Battery cluster equalization charging management system based on time series prediction

Abstract

The application relates to the technical field of battery swap cabinet charging management, and discloses a battery swap cabinet battery cluster equalization charging management system based on time sequence prediction, which comprises a data evaluation module, a thermal recognition module and a polarization constraint module. The data evaluation module reads state of charge, terminal voltage and surface temperature data, receives an expected swap-out timestamp, and calculates a battery swap urgency index. The thermal recognition module extracts effective temperature rise data and updates a spatial thermal coupling impedance matrix. The polarization constraint module extracts cross-polarization relaxation and generates a charging phase-locked constraint flag before a stable slope threshold is reached. The decision execution module collects the urgency index, the impedance matrix and the flag, solves a mixed integer quadratic programming model, and outputs an optimal discrete control vector to drive a direct current power conversion device. The application incorporates physical and electrochemical constraints into global mathematical optimization, and realizes the suppression of local temperature rise and polarization damage of the battery cluster under the premise of meeting the multi-position concurrent battery swap demand.

Description

Technical Field

[0001] This invention relates to the field of battery swapping cabinet charging management technology, specifically to a battery swapping cabinet battery cluster equalization charging management system based on time-series prediction. Background Technology

[0002] As an energy replenishment device for micro electric vehicles, the battery swapping cabinet has multiple physical compartments inside. It replenishes the batteries in the compartments through a power conversion module. When performing cluster charging tasks, the existing battery swapping cabinets generally allocate charging power based on the current state of charge of a single battery and the business priority set by the system, so as to ensure that low-charge batteries can be given priority for energy replenishment.

[0003] Existing charging management and control algorithms have technical limitations when handling concurrent charging services in multiple compartments. The control system treats each battery as an independent control node and does not consider the spatial heat transfer superposition effect between adjacent compartments in conjunction with the physical structure. When multiple compartments perform high-power charging at the same time, severe heat accumulation will occur inside the cabinet. Furthermore, the existing system does not take into account the electrochemical polarization state inside the battery when outputting the charging current for mutual exclusion control. Some batteries are continuously subjected to current excitation before they have completed static depolarization, causing polarization damage inside the battery cells.

[0004] Meanwhile, under conditions of excessively high ambient temperature, existing algorithms may fail due to the conflict between temperature constraints and business requirements, and the system lacks fault tolerance and degradation control mechanisms for the underlying computational logic. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a time-series prediction-based battery cluster equalization charging management system for battery swapping cabinets. This system solves the problem that existing battery swapping cabinets, when charging concurrently, do not consider spatial thermal coupling effects and electrochemical polarization states, leading to localized heat accumulation and battery polarization damage.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a battery cluster equalization charging management system for battery swapping cabinets based on time-series prediction, the system comprising: The data evaluation module is used to read state of charge data, terminal voltage data and surface temperature data, and receive the expected swap timestamp to calculate the swap urgency index. The thermal identification module is used to receive the surface temperature data, extract the effective temperature rise data, and update the spatial thermal coupling impedance matrix. The polarization constraint module is used to receive the spatial thermal coupling impedance matrix and the terminal voltage data, extract the cross-polarization relaxation, and generate a charging phase-locked constraint flag before reaching the steady slope threshold defined by the static characteristics. The decision execution module is used to combine the battery swapping urgency index, the spatial thermal coupling impedance matrix, and the charging phase-locked constraint flag to solve a mixed integer quadratic programming model, and output the optimal discrete control vector to drive the DC power conversion device.

[0007] The system establishes a battery swapping urgency and spatial thermal coupling impedance model, and combines the cross-polarization relaxation state to incorporate physical and electrochemical constraints into a mixed integer quadratic programming model for global optimization. This allows the system to suppress battery cluster temperature rise and polarization damage while ensuring the continuity of battery swapping services.

[0008] Furthermore, the data evaluation module subtracts the current actual state of charge data from the target full charge state threshold defined by the charge and discharge safety interval specification to obtain the power gap value, and uses the difference between the expected swap timestamp and the system real time as the physical remaining time window. The data evaluation module performs a division operation on the power gap value and the physical remaining time window to obtain the battery swap urgency index.

[0009] By quantifying the power shortage that needs to be replenished within a unit time window, a mapping relationship between time and charging demand is constructed, providing numerical input for business priorities to the decision-making model.

[0010] Furthermore, when the data evaluation module determines that it has failed to receive the expected replacement time series sent from the cloud, resulting in the lack of the expected replacement timestamp, it calls the locally pre-stored fixed reference duration and adds it to the system real-time time to obtain the expected replacement timestamp. When the system real-time time is greater than or equal to the expected replacement timestamp, resulting in the physical remaining time window being less than or equal to zero, it executes a jump instruction to intercept the division operation logic and assigns the battery replacement urgency index to the maximum floating-point constant peak point supported by the hardware.

[0011] The configuration is used to handle communication interruptions or timestamp inversion anomalies. Through extreme value substitution and logical interception, it prevents the microprocessor from generating division-by-zero overflow errors and maintains the continuity of system operation.

[0012] Furthermore, the thermal identification module sets the pulse charging command issued by the decision execution module as the trigger condition, records the reference temperature at the start time of the physical compartment charging pulse, starts an asynchronous thermal observation time window containing a physical hysteresis constant, continuously traverses the input data stream within the asynchronous thermal observation time window to extract the peak data of the surface temperature, and subtracts the peak data from the reference temperature to obtain the effective temperature rise data.

[0013] By taking advantage of the physical hysteresis characteristics of heat conduction, the system uses an asynchronous time window to capture the peak value of the thermal response, eliminates background interference from ambient temperature fluctuations, and obtains effective temperature rise data that reflects the internal heating state.

[0014] Furthermore, the thermal identification module extracts the charging pulse excitation power within the asynchronous thermal observation time window as input vector data, extracts the corresponding effective temperature rise data as tag output data, accumulates the input vector data and the tag output data in the system memory to form a historical observation dataset with time series correlation, and continuously substitutes the historical observation dataset into the gain matrix and covariance matrix for iterative feedback calculation to obtain new spatial thermal coupling impedance matrix elements.

[0015] The configuration utilizes closed-loop feedback between input excitation and output response for online parameter identification, enabling the impedance matrix to dynamically track the shift in heat transfer characteristics caused by environmental changes or equipment aging.

[0016] Furthermore, the polarization constraint module traverses each element in the spatial thermal coupling impedance matrix and compares the extracted element values ​​with the strong coupling physical threshold set in conjunction with the battery swapping cabinet hardware structure. When the element value between any two physical compartments is greater than or equal to the strong coupling physical threshold, it is determined that there is a strong coupling relationship between the two physical compartments. For dormant batteries that are determined to have a strong coupling relationship and are in the charging intermittent period, the terminal voltage data is continuously read and stored in the first-in-first-out storage queue in chronological order to construct a sliding sampling window.

[0017] By quantifying the heat transfer resistance between adjacent physical compartments, nodes with heat transfer risks are identified, and the status of associated dormant cells is monitored, providing basic data for polarization analysis.

[0018] Furthermore, after determining that the sliding sampling window is filled with the terminal voltage data, the polarization constraint module reads the discrete-time sampling points and the corresponding terminal voltage data as the basic original data nodes in the two-dimensional coordinate system, calls the univariate linear regression algorithm to extract the slope value of the fitted line as the cross-polarization relaxation, and sets the charging phase-locked constraint flag to a logic high level when the absolute value of the cross-polarization relaxation is greater than or equal to the stationary slope threshold.

[0019] The rate of elimination of electrochemical polarization is characterized by the voltage drop slope. By setting phase-locked constraints, the recharging of batteries that have not been completely depolarized is restricted, thus preventing battery damage caused by continuous charging and discharging.

[0020] Furthermore, the decision execution module maps the spatial thermal coupling impedance matrix to the basic thermal penalty term parameter of the mixed integer quadratic programming model, which penalizes the superposition effect of heat generation. The objective function of the mixed integer quadratic programming model is composed of the subtraction of the urgency term representing the priority of business needs and the spatial thermal coupling quadratic penalty term representing the risk of physical temperature rise. The internal microprocessor calls a pre-set mathematical analysis library to establish a mathematical optimization space for discrete control variables.

[0021] The multi-objective management problem is transformed into a mathematical analytical model. The charging progress is guaranteed by the first term, and the heat accumulation caused by the concurrent high-power charging of adjacent warehouses is suppressed by the second term.

[0022] Furthermore, the decision execution module forcibly constrains the total charging power of each operating physical compartment within the maximum available power supply capacity of the distribution transformer. Based on the node safety temperature derating upper limit threshold read by the battery management system and the charging phase-locked loop constraint flag, it establishes temperature prediction constraints and polarization mutual exclusion constraints and pushes them into the solution domain of the mixed integer quadratic programming model. When the charging phase-locked loop constraint flag is at a logic high level, it forcibly restricts the discrete control variable corresponding to the physical compartment to take the value of a digital zero.

[0023] Physical and electrochemical boundary conditions are applied in the solution domain to ensure that the output control strategy meets the global grid capacity limit while taking into account the temperature and state safety of individual battery cells.

[0024] Furthermore, the decision execution module calls the branch and bound algorithm to output the optimal discrete control vector within a limited computation time. When the solver returns an algorithm dead zone problem where no feasible solution is found, the module calls the safety rollback instruction written in the program to force all the discrete control variables to be assigned to zero and execute a global standby cooling action. The optimal discrete control vector is then parsed and converted into a communication control message to control the DC power conversion device.

[0025] By introducing dead zone handling and safety rollback mechanisms, the system can be degraded to standby mode under boundary conditions where the mathematical model has no feasible solution, thus avoiding the risk of equipment damage in high-temperature environments.

[0026] This invention provides a time-series prediction-based battery cluster equalization charging management system for battery swapping cabinets. It offers the following advantages: 1. This invention calculates the division result of the power shortage and the physical remaining time window through the data evaluation module, objectively quantifying the urgency of battery replacement for each physical compartment. In response to abnormal situations such as communication interruption leading to missing expected replacement timestamps or system time inversion, the system introduces extreme value substitution and jump instruction interception logic to prevent division-by-zero overflow errors in the underlying microprocessor and ensure the continuity of operation of the charging management system under abnormal communication conditions.

[0027] 2. This invention extracts effective temperature rise data through a thermal identification module to update the spatial thermal coupling impedance matrix, and extracts cross-polarization relaxation through a polarization constraint module to generate a charging phase-locked constraint flag. The control strategy quantifies the superimposed heat transfer effect between adjacent physical compartments and uses the voltage drop slope to limit the recharging action of batteries that have not completed depolarization. While scheduling concurrent charging services in multiple compartments, it suppresses local heat accumulation and electrochemical polarization damage within the battery cluster.

[0028] 3. This invention introduces the urgency index, thermal coupling impedance, and phase-locked loop constraint flag into the mixed integer quadratic programming model through the decision execution module. Under the boundary conditions of total power supply and node safety temperature, it outputs discrete control vectors. At the same time, the system is configured with a safety backoff command. When the operating conditions cause the solver to enter a dead zone state with no feasible solution, it forces the global standby cooling to avoid the risk of equipment thermal runaway caused by the system continuing to run in the state of algorithm failure. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the system timing interaction of the present invention; Figure 2 This is a schematic diagram of the method flow of the present invention; Figure 3 The following is a timing diagram of the thermally coupled asynchronous identification logic of the present invention, wherein Figure (a) is a timing diagram of the trigger signal of the control execution end, and Figure (b) is a timing diagram of the effective temperature rise data extraction of the state observation end; Figure 4 This is a schematic diagram of the polarization state monitoring and phase-locked constraint process of the present invention; Figure 5 Figure 1 is a timing diagram of the control decision and execution interaction of the present invention. Figure 2 is a timing diagram of the charging phase-locked constraint flag output by the polarization constraint module, Figure 3 is a timing diagram of the optimal discrete control vector output by the decision execution module, and Figure 4 is a schematic diagram of the external physical constraints and the superposition of heat generation penalty verification. Figure 6 The system equalization scheduling timing evolution diagram of the present invention is shown in Figure (a), which is the cross-polarization relaxation state monitoring diagram of physical warehouse B, Figure (b) is the logic state diagram of the charging phase-locked constraint flag bit for physical warehouse A, and Figure (c) is the timing diagram of the output of discrete control variables corresponding to physical warehouse A. Detailed Implementation

[0030] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] See attached document Figure 1 This invention provides a battery cluster equalization charging management system for battery swapping cabinets based on time-series prediction. The system includes: a data evaluation module, a thermal identification module, a polarization constraint module, and a decision execution module.

[0032] The data evaluation module is used to periodically read the state of charge data, terminal voltage data, and surface temperature data of the batteries in the physical compartments inside the battery swapping cabinet. It receives the expected swapping timestamp issued by the scheduling system, calculates the battery swapping urgency index of the physical compartment batteries within the current control time window, and transmits the terminal voltage data to the polarization constraint module. At the same time, it transmits the surface temperature data to the thermal identification module and the battery swapping urgency index to the decision execution module, serving as the data prerequisite for subsequent state identification and scheduling decisions.

[0033] The thermal identification module receives surface temperature data transmitted by the data evaluation module and uses the pulse charging command issued by the decision execution module as the trigger condition to establish a data feedback flow from the control execution end to the state observation end. It records the reference temperature at the start of the charging pulse of the physical compartment, starts an asynchronous thermal observation time window containing the physical hysteresis constant to extract the effective temperature rise data of adjacent physical compartments, and uses the recursive least squares method to update the spatial thermal coupling impedance matrix that characterizes the internal spatial transmission characteristics in real time. The updated spatial thermal coupling impedance matrix is ​​distributed to the polarization constraint module as the basic input for thermal coupling relationship determination, and transmitted to the decision execution module as the thermal penalty term parameter in the decision model.

[0034] The polarization constraint module receives the spatial thermal coupling impedance matrix output by the thermal identification module and the terminal voltage data output by the data evaluation module. Based on the spatial thermal coupling impedance matrix, it determines the thermal coupling relationship between adjacent physical compartments. When a strong coupling relationship is determined, it collects the discrete time series of terminal voltage of the battery during the charging interval and extracts the cross-polarization relaxation, which represents the electrochemical state, using a univariate linear regression algorithm. Before the cross-polarization relaxation reaches the stable slope threshold, it generates a charging phase-locked constraint flag for the coupled physical compartment. When the cross-polarization relaxation reaches the stable slope threshold, it automatically clears the charging phase-locked constraint flag. Then, the generated or cleared charging phase-locked constraint flag is input to the decision execution module in real time to form a dynamic safety scheduling boundary.

[0035] The decision execution module is used to collect the battery swapping urgency index output by the data evaluation module, the spatial thermal coupling impedance matrix output by the thermal identification module as a thermal penalty term parameter, and the charging phase-locked constraint flag output by the polarization constraint module. It constructs a mixed integer quadratic programming model with the goal of maximizing the battery swapping urgency index and minimizing the superposition of heat generation. It substitutes the grid capacity limit, the node safety temperature derating upper limit threshold, and the charging phase-locked constraint flag to solve the problem. The output optimal discrete control vector is converted into a communication control message, which drives the DC power conversion equipment corresponding to each physical compartment to perform pulse charging or command shutdown. When pulse charging is performed, a pulse charging command is generated and sent to the thermal identification module. The output state of the DC power conversion equipment changes, which in turn causes changes in the physical environment parameters. These changes are read by the data evaluation module in the next cycle, thus forming a complete closed data processing and physical control chain.

[0036] See attached document Figure 2 This invention provides a time-series prediction-based method for balanced charging management of battery clusters in battery swapping cabinets. The method includes the following steps: S1, the data evaluation module collects the sensor data of the batteries in the battery swapping cabinet according to the preset basic clock cycle, extracts the expected swapping timestamp of each battery to calculate the swapping urgency index of the uncharged batteries, then transmits the terminal voltage data to the polarization constraint module, transmits the surface temperature data to the thermal identification module, and transmits the swapping urgency index to the decision execution module to complete the distribution and mapping of the underlying parameters. S2, the thermal identification module extracts the temperature response of each physical compartment within the asynchronous thermal observation time window based on surface temperature data and combined with the pulse charging command issued by the decision execution module. It uses the observed effective temperature rise data to smoothly update the transfer coefficient in the spatial thermal coupling impedance matrix online, and outputs the updated spatial thermal coupling impedance matrix to the polarization constraint module and the decision execution module respectively. S3, the polarization constraint module traverses the updated spatial thermal coupling impedance matrix to determine the physical compartment pairs with thermal coupling constraints. Combined with the end sliding sampling window, it calculates the slope of the fitted line as the cross-polarization relaxation. Combined with the stable slope threshold, it generates a mandatory charging phase-locked constraint flag or performs a flag reset action. The current state charging phase-locked constraint flag is then input to the decision execution module in real time as a constraint condition. S4, at the beginning of each control time window, the decision execution module integrates the battery swapping urgency index, the spatial thermal coupling impedance matrix as a thermal penalty term parameter, and the charging phase-locked loop constraint flag to construct and solve a mixed integer quadratic programming model. It then sends communication control messages through the bus to control the DC power conversion equipment to perform time-division multiplexing scheduling actions in an interleaved manner. The pulse charging command generated during the scheduling process is fed back to the thermal identification module as a trigger condition. The data evaluation module captures the environmental and electrical feedback generated by the scheduling actions through periodic reading actions, thus completing the closed-loop iteration of the method process on the time axis.

[0037] See attached document Figure 3 The data evaluation module establishes a communication connection with the battery management system corresponding to each physical compartment in the battery swapping cabinet through the internal control LAN bus. Within the set basic clock cycle, it initiates a polling communication command to each physical compartment to read the corresponding state of charge data, terminal voltage data, and surface temperature data. For the cell voltage analog-to-digital conversion process and the resistance-temperature conversion mechanism of the temperature sensor, hardware selection and underlying program development can be carried out according to general battery management technical specifications. After the data evaluation module collects the state data, it performs a structured binding operation between the state data and the corresponding physical compartment number in the system memory.

[0038] The edge computing gateway built into the battery swapping cabinet receives the expected swapping time series generated by the cloud scheduling system based on user reservation information and the site operation model constructed using existing time series prediction algorithms through a wireless communication link. It then sends the expected swapping time series to the local network of the battery swapping cabinet. After receiving the expected swapping time series, the data evaluation module parses and extracts the expected swapping timestamp corresponding to each battery in the expected swapping time series. By reading the unique identification code inside the battery hardware and comparing it with the current physical location status, the expected swapping timestamp is mapped to the specific physical location to complete the initial mapping and association of the underlying data parameters.

[0039] The data evaluation module calculates the battery swapping urgency index of each uncharged battery in the battery swapping cabinet within the current control time window based on existing general strategies. In this embodiment, the control time window ranges from 1 minute to 5 minutes. The data evaluation module subtracts the pre-configured target full charge state threshold from the current actual charge state data to obtain the power gap value and uses the difference between the expected swapping timestamp and the system real-time time as the physical remaining time window. The data evaluation module divides the power gap value by the physical remaining time window to obtain the battery swapping urgency index. In combination with the charging and discharging safety interval specification, the target full charge state threshold ranges from 90% to 100% in this embodiment.

[0040] In response to data anomalies, the data evaluation module incorporates a numerical error correction branch instruction. When cloud communication is interrupted, resulting in the failure to receive the expected swap time series, the data evaluation module adds the locally pre-stored fixed reference duration to the system real-time time to obtain the expected swap timestamp. In this embodiment, the value range of the pre-stored fixed reference duration is 3600s to 7200s. Before performing the division operation, the data evaluation module performs a size judgment logic calculation on the difference between the expected swap timestamp and the system real-time time. When the system real-time time is greater than or equal to the expected swap timestamp, causing the physical remaining time window to be less than or equal to zero, the data evaluation module executes a jump instruction to intercept the division operation logic and assigns the battery swap urgency index to the maximum floating-point constant peak point supported by the hardware to complete the logic closure action of the algorithm dead zone.

[0041] After a series of steps including reading and quantifying underlying parameters and handling abnormal branches, the data evaluation module packages and prepares the calculated indicators in the system memory space for distribution. The data evaluation module then uses the built-in data distribution bus to synchronously send the extracted and parsed terminal voltage data and surface temperature data to the polarization constraint module node and the thermal identification module node according to the physical location number. At the same time, the data evaluation module transmits the battery swapping urgency index output by the general calculation logic to the decision execution module port through memory sharing to complete the input deployment of initial physical parameters and priority weight parameters for multi-node physical model calculation and constraint relationship verification.

[0042] See attached document Figure 3 The thermal identification module receives surface temperature data periodically transmitted by the data evaluation module and maintains a background monitoring state. The thermal identification module uses the pulse charging command issued by the decision execution module as the trigger condition to establish a data flow from the control execution end to the state observation end. The heat transfer in the physical space is limited by the thermal resistance of the air medium and the thermal capacity of the battery swapping cabinet structure, which will produce pure hysteresis physical characteristics. The thermal identification module sets a physical hysteresis constant based on the physical distance inside the battery swapping cabinet and the ambient wind speed, and starts an asynchronous thermal observation time window independent of the electrical control cycle accordingly.

[0043] The thermal identification module quantifies the actual heat conduction capacity between hardware structures based on the physical phenomenon of temperature rise between adjacent nodes caused by the charging action. In this embodiment, the physical lag constant ranges from 30s to 120s, taking into account the structural size and passive heat dissipation characteristics. At the start of the pulse charging command, the thermal identification module records the real-time surface temperature of the adjacent observation physical compartment as the reference temperature. Within the asynchronous thermal observation time window defined by the physical lag constant, the thermal identification module continuously traverses the input data stream and extracts the peak data of the surface temperature. The thermal identification module subtracts the peak data from the reference temperature to obtain the effective temperature rise data. When it is determined that the peak data is lower than the reference temperature, causing the effective temperature rise data to be negative, the thermal identification module forcibly clears the effective temperature rise data to zero.

[0044] The thermal identification module uses a recursive least squares method with a forgetting factor to update the spatial thermal coupling impedance matrix, which characterizes the internal spatial transmission features, in real time based on the effective temperature rise data. The elements in the spatial thermal coupling impedance matrix represent the thermal impedance coefficients between different physical compartments, and these coefficients dynamically drift with changes in ambient temperature and cooling fan speed. The recursive least squares method with a forgetting factor introduces attenuation weights to reduce the weight of historical data in the current parameter estimation. The thermal identification module uses a parameter update algorithm to fit and correct the latest calculated effective temperature rise data with the corresponding charging pulse excitation power in the issued command to obtain new spatial thermal coupling impedance matrix elements. The specific mathematical model for the recursive update of matrix elements is defined as follows: ; ; ; In the formula, The system gain matrix; The discrete system time step index is used to perform a recursive update algorithm for the heat identification module; This represents the historical covariance matrix of the previous period; This represents the parameter vector currently calculated to characterize the thermal impedance properties of the observed physical compartment; Represents the parameter vector of the previous period; This represents the scalar value of the effective temperature rise data extracted from the observed physical compartment; This represents an input vector containing the excitation power of the charging pulse corresponding to each physical compartment; For the input vector The transpose vector after performing a matrix transpose operation; Represents the system gain matrix; This represents the covariance matrix for the current calculation period; The forgetting factor is used in this embodiment, and its value ranges from 0.95 to 0.99. Before executing the recursive update algorithm, the thermal identification module needs to dynamically construct an online training dataset for system identification. The thermal identification module extracts the charging pulse excitation power within the asynchronous thermal observation time window as input vector data. Simultaneously, the corresponding effective temperature rise data is extracted as the label output data. As the control time window progresses, the input vector data and the label output data are continuously accumulated in the system memory to form a historical observation dataset with time series correlation. The thermal identification module continuously substitutes the historical observation dataset into the gain matrix and covariance matrix for iterative feedback calculation, so that the parameter vector characterizing the thermal impedance characteristics of the physical compartment continuously converges adaptively with the input of real observation data, thereby completing the online real-time parameter training action of recursively updating the mathematical model of matrix elements. After completing the parameter update calculation, the thermal identification module distributes the spatial thermal coupling impedance matrix to the polarization constraint module as input and synchronously transmits it to the decision execution module as the thermal penalty term parameter.

[0045] See attached document Figure 4 The polarization constraint module receives the spatial thermal coupling impedance matrix output by the thermal identification module and the terminal voltage data synchronously transmitted by the data evaluation module. The polarization constraint module traverses each element in the spatial thermal coupling impedance matrix and compares the extracted element values ​​with the preset strong coupling physical threshold. In this embodiment, combined with the battery swapping cabinet structure, the strong coupling physical threshold ranges from 0.5℃ / kW to 1.5℃ / kW. When the element value between any two physical compartments is greater than or equal to the strong coupling physical threshold, the polarization constraint module determines that there is a strong coupling relationship between the two physical compartments and activates the construction process of the terminal voltage discrete time series sliding sampling window.

[0046] For dormant batteries that are determined to have a strong coupling relationship and are currently in a charging intermittent period, the polarization constraint module continuously reads the terminal voltage data and stores it in the first-in-first-out storage queue according to the time sequence to construct a sliding sampling window. Before the sliding sampling window is filled with data, the polarization constraint module remains in a waiting state and outputs a safe default value to the decision execution module to prevent empty address read errors. After the battery disconnects the charging current, the internal chemical reaction does not stop immediately, causing the terminal voltage to undergo a slow relaxation process. When the sliding sampling window is filled with terminal voltage data, the polarization constraint module calls the existing univariate linear regression algorithm to perform least squares fitting on the discrete time series of terminal voltage within the window to remove high-frequency electrical white noise and calculates the slope of the fitted line as the cross-polarization relaxation.

[0047] The polarization constraint module reads discrete-time sampling points arranged chronologically within a sliding sampling window and their corresponding terminal voltage data as the basic raw data nodes in a two-dimensional coordinate system. The system's underlying microcontroller calls a pre-defined univariate linear regression algorithm to continuously execute iterative calculation logic on the basic raw data nodes in the two-dimensional coordinate system. The slope value of the fitted line generated by the polarization constraint module is directly used as the cross-polarization relaxation, which characterizes the intensity of the internal reaction. The specific cross-polarization relaxation calculation model is defined as follows: ; In the formula, Cross-polarization relaxation is used to characterize the stability of the internal electrochemical state and is forcibly set to the maximum floating-point value supported by the hardware when the result of the denominator is zero. The total number of discrete-time sampling points N within the sliding sampling window is 10 to 50 in this embodiment. For the first The timestamp value of the sampling point; For the first Terminal voltage data corresponding to each sampling point; Starting from the number 1 to the value The index of the last consecutive positive integers.

[0048] The polarization constraint module compares the calculated absolute value of cross-polarization relaxation with the system's preset steady slope threshold to determine whether the micro-electrochemical polarization elimination process is stabilizing. Based on the electrochemical relaxation characteristics of lithium batteries, the steady slope threshold in this embodiment ranges from 1mV / s to 5mV / s. Before the absolute value of cross-polarization relaxation is greater than or equal to the steady slope threshold, the polarization constraint module generates a charging phase-locked constraint flag for the coupled physical compartment and sets the charging phase-locked constraint flag to a logic high level. When the absolute value of the cross-polarization relaxation is less than the stability slope threshold, the micro-electrochemical polarization elimination is completed. The polarization constraint module automatically clears the charging phase-locked constraint flag for the coupled physical compartment and resets the charging phase-locked constraint flag to a logic low level. The polarization constraint module uses the generated or reset charging phase-locked constraint flag as a discrete constraint condition to input to the decision execution module in real time, thus completing the dynamic construction and release of the cross-node phase-locked safety boundary.

[0049] See attached document Figure 5 At the beginning of each control time window, the decision execution module reads data from the shared memory area of ​​the control local area network bus. The decision execution module collects the battery swapping urgency index periodically output by the data extraction evaluation module and the spatial thermal coupling impedance matrix updated online by the thermal identification module. At the same time, the decision execution module receives the charging phase-locked constraint flag bit transmitted by the polarization constraint module through the hardware interrupt pin. The decision execution module directly maps the received spatial thermal coupling impedance matrix to the basic thermal penalty term parameters of the penalty and heating superposition effect in the mixed integer quadratic programming model.

[0050] The battery swapping station scheduling system needs to find a balance between prioritizing power supply for battery swapping services and preventing thermal runaway caused by simultaneous high-power charging of multiple adjacent physical bays. The decision execution module constructs a mixed-integer quadratic programming model with time-division multiplexing characteristics based on acquired physical state data and logical constraint state data. The objective function of the mixed-integer quadratic programming model is formed by subtracting a linear term representing the urgency of service demand priority from a spatial thermal coupling quadratic penalty term representing the risk of physical temperature rise. The decision execution module calls a pre-built mathematical analysis library in its internal microprocessor to establish a mathematical optimization space for discrete control vectors. The specific objective function of the mixed-integer quadratic programming model is defined as follows: ; In the formula, The overall scheduling benefit value is output to optimize the objective function; The total number of physical warehouses is 8 to 24 in this embodiment. The positive integer traversal index is the target physical location within the battery swapping cabinet; For each physical location in the battery swapping cabinet other than the target physical location, use a positive integer traversal index. In order to target the The discrete control variables for solving physical storage locations are limited to taking only the values ​​0 or 1; In order to target the The discrete control variables for solving physical storage locations are limited to taking only the values ​​0 or 1; The battery swapping urgency index is input to the data evaluation module; These are the numerical values ​​of the elements representing the thermal coupling strength in the space thermal coupling impedance matrix; The business requirement weighting coefficient is used, and in this embodiment, the value range of the business requirement weighting coefficient is 0.6 to 0.8. The thermal safety penalty weighting coefficient, in this embodiment, ranges from 0.2 to 0.4. The sum of them is always a constant 1.

[0051] The underlying distribution transformer has a physical output limit, and lithium battery cell materials undergo irreversible thermal runaway chemical side reactions at high temperatures. The decision execution module transforms external physical constraints and internal state determination flags into algebraic constraints and injects them into the solution domain of a mixed-integer quadratic programming model. Addressing the conventional grid capacity limitation rules in existing technologies, the decision execution module, based on fundamental algebraic logic, forcibly constrains the total charging power of each operating physical location within the maximum available feed power of the distribution transformer. In this embodiment, the maximum available feed power ranges from 100kW to 500kW. Simultaneously, the decision execution module establishes innovative temperature prediction constraints and polarization mutual exclusion constraints based on the read node safety temperature derating limit threshold and the charging phase-locked loop constraint flag. The specific mathematical expression of the algebraic constraint group is defined as follows: ; ; In the formula, The number read at the current moment Physical compartment surface temperature data; The diagonal element in the spatial thermal coupling impedance matrix represents the thermal impedance of the i-th physical compartment itself. For the first The charging pulse excitation power set for the physical compartment is, in this embodiment, ranging from 10kW to 30kW. For each physical location in the battery swapping cabinet other than the target physical location, use a positive integer traversal index. The positive integer traversal index is the target physical location within the battery swapping cabinet; The total number of physical warehouses is 8 to 24 in this embodiment. In order to target the The discrete control variables for solving physical storage locations are limited to taking only the values ​​0 or 1; In order to target the The discrete control variables for solving physical storage locations are limited to taking only the values ​​0 or 1; For the first The charging pulse excitation power set for the DC power conversion equipment in the physical compartment; The node safety temperature derating upper limit threshold is set by the battery management system based on the chemical system. In this embodiment, the value range of the node safety temperature derating upper limit threshold is 45°C to 55°C. The first one in the polarization relaxation period Physical storage units are for units with thermal coupling relationships. When the charging phase-locked constraint flag bit generated by the physical compartment is high (i.e., the value is 1), the controller forcibly restricts the first... Discrete control variables corresponding to physical warehouse locations The value is set to 0, thereby blocking the input of charging action for adjacent coupled physical compartments, and completing the thermal isolation and polarization safety waiting action across physical compartments.

[0052] Before performing mathematical operations on the mixed-integer quadratic programming model, the decision execution module needs to dynamically construct the solution model based on the multi-dimensional sensor data stream. The decision execution module sequentially extracts the battery swapping urgency index as the target benefit coefficient, extracts the spatial thermal coupling impedance matrix elements as the target penalty coefficient, and extracts the real-time temperature and phase-locked loop constraint flags as constant boundaries of the constraint equation. The decision execution module substitutes the real-time values ​​into the matrix equation framework to complete the instantiation data loading action of the abstract operations research mathematical model. For the non-convex optimization space that has been constructed and contains discrete binary variables, the decision execution module internally calls the existing branch and bound algorithm as the core mathematical solution engine. The branch and bound algorithm continuously relaxes and splits the complex discrete optimization region into multiple linear continuous sub-problem trees with clear upper and lower bounds. During the calculation iteration process, it quickly eliminates inferior solution branches through pruning operations. Finally, within the limited calculation time, it outputs the optimal discrete control vector that satisfies all external boundary conditions and maximizes the benefit of the objective function.

[0053] To address the algorithmic dead zone problem where the solver returns no feasible solution, the decision execution module writes a safe rollback instruction within the program to force all discrete control variables to be assigned digital zeros and execute a global standby cooling action. The decision execution module analyzes and converts the optimal discrete control vector into communication control messages and sends them to each DC power conversion device via the bus. When the communication bus suffers electromagnetic interference that causes the communication control message to fail to be sent, the decision execution module immediately triggers a maximum of three communication control message retransmission logic to prevent execution stagnation dead zones caused by communication packet loss.

[0054] The communication control message drives the DC power conversion equipment to perform pulse charging action or performs electronic soft shutdown by adjusting the duty cycle of the pulse width modulation signal to zero value to replace the traditional physical contactor opening and avoid mechanical arcing. For the specific drive operation of the DC power conversion equipment based on the high-frequency switching modulation of the insulated gate bipolar transistor, you can refer to the general power electronic converter manual for program writing. The decision execution module generates a pulse charging command at the synchronous moment when driving the DC power conversion equipment to perform pulse charging and sends it back to the thermal identification module as the hardware trigger condition for the asynchronous observation time window.

[0055] Changes in the output state of DC power conversion equipment directly cause real dynamic changes in physical environmental parameters such as heat accumulation inside the battery and increased cell charge. The data evaluation module rereads the updated physical environmental parameters such as surface temperature and terminal voltage according to the polling instruction in the next set basic clock cycle. The system completes the underlying data update and continuously drives the hardware state transition by repeatedly executing the serial program of acquisition, evaluation, constraint identification and optimization decision-making. The collaborative workflow between various modules ultimately forms a closed-loop iterative chain of computation processing and underlying physical control spanning multiple control nodes in physical space and logical memory.

[0056] Application Examples: Reference Appendix Figure 6 This embodiment uses two adjacent physical compartments, A and B, within the battery swapping cabinet as examples. Compartment A holds batteries that are not fully charged and are expected to be swapped out soon; compartment B holds batteries that are half-charged during a charging interval. The two compartments are physically thermally coupled and share a localized heat dissipation space within the battery swapping cabinet.

[0057] The data evaluation module collects the state of charge and terminal voltage data of bay A and bay B according to the basic clock cycle. The module, combined with the expected swapping timestamp issued by the scheduling system, calculates that the swapping urgency index for bay A is higher than that for bay B.

[0058] The thermal identification module extracts surface temperature data from historical charging processes and updates the spatial thermal coupling impedance matrix online using a recursive least squares method with a forgetting factor. The calculated thermal coupling impedance matrix elements between compartment A and compartment B are greater than the system-set strong coupling physical threshold, thus the system determines that the two have a strong thermal coupling relationship.

[0059] After receiving the judgment result, the polarization constraint module extracts the discrete-time series of terminal voltage for battery B in the charging intermittent period and constructs a sliding sampling window. The polarization constraint module calls a univariate linear regression algorithm to calculate the slope of the fitted line as the cross-polarization relaxation. Based on the current sampling data, the cross-polarization relaxation is greater than the stationary slope threshold, indicating that the electrochemical polarization inside battery B has not been eliminated. The polarization constraint module then generates a charging phase-locked constraint flag for battery A and outputs a high-level signal.

[0060] The decision execution module constructs a mixed-integer quadratic programming model at the start of the current control time window. During the model solution process, the high battery swapping urgency index of bay A is used as the incremental revenue in the objective function. However, due to the presence of a charging phase-locked loop constraint flag for bay A, algebraic constraints force the discrete control variable corresponding to bay A to take a value of zero. Based on this, the decision execution module generates a communication control message, instructing the corresponding DC power conversion equipment to stop pulse charging bay A, thereby preventing the heat generated by charging bay A from causing heat accumulation in bay B.

[0061] As time progresses, when the polarization constraint module calculates that the cross-polarization relaxation of bay B is lower than the stability slope threshold, it automatically clears the charging phase-locked constraint flag for bay A. In the decision-making operation of the next cycle, due to the removal of the polarization mutual exclusion constraint, the decision execution module, based on the urgency advantage of bay A, outputs the optimal discrete control vector to drive the DC power conversion device to perform pulse charging action on bay A. The system uses this logic to complete time-division multiplexing equalization scheduling in a loop.

[0062] While specific embodiments of the present invention have been described above, those skilled in the art should understand that these specific embodiments are merely illustrative. Those skilled in the art can omit, substitute, and modify the details of the above methods and systems in various ways without departing from the principles and essence of the present invention. For example, combining the above method steps to perform substantially the same function and achieve substantially the same result using substantially the same method falls within the scope of the present invention.

Claims

1. A battery cluster equalization charging management system for battery swapping cabinets based on time-series prediction, characterized in that, The system includes: The data evaluation module is used to read state of charge data, terminal voltage data and surface temperature data, and receive the expected swap timestamp to calculate the swap urgency index. The thermal identification module is used to receive the surface temperature data, extract the effective temperature rise data, and update the spatial thermal coupling impedance matrix. The polarization constraint module is used to receive the spatial thermal coupling impedance matrix and the terminal voltage data, extract the cross-polarization relaxation, and generate a charging phase-locked constraint flag before reaching the steady slope threshold defined by the static characteristics. The decision execution module is used to combine the battery swapping urgency index, the spatial thermal coupling impedance matrix, and the charging phase-locked constraint flag to solve a mixed integer quadratic programming model, and output the optimal discrete control vector to drive the DC power conversion device.

2. The battery cluster equalization charging management system for battery swapping cabinets based on time-series prediction as described in claim 1, characterized in that, The data evaluation module subtracts the target full charge state threshold defined by the charging and discharging safety interval specification from the current actual charge state data to obtain the power gap value. The difference between the expected swap timestamp and the system real-time time is used as the physical remaining time window. The data evaluation module performs a division operation on the power gap value and the physical remaining time window to obtain the battery swap urgency index.

3. The battery cluster equalization charging management system for battery swapping cabinets based on time-series prediction as described in claim 2, characterized in that, When the data evaluation module determines that it has failed to receive the expected replacement time series sent from the cloud, resulting in the lack of the expected replacement timestamp, it calls the locally pre-stored fixed reference duration and adds it to the system real-time time to obtain the expected replacement timestamp. When the system real-time time is greater than or equal to the expected replacement timestamp, resulting in the physical remaining time window being less than or equal to zero, it executes a jump instruction to intercept the division operation logic and assigns the battery replacement urgency index to the maximum floating-point constant peak point supported by the hardware.

4. The battery cluster equalization charging management system for battery swapping cabinets based on time-series prediction as described in claim 1, characterized in that, The thermal identification module sets the pulse charging command issued by the decision execution module as the trigger condition, records the reference temperature at the start time of the physical compartment charging pulse, starts an asynchronous thermal observation time window containing a physical hysteresis constant, continuously traverses the input data stream within the asynchronous thermal observation time window to extract the peak data of the surface temperature, and subtracts the peak data from the reference temperature to obtain the effective temperature rise data.

5. The battery cluster equalization charging management system for battery swapping cabinets based on time-series prediction as described in claim 4, characterized in that, The thermal identification module extracts the charging pulse excitation power within the asynchronous thermal observation time window as input vector data, extracts the corresponding effective temperature rise data as tag output data, accumulates the input vector data and the tag output data in the system memory to form a historical observation dataset with time series correlation, and continuously substitutes the historical observation dataset into the gain matrix and covariance matrix for iterative feedback calculation to obtain new spatial thermal coupling impedance matrix elements.

6. The battery cluster equalization charging management system for battery swapping cabinets based on time-series prediction according to claim 1, characterized in that, The polarization constraint module traverses each element in the spatial thermal coupling impedance matrix and compares the extracted element values ​​with the strong coupling physical threshold set in conjunction with the battery swapping cabinet hardware structure. When the element value between any two physical compartments is greater than or equal to the strong coupling physical threshold, it is determined that there is a strong coupling relationship between the two physical compartments. For dormant batteries that are determined to have a strong coupling relationship and are in the charging intermittent period, the terminal voltage data is continuously read and stored in the first-in-first-out storage queue in chronological order to construct a sliding sampling window.

7. The battery cluster equalization charging management system for battery swapping cabinets based on time-series prediction according to claim 6, characterized in that, After determining that the sliding sampling window is filled with the terminal voltage data, the polarization constraint module reads the discrete-time sampling points and the corresponding terminal voltage data as the basic original data nodes in the two-dimensional coordinate system, calls the univariate linear regression algorithm to extract the slope value of the fitted line as the cross-polarization relaxation, and sets the charging phase-locked constraint flag to a logic high level when the absolute value of the cross-polarization relaxation is greater than or equal to the stationary slope threshold.

8. The battery cluster equalization charging management system for battery swapping cabinets based on time-series prediction according to claim 1, characterized in that, The decision execution module maps the spatial thermal coupling impedance matrix to the basic thermal penalty term parameter of the mixed integer quadratic programming model, which penalizes the superposition effect of heat generation. The objective function of the mixed integer quadratic programming model is formed by subtracting the urgency term representing the priority of business needs from the spatial thermal coupling quadratic penalty term representing the risk of physical temperature rise. The module calls a pre-set mathematical analysis library in the internal microprocessor to establish a mathematical optimization space for discrete control variables.

9. The battery cluster equalization charging management system for battery swapping cabinets based on time-series prediction as described in claim 8, characterized in that, The decision execution module forcibly constrains the total charging power of each operating physical compartment within the maximum available power supply capacity of the distribution transformer. Based on the node safety temperature derating upper limit threshold read by the battery management system and the charging phase-locked loop constraint flag, it establishes temperature prediction constraints and polarization mutual exclusion constraints and pushes them into the solution domain of the mixed integer quadratic programming model. When the charging phase-locked loop constraint flag is at a logic high level, it forcibly restricts the discrete control variable corresponding to the physical compartment to take the value of zero.

10. The battery cluster equalization charging management system for battery swapping cabinets based on time-series prediction according to claim 9, characterized in that, The decision execution module calls the branch and bound algorithm to output the optimal discrete control vector within a limited computation time. When the solver returns an algorithm dead zone problem where no feasible solution is found, it calls the safety rollback instruction written in the program to force all the discrete control variables to be assigned to zero and execute a global standby cooling action. The optimal discrete control vector is parsed and converted into a communication control message to control the DC power conversion device.

Images