devices generated during charging and discharging and continuous operation, various heat dissipation components such as fans, guide plates and cooling pipelines need to be arranged in the limited space. Due to the diverse and dense device layout, the airflow channel often forms a complex three-dimensional turbulent flow structure, and local vortex and dead zones can easily cause heat spot accumulation, resulting in uneven temperature distribution and reduced heat dissipation efficiency. In addition, environmental conditions (such as load fluctuations and external temperature changes) and frequent internal layout adjustments exacerbate the dynamic nature and uncertainty of the adaptive thermal management of the energy storage container, and pose stringent requirements on the real-time response capability and prediction accuracy of the heat dissipation system.

[0003] Most of the prior art relies on offline CFD simulation or empirical formula to evaluate the airflow and temperature field under static layout, and uses fixed PID control, preset fan curve or single-target optimization strategy to adjust the heat dissipation components. When facing layout changes, new devices or sudden conditions, these methods are difficult to capture transient vortex characteristics and local temperature fluctuations in a timely manner, and have problems such as large prediction error, control lag and uncontrollable energy consumption. At the same time, there is a lack of accurate quantification of sensor blind area uncertainty, and it is also impossible to achieve dynamic trade-off between energy consumption and temperature balance, and there is no perfect online self-learning and verification mechanism, so it is difficult to effectively prevent the risk of local thermal runaway caused by chaotic airflow disturbance. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the present application provides an adaptive thermal management system for an energy storage container, which realizes dynamic airflow regulation through multi-fidelity data fusion, adaptive chaos detection and multi-objective optimization control, to solve the problem of thermal runaway risk caused by chaotic disturbance in the background technology.

[0005] To achieve the above purpose, the present application provides the following technical scheme: an adaptive thermal management system for an energy storage container, comprising:

[0006] A data fusion and uncertainty quantification module synchronously collects multi-source data in the energy storage container when detecting changes in device layout or local temperature abnormalities in the energy storage container, and the multi-source data at least includes real-time airflow data, CFD simulation data and temperature distribution data. The multi-source data is fused through a Gaussian process regression algorithm, and the interpolation error of the sensor blind area is quantified using a Bayesian method, and a fused flow field map with a confidence interval is output.

[0007] a chaos intensity analysis module, which divides the fusion flow field map into a plurality of grid cells, uses a time series convolution network to extract multi-scale chaos features for a grid cell satisfying preset requirements of a confidence interval, and divides a chaos intensity level;

[0008] an abnormal area identification module, which generates a dynamic temperature threshold value in combination with a temperature gradient standard deviation of the grid cells, constructs an error thermal map of the energy storage container based on the fusion flow field map and the dynamic temperature threshold value, and takes a difference between the dynamic temperature threshold value and a predicted temperature in the fusion flow field map as an error thermal parameter of each grid cell; and when the chaos intensity level and the error thermal parameter jointly exceed a limit, marks an abnormal area;

[0009] an environment control instruction generation module, which separates real-time airflow data of the abnormal area into a low-frequency steady-state base flow component and a high-frequency chaotic disturbance component through variational mode decomposition when a duration of the abnormal area exceeds a limit, constructs a multi-objective optimization model with the lowest energy consumption and the smallest temperature fluctuation as targets, uses an improved particle swarm algorithm to obtain a Pareto optimal solution of a deflection angle of a deflector and a rotating speed of a fan, and generates an environment control instruction balancing energy consumption and heat dissipation efficiency;

[0010] a reliability verification module, which collects airflow field residuals in real time after executing the environment control instruction, takes a difference between actually measured airflow data and CFD simulation data as the airflow field residuals, and triggers a closed-loop verification process if the airflow field residuals exceed a preset threshold value;

[0011] a meta-reinforcement learning optimization module, which builds a digital twin platform of the energy storage container, randomly generates multi-working-condition disturbance data (including temperature disturbance, load disturbance and layout disturbance) in the digital twin platform, trains a meta-reinforcement learning model to quickly adapt to multi-working-condition disturbance scenarios (to avoid hysteresis of manual debugging) under the premise of meeting safety constraints, searches for the environment control instruction within a safety constraint boundary, and outputs the meta-reinforcement learning model after safety verification for subsequent calling.

[0012] Preferably, the multi-source data fusion manner is that the real-time airflow data, the temperature distribution data and the CFD simulation data are fused through a Gaussian process regression algorithm to construct a joint probability model.

[0013] The joint probability model is used to complementarily fuse the CFD simulation data with the real-time airflow data and the temperature distribution data to generate a fusion flow field map (i.e., a prediction of a flow field-temperature field joint distribution) of the energy storage container.

[0014] The space in the energy storage container is divided into a plurality of grid cells, and the fusion flow field map contains airflow velocity, airflow direction angle and predicted temperature of each grid cell.

[0015] The kernel function of the joint probability model is defined and the hyperparameters are optimized in the following manner: setting the optimization target as maximizing the likelihood probability of the joint probability model output fusion flow field map and measured data; using the conjugate gradient method to iteratively adjust the kernel function parameters until the joint probability model meets the optimization target.

[0016] Preferably, the chaotic intensity analysis module comprises:

[0017] Based on the fusion flow field map, the grid cells with confidence intervals meeting the requirements are defined as high confidence areas; the gas flow rate time series data and temperature time series data in the high confidence areas are extracted;

[0018] A time series convolution network is constructed, which includes 4 layers of dilated convolution layers with dilated factors of 1, 2, 4, and 8, respectively, a convolution kernel size of 3, and a channel number of 64;

[0019] The airflow flow rate time series data of the high confidence area is input, and a multi-scale chaotic feature vector is output. The dilated convolution structure of the time series convolution network covers time scales from seconds to minutes, which is suitable for the battery charging and discharging period (typical period 30-60 seconds), ensuring that the feature extraction result reflects the dynamic balance of battery heat generation and dissipation, and avoiding transient noise interference;

[0020] The chaotic feature vectors of all grid cells are input to obtain the chaotic intensity level of each area.

[0021] Preferably, when the chaotic intensity level of the grid cell exceeds the preset value and the predicted temperature exceeds the dynamic temperature threshold, it is marked as an abnormal grid;

[0022] Perform morphological closing operation (dilation and erosion operation) on the abnormal grids that meet the triggering conditions to generate a connected region coordinate set, which provides spatial positioning input for the target control of the environment control instruction generation module.

[0023] Preferably, the environment control instruction generation process comprises the following steps:

[0024] When the duration of the abnormal area exceeds the limit, use the variational mode decomposition technique to separate the low-frequency steady-state component and the high-frequency steady-state component of the abnormal area airflow, the low-frequency steady-state component represents the average motion of the airflow maintaining heat dissipation, and the high-frequency steady-state component represents the vortex pulsation and noise in the battery cluster gap.

[0025] A multi-objective optimization model is constructed with the goal of minimizing energy consumption and temperature fluctuations. The safety constraints of the multi-objective optimization model include the airflow velocity safety range, the airflow velocity safety range, and the upper limit of the fan speed.

[0026] The deflector angle and fan speed are solved by improving the particle swarm algorithm to obtain the Pareto optimal solution, and environmental control instructions are constructed based on the Pareto optimal solution; the multi-objective optimization model is used to balance energy consumption and heat dissipation efficiency, and to suppress the risk of thermal runaway caused by chaotic disturbance.

[0027] Preferably, when the closed-loop verification process is triggered, it includes:

[0028] Based on the proportional relationship between the current airflow field residual error and the historical residual error mean, the residual stability coefficient is calculated;

[0029] The historical residual baseline is calculated based on the average value of the historical residual data in the sliding time window, and the residual stability coefficient is dynamically generated by the ratio of the current residual error to the historical residual baseline. The residual stability coefficient is used to quantify the transient degree of the energy storage container environment deviating from the stable state;

[0030] According to the fluctuation amplitude of the residual stability coefficient, the confidence interval threshold and the strategy search direction of the meta-reinforcement learning model are dynamically adjusted;

[0031] When the residual stability coefficient exceeds the preset threshold, the confidence interval threshold is tightened and the exploration weight of the meta-reinforcement learning model for new strategies is increased; when the residual stability coefficient is lower than the preset threshold, the confidence interval threshold is relaxed and the verified strategy is preferentially called.

[0032] Preferably, the system further comprises:

[0033] The condensation risk disturbance module is used to identify the temperature and humidity data of potential condensation risk points in the energy storage container, and when the condensation risk score is detected to exceed the preset value, the sensor data is input into the condensation risk assessment model. Based on the boundary reference conditions generated by CFD simulation, combined with the local dew point proximity rate, airflow disturbance decay coefficient and temperature difference change rate, the condensation risk score of each condensation risk point is output, and the high-risk area identification is generated. The condensation disturbance instruction of the high-risk area is generated to enhance air flow and suppress condensation formation.

[0034] Preferably, the condensation risk disturbance module includes a temperature and humidity sensing unit, a condensation risk assessment model, a local behavior baseline construction unit and an anomaly detection model.

[0035] The temperature and humidity sensing unit is used to collect temperature and humidity data in different spatial positions in the energy storage container in real time.

[0036] The condensation risk assessment model refers to a function model that takes local temperature and humidity data as input, combines simulation boundary conditions, and outputs a condensation risk score, which is used to dynamically assess the possibility of condensation water formation in the local space.

[0037] The local behavior baseline construction unit is configured to construct a local environmental behavior baseline of each monitoring area in the energy storage container under a normal operation condition, and the local environmental behavior baseline serves as a contrast reference for risk scoring.

[0038] The abnormality detection model is configured to identify a condensation trigger mode (obtained based on a dew point approximation behavior and a humidity sudden increase feature) and assist in judging whether the condensation risk score is rapidly increased due to non-periodic fluctuations.

[0039] Preferably, the operation process of the condensation risk assessment model comprises the following steps:

[0040] The local dew point approximation rate, the airflow disturbance decay coefficient and the temperature difference change rate of each risk point are obtained, and are respectively mapped to a unified scoring scale through a normalization function; the local dew point approximation rate refers to the approximation degree between the real-time air temperature of a risk point and the current dew point temperature thereof, and is used to measure the trend of water vapor in the air approaching a saturated state; the airflow disturbance decay coefficient refers to the decay amplitude of the local airflow disturbance intensity per unit time, and is used to describe the natural weakening of the local air disturbance degree in a dead angle; and the temperature difference change rate refers to the change speed of the difference between the local air temperature and the temperature of an adjacent area per unit time, and is used to capture the temperature mutation trend of a micro environment.

[0041] The relative humidity is combined with the local air temperature to calculate the local absolute humidity by using an ideal gas state equation.

[0042] The weighted sum of the standardized local dew point approximation rate, the airflow disturbance decay coefficient, the temperature difference change rate and the local absolute humidity is obtained by using an experience or training derived weight parameter, so as to obtain the condensation risk score of each risk point.

[0043] Preferably, the execution of the condensation disturbance instruction process is limited by the upper limit of power consumption and the constraint of the main airflow structure, and the disturbance operation needs to dynamically compensate through a disturbance amplitude adjustment mechanism while locally enhancing the airflow disturbance, so as to control the overall main airflow field deviation within an acceptable range.

[0044] Technical effects and advantages of the present application:

[0045] (1) The energy storage container adaptive thermal management system provided by the present application realizes high-confidence dynamic modeling and accurate identification of abnormal areas of complex airflow and temperature field in the energy storage container by constructing a joint probability model integrating multi-source data, introducing Gaussian process regression and multi-precision CFD simulation fusion, combining chaotic feature extraction and abnormality identification mechanism, effectively solves the problems of slow response to transient disturbance, inaccurate local thermal runaway prediction and rough energy consumption control in the prior art, and improves the adaptive regulation and control ability and operation safety of the thermal management system under dynamic conditions.

[0046] (2) The energy storage container adaptive thermal management system provided by the application can realize real-time evaluation of local condensation risk, build a high-precision condensation early warning mechanism, and identify a condensation trigger mode based on edge computing, trigger a local disturbance control unit to implement micro-airflow excitation adjustment, and realize dynamic intervention on potential cold areas, by collecting energy storage container temperature and humidity data, combining CFD simulation and a multi-parameter coupling model, and introducing a reinforcement learning mechanism to optimize control strategy parameters, forming a data-driven closed-loop adaptive thermal management system, effectively solving the problem that the prior art cannot identify and intervene in local condensation water accumulation, improving the response speed and local adaptability of the thermal management system in complex environments, and ensuring the safety and reliability of the operation of the energy storage equipment. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 The energy storage container adaptive thermal management system structure diagram of the application is shown in FIG. 1.

[0048] Figure 2 The energy storage container adaptive thermal management system structure diagram of the application is shown in FIG. 1.

[0049] Figure 3 The condensation risk disturbance module structure diagram of the application is shown in FIG. 4. DETAILED DESCRIPTION

[0050] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. Although exemplary 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 limited by the embodiments described herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be accurately conveyed to those skilled in the art.

[0051] It should be understood that the size of each part shown in the drawings is not drawn in accordance with the actual proportional relationship for the convenience of description.

[0052] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way limiting to the application and use of the application.

[0053] Techniques, methods, and equipment known to those of ordinary skill in the relevant art can not be discussed in detail, but should be considered part of the specification where appropriate.

[0054] Embodiment 1, refer to Figure 1 The energy storage container adaptive thermal management system structure diagram of the application and Figure 2 The energy storage container adaptive thermal management system structure diagram of the application and Figure 1 The energy storage container adaptive thermal management system structure diagram of the application and

[0055] The data fusion and uncertainty quantification module synchronously collects multi-source data in the energy storage container when detecting the change of equipment layout or local temperature anomaly, the multi-source data at least including real-time airflow data, CFD simulation data and temperature distribution data; the multi-source data is fused through a Gaussian process regression algorithm, and the interpolation error of the sensor blind area is quantified by using a Bayesian method, and a fused flow field map with a confidence interval is output;

[0056] The chaotic intensity analysis module divides the fused flow field map into a plurality of grid cells, extracts multi-scale chaotic features using a time series convolution network for the grid cells whose confidence intervals meet preset requirements, and divides the chaotic intensity levels;

[0057] The abnormal area identification module generates a dynamic temperature threshold value in combination with the temperature gradient standard deviation of the grid cells, constructs an error thermal map of the energy storage container based on the fused flow field map and the dynamic temperature threshold value, and the error thermal parameter of each grid cell is the difference between the dynamic temperature threshold value and the predicted temperature in the fused flow field map; when the chaotic intensity level and the error thermal parameter jointly exceed the limit, it is marked as an abnormal area;

[0058] Explanations, the chaotic intensity level and the error thermal parameter jointly exceeding the limit means:

[0059] After normalizing the difference between the dynamic temperature threshold value and the predicted temperature in the fused flow field map, the chaotic intensity level is weighted and summed to generate an abnormal risk coefficient, and the abnormal risk coefficient exceeds the preset value.

[0060] The environmental control instruction generation module separates the real-time airflow data of the abnormal area into a low-frequency steady-state base flow component and a high-frequency chaotic disturbance component through variational mode decomposition when the duration of the abnormal area exceeds the limit, constructs a multi-objective optimization model with the lowest energy consumption and the smallest temperature fluctuation as the target, and adopts an improved particle swarm optimization algorithm to obtain the Pareto optimal solution of the deflection angle of the deflector and the rotating speed of the fan, and generates an environmental control instruction balancing energy consumption and heat dissipation efficiency;

[0061] The reliability verification module collects airflow field residuals in real time after executing the environmental control instruction, the airflow field residuals representing the difference between the actual measured airflow data and the CFD simulation data, and being used to determine whether the control instruction is effective, if not, re-verify, if the airflow field residuals exceed the preset threshold, trigger the early warning and closed-loop verification process; the closed-loop verification process is used to verify whether the execution of the environmental control instruction is abnormal, if the execution of the environmental control instruction is normal, enter the next module;

[0062] The meta-reinforcement learning optimization module builds a digital twin platform of the energy storage container, randomly generates multi-working-condition disturbance data (including temperature disturbance, load disturbance, and layout disturbance) in the digital twin platform, trains the meta-reinforcement learning model to quickly adapt to the multi-working-condition disturbance scene under the premise of meeting the safety constraint (to avoid the hysteresis of manual debugging), controls the search of the environmental control instruction within the safety constraint boundary, and outputs the meta-reinforcement learning model verified in safety for subsequent calling.

[0063] It needs to be further explained in the embodiments of the present application that the collection mode of the multi-source data is that a pressure sensor is arranged on the top of the energy storage container and in the gap between the battery clusters, real-time airflow data including airflow speed and airflow direction angle are acquired through the pressure sensor array, a CFD simulation model is called to output CFD simulation data, temperature distribution data are acquired through an infrared thermal imager to generate a two-dimensional temperature matrix, and the pressure sensor coordinates are mapped with the CFD simulation data to ensure the spatial alignment of the real-time airflow data and the CFD simulation data and ensure the spatial correspondence during data fusion.

[0064] In a possible embodiment, the CFD simulation model is a multi-precision CFD simulation model outputting CFD simulation data with low, medium and high precision; the hierarchical setting of the multi-precision CFD simulation model provides a cross-scale data basis for Gaussian process regression through precision-speed grading and physical feature complementarity, meets the real-time control requirement, improves the prediction reliability through fusion, and solves the conflict between dynamic layout and limited computing resources in the energy storage container scene; if the high-precision CFD simulation model takes too long to calculate (hour level) and cannot participate in online fusion in real time, the high-precision CFD simulation model is only used for offline calibration of weights, and only low / medium / high-precision CFD simulation models are used for online fusion; the low, medium and high-precision CFD simulation models are defined according to the following rules:

[0065] The low-precision model: the minimum grid size is 1 / 100 to 1 / 50 of the characteristic length L of the energy storage container, the RANS equation and the k-∈ turbulence model are used, and the single simulation time Tcalc≤10s;

[0066] The medium-precision model: the minimum grid size is L / 200 to L / 100, the LES or TRANS equation is used, and the single simulation time 10s<Tcalc≤300s;

[0067] The high-precision model: the minimum grid size is L / 500 to L / 200, the DNS or fully analytical LES is used, and the single simulation time Tcalc>300s.

[0068] It should be further explained in the embodiments of the present invention that the confidence interval is calculated as follows: using the covariance matrix output by Gaussian process regression, the prediction standard deviation of each grid cell is calculated (reflecting the credibility of the grid cell fusion result), based on the normal distribution assumption of the Gaussian process, with the mean ± 2 times the standard deviation as the confidence interval boundary, including the confidence interval of flow rate and temperature;

[0069] Explanation: By only processing areas with high confidence and ignoring areas with low confidence, error propagation can be avoided; if a certain CFD simulation model has systematic deviations in specific areas (such as high temperature areas), the contribution of high-error CFD simulation models can be suppressed by inverse weighting of the Bayesian posterior covariance matrix to avoid incorrect predictions affecting the fusion results.

[0070] What needs to be further explained in the embodiments of the present invention is that the real-time airflow data, temperature distribution data and CFD simulation data are fused through the Gaussian process regression algorithm to construct a joint probability model; the joint probability model is used to complementarily fuse the CFD simulation data with the real-time airflow data and temperature distribution data to generate a fused flow field map of the energy storage container (i.e., a prediction of the joint distribution of the flow field-temperature field); the space inside the energy storage container is divided into a number of grid units, and the fused flow field map contains the airflow velocity, airflow direction angle and predicted temperature of each grid unit; the kernel function of the joint probability model is defined and the hyperparameters are optimized, and the method of optimizing the hyperparameters is: setting the optimization goal to maximize the likelihood probability of the fused flow field map output by the joint probability model and the measured data; using the conjugate gradient method to iteratively adjust the kernel function parameters until the joint probability model meets the optimization goal; the gas flow velocity time series data and the temperature time series data

[0071] Explanation: The kernel function is a joint kernel function formed by the Matern kernel function and the linear kernel function; the Matern kernel function is used to capture the smooth change characteristics of airflow velocity in adjacent areas (for example, the closer the distance, the stronger the velocity correlation); the linear kernel function is used to characterize the relationship between the prediction deviation of CFD simulation models with different precision and position changes; the Matern kernel function is used to force the airflow field to conform to spatial continuity and avoid the anti-physical prediction of pure data-driven methods; the linear kernel function is used to make the advantages of multi-precision CFD simulation models complement each other, the low-fidelity CFD simulation model is used to maintain the global trend, and the high-fidelity CFD simulation model is used to extract local precision; the joint kernel function meets the time constraints of online control through analytical or fast approximate calculations.

[0072] It should be further explained in the embodiments of the present invention that the multi-scale chaos feature refers to the extraction of chaotic characteristic parameters of different time scales through a temporal convolutional network. The multi-scale chaos feature covers a time window from seconds to minutes, adapts to the battery charge and discharge cycle, and ensures the capture of transient eddy currents and periodic fluctuations, including:

[0073] Spectral entropy: based on singular value decomposition of flow velocity signal, the higher the entropy value, the stronger the energy dispersion of the flow field (the higher the chaos degree);

[0074] Recurrence plot density: generate recurrence matrix through phase space reconstruction, and statistically quantify the repeatability of the trajectory to determine the determinacy (the lower the density, the worse the determinacy);

[0075] Permutation entropy: used to quantify the randomness of time series, the higher the value, the stronger the chaos;

[0076] Recurrence quantification analysis-determinacy: the proportion of deterministic structure in the recurrence matrix, the lower the value, the stronger the chaos.

[0077] Further explained in the embodiments of the present application is that the chaos intensity analysis module comprises:

[0078] Based on the fusion flow field map, the grid cells meeting the requirements of the confidence interval are defined as high confidence areas; the gas flow velocity time series data and temperature time series data in the high confidence areas are extracted, and in a possible embodiment, the time surface length is set to 60 seconds and the sliding step is 10 seconds;

[0079] Explanation: The high confidence area is the core area of the air flow chaos detection, and by eliminating the low confidence area (such as the corner of the container or the blind area blocked by the equipment), the interference of the interpolation error on the subsequent analysis is avoided.

[0080] A time series convolution network is constructed, which includes 4 layers of dilated convolution layers with dilated factors of 1, 2, 4 and 8, a convolution kernel size of 3 and a channel number of 64; the gas flow velocity time series data in the high confidence area is input, and a multi-scale chaos feature vector is output; the dilated convolution structure of the time series convolution network covers the time scale from seconds to minutes, which is suitable for the battery charging and discharging period (typical period 30-60 seconds), ensures that the feature extraction result reflects the dynamic balance of battery heat generation and dissipation, and avoids transient noise interference;

[0081] The chaos feature vectors of all grid cells are input, and the chaos intensity level of each area is obtained.

[0082] Explanation: Calculate the temperature gradient amplitude of the high confidence area The corresponding 90% quantile Q is counted 90 (90% quantile means that 90% of the grid cells have a temperature gradient lower than this value, and 10% of the grid cells have a temperature gradient higher than this value), and the dynamic temperature threshold is defined as the sum of the average temperature of the grid area and 2 times Q 90 When the chaos intensity level of the grid cell exceeds the preset value and the predicted temperature exceeds the dynamic temperature threshold, it is marked as an abnormal grid.

[0083] Perform morphological closing operation (dilation, erosion operation) on the abnormal grid meeting the trigger condition to generate a connected region coordinate set to provide spatial positioning input for the target control of the environment control instruction generation module.

[0084] Further explained in the embodiments of the present application is that the environment control instruction generation process includes the following steps:

[0085] When the abnormal region duration exceeds the limit, use variational mode decomposition to separate the low-frequency steady-state component and the high-frequency steady-state component of the abnormal region airflow, the low-frequency steady-state component represents the average movement of the airflow maintaining heat dissipation, and the high-frequency steady-state component represents the vortex pulsation and noise of the battery cluster gap;

[0086] A multi-objective optimization model is constructed to minimize energy consumption and temperature fluctuation, and the safety constraints of the multi-objective optimization model include the airflow velocity safety range, the airflow velocity safety range, and the upper limit of the fan speed;

[0087] The Pareto optimal solution of the deflector angle and the fan speed is solved by improving the particle swarm algorithm, and the environment control instruction is constructed based on the Pareto optimal solution. The multi-objective optimization model is used to balance energy consumption and heat dissipation efficiency, and to suppress the risk of thermal runaway caused by chaotic disturbance;

[0088] In a possible embodiment, the high-power discharge is effective to minimize temperature fluctuation, and the low-discharge power is preferred to reduce total power consumption, and the heat dissipation demand of different working conditions is adapted.

[0089] The improved particle swarm algorithm searches for the Pareto optimal solution set by dynamically adjusting the inertia weight parameter, fusing the multi-objective optimization mechanism, and adopting the adaptive constraint processing strategy, and specifically includes:

[0090] The inertia weight is dynamically adjusted based on the number of iterations and the particle swarm distribution density. A larger weight value is maintained in the early stage to enhance the global search ability, and gradually reduced in the later stage to improve the local optimization accuracy;

[0091] Fast non-dominated sorting and congestion evaluation are used to stratify and screen the solution set, prevent local optimization by maintaining diversity, and introduce an elite reservation strategy to inject historical optimal solutions into the iteration process to accelerate convergence;

[0092] A double-objective optimization function is constructed with the fan energy consumption and the battery cluster temperature fluctuation amplitude as the core, and the energy proportion of the low-frequency steady-state component and the high-frequency transient component of the airflow signal extracted by variational mode decomposition is used to dynamically allocate target weight coefficients. When the low-frequency component is dominant, energy consumption optimization is emphasized, and when the high-frequency component is significant, temperature suppression demand is strengthened;

[0093] A penalty function mechanism is constructed for the airflow velocity safety range and the upper limit of the fan speed, and a constraint violation penalty term is added to the objective function to ensure that the solution set meets the physical limitations;

[0094] The Pareto frontier solution set is prioritized based on the normalized comprehensive decision index, and the solution with the highest balance between energy consumption and temperature fluctuation is selected as the global optimal control instruction of the deflector deflection angle and the fan rotating speed, so that the adaptive collaborative optimization of the heat dissipation performance and the energy efficiency is realized.

[0095] It needs to be further explained in the embodiments of the application that when the closed-loop verification process is triggered, it includes:

[0096] Based on the proportional relationship between the current air flow field residual error and the historical residual error average, a residual stability coefficient is calculated;

[0097] The historical residual baseline is calculated based on the average value of the historical residual data in the sliding time window, and the residual stability coefficient is dynamically generated through the ratio of the current residual error to the historical residual baseline. The residual stability coefficient is used to quantify the transient degree of the energy storage container environment deviating from the stable state;

[0098] According to the fluctuation amplitude of the residual stability coefficient, the confidence interval threshold and the strategy search direction of the meta-reinforcement learning model are dynamically adjusted;

[0099] When the residual stability coefficient exceeds the preset threshold, the confidence interval threshold is tightened and the exploration weight of the meta-reinforcement learning model for new strategies is increased; when the residual stability coefficient is lower than the preset threshold, the confidence interval threshold is relaxed and the verified strategy is preferentially called.

[0100] Background, the existing energy storage container is usually equipped with temperature control equipment, which adjusts the temperature and humidity inside the container through overall air flow to avoid the risk of performance degradation or thermal runaway of the battery caused by changes in environmental conditions; however, due to the complex internal structure of the container, especially in the local areas such as pipe bends, thermal insulation interfaces, and water collection nodes, air flow dead angles are easily formed, resulting in deviations between local temperature and humidity environment and overall control environment; small flow stagnation areas are extremely easy to trigger local condensation under certain conditions, forming condensate accumulation, which not only reduces the local heat exchange efficiency, but also may cause serious hidden dangers such as corrosion of heat conduction medium and degradation of electrical insulation performance, threatening the long-term stable operation of the energy storage system.

[0101] In view of the above phenomenon, the existing technology mainly implements global temperature and humidity regulation by arranging temperature control equipment at the top of the container, lacking the ability of fine-grained recognition and dynamic intervention of local micro-environment. When condensate accumulation occurs in the local area, the system cannot sense and respond in real time, resulting in continuous diffusion of local cold area effect, ultimately affecting the overall thermal management effect and increasing the system failure risk. In addition, the existing control strategy fails to effectively combine local climate characteristics and dynamic risk assessment, lacking an adaptive adjustment mechanism based on data driving, based on which the content of embodiment 2 is set.

[0102] Embodiment 2, refer toFigure 3 The condensation risk disturbance module structure block diagram of the application embodiment is different from that of embodiment 1 in that it further comprises:

[0103] The condensation risk disturbance module is used for identifying temperature and humidity data of potential condensation risk points in the energy storage container, and when detecting that the condensation risk score exceeds a preset value, inputting the sensing data into a condensation risk assessment model, outputting a condensation risk score of each condensation risk point based on boundary reference conditions generated by CFD simulation, combining a local dew point approach rate, an air flow disturbance decay coefficient and a temperature difference change rate, and generating a high-risk area identification, and generating a condensation disturbance instruction of the high-risk area to enhance air flow and inhibit condensation formation.

[0104] The temperature and humidity sensing unit is used for collecting temperature and humidity data of different spatial positions in the energy storage container in real time.

[0105] The condensation risk assessment model refers to a function model taking local temperature and humidity data as input, combining simulation boundary conditions, and outputting a condensation risk score, which is used for dynamically evaluating the possibility of forming condensate in the local space.

[0106] The local behavior baseline construction unit is used for constructing a local environmental behavior baseline of each monitoring area in the energy storage container under normal operating conditions, and serving as a comparison reference for the risk score.

[0107] The abnormality detection model is used for identifying the baseline deviation degree, identifying a condensation triggering mode (obtained based on dew point approach behavior and humidity sudden increase characteristics), and assisting in judging whether the condensation risk score is rapidly increased due to non-periodic fluctuations.

[0108] It is explained that the boundary reference condition refers to a preset operating boundary data set based on computational fluid dynamics simulation of internal air flow and heat transfer state of the energy storage container, which is used as a theoretical reference benchmark for local microclimate monitoring data analysis.

[0109] It is explained that the local environmental behavior baseline is used for constructing a temperature and humidity change baseline curve of each monitoring area in the energy storage container under normal operating conditions, and serving as a comparison reference for the condensation risk score; the construction of the local environmental behavior baseline relies on a multi-scale sliding window method to identify stable patterns and fluctuation boundaries in the time series, form a behavior mapping of the "normal" climate behavior of the area, and is used for subsequent condensation triggering condition judgment and abnormality detection.

[0110] In the application embodiment, the operation process of the condensation risk assessment model further needs to be explained.

[0111] The local dew point approach rate, airflow disturbance decay coefficient and temperature difference change rate of each risk point are acquired, and are respectively mapped to a unified scoring scale through a normalization function; the local dew point approach rate refers to the closeness between the real-time air temperature and the current dew point temperature of a risk point, and is used to measure the trend of water vapor in the air approaching the saturated state; the airflow disturbance decay coefficient refers to the decay amplitude of the local airflow disturbance intensity per unit time, and is used to describe the natural weakening of the local air disturbance degree at the dead angle; and the temperature difference change rate refers to the change speed of the difference between the local air temperature and the temperature of the adjacent region per unit time, and is used to capture the temperature mutation trend of the micro environment.

[0112] The relative humidity is combined with the local air temperature to calculate the local absolute humidity by using the ideal gas state equation.

[0113] The standardized local dew point approach rate, airflow disturbance decay coefficient, temperature difference change rate and local absolute humidity are weighted and summed by using the weight parameters obtained through experience or training to obtain the condensation risk score of each risk point.

[0114] It needs to be further explained in the embodiments of the present application that when the condensation risk score of the condensation risk point exceeds the preset requirement and the abnormality detection model identifies the condensation trigger mode, a condensation disturbance instruction is generated; the condensation disturbance instruction includes the wind speed, wind direction and disturbance duration parameters, and is used to control the local guide vane or air supply unit to implement the airflow disturbance operation, and the change trend of the risk score is tracked after the disturbance is executed, and the disturbance parameters are optimized through the reinforcement learning mechanism to improve the intervention effect and energy efficiency ratio.

[0115] It is explained that the execution of the condensation disturbance instruction process is limited by the upper limit of power consumption and the constraint of the main airflow structure, and the disturbance operation needs to dynamically compensate through the disturbance amplitude adjustment mechanism while locally enhancing the airflow disturbance, so that the overall main airflow field offset is controlled within an acceptable range.

[0116] It needs to be further explained in the embodiments of the present application that the environmental data of the corresponding region after the execution of the condensation disturbance instruction is collected, which is used to evaluate the disturbance response quality coefficient and update the condensation risk evaluation model; the change trend of the condensation risk score in the high-risk area is tracked, and if the condensation risk score decreases, it is marked as a valid strategy, and the current disturbance control parameters are recorded, otherwise it is invalid; if the strategy is invalid, the control parameter weight is iteratively updated through the reinforcement learning method, including the air supply direction, speed and action time, and the reinforcement learning process is based on the limited exploration mechanism, that is, under the safety constraint, the disturbance amplitude and time window are limited, and the control parameters are iteratively updated in the controllable range.

[0117] Summary: The embodiment of the present application collects energy storage container temperature and humidity data, combines CFD simulation and multi-parameter coupling model, evaluates local condensation risk in real time, constructs high-precision condensation early warning mechanism, identifies condensation trigger mode based on edge computing, triggers local disturbance control unit to implement micro air flow excitation adjustment, and realizes dynamic intervention on potential cold area; By introducing the reinforcement learning mechanism to optimize the control strategy parameters, a closed-loop adaptive thermal management system driven by data is formed, which effectively solves the problem of unable to identify and intervene local condensation water accumulation in the prior art, improves the response speed and local adaptability of the thermal management system in complex environment, and ensures the safety and reliability of the operation of the energy storage equipment.

[0118] Finally: The above only describes the preferred embodiments of the present application and is not used to limit the present application. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. The adaptive thermal management system of energy storage container is characterized by: include: The data fusion and uncertainty quantification module, when detecting changes in the equipment layout or local temperature anomalies in the energy storage container, synchronously collects multi-source data from the energy storage container. The multi-source data includes at least real-time airflow data, CFD simulation data, and temperature distribution data. It fuses the multi-source data using a Gaussian process regression algorithm and uses the Bayesian method to quantify the interpolation error in the sensor blind area, outputting a fused flow field map with confidence intervals. The chaos intensity analysis module divides the fused flow field map into several grid cells. For grid cells whose confidence intervals meet preset requirements, a temporal convolutional network is used to extract multi-scale chaos features and classify the chaos intensity levels. The abnormal area identification module generates a dynamic temperature threshold based on the standard deviation of the temperature gradient of the grid cells. Based on the fused flow field map and the dynamic temperature threshold, it constructs an error thermal map of the energy storage container. The error thermal parameter of each grid cell is the difference between the dynamic temperature threshold and the predicted temperature in the fused flow field map. When the chaos intensity level and the error thermal parameter jointly exceed the limit, the area is marked as abnormal. The environmental control instruction generation module, when the duration of the abnormal area exceeds the limit, separates the real-time airflow data of the abnormal area into a low-frequency steady-state base flow component and a high-frequency chaotic disturbance component through variational mode decomposition, constructs a multi-objective optimization model with the goals of minimizing energy consumption and temperature fluctuation, and uses an improved particle swarm algorithm to obtain the Pareto optimal solution of the guide plate deflection angle and the fan speed, generating environmental control instructions that balance energy consumption and heat dissipation efficiency.

2. The adaptive thermal management system for energy storage container according to claim 1, characterized in that: The multi-source data fusion method is: real-time airflow data, temperature distribution data and CFD simulation data are fused through Gaussian process regression algorithm to construct a joint probability model; The joint probability model is used to complementarily fuse CFD simulation data with real-time airflow data and temperature distribution data to generate a fused flow field map of the energy storage container; The space inside the energy storage container is divided into several grid units, and the fused flow field map contains the airflow velocity, airflow direction angle and predicted temperature of each grid unit; The kernel function of the joint probability model is defined and the hyperparameters are optimized. The optimization method is as follows: the optimization objective is set to maximize the likelihood probability of the output fusion flow field map of the joint probability model and the measured data; the conjugate gradient method is used to iteratively adjust the kernel function parameters until the joint probability model meets the optimization objective.

3. The adaptive thermal management system for energy storage container according to claim 2, characterized in that: When the chaos intensity level of a grid cell exceeds the preset value and the predicted temperature exceeds the dynamic temperature threshold, it is marked as an abnormal grid; Morphological closing operations are performed on abnormal grids that meet the trigger conditions to generate a set of connected region coordinates, providing spatial positioning input for the targeted control of the environmental control instruction generation module.

4. The adaptive thermal management system for energy storage container according to claim 1, characterized in that: The process of generating environmental control instructions includes the following steps: When the duration of the abnormal area exceeds the limit, the variational mode decomposition technology is used to separate the low-frequency steady-state base flow component and the high-frequency steady-state base flow component of the airflow in the abnormal area. The low-frequency steady-state base flow component represents the average motion of the airflow to maintain heat dissipation; the high-frequency steady-state base flow component represents the eddy current pulsation and noise in the gap between the battery clusters. A multi-objective optimization model with the goal of minimizing energy consumption and temperature fluctuation was constructed. The safety constraints of the multi-objective optimization model included the airflow velocity safety range, the airflow velocity safety range, and the upper limit of the fan speed; An improved particle swarm algorithm is used to solve the Pareto optimal solution of the guide plate deflection angle and the fan speed, and environmental control instructions are constructed based on the Pareto optimal solution. A multi-objective optimization model is used to balance energy consumption and heat dissipation efficiency, and to suppress the risk of thermal runaway caused by chaotic disturbances.

5. The adaptive thermal management system for energy storage container according to claim 1, characterized in that: Also includes: The reliability verification module collects the airflow field residual in real time after executing the environmental control command. If the airflow field residual exceeds the preset threshold, the closed-loop verification process is triggered; The meta-reinforcement learning optimization module builds a digital twin platform for energy storage containers, randomly generates multi-operating condition disturbance data in the digital twin platform, trains the meta-reinforcement learning model to quickly adapt to multi-operating condition disturbance scenarios while meeting safety constraints, controls the search for environmental control instructions within the safety constraint boundary, and outputs the meta-reinforcement learning model after safety verification for subsequent calls.

6. The adaptive thermal management system for energy storage container according to claim 5, characterized in that: When the closed-loop verification process is triggered, it includes: Based on the proportional relationship between the current airflow field residual and the historical residual mean, the residual stability coefficient is calculated; The historical residual baseline is calculated based on the average value of the historical residual data within the sliding time window. The residual stability coefficient is dynamically generated by the ratio of the current residual to the historical residual baseline. The residual stability coefficient is used to quantify the transient degree of the energy storage container environment deviating from the stable state. Dynamically adjust the confidence interval threshold and the strategy search direction of the meta-reinforcement learning model according to the fluctuation range of the residual stability coefficient; When the residual stability coefficient exceeds the preset threshold, the confidence interval threshold is tightened and the exploration weight of the meta-reinforcement learning model for the new strategy is increased; When the residual stability coefficient is lower than the preset threshold, the confidence interval threshold is relaxed and the verified strategy is called first.

7. The adaptive thermal management system for energy storage containers according to any one of claims 1 or 6, characterized in that: The system further comprises: The condensation risk disturbance module is used to identify the temperature and humidity data of potential condensation risk points in the energy storage container. When it detects that the condensation risk score exceeds the preset value, the sensor data is input into the condensation risk assessment model. Based on the boundary reference conditions generated by CFD simulation, combined with the local dew point approach rate, airflow disturbance attenuation coefficient and temperature difference change rate, the condensation risk score of each condensation risk point is output, and a high-risk area identification is generated. The condensation disturbance instructions for the high-risk area are generated to enhance air flow and inhibit condensation formation.

8. The adaptive thermal management system for energy storage container according to claim 7, characterized in that: The condensation risk disturbance module includes a temperature and humidity sensing unit, a condensation risk assessment model, a local behavior baseline construction unit and an anomaly detection model; The temperature and humidity sensing unit is used to collect temperature and humidity data at different spatial locations in the energy storage container in real time; The condensation risk assessment model is a function model that takes local temperature and humidity data as input, combines simulation boundary conditions, and outputs a condensation risk score, which is used to dynamically assess the possibility of condensation water forming in a local space; The local behavior baseline construction unit is used to construct the local environmental behavior baseline of each monitoring area inside the energy storage container under normal operating conditions, and serve as a comparison reference for the risk score; The anomaly detection model is used to identify the degree of baseline deviation and the condensation triggering mode, and to assist in determining whether the condensation risk score increases rapidly due to non-periodic fluctuations.

9. The adaptive thermal management system for energy storage container according to claim 8, characterized in that: The operation process of the condensation risk assessment model includes: The local dew point approach rate, airflow disturbance attenuation coefficient, and temperature difference change rate are obtained for each risk point and mapped to a unified scoring scale through a normalized function. The local dew point approach rate refers to the degree of proximity between the real-time air temperature at a risk point and its current dew point temperature, and is used to measure the tendency of water vapor in the air to approach saturation. The airflow disturbance attenuation coefficient refers to the attenuation amplitude of the local airflow disturbance intensity per unit time, and is used to describe the natural weakening of the local air disturbance degree in the blind spots of the space. The temperature difference change rate refers to the rate of change of the difference between the local air temperature and the temperature of the adjacent area per unit time, and is used to capture the trend of temperature mutation in the microenvironment. The local absolute humidity is calculated using the ideal gas state equation by combining relative humidity with the local air temperature; Using weight parameters derived from experience or training, the standardized local dew point approach rate, airflow disturbance attenuation coefficient, temperature difference change rate and local absolute humidity are weighted and summed to obtain the condensation risk score for each risk point.

10. The adaptive thermal management system for energy storage container according to claim 9, characterized in that: The process of executing the condensation disturbance instruction is limited by the power consumption upper limit and the main airflow structure constraints. The disturbance operation needs to enhance the airflow disturbance locally while dynamically compensating through the disturbance amplitude adjustment mechanism to control the overall main airflow field offset within an acceptable range.

Images