A multi-parameter computing power cooperative optimization UPS battery pack management control method

Abstract

The application relates to the technical field of UPS and specifically discloses a UPS battery pack management control method based on multi-parameter computing power collaborative optimization, which collects multi-dimensional state parameters such as cell terminal voltage, surface temperature and internal resistance, constructs a composite state matrix capable of reflecting the state change of the cell, analyzes the continuous operation state by using a sliding time window, extracts the dominant features in the state change of the battery pack by singular value decomposition, calculates the comprehensive deviation degree according to the contribution degree of different cells in the dominant features, converts the comprehensive deviation degree into an equalization control weight, and drives the equalization circuit to perform differentiated shunt control. The method can comprehensively consider the multi-parameter coupling change of the cell, improve the state difference identification capability of the cell, realize more accurate and dynamic equalization adjustment, and improve the operation consistency and reliability of the UPS battery pack.

Description

Technical Field

[0001] This invention relates to the field of UPS technology, and specifically to a UPS battery pack management and control method with multi-parameter computing power collaborative optimization. Background Technology

[0002] UPS battery packs typically consist of multiple cells connected in series and parallel. By collecting state parameters such as voltage, temperature, and internal resistance during cell operation, the differences in state between cells are analyzed. A balancing circuit then adjusts the energy distribution among the different cells to maintain good operational consistency within the battery pack. In actual operation, cell electrical parameters and thermal states gradually differ due to manufacturing variations, aging processes, and the operating environment. Therefore, dynamic balancing control is necessary based on these changes in cell state.

[0003] Existing UPS battery management methods typically determine whether a cell needs balancing based on threshold changes of a single state parameter or multiple independent parameters, and adjust the operating state of the balancing circuit accordingly. However, cell state changes often exhibit multi-parameter coupling characteristics, and changes in a single parameter may not fully reflect the overall state deviation of the cell. Furthermore, the state differences between different cells continuously change over time, making it difficult for fixed thresholds or simple parameter comparisons to accurately distinguish which cells require priority balancing. This can easily lead to a deviation between the balancing resource allocation and the actual state of the cells. Summary of the Invention

[0004] The purpose of this invention is to provide a UPS battery pack management and control method with multi-parameter computing power collaborative optimization, thereby solving the above-mentioned technical problems.

[0005] The objective of this invention can be achieved through the following technical solutions: A UPS battery pack management and control method with multi-parameter computing power collaborative optimization includes the following steps: S1. At each sampling moment, the terminal voltage and surface temperature of each cell are collected, and the internal resistance value of each cell is collected periodically. The terminal voltage, surface temperature and internal resistance value of a single cell at the same moment are combined into a multi-parameter state vector. S2. Preprocess the multi-parameter state vectors of all cells to construct a composite state matrix. The rows of the composite state matrix correspond to the sampling time, and the columns are arranged sequentially according to the cell number. S3. Set a sliding time window of a predetermined length. Whenever a new sampling time arrives, add a new row of the composite state matrix to the sliding time window and discard the earliest row in the sliding time window. Perform singular value decomposition on the composite state matrix in the sliding time window and extract the right singular vector corresponding to the maximum singular value. S4. Group the right singular vectors according to the cell. The number of elements in each group is equal to the total number of features of the cell. Calculate the square root of the sum of squares of all elements in each group as the comprehensive deviation of the cell. S5. The overall deviation is converted into the cell's equalization control weight through a piecewise linear mapping with dead zone threshold and saturation threshold. The equalization control weight is then input into the battery pack equalization circuit to perform current shunt regulation.

[0006] As a further aspect of the present invention: S1 specifically includes: At each sampling moment, the analog-to-digital converter is triggered to quantize the potential difference between the positive and negative terminals of each cell, and the quantization result is used as the terminal voltage. The temperature acquisition channel reads the voltage division value of the thermistor attached to the surface of each cell housing, and converts it into surface temperature after analog-to-digital conversion. After each predetermined sampling interval, an AC test current with a constant amplitude and fixed duration is injected into each cell, and the AC response voltage at both ends of each cell is collected synchronously. The ratio of the AC response voltage amplitude to the AC test current amplitude is used as the internal resistance value at that sampling moment. For each sampling time of unmeasured internal resistance between two adjacent internal resistance measurement times, the internal resistance values ​​of the previous measurement time and the next measurement time are used as endpoints. Linear interpolation is performed according to the time ratio of the sampling time to the two measurement times before and after to obtain the internal resistance value at that sampling time.

[0007] As a further aspect of the present invention, the process of constructing the composite state matrix is ​​as follows: At each sampling time, the average terminal voltage, average surface temperature, and average internal resistance of all cells at that sampling time are calculated. The terminal voltage relative deviation is obtained by subtracting the average terminal voltage from the terminal voltage in the multi-parameter state vector of each cell, the surface temperature relative deviation is obtained by subtracting the average surface temperature from the surface temperature, and the internal resistance relative deviation is obtained by subtracting the average internal resistance from the internal resistance. The global average and global standard deviation of the relative deviation of the terminal voltage of a single cell are statistically analyzed. The standardized relative deviation of the terminal voltage is obtained by subtracting the global average of the relative deviation of the terminal voltage sequence from each relative deviation of the terminal voltage and dividing by the global standard deviation of the relative deviation of the terminal voltage sequence. Obtain the relative deviation of standardized surface temperature and the relative deviation of standardized internal resistance.

[0008] As a further aspect of the present invention, the process of constructing the composite state matrix further includes: The standardized terminal voltage relative deviation of a single cell constitutes a standardized terminal voltage relative deviation sequence. The differential operation between adjacent sampling times is performed on this sequence, and the result of the differential operation is divided by the sampling time interval to obtain the standardized terminal voltage change rate. Obtain the standardized surface temperature change rate and the standardized internal resistance change rate; For a single cell, the standardized relative deviation of terminal voltage, standardized relative deviation of surface temperature, standardized relative deviation of internal resistance, standardized rate of change of terminal voltage, standardized rate of change of surface temperature, and standardized rate of change of internal resistance at the same sampling time are concatenated into a single cell composite feature vector. The single-cell composite feature vectors of all battery cells are arranged and stacked in order of cell number to form a composite state matrix. The number of rows in the composite state matrix is ​​equal to the total number of sampling times, and the number of columns is equal to the total number of battery cells multiplied by six.

[0009] As a further aspect of the present invention: the process of extracting the right singular vector is as follows: The composite state matrix within the sliding time window is denoted as G. The row index of G corresponds to the sampling time, and the column index of G is arranged sequentially according to the cell number. Each cell corresponds to six consecutive columns, which are, in order, the relative deviation of standardized terminal voltage, the relative deviation of standardized surface temperature, the relative deviation of standardized internal resistance, the rate of change of standardized terminal voltage, the rate of change of standardized surface temperature, and the rate of change of standardized internal resistance. The dimension of G is M rows multiplied by six times the total number of cells and columns, where M is the total number of sampling times included in the sliding time window. Perform singular value decomposition on G to extract the maximum singular value and its corresponding right singular vector.

[0010] As a further aspect of the present invention, the extraction of the right singular vector also includes: At each new sampling moment, update the composite state matrix G within the sliding time window, record the maximum singular value in the singular value decomposition result corresponding to the updated G, and obtain the new right singular vector. Collect the processor utilization rate of the battery management main controller, set the computing power utilization threshold, the first decomposition interval and the second decomposition interval, and the first decomposition interval is less than the second decomposition interval; when the processor utilization rate is lower than the computing power utilization threshold, the first decomposition interval is used, and when the processor utilization rate is greater than or equal to the computing power utilization threshold, the second decomposition interval is used; when the cumulative number of incremental singular value decompositions reaches the currently used decomposition interval, perform a complete singular value decomposition on the composite state matrix G within the sliding time window, and replace the singular value decomposition result retained at the previous sampling time with the result of this complete singular value decomposition.

[0011] As a further aspect of the present invention, the specific process for calculating the overall deviation is as follows: According to the cell number, six consecutive elements are taken from the starting position of the right singular vector as a group, and each group corresponds to one cell. The right singular vector, which is six times the length of the total number of cells, is divided into groups equal to the total number of cells. For each group of six elements, calculate the sum of squares of the six elements, and take the square root of the sum of squares to obtain the overall deviation of the corresponding cell in that group. The overall deviation of each cell constitutes an overall deviation vector, and the length of the overall deviation vector is equal to the total number of cells.

[0012] As a further aspect of the present invention: The specific process for obtaining the balance control weights is as follows: Set a dead zone threshold K1 and a saturation threshold K2, where K1 < K2; For any cell's overall deviation P, if P < K1, then the balance control weight X = 0; If K1≤P≤K2, then the balance control weight = (P-K1) / (K2-K1). If P > K2, then the balance control weight = 1.

[0013] As a further aspect of the present invention, the specific process of diversion regulation is as follows: The battery pack balancing circuit includes a controlled shunt branch connected in parallel with each cell. The controlled shunt branch includes an adjustable resistor and a switching transistor connected in series. The equalization control weight of each cell is used as the duty cycle of the pulse width modulation signal, and the conduction time is equal to the duty cycle command multiplied by the total period of the pulse width modulation signal and then rounded down. Each controlled shunt branch's switch receives a corresponding pulse width modulation signal at its gate. The switch is turned on during the high-level range of the pulse width modulation signal and turned off during the low-level range. The on-time is determined by the duty cycle command. When on, the controlled shunt branch forms a shunt path for the corresponding cell.

[0014] The beneficial effects of this invention compared to the prior art are as follows: This invention comprehensively analyzes the multi-dimensional state changes of battery cells, enabling the balancing control to reflect the relative deviation of the cells in the overall operating state. This reduces the impact of single parameter fluctuations on balancing judgment and improves the accuracy of cell state identification. Simultaneously, it dynamically adjusts the balancing control intensity based on the state differences of different cells, prioritizing balancing resources for cells with larger state deviations. This avoids over- or under-adjustment during the balancing process, helping to improve the internal consistency of the battery pack, extend its lifespan, and enhance the stability and reliability of UPS power supply. This method can adapt to changes in cell state over operating time, giving the balancing control better dynamic adaptability. Attached Figure Description

[0015] The invention will now be further described with reference to the accompanying drawings.

[0016] Figure 1 This is a flowchart illustrating a UPS battery pack management and control method based on multi-parameter computing power collaborative optimization according to the present invention. Detailed Implementation

[0017] The technical solutions of 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.

[0018] Please see Figure 1 As shown, this invention is a UPS battery pack management and control method with multi-parameter computing power collaborative optimization, comprising the following steps: S1. At each sampling moment, the terminal voltage and surface temperature of each cell are collected, and the internal resistance value of each cell is collected periodically. The terminal voltage, surface temperature and internal resistance value of a single cell at the same moment are combined into a multi-parameter state vector. In a preferred embodiment of the present invention, S1 specifically includes: When collecting cell status parameters, the analog-to-digital converter (ADC) is first triggered according to a preset sampling period to convert the voltage sampling signal between the positive and negative terminals of a single cell into a digital value. The terminal voltage value of the corresponding cell at the current sampling moment is obtained based on the conversion result. Simultaneously, the output signal of the thermistor located on the cell casing surface is read through the temperature acquisition channel. The thermistor's resistance changes accordingly with the cell surface temperature, causing its voltage division value to change synchronously. The acquired voltage division signal is input to the ADC for quantization, and the cell surface temperature is calculated based on the correlation between the thermistor resistance and temperature. Since the internal resistance changes more slowly than the terminal voltage and surface temperature, internal resistance detection is performed at preset sampling intervals. During internal resistance detection, a constant amplitude and duration AC test current is applied to the corresponding cell, and the AC response voltage generated across the cell under this AC test current is collected. The current internal resistance value is calculated based on the proportional relationship between the AC response voltage amplitude and the AC test current amplitude. Since the internal resistance detection period is longer than the sampling periods for terminal voltage and surface temperature, there are sampling moments where the internal resistance value is not acquired between two adjacent internal resistance detection moments. Therefore, the internal resistance values ​​corresponding to the two adjacent internal resistance detection moments are used as interpolation endpoints. The interpolation ratio is determined according to the time position of the sampling moment to be supplemented between the two internal resistance detection moments. The internal resistance values ​​of the two endpoints are linearly interpolated to obtain the estimated internal resistance value of the corresponding sampling moment, so that each sampling moment has the corresponding terminal voltage, surface temperature and internal resistance value.

[0019] S2. Preprocess the multi-parameter state vectors of all cells to construct a composite state matrix. The rows of the composite state matrix correspond to the sampling time, and the columns are arranged sequentially according to the cell number. In a preferred embodiment of the present invention, the process of constructing the composite state matrix is ​​as follows: At each sampling time, the average terminal voltage, average surface temperature, and average internal resistance of all cells at that sampling time are calculated. The terminal voltage relative deviation is obtained by subtracting the average terminal voltage from the terminal voltage in the multi-parameter state vector of each cell, the surface temperature relative deviation is obtained by subtracting the average surface temperature from the surface temperature, and the internal resistance relative deviation is obtained by subtracting the average internal resistance from the internal resistance. The global average and global standard deviation of the relative deviation of the terminal voltage of a single cell are statistically analyzed. The standardized relative deviation of the terminal voltage is obtained by subtracting the global average of the relative deviation of the terminal voltage sequence from each relative deviation of the terminal voltage and dividing by the global standard deviation of the relative deviation of the terminal voltage sequence. Obtain the relative deviation of standardized surface temperature and the relative deviation of standardized internal resistance.

[0020] The process of obtaining the relative deviation of standardized surface temperature and the relative deviation of standardized internal resistance can be analogous to the relative deviation of standardized terminal voltage.

[0021] In a preferred embodiment of this invention, the process of constructing the composite state matrix further includes: The standardized terminal voltage relative deviations of a single cell at each sampling time are arranged in chronological order to form a standardized terminal voltage relative deviation sequence. For any sampling time other than the first sampling time, the standardized terminal voltage relative deviation of the current sampling time is subtracted from the standardized terminal voltage relative deviation of the previous sampling time to obtain a first-order backward difference value. The first-order backward difference value is divided by the sampling time interval between the current sampling time and the previous sampling time to obtain the standardized terminal voltage change rate of the current sampling time. Since there is no previous sampling time for the first sampling time, the standardized terminal voltage change rate of the first sampling time is set to zero. Obtain the standardized surface temperature change rate and the standardized internal resistance change rate; For a single cell, the standardized relative deviation of terminal voltage, standardized relative deviation of surface temperature, standardized relative deviation of internal resistance, standardized rate of change of terminal voltage, standardized rate of change of surface temperature, and standardized rate of change of internal resistance at the same sampling time are concatenated into a single cell composite feature vector. The single-cell composite feature vectors of all battery cells are arranged and stacked in order of cell number to form a composite state matrix. The number of rows in the composite state matrix is ​​equal to the total number of sampling times, and the number of columns is equal to the total number of battery cells multiplied by six.

[0022] The process of obtaining the standardized surface temperature change rate and the standardized internal resistance change rate can be analogous to the standardized terminal voltage change rate.

[0023] S3. Set a sliding time window of a predetermined length. Whenever a new sampling time arrives, add a new row of the composite state matrix to the sliding time window and discard the earliest row in the sliding time window. Perform singular value decomposition on the composite state matrix in the sliding time window and extract the right singular vector corresponding to the maximum singular value. In a preferred embodiment of the present invention, the process of extracting the right singular vector is as follows: The composite state matrix within the sliding time window is denoted as G. The row index of G corresponds to the sampling time, and the column index of G is arranged sequentially according to the cell number. Each cell corresponds to six consecutive columns, which are, in order, the relative deviation of standardized terminal voltage, the relative deviation of standardized surface temperature, the relative deviation of standardized internal resistance, the rate of change of standardized terminal voltage, the rate of change of standardized surface temperature, and the rate of change of standardized internal resistance. The dimension of G is M rows multiplied by six times the total number of cells and columns, where M is the total number of sampling times included in the sliding time window. Perform singular value decomposition on G to extract the maximum singular value and its corresponding right singular vector.

[0024] In a preferred embodiment, the extraction of the right singular vector further includes: At each new sampling moment, update the composite state matrix G within the sliding time window, record the maximum singular value in the singular value decomposition result corresponding to the updated G, and obtain the new right singular vector. Collect the processor utilization rate of the battery management main controller, set the computing power utilization threshold, the first decomposition interval and the second decomposition interval, and the first decomposition interval is less than the second decomposition interval; when the processor utilization rate is lower than the computing power utilization threshold, the first decomposition interval is used, and when the processor utilization rate is greater than or equal to the computing power utilization threshold, the second decomposition interval is used; when the cumulative number of incremental singular value decompositions reaches the currently used decomposition interval, perform a complete singular value decomposition on the composite state matrix G within the sliding time window, and replace the singular value decomposition result retained at the previous sampling time with the result of this complete singular value decomposition.

[0025] Each update of the sliding time window involves deleting only the earliest row and adding a new row, while the rest of the data within the window remains unchanged. Therefore, it's unnecessary to re-perform the full singular value decomposition (SVD) on the entire composite state matrix at every sampling time. First, the deletion update eliminates the influence of the earliest row on the original SVD result. Then, incremental SVD adds the influence of the new row, allowing the current update to utilize the previous calculation results. Continuous incremental updates may introduce numerical errors; therefore, full SVD is performed intermittently to recalibrate the results. The full SVD interval is shortened when processor utilization is low and lengthened when it is high, allowing the SVD task to adjust its computation method based on available computing power.

[0026] S4. Group the right singular vectors according to the cell. The number of elements in each group is equal to the total number of features of the cell. Calculate the square root of the sum of squares of all elements in each group as the comprehensive deviation of the cell. In a preferred embodiment of the present invention, the specific process for calculating the overall deviation is as follows: According to the cell number, six consecutive elements are taken from the starting position of the right singular vector as a group, and each group corresponds to one cell. The right singular vector, which is six times the length of the total number of cells, is divided into groups equal to the total number of cells. For each group of six elements, calculate the sum of squares of the six elements, and take the square root of the sum of squares to obtain the overall deviation of the corresponding cell in that group. The overall deviation of each cell constitutes an overall deviation vector, and the length of the overall deviation vector is equal to the total number of cells.

[0027] In a preferred embodiment, the specific process for obtaining the balance control weights is as follows: Set a dead zone threshold K1 and a saturation threshold K2, where K1 < K2; For any cell's overall deviation P, if P < K1, then the balance control weight X = 0; If K1≤P≤K2, then the balance control weight = (P-K1) / (K2-K1). If P > K2, then the balance control weight = 1.

[0028] S5. The overall deviation is converted into the cell's equalization control weight through a piecewise linear mapping with dead zone threshold and saturation threshold. The equalization control weight is then input into the battery pack equalization circuit to perform current shunt regulation.

[0029] In a preferred embodiment of the present invention, the specific process of diversion control is as follows: The equalization circuit performs current shunting regulation based on the equalization control weight corresponding to each cell. The battery pack equalization circuit has multiple controlled shunting branches, each connected in parallel with one cell to provide an additional discharge path for the corresponding cell when equalization is required. A single controlled shunting branch includes an adjustable resistor and a switching transistor connected in series. The adjustable resistor limits the current during the shunting process to prevent excessive discharge current from the cell during equalization, while the switching transistor switches the conduction state of the shunting branch according to the control signal.

[0030] During equalization control, the equalization control weights for each cell are first obtained, and these weights are then converted into the duty cycle of the pulse width modulation (PWM) signal for the corresponding controlled shunt branch. For example, when the overall deviation of a cell is low, its corresponding equalization control weight is small, resulting in a shorter high-level duration in the generated PWM signal and a shorter on-time for the switching transistor. Conversely, when the overall deviation of a cell is high, its corresponding equalization control weight increases, and the high-level duration of the PWM signal increases accordingly, allowing the switching transistor to remain on for a longer period. Specifically, the on-time of the switching transistor within a single cycle is calculated based on the total period of the PWM signal and the corresponding duty cycle command, and the calculated on-time is converted into the actual on-time within the control cycle.

[0031] Subsequently, the pulse width modulation (PWM) signal corresponding to each cell is input to the gate control terminal of the switching transistor in the corresponding controlled shunt branch. When the PWM signal is high, the switching transistor is turned on, forming a closed shunt circuit between the corresponding cell and the adjustable resistor. Part of the current inside the cell is released through the controlled shunt branch, achieving balanced discharge of the cell. When the PWM signal is low, the switching transistor is turned off, the shunt branch is disconnected, and the cell returns to normal operation. By continuously adjusting the duty cycle of the PWM signal, the shunt duration corresponding to different cells is matched to their deviation from the state, thereby achieving dynamic adjustment of the energy differences among the cells within the battery pack.

[0032] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the present invention should still fall within the scope of the present invention.

Claims

1. A UPS battery pack management and control method with multi-parameter computing power collaborative optimization, characterized in that, Includes the following steps: S1. At each sampling moment, the terminal voltage and surface temperature of each cell are collected, and the internal resistance value of each cell is collected periodically. The terminal voltage, surface temperature and internal resistance value of a single cell at the same moment are combined into a multi-parameter state vector. S2. Preprocess the multi-parameter state vectors of all cells to construct a composite state matrix. The rows of the composite state matrix correspond to the sampling time, and the columns are arranged sequentially according to the cell number. S3. Set a sliding time window of a predetermined length. Whenever a new sampling time arrives, add a new row of the composite state matrix to the sliding time window and discard the earliest row in the sliding time window. Perform singular value decomposition on the composite state matrix in the sliding time window and extract the right singular vector corresponding to the maximum singular value. S4. Group the right singular vectors according to the cell. The number of elements in each group is equal to the total number of features of the cell. Calculate the square root of the sum of squares of all elements in each group as the comprehensive deviation of the cell. S5. The overall deviation is converted into the cell's equalization control weight through a piecewise linear mapping with dead zone threshold and saturation threshold. The equalization control weight is then input into the battery pack equalization circuit to perform current shunt regulation.

2. The UPS battery pack management and control method based on multi-parameter computing power collaborative optimization according to claim 1, characterized in that, S1 specifically includes: At each sampling moment, the analog-to-digital converter is triggered to quantize the potential difference between the positive and negative terminals of each cell, and the quantization result is used as the terminal voltage. The temperature acquisition channel reads the voltage division value of the thermistor attached to the surface of each cell housing, and converts it into surface temperature after analog-to-digital conversion. After each predetermined sampling interval, an AC test current with a constant amplitude and fixed duration is injected into each cell, and the AC response voltage at both ends of each cell is collected synchronously. The ratio of the AC response voltage amplitude to the AC test current amplitude is used as the internal resistance value at that sampling moment. For each sampling time of unmeasured internal resistance between two adjacent internal resistance measurement times, the internal resistance values ​​of the previous measurement time and the next measurement time are used as endpoints. Linear interpolation is performed according to the time ratio of the sampling time to the two measurement times before and after to obtain the internal resistance value at that sampling time.

3. The UPS battery pack management and control method based on multi-parameter computing power collaborative optimization according to claim 1, characterized in that, The process of constructing the composite state matrix is ​​as follows: At each sampling time, the average terminal voltage, average surface temperature, and average internal resistance of all cells at that sampling time are calculated. The terminal voltage relative deviation is obtained by subtracting the average terminal voltage from the terminal voltage in the multi-parameter state vector of each cell, the surface temperature relative deviation is obtained by subtracting the average surface temperature from the surface temperature, and the internal resistance relative deviation is obtained by subtracting the average internal resistance from the internal resistance. The global average and global standard deviation of the relative deviation of the terminal voltage of a single cell are statistically analyzed. The standardized relative deviation of the terminal voltage is obtained by subtracting the global average of the relative deviation of the terminal voltage sequence from each relative deviation of the terminal voltage and dividing by the global standard deviation of the relative deviation of the terminal voltage sequence. Obtain the relative deviation of standardized surface temperature and the relative deviation of standardized internal resistance.

4. The UPS battery pack management and control method based on multi-parameter computing power collaborative optimization according to claim 3, characterized in that, The process of constructing the composite state matrix also includes: The standardized terminal voltage relative deviation of a single cell constitutes a standardized terminal voltage relative deviation sequence. The differential operation between adjacent sampling times is performed on this sequence, and the result of the differential operation is divided by the sampling time interval to obtain the standardized terminal voltage change rate. Obtain the standardized surface temperature change rate and the standardized internal resistance change rate; For a single cell, the standardized relative deviation of terminal voltage, standardized relative deviation of surface temperature, standardized relative deviation of internal resistance, standardized rate of change of terminal voltage, standardized rate of change of surface temperature, and standardized rate of change of internal resistance at the same sampling time are concatenated into a single cell composite feature vector. The single-cell composite feature vectors of all battery cells are arranged and stacked in order of cell number to form a composite state matrix. The number of rows in the composite state matrix is ​​equal to the total number of sampling times, and the number of columns is equal to the total number of battery cells multiplied by six.

5. The UPS battery pack management and control method based on multi-parameter computing power collaborative optimization according to claim 4, characterized in that, The process of extracting the right singular vector is as follows: The composite state matrix within the sliding time window is denoted as G. The row index of G corresponds to the sampling time, and the column index of G is arranged sequentially according to the cell number. Each cell corresponds to six consecutive columns, which are, in order, the relative deviation of standardized terminal voltage, the relative deviation of standardized surface temperature, the relative deviation of standardized internal resistance, the rate of change of standardized terminal voltage, the rate of change of standardized surface temperature, and the rate of change of standardized internal resistance. The dimension of G is M rows multiplied by six times the total number of cells and columns, where M is the total number of sampling times included in the sliding time window. Perform singular value decomposition on G to extract the maximum singular value and its corresponding right singular vector.

6. The UPS battery pack management and control method based on multi-parameter computing power collaborative optimization according to claim 5, characterized in that, Extracting right singular vectors also includes: At each new sampling moment, update the composite state matrix G within the sliding time window, record the maximum singular value in the singular value decomposition result corresponding to the updated G, and obtain the new right singular vector. Collect the processor utilization rate of the battery management main controller, set the computing power utilization threshold, the first decomposition interval and the second decomposition interval, and the first decomposition interval is less than the second decomposition interval; when the processor utilization rate is lower than the computing power utilization threshold, the first decomposition interval is used, and when the processor utilization rate is greater than or equal to the computing power utilization threshold, the second decomposition interval is used; when the cumulative number of incremental singular value decompositions reaches the currently used decomposition interval, perform a complete singular value decomposition on the composite state matrix G within the sliding time window, and replace the singular value decomposition result retained at the previous sampling time with the result of this complete singular value decomposition.

7. The UPS battery pack management and control method for multi-parameter computing power collaborative optimization according to claim 6, characterized in that, The specific process for calculating the overall deviation is as follows: According to the cell number, six consecutive elements are taken from the starting position of the right singular vector as a group, and each group corresponds to one cell. The right singular vector, which is six times the length of the total number of cells, is divided into groups equal to the total number of cells. For each group of six elements, calculate the sum of squares of the six elements, and take the square root of the sum of squares to obtain the overall deviation of the corresponding cell in that group. The overall deviation of each cell constitutes an overall deviation vector, and the length of the overall deviation vector is equal to the total number of cells.

8. The UPS battery pack management and control method based on multi-parameter computing power collaborative optimization according to claim 1, characterized in that, The specific process for obtaining the balance control weights is as follows: Set a dead zone threshold K1 and a saturation threshold K2, where K1 < K2; For any cell's overall deviation P, if P < K1, then the balance control weight X = 0; If K1≤P≤K2, then the balance control weight = (P-K1) / (K2-K1). If P > K2, then the balance control weight = 1.

9. The UPS battery pack management and control method based on multi-parameter computing power collaborative optimization according to claim 1, characterized in that, The specific process of diversion and regulation is as follows: The battery pack balancing circuit includes a controlled shunt branch connected in parallel with each cell. The controlled shunt branch includes an adjustable resistor and a switching transistor connected in series. The equalization control weight of each cell is used as the duty cycle of the pulse width modulation signal, and the conduction time is equal to the duty cycle command multiplied by the total period of the pulse width modulation signal and then rounded down. Each controlled shunt branch's switch receives a corresponding pulse width modulation signal at its gate. The switch is turned on during the high-level range of the pulse width modulation signal and turned off during the low-level range. The on-time is determined by the duty cycle command. When on, the controlled shunt branch forms a shunt path for the corresponding cell.

Images