Power distribution method and system of liquid cooling string PCS energy storage converter

Abstract

The invention relates to the technical field of power distribution, and discloses a power distribution method and system for a liquid cooling string PCS energy storage converter, and the method comprises the steps: constructing a multi-dimensional state matrix of each energy storage unit in the liquid cooling string PCS energy storage converter; establishing a thermal coupling model among the energy storage units according to the temperature parameters in the multi-dimensional state matrix, and constructing a temperature constraint projection mapping function; constructing a power optimization objective function considering the thermal coupling effect; calculating temperature change caused by disturbance through a thermal coupling model to obtain a temperature sensitive gradient vector; the power distribution weight of each energy storage unit is dynamically adjusted, and when the iteration change value of the power optimization objective function is smaller than a preset threshold value and the thermal coupling model is stable, a target power distribution strategy is output, collaborative optimization of power distribution and temperature control is achieved, the problem of thermal interference between the energy storage units is effectively solved, and the reliability of power distribution and temperature control is improved. And the overall operation efficiency of the liquid cooling string PCS energy storage converter is obviously improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power distribution, and in particular to a power distribution method and system for a liquid-cooled string PCS energy storage converter. Background Art

[0002] With the rapid development of renewable energy power generation and energy storage technology, liquid-cooled string PCS energy storage converters are widely used due to their high power density, high efficiency and modularity. However, in actual operation, due to the complex thermal coupling relationship and power coupling effect between energy storage units, traditional power allocation methods are difficult to take into account both system efficiency and thermal management requirements.

[0003] Existing power allocation strategies are mainly based on simplified thermal models and fixed allocation rules, which cannot accurately describe the temperature dynamic characteristics of the liquid cooling system, resulting in low system efficiency and even local overheating. In addition, traditional methods usually use model-based optimization algorithms, which require accurate system parameters and complex mathematical models, which are often difficult to obtain or have large errors in practical applications. Summary of the invention

[0004] The present invention provides a power distribution method and system for a liquid-cooled string-type PCS energy storage converter, which realizes the coordinated optimization of power distribution and temperature control, effectively solves the thermal interference problem between energy storage units, and significantly improves the overall operating efficiency of the liquid-cooled string-type PCS energy storage converter.

[0005] In a first aspect, the present invention provides a power distribution method for a liquid-cooled string-type PCS energy storage converter, the power distribution method for the liquid-cooled string-type PCS energy storage converter comprising: Collect the output voltage, output current, coolant inlet temperature, outlet temperature and flow data of each energy storage unit in the liquid-cooled string PCS energy storage converter to build a multi-dimensional state matrix; Establishing a thermal coupling model between the energy storage units according to the temperature parameters in the multidimensional state matrix, and constructing a temperature constraint projection mapping function based on the thermal coupling model; The temperature constraint projection mapping function is combined with the power loss of each energy storage unit to construct a power optimization objective function considering the thermal coupling effect; Performing random perturbation sampling on the initial power allocation strategy of the energy storage unit, and calculating the temperature change caused by the perturbation through the thermal coupling model to obtain a temperature sensitive gradient vector; The power allocation weight of each energy storage unit is dynamically adjusted based on the temperature sensitive gradient vector, and when the iterative change value of the power optimization objective function is less than a preset threshold and the thermal coupling model is stable, a target power allocation strategy is output.

[0006] In a second aspect, the present invention provides a power distribution system for a liquid-cooled string-type PCS energy storage converter, the power distribution system for the liquid-cooled string-type PCS energy storage converter comprising: The acquisition module is used to collect the output voltage, output current, coolant inlet temperature, outlet temperature and flow data of each energy storage unit in the liquid-cooled string PCS energy storage converter to construct a multi-dimensional state matrix; An establishment module is used to establish a thermal coupling model between each energy storage unit according to the temperature parameters in the multidimensional state matrix, and to construct a temperature constraint projection mapping function based on the thermal coupling model; A construction module is used to combine the temperature constraint projection mapping function with the power loss of each energy storage unit to construct a power optimization objective function considering thermal coupling effect; A calculation module, used for performing random perturbation sampling on the initial power allocation strategy of the energy storage unit, and calculating the temperature change caused by the perturbation through the thermal coupling model to obtain a temperature sensitive gradient vector; An output module is used to dynamically adjust the power allocation weight of each energy storage unit based on the temperature sensitive gradient vector, and output a target power allocation strategy when the iterative change value of the power optimization objective function is less than a preset threshold and the thermal coupling model is stable.

[0007] In the technical solution provided by the present invention, by constructing a temperature constraint projection mapping function, the thermal management constraint is directly integrated into the optimization process, avoiding the convergence problem caused by the use of penalty functions in traditional methods, and ensuring that the energy storage unit always operates within a safe temperature range. The optimization method without model feedback is adopted, and the optimal power distribution can be achieved only by relying on the real-time feedback data of the system, which overcomes the shortcomings of the traditional model-based method that is highly dependent on system parameters and improves the applicability and robustness of the algorithm. The introduction of thermal coupling model and temperature-sensitive gradient vector realizes the coordinated optimization of power distribution and temperature control, effectively solves the thermal interference problem between energy storage units, and significantly improves the overall operation efficiency of the system. A gradient estimation method based on random perturbation sampling is designed to avoid complex derivative calculation processes, reduce the computational complexity of the algorithm, and make it more suitable for real-time control applications. The dynamic weight adjustment mechanism is adopted, combined with the temperature stability criterion, to ensure the convergence of the optimization process and the stability of temperature control, and improve the reliability and safety of the system. Through the iterative optimization of the power optimization objective function, a significant reduction in system loss is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0009] Figure 1 Schematic diagram of the steps of a power distribution method for a liquid-cooled string PCS energy storage converter according to an embodiment of the present invention; Figure 2 Schematic diagram of the structure of the power distribution system of the liquid-cooled string PCS energy storage converter in an embodiment of the present invention. DETAILED DESCRIPTION

[0010] An embodiment of the present invention provides a power distribution method and system for a liquid-cooled string PCS energy storage converter. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0011] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , an embodiment of a power distribution method for a liquid-cooled string PCS energy storage converter in an embodiment of the present invention includes: Step S1, collecting the output voltage, output current, coolant inlet temperature, outlet temperature and flow data of each energy storage unit in the liquid-cooled string PCS energy storage converter, and constructing a multi-dimensional state matrix; It is understandable that the execution subject of the present invention may be a power distribution system of a liquid-cooled string PCS energy storage converter, or a terminal or a server, which is not specifically limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0012] Specifically, the voltage and current signals of all energy storage units are collected at the same time point through a high-precision ADC (analog-to-digital converter) to obtain the output voltage sampling sequence and the output current sampling sequence. The sampled voltage and current signal sequences are digitally filtered to eliminate high-frequency noise and random errors in the sampling process, and obtain smoother and more stable filtered voltage value sequences and filtered current value sequences. Through low-pass filtering algorithm, Kalman filtering or adaptive filtering algorithm, the signal can reduce fluctuations caused by transient disturbances or external interference while retaining key information. At the same time, the temperature of the coolant pipeline of the liquid-cooled string PCS energy storage converter is collected. By installing a temperature sensor on the cooling pipeline of each energy storage unit, the inlet temperature and outlet temperature data of the coolant are obtained respectively, reflecting the heat exchange of the coolant when it flows through the energy storage unit. By differentially calculating the inlet temperature and outlet temperature of the coolant, a temperature gradient sequence is obtained to characterize the local temperature change of the energy storage unit. Heat transfer calculation is performed based on the coolant flow data and the temperature gradient sequence. The size of the coolant flow directly affects the efficiency of heat transfer, so it is necessary to ensure that the flow sensor can accurately measure the real-time flow value of the coolant, so as to calculate the heat dissipation power of each energy storage unit through the heat transfer formula. According to the basic principle of heat transfer, the heat absorbed by the coolant per unit time is expressed as the flow multiplied by the specific heat capacity and then multiplied by the temperature gradient, that is, the heat dissipation power is equal to the product of the coolant mass flow rate, the specific heat capacity and the temperature gradient, where the mass flow rate is calculated by the volume flow rate and the liquid density. Through this calculation process, the heat dissipation of each energy storage unit is estimated and used as part of the multidimensional state matrix. The filtered voltage value sequence, the filtered current value sequence and the heat dissipation power data are concatenated to form an initial state matrix. The initial state matrix is ​​normalized, and the minimum-maximum normalization method or the standardization method is used to process the data of different physical quantities at the same scale, thereby avoiding calculation deviations caused by different data dimensions. In order to ensure the time series of the data, timestamp information is added to the multidimensional state matrix, so that the state evolution of the energy storage unit can be analyzed based on the time series when the power allocation strategy is optimized later, so as to adjust the power allocation strategy more accurately.

[0013] Step S2, establishing a thermal coupling model between the energy storage units according to the temperature parameters in the multidimensional state matrix, and constructing a temperature constraint projection mapping function based on the thermal coupling model; Specifically, the temperature parameters in the multidimensional state matrix are analyzed in time series to calculate the heat transfer relationship between the energy storage units. By analyzing the temperature change trend at different time steps, combined with the flow of coolant and environmental factors, the temperature conduction coefficient and heat capacity coefficient between adjacent energy storage units are calculated. The temperature conduction coefficient characterizes the heat transfer efficiency between energy storage units, while the heat capacity coefficient reflects the thermal inertia of a single energy storage unit during temperature changes. Therefore, these two parameters are accurately solved based on the heat conduction equation and experimental data regression method. The heat balance equation is used to characterize the evolution of unit temperature over time, in which the temperature change rate of each unit is jointly determined by its own heat capacity and the heat transfer of the surrounding units, so that the corresponding heat transfer coefficient is obtained by least squares fitting or machine learning methods. The thermal network topology of the energy storage unit is established based on the temperature conduction coefficient and the heat capacity coefficient, and a thermal coupling relationship matrix is ​​constructed. Since the energy storage units are arranged in series or parallel in physical structure, and the flow path of the liquid cooling system determines the way of heat exchange, by analyzing the path of the coolant flowing through each unit and combining the geometric distribution information of the energy storage unit, a network topology with energy storage units as nodes and heat conduction paths as edges is constructed, and a thermal coupling relationship matrix is ​​formed based on this topology. In this matrix, each element represents the heat exchange intensity between two energy storage units, and the diagonal elements of the matrix represent the heat capacity effect of the unit itself, while the non-diagonal elements represent the coupling intensity between different units. By constructing a thermal coupling relationship matrix, the temperature evolution law between energy storage units is systematically described. According to the thermal coupling relationship matrix, the temperature state space equation is constructed, and the safe operating temperature range of the energy storage unit is set. The basic form of the temperature state space equation is described based on the differential equation of heat transfer, that is, the temperature change rate of the unit is equal to the combined effect of the input power, the heat dissipation power and the thermal coupling effect, and the equation is expressed as a linear state equation, in which the state variable is the temperature of each energy storage unit, the input variable is the power distribution, and the system matrix is ​​determined by the thermal coupling relationship matrix. Based on this state equation, the temperature safety operating range of each energy storage unit is set, that is, the temperature must be kept within a reasonable range to ensure the stable operation of the energy storage system. The safety range is determined by the working characteristics of the battery and the heat dissipation capacity of the cooling system. Therefore, reasonable upper and lower bounds are set through experimental data and industry standards. The temperature state space equation is converted into an inequality constraint group for optimization calculation. In the state space expression, the temperature of each energy storage unit is affected by its own state and the state of the surrounding units. The temperature constraint conditions are converted into a linear inequality group through matrix operations, and then the temperature constraint boundary conditions are defined. These boundary conditions limit the range of influence of power distribution on temperature evolution, so that the optimization calculation is carried out within a reasonable temperature range. The temperature constraint boundary conditions are transformed into a convex set mapping, and a projection operator is constructed based on this. The convex set mapping transformation is to project complex inequality constraints onto an easy-to-calculate convex set so that they can converge quickly during the optimization calculation process.The projection operator is constructed by least squares projection or convex optimization method, which projects any power allocation scheme that does not meet the temperature constraint into the feasible solution space, so that the optimization algorithm can generate a solution that meets the temperature constraint at each iteration. By combining the projection operator with the thermal coupling model, a temperature constraint projection mapping function is constructed, which adjusts the power allocation scheme at each step of the optimization calculation, so that it can meet the temperature safety constraint while maximizing the overall efficiency of the energy storage system.

[0014] The temperature constraint boundary conditions are subjected to convex set decomposition operations, and the entire temperature constraint space is divided into multiple non-intersecting convex subsets. Since the temperature of the energy storage unit is affected by the thermal coupling effect, the temperature distribution is not a single linear constraint, but presents a complex high-dimensional structure. Therefore, the convex analysis method is used to represent the temperature constraint space as a union of a set of convex sets, so as to facilitate efficient optimization in the subsequent calculation process. By performing feature analysis on the temperature state variables, the Gaussian mixture model or hierarchical clustering algorithm is used to identify different temperature regions and partition them to construct multiple local convex subsets, each of which corresponds to the feasible temperature range of the energy storage unit under specific thermal conditions. The objective function of the quadratic programming problem is constructed based on the convex subsets, and the temperature constraint boundary conditions are converted into a standard constraint matrix form. The objective function of the quadratic programming problem is defined as minimizing the power loss while ensuring that the temperature state satisfies the defined convex subset constraints. The objective function is in the form of the quadratic norm of the power allocation variable, and the constraint matrix is ​​used to characterize the influence of temperature on power allocation. The constraints are introduced into the objective function by the Lagrange multiplier method or the augmented Lagrangian method, so that the optimization problem can obtain the optimal solution under the convex set constraints. The construction of the constraint matrix is ​​based on the thermal coupling relationship of the energy storage unit. By analyzing the heat transfer path, the temperature constraint condition is expressed as a set of linear inequalities, and the matrix representation is used to describe the feasible range of the temperature state variables to ensure that the optimization problem can maintain numerical stability during the solution process. The constraint matrix is ​​subjected to singular value decomposition to obtain an orthogonal basis matrix and an eigenvalue matrix. Singular value decomposition decomposes the original high-dimensional constraint matrix into an orthogonal basis matrix and a diagonal eigenvalue matrix, so that the optimization problem can be calculated in a low-dimensional space, improving computational efficiency and reducing redundant degrees of freedom. Let the constraint matrix be A, then it can be expressed as , where U and V are the left and right orthogonal basis matrices, respectively, and Σ is a diagonal matrix whose diagonal elements are singular values. In the temperature constraint optimization process, the orthogonal basis matrix is ​​used to transform the coordinate system of the temperature state variable, while the eigenvalue matrix is ​​used to measure the temperature change amplitude in different directions, so as to accurately project the temperature state. Based on the orthogonal basis matrix and eigenvalue matrix obtained by singular value decomposition, a projection space transformation matrix is ​​constructed, and the projection space transformation matrix is ​​subjected to tensor product operation with the thermal coupling model to obtain the tensor product operation result. The tensor product operation is used to combine the thermal coupling model with the projection space so that the thermal coupling relationship remains consistent in the coordinate system after the projection transformation, thereby ensuring the physical rationality of the optimization calculation. By calculating the tensor product, the influence of the thermal coupling model is incorporated into the optimization calculation framework, so that the power allocation strategy fully considers the heat transfer effect between the energy storage units and optimizes the overall temperature distribution. The tensor product operation result is subjected to singular value threshold processing to obtain the temperature constraint projection kernel matrix. By setting a threshold, the eigenvalues ​​in the singular value matrix that are less than the threshold are set to zero to eliminate the degrees of freedom that have little impact on the temperature state and ensure that the optimization calculation is performed under numerical stability. The temperature constraint projection kernel matrix is ​​convolved with the thermal coupling model to obtain the temperature constraint projection mapping function. The convolution operation is to smooth the changes in the temperature state during the optimization process, so that the power allocation adjustment can more gently affect the temperature evolution of the energy storage unit and avoid excessive temperature fluctuations due to sudden power adjustments.

[0015] Step S3, combining the temperature constraint projection mapping function with the power loss of each energy storage unit to construct a power optimization objective function considering the thermal coupling effect; Specifically, the conduction loss and switching loss of each energy storage unit are calculated based on the filtered voltage value sequence and the filtered current value sequence to obtain the total power loss of the entire system. In the power conversion process, the conduction loss mainly comes from the energy loss caused by the current flowing through the on-resistance of the power semiconductor device in the on state, while the switching loss is the energy dissipation generated by the power device during the switching conversion process. The conduction loss and switching loss are calculated respectively using the current and voltage data of the energy storage unit, the switching frequency of the power converter and the switching characteristic parameters of the semiconductor device to obtain the total power loss. The temperature constraint projection mapping function is Taylor expanded to obtain a locally linearized temperature constraint function, thereby simplifying the optimization calculation and improving the solution efficiency. In this process, an appropriate expansion point is selected, that is, the temperature value corresponding to the current power distribution state, and the projection mapping function is approximated in the form of Taylor expansion to obtain a first-order or high-order linearized temperature constraint function. The total power loss is combined with the locally linearized temperature constraint function to construct a loss function with constraints, where the total power loss is used to characterize the energy efficiency of the system, and the linearized temperature constraint function is used to ensure that the power distribution does not cause the temperature to exceed the safe range. In order to effectively constrain the impact of power loss on temperature during the optimization calculation process, the penalty factor is set according to the constrained loss function to obtain the augmented Lagrangian function. The augmented Lagrangian function is an optimization method that can incorporate constraints into the objective function. By introducing penalty factors and Lagrangian multipliers, the optimization problem can be solved in an unconstrained optimization framework. By defining a penalty term, when the temperature constraint is violated due to power allocation, the penalty term will increase significantly, forcing the optimization algorithm to automatically adjust the power allocation strategy to meet the temperature constraint. The augmented Lagrangian function is subjected to variational calculation to solve the optimality conditions of power allocation, and based on this, the power optimization objective function is constructed. Variational calculation is a necessary condition for obtaining the optimal solution by deriving the optimization function and analyzing its stationary point conditions. In the framework of augmented Lagrangian optimization, the power allocation variables are derived, and combined with the temperature constraints, the KKT (Karush-Kuhn-Tucker) optimality conditions of the optimization problem are derived. These conditions include gradient equations, complementary relaxation conditions, and constraint satisfaction conditions, thereby systematically characterizing the optimal solution structure of power allocation. Based on the obtained optimality conditions, the final power optimization objective function is constructed. This objective function minimizes the power loss of the entire system while satisfying the temperature safety constraints, thereby improving the overall operating efficiency of the energy storage system.

[0016] Step S4, performing random disturbance sampling on the initial power allocation strategy of the energy storage unit, and calculating the temperature change caused by the disturbance through a thermal coupling model to obtain a temperature sensitive gradient vector; Specifically, a uniformly distributed random perturbation vector is generated based on the power optimization objective function to perform perturbation sampling on the initial power allocation strategy of the energy storage unit and obtain multiple power allocation schemes after perturbation. The generation of random perturbation vectors needs to ensure a certain uniformity so that the power perturbations of different energy storage units cover the entire feasible range, improving the comprehensiveness of sampling and the accuracy of calculation. The amplitude of the random perturbation vector is controlled within a certain range to prevent excessive perturbations from causing calculation instability, while ensuring that the power allocation after each perturbation still meets the basic operating constraints of the system. The random perturbation vector is superimposed on the initial power allocation scheme to form a power allocation strategy after perturbation, thereby simulating the temperature change trend under different power allocation conditions. The power allocation strategy after perturbation is substituted into the thermal coupling model to calculate the temperature response caused by the perturbation. The function of the thermal coupling model is to describe the heat transfer relationship between energy storage units, where the temperature change of each energy storage unit is not only affected by its own power loss, but also by the heat transfer effect of the adjacent units. When calculating the temperature response after perturbation, the influence of the overall heat exchange is considered. In order to obtain accurate temperature response results, the heat transfer equation is solved numerically to obtain the temperature change of each energy storage unit after the disturbance. The calculated temperature response data reflects the temperature change trend of each energy storage unit under different power disturbances. The temperature response after the disturbance is compared with the initial temperature, and the temperature change is obtained by calculating the difference between the two, which reflects the degree of response of the temperature of the energy storage unit to the power disturbance. The temperature sensitivity matrix is ​​constructed according to the temperature change and the random disturbance vector. This matrix is ​​used to describe the correlation between power distribution and temperature change, in which each element represents the impact of the power change of a certain energy storage unit on the temperature change of itself or other units. By constructing this matrix, how power adjustment affects the thermal stability of the energy storage system is systematically analyzed. The temperature sensitivity matrix is ​​subjected to singular value decomposition to extract its main eigenvector, and the temperature sensitive gradient vector is constructed based on the eigenvector. Singular value decomposition decomposes the temperature sensitivity matrix into eigenvectors in different directions and sorts them according to importance. Through this method, the direction in which the temperature is most sensitive to power distribution is extracted, and the temperature sensitive gradient vector is constructed using this information.

[0017] Step S5: dynamically adjust the power allocation weight of each energy storage unit based on the temperature sensitive gradient vector, and output the target power allocation strategy when the iterative change value of the power optimization objective function is less than a preset threshold and the thermal coupling model is stable.

[0018] Specifically, the temperature-sensitive gradient vector is used to perform gradient descent calculation on the power allocation weight to obtain the updated power allocation weight. The gradient descent method is used to optimize the power allocation strategy so that it can gradually converge to the optimal solution, wherein the temperature-sensitive gradient vector provides the direction of power allocation adjustment, so that the power allocation is optimized in the direction of reducing temperature imbalance. In each iteration, the power allocation weight of the previous round is used, and the temperature-sensitive gradient vector obtained by the current calculation is combined to update the power allocation weight so that the power adjustment can gradually approach the optimal allocation state. In order to ensure the stability of the calculation, the step size of the gradient descent is appropriately set so that the optimization process can converge quickly and avoid oscillation or overshoot, thereby ensuring the rationality and effectiveness of the power allocation adjustment. The updated power allocation weight is substituted into the power optimization objective function to calculate the iterative change value of the objective function, so as to determine whether the power optimization process converges. The iterative change value of the power optimization objective function is used to measure the progress of the optimization process. If the change amplitude of the objective function gradually decreases, it means that the optimization process is converging, and if the change amplitude is still large, it means that the current power allocation strategy still needs further adjustment. After each round of iteration, the new value of the optimization objective function is calculated and compared with the value of the previous round to obtain the iterative change value, and the value is used to judge the convergence trend of the power allocation optimization. At the same time, in order to ensure that the optimized power allocation strategy maintains the stability of temperature while satisfying the power loss optimization, the temperature change rate is calculated based on the thermal coupling model, and it is judged whether the temperature enters a stable state. Since the thermal coupling model describes the heat transfer relationship between energy storage units, the model is used to calculate the temperature change of each energy storage unit under the current power allocation strategy, and the temperature change rate, that is, the rate of change of the temperature of each energy storage unit over time, is obtained. If the temperature change rate is large, it means that the temperature of the energy storage unit is still in a dynamic change stage, and the power allocation strategy still needs to be adjusted. If the norm of the temperature change rate is less than the preset temperature stability threshold, it means that the temperature has stabilized and it is considered that the thermal state of the system has reached equilibrium. In the process of calculating the temperature change rate, the temperature data is processed by numerical differentiation or sliding window method to avoid the influence of noise interference on the stability judgment, thereby improving the accuracy of the calculation. The iterative change value of the objective function and the temperature stability flag are combined to generate a convergence judgment result. After the convergence judgment result shows that the optimization process has reached a stable state, the initial power allocation strategy is modified based on the final optimization result to obtain the final target power allocation strategy, and the strategy is decomposed to generate the power control instruction set for each energy storage unit. The power control instruction set is used to guide the energy storage converter to perform specific power allocation control on each energy storage unit to ensure that the system performs energy scheduling according to the optimized strategy. In the process of generating power control instructions, the voltage, current and charge and discharge status of the energy storage unit are combined to ensure that the generated power instructions meet the operating limitations of the equipment and can effectively improve the overall efficiency of the energy storage system.

[0019] In the embodiment of the present invention, by constructing a temperature constraint projection mapping function, the thermal management constraint is directly integrated into the optimization process, avoiding the convergence problem caused by the use of penalty functions in traditional methods, and ensuring that the energy storage unit always operates within a safe temperature range. The optimization method without model feedback is adopted, and the optimal power distribution can be achieved only by relying on the real-time feedback data of the system, which overcomes the shortcomings of the traditional model-based method that is highly dependent on system parameters and improves the applicability and robustness of the algorithm. The introduction of thermal coupling model and temperature-sensitive gradient vector realizes the coordinated optimization of power distribution and temperature control, effectively solves the thermal interference problem between energy storage units, and significantly improves the overall operation efficiency of the system. A gradient estimation method based on random perturbation sampling is designed to avoid complex derivative calculation processes, reduce the computational complexity of the algorithm, and make it more suitable for real-time control applications. The dynamic weight adjustment mechanism is adopted, combined with the temperature stability criterion, to ensure the convergence of the optimization process and the stability of temperature control, and improve the reliability and safety of the system. Through the iterative optimization of the power optimization objective function, a significant reduction in system loss is achieved.

[0020] In a specific embodiment, the process of executing step S1 may specifically include the following steps: The voltage sampling channel and the current sampling channel of each energy storage unit in the liquid-cooled string PCS energy storage converter are synchronously sampled to obtain an output voltage sampling sequence and an output current sampling sequence; Performing digital filtering on the output voltage sampling sequence and the output current sampling sequence to obtain a filtered voltage value sequence and a filtered current value sequence; Collect temperature sensor data on the liquid cooling pipeline of each energy storage unit, perform differential calculation on the coolant inlet temperature and outlet temperature, and obtain a temperature gradient sequence; Based on the coolant flow data and the temperature gradient sequence, the heat transfer calculation is performed to obtain the heat dissipation power of each energy storage unit; The filtered voltage value sequence, the filtered current value sequence and the heat dissipation power are concatenated in a matrix to construct an initial state matrix, and the initial state matrix is ​​normalized and timestamp information is added to obtain a multi-dimensional state matrix.

[0021] Specifically, the voltage sampling channel and current sampling channel of each energy storage unit are sampled synchronously to ensure that the collected data is consistent on the time axis, thereby avoiding calculation errors caused by asynchronous sampling timing. A high-precision ADC (analog-to-digital converter) is used to synchronously sample the voltage and current signals of each energy storage unit, and the sampling frequency is set to To ensure sufficient time resolution, the obtained voltage sampling sequence is expressed as: in, For the first energy storage unit in The voltage value of the sampling point, is the actual voltage signal, is the sampling period, is the sampling point index. Similarly, the current sampling sequence is: in, is the current sampling value of the first energy storage unit. The voltage and current sampling sequences are digitally filtered to eliminate errors caused by noise interference or high-frequency signals during the sampling process and improve the smoothness and accuracy of the data. In the filtering process, a low-pass filter is used, such as a discrete-time finite impulse response (FIR) filter, and the output of the filtered voltage value sequence is expressed as: in, is the voltage value after filtering, are the coefficients of the filter, is the order of the filter. Similarly, the current signal is filtered to obtain the filtered current value sequence: After completing the digital filtering of the voltage and current signals, collect the temperature sensor data on the liquid cooling pipeline to obtain the inlet and outlet temperatures of the coolant and calculate the temperature gradient. Install a temperature sensor at the coolant flow point of each energy storage unit to collect the inlet temperature. and outlet temperature , the temperature gradient is calculated as follows: in, Indicates the temperature change of the first energy storage unit. This value directly reflects the amount of heat absorbed by the coolant when it flows through the energy storage unit. After obtaining the temperature gradient, the heat dissipation power of each energy storage unit is calculated in combination with the coolant flow data. Assume that the volume flow rate of the coolant is , the specific heat capacity is , the density is , then the mass flow rate is expressed as: The calculation formula for heat dissipation power is as follows: in, is the heat dissipation power of the first energy storage unit, which represents the heat taken away by the coolant per unit time. The heat dissipation power of each energy storage unit is obtained through this calculation formula. The filtered voltage value sequence, the filtered current value sequence and the heat dissipation power are concatenated in a matrix to construct the initial state matrix. The state data of each energy storage unit is set to include three parts: voltage, current and heat dissipation power, then the initial state matrix is ​​expressed as: Each row of the matrix represents the state data of an energy storage unit. The initial state matrix is ​​normalized to eliminate the dimensional effects between different physical quantities and improve the consistency of the data and the stability of the calculation. The normalization process uses the minimum-maximum normalization or Z-score normalization method, for example, the minimum-maximum normalization: in, is the normalized state value, is the original state value, and are the minimum and maximum values ​​of the matrix respectively. Timestamp information is added to the normalized multidimensional state matrix so that it can be tracked and analyzed in the time dimension. The final multidimensional state matrix is ​​as follows: in, Represents the timestamp corresponding to each data point, making the entire dataset useful for time series analysis.

[0022] In a specific embodiment, the process of executing step S2 may specifically include the following steps: Perform time series analysis on the temperature parameters in the multi-dimensional state matrix and calculate the temperature conductivity and heat capacity coefficient between adjacent energy storage units; Based on the temperature conductivity coefficient and the heat capacity coefficient, the thermal network topology of the energy storage unit is established to obtain the thermal coupling relationship matrix; Construct the temperature state space equation according to the thermal coupling relationship matrix and set the temperature safe operation range; The temperature state space equation is transformed into an inequality constraint group to obtain the temperature constraint boundary conditions; The temperature constraint boundary conditions are transformed into convex sets, a projection operator is constructed, and a temperature constraint projection mapping function is constructed based on the projection operator and the thermal coupling model.

[0023] Specifically, the temperature data of the energy storage unit is analyzed in time series to calculate the temperature conductivity and heat capacity coefficient between adjacent energy storage units. The temperature is By analyzing its temperature change trend, the temperature conduction coefficient is calculated using the heat conduction equation and heat capacity coefficient According to Fourier's law of heat conduction, the energy storage unit per unit time With adjacent units The heat transfer between is expressed as: in, is the total energy of heat transfer, is the effective heat transfer area of ​​the energy storage unit, and are the temperatures of adjacent cells, is the sampling time interval, and Represents the thermal conductivity, which determines the efficiency of heat transfer between energy storage units. The heat capacity of an energy storage unit is determined by the specific heat capacity of its material. and quality The calculation results are: in, Represents the total heat capacity of the energy storage unit and determines how quickly its temperature responds to thermal disturbances. By performing regression analysis on data from multiple time steps, the heat capacity of each energy storage unit is estimated based on the measured temperature data. and The thermal network topology of the energy storage unit is established based on the temperature conductivity coefficient and heat capacity coefficient of the energy storage unit to form the overall thermal coupling relationship matrix. energy storage units, each unit interacts with the adjacent unit through heat transfer, then define a Thermal coupling matrix : where each off-diagonal element Represents energy storage unit The temperature conduction coefficient between each unit and its adjacent units, while the diagonal elements represent the inverse of the sum of the heat transfer from each unit to all adjacent units. This matrix describes the thermal coupling relationship in the entire energy storage system and is used for subsequent temperature state modeling. Based on the constructed thermal coupling relationship matrix, a temperature state space equation is established to describe the temperature change of each energy storage unit and set the temperature safety operating range. The dynamic temperature change of the energy storage unit is described by the thermal balance equation, that is, the temperature change per unit time is determined by the input power, heat transfer loss and heat dissipation process, and its state equation is expressed as: in, represents the input power, and Represents the temperature change term caused by thermal coupling. In practical applications, this equation is simulated and calculated by numerical solution to predict the temperature change trend of the energy storage unit at any time point. In order to ensure that the system operates within a safe temperature range, a temperature safety operating range is set, that is, the maximum allowable temperature is defined. and the minimum allowable temperature , thereby constraining the temperature variation range of the system: In order to apply the temperature state space equation to the optimization calculation, it is converted into an inequality constraint group to obtain the temperature constraint boundary conditions. The state equation is discretized and converted into the constraint conditions of the optimization problem. The discretized temperature equation is expressed as: Combined with the temperature safety interval, the temperature inequality constraint group is obtained: The temperature constraint boundary conditions are transformed by convex set mapping to construct a projection operator, and a temperature constraint projection mapping function is constructed based on the projection operator and the thermal coupling model. Define the projection operator , which is used to project any power allocation scheme that does not meet the temperature constraint into the feasible solution space. The definition of the projection operator is as follows: in, is the projected temperature value, and the goal of this optimization problem is to find the closest By applying the projection operator to the thermal coupling model, the temperature constraint projection mapping function is obtained: in, For the final optimized temperature distribution, is the adjustment coefficient, which is used to control the step size of temperature adjustment. This mapping function can ensure that the power allocation always meets the temperature safety constraint during the optimization calculation process, and takes into account the thermal coupling effect between energy storage units.

[0024] In a specific embodiment, the process of performing a convex set mapping transformation on the temperature constraint boundary condition, constructing a projection operator, and constructing a temperature constraint projection mapping function based on the projection operator and the thermal coupling model may specifically include the following steps: Perform convex set decomposition operation on the temperature constraint boundary conditions to divide the temperature constraint space into multiple non-intersecting convex subsets; The objective function of the quadratic programming problem is constructed based on convex subsets, and the temperature constraint boundary conditions are transformed into constraint matrix form; Perform singular value decomposition on the constraint matrix to obtain an orthogonal basis matrix and an eigenvalue matrix; A projection space transformation matrix is ​​constructed based on the orthogonal basis matrix and the eigenvalue matrix, and a tensor product operation is performed on the projection space transformation matrix and the thermal coupling model to obtain a tensor product operation result; The singular value threshold processing is performed on the tensor product operation result to obtain the temperature constraint projection kernel matrix, and the temperature constraint projection kernel matrix is ​​convolved with the thermal coupling model to obtain the temperature constraint projection mapping function.

[0025] Specifically, the temperature constraint space is analyzed and divided into multiple non-intersecting convex subsets to facilitate efficient solution of subsequent optimization calculations. Suppose the entire temperature constraint space is , which contains all possible temperature states The following inequality constraints are satisfied: in, and They represent the lower and upper limits of the temperature respectively. Since the temperature state is affected by the thermal coupling relationship, there are non-convex areas, which are divided into multiple non-intersecting convex subsets , so that the temperature distribution in each convex subset is solved by the convex optimization method. Through the set decomposition algorithm, the original temperature constraint region is decomposed into multiple sub-regions, each of which satisfies the convexity requirement, that is, for any , both have: After completing the convex set decomposition of the temperature constraint space, the objective function of the quadratic programming problem is constructed based on these convex subsets, and the temperature constraint boundary conditions are converted into constraint matrix form. The objective function is defined as minimizing the sum of squares of temperature deviations from its ideal distribution while ensuring that the constraints of the thermal coupling model are met. The objective function is expressed as: in, For ideal temperature distribution, the optimization variables The thermal coupling constraint needs to be satisfied. In order to convert the temperature constraint into a matrix form, the thermal coupling relationship matrix is ​​introduced. , we get the following linear constraint equation: in, Represents the input power distribution. Combined with the constraint boundaries, the complete constraint matrix is ​​defined and the right-hand vector , so that the optimization problem is expressed in standard quadratic programming form: subject to in, is a positive definite matrix, representing the quadratic term in the objective function, is the coefficient of the first-order term, and Describe the temperature constraints together. Perform singular value decomposition on the constraint matrix to obtain the orthogonal basis matrix and eigenvalue matrix. Assume that the constraint matrix The size is , decomposed as follows: in, and are orthogonal matrices, and is a diagonal matrix whose diagonal elements are singular values. Through this decomposition, the original temperature constraint problem is projected into a low-dimensional feature space to reduce the computational complexity. Based on the orthogonal basis matrix and eigenvalue matrix obtained by singular value decomposition, a projection space transformation matrix is ​​constructed, and the transformation matrix is ​​tensor-producted with the thermal coupling model. Let the transformation matrix be , then its construction is as follows: This matrix is ​​used to map the original temperature variables into a low-dimensional space, making the optimization calculation more efficient. Perform tensor product operations to integrate thermal coupling information and obtain tensor product operation results : in, Represents the tensor product operation, which is used to fuse information from different spaces so that the optimization problem can be calculated in the reduced-dimensional space. The result of the tensor product operation is subjected to singular value threshold processing to obtain the temperature constraint projection kernel matrix. The purpose of singular value threshold processing is to remove eigenvalues ​​with less influence, thereby reducing the computational complexity and retaining the main temperature influence mode. Set a threshold , and prune the singular value matrix: in, Represents the processed singular value matrix. The temperature constraint projection kernel matrix Defined as: The temperature constraint projection kernel matrix is ​​convolved with the thermal coupling model to obtain the temperature constraint projection mapping function. Let the convolution kernel be , then the convolution operation is expressed as: in, For the final optimized temperature distribution, the mapping function is used to ensure that the temperature state always meets the constraints during the optimization process and optimize the power distribution under the action of thermal coupling relationship.

[0026] In a specific embodiment, the process of executing step S3 may specifically include the following steps: Calculate the conduction loss and switching loss of each energy storage unit based on the filtered voltage value sequence and the filtered current value sequence to obtain the total power loss; Taylor expansion is performed on the temperature constraint projection mapping function to obtain a locally linearized temperature constraint function, and the total power loss is combined with the locally linearized temperature constraint function to construct a constrained loss function; The penalty factor is set according to the constrained loss function to obtain the augmented Lagrangian function; The augmented Lagrangian function is varied to obtain the optimality condition of power allocation, and the power optimization objective function is constructed based on the optimality condition.

[0027] Specifically, the conduction loss and switching loss of each energy storage unit are calculated to obtain the total power loss. The conduction loss mainly comes from the equivalent on-resistance of the power semiconductor device in the on state. and the current flowing through the device , and its loss is expressed as: in, is the conduction loss, is the filtered current value sequence. The switching loss comes from the voltage and current overlap of the power device at the moment of opening and closing, resulting in instantaneous energy loss. Assume the switching frequency is , the device’s turn-on time and turn-off time are and , then the switching loss is expressed as: in, is the voltage value sequence after filtering. Considering the conduction loss and switching loss, the total power loss is expressed as: in, is the total loss of each energy storage unit. Taylor expansion is performed on the temperature constraint projection mapping function to obtain a local linearized temperature constraint function, thereby simplifying the optimization calculation and improving the solution efficiency. The Taylor expansion method is used to approximate it in a certain local area so that the temperature constraint can be expressed as a linear combination of power allocation variables. Let the temperature value corresponding to the current power allocation state be , then in Nearby Perform a first-order Taylor expansion: in, is the gradient of the temperature constraint mapping function. After local linearization, the total power loss is Combined with the linearized temperature constraint function, a constrained loss function is constructed: in, is a constrained loss function, is the adjustment coefficient used to weigh the impact of power loss and temperature constraints. After constructing the constrained loss function, the penalty factor is introduced according to the characteristics of the optimization problem to construct the augmented Lagrangian function. The augmented Lagrangian function is an optimization method that can internalize the constraints into the objective function. By introducing the Lagrangian multiplier and the quadratic penalty term, the optimization problem can be solved in the unconstrained optimization framework. Assume is the Lagrange multiplier, is the penalty factor, then the augmented Lagrangian function is defined as: The second term is the standard Lagrangian term, and the third term is an additional quadratic penalty term to ensure that the temperature constraint is not violated excessively. In order to solve the optimal power allocation strategy, the augmented Lagrangian function is varied to derive the optimality condition of power allocation. The core idea of ​​variational calculation is to optimize the variables. Take the derivative and analyze its stationary point condition. For the optimization variable, i.e., the power allocation vector of each energy storage unit, the optimality condition is given by the following equation: After expanding the calculation, we can get: This equation describes the optimal condition for power allocation, that is, how the power allocation should be adjusted to minimize the loss under a given temperature constraint. Based on this optimality condition, the final power optimization objective function is constructed. The objective function needs to minimize the power loss while satisfying the temperature constraint, so it is expressed as: In a specific embodiment, the process of executing step S4 may specifically include the following steps: Generate a uniformly distributed random disturbance vector based on the power optimization objective function, perform disturbance sampling on the initial power allocation strategy of the energy storage unit, and obtain the power allocation after disturbance; Substitute the power distribution after the disturbance into the thermal coupling model to calculate the temperature response caused by the disturbance; Perform differential operation on the temperature response and the initial temperature to obtain the temperature change, and construct a temperature sensitivity matrix based on the temperature change and the random disturbance vector; The temperature sensitivity matrix is ​​subjected to singular value decomposition to extract the principal eigenvector, and the temperature sensitive gradient vector is constructed based on the principal eigenvector.

[0028] Specifically, the initial power allocation strategy of the energy storage unit is perturbed and sampled to simulate the impact of different power allocation schemes on temperature distribution. Assume that the initial power allocation scheme is expressed as ,in is a vector containing the power distribution of all energy storage units, namely: in, Representative The initial power allocation value of each energy storage unit. In order to perform disturbance sampling, a uniformly distributed random disturbance vector is generated , each element of which obeys a uniform distribution within a certain range: in, is the disturbance range, which is used to control the disturbance intensity. The disturbance vector is superimposed on the initial power allocation vector to obtain the power allocation vector after disturbance: in, represents the power allocation scheme after the disturbance. In this way, multiple different power allocation schemes are simulated and their effects on the temperature of the energy storage unit are analyzed. The power allocation scheme after the disturbance is substituted into the thermal coupling model to calculate the temperature response caused by the disturbance. The temperature change of the energy storage unit is affected by the power input. Due to the influence of thermal coupling effect, the thermal balance equation is used to describe the dynamic change of temperature. is the temperature vector before the disturbance, is the temperature vector after disturbance, the thermal coupling model is expressed as: in, is the thermal capacity matrix of the energy storage unit, is the thermal coupling coefficient matrix, which describes the heat transfer relationship between adjacent energy storage units. The equation is solved numerically to obtain the steady-state temperature after disturbance. The temperature response after the disturbance is differentiated from the initial temperature to calculate the temperature change : in, Represents the temperature change caused by power disturbance. In order to analyze the sensitivity of power disturbance to temperature, a temperature sensitivity matrix is ​​constructed , which is defined as: in, Each element of Energy storage unit The power change of the energy storage unit The influence of temperature. This matrix describes the sensitivity of power distribution to temperature changes. The temperature sensitivity matrix is ​​subjected to singular value decomposition to extract the main eigenvector and construct the temperature sensitive gradient vector based on the eigenvector. Singular value decomposition is a matrix decomposition method that converts the temperature sensitivity matrix This breaks down into: in, and are the left and right orthogonal basis matrices, It is a diagonal matrix whose diagonal elements are singular values. The singular values ​​represent the main direction of change of the matrix, so the main eigenvector corresponding to the largest singular value is extracted , to characterize the main impact direction of temperature on power distribution. By selecting the first few main eigenvectors and combining their corresponding singular values, the temperature sensitive gradient vector is constructed: in, is the number of main singular values ​​selected, is the corresponding eigenvector, is the corresponding singular value. The temperature-sensitive gradient vector is used to characterize the main influence direction of power distribution on temperature change. Therefore, in the subsequent optimization process, this vector is used to guide the power adjustment strategy to minimize power loss while keeping the temperature of the energy storage unit within a safe range.

[0029] In a specific embodiment, the process of executing step S5 may specifically include the following steps: Performing gradient descent calculation on the power allocation weight based on the temperature sensitive gradient vector to obtain an updated power allocation weight; Substitute the updated power allocation weight into the power optimization objective function and calculate the iterative change value of the objective function; The temperature change rate is calculated according to the thermal coupling model, and when the norm of the temperature change rate is less than the temperature stability threshold, a temperature stability mark is obtained; Combined judgment is performed on the iterative change value of the objective function and the temperature stability mark to generate a convergence judgment result; The initial power allocation strategy is modified based on the convergence judgment result to obtain the target power allocation strategy, and the target power allocation strategy is decomposed and calculated to obtain the power control instruction set of each energy storage unit.

[0030] Specifically, the temperature-sensitive gradient vector is used to perform gradient descent calculation on the power allocation weight to obtain the updated power allocation weight. Assume that the current power allocation weight is , the temperature sensitive gradient vector is , the learning rate is , then the updated power allocation weight is expressed by the gradient descent formula as: in, Assign weights to the updated power, Control the optimization step size to ensure that the optimization process does not experience violent fluctuations or slow convergence. The role of the gradient descent method is to gradually adjust the power allocation strategy so that it can optimize the temperature distribution and thus improve the overall operating efficiency of the energy storage system. Substitute the updated power allocation weight into the power optimization objective function to calculate the iterative change value of the objective function. Suppose the optimization objective function is , then in The objective function value during round iteration is expressed as: in, Represents the iterative change value of the optimization objective function, which is used to determine whether the optimization process tends to converge. It gradually decreases, indicating that the power allocation strategy is converging to the optimal solution, otherwise the power allocation weight needs to be further adjusted. At the same time, in order to ensure that the optimized power allocation strategy can maintain temperature stability while satisfying power loss optimization, the temperature change rate is calculated based on the thermal coupling model, and it is determined whether the temperature enters a stable state. Assume that the temperature vector of the current energy storage unit is , the time step is , then the temperature change rate is expressed as: in, represents the temperature change rate vector, is the updated temperature value, is the temperature value of the previous time step. In order to determine whether the temperature is stable, calculate the norm of the temperature change rate: if Less than the preset temperature stability threshold , the system is considered to have reached a thermally stable state, and the temperature stability flag is defined at this time: Among them, when When , it indicates that the temperature change of the system has stabilized, and further optimization calculation stops. When , the power allocation strategy is continuously optimized to ensure that the temperature distribution reaches a steady state. and temperature stable mark Perform combined judgment to generate convergence judgment results. Less than a certain threshold And the temperature is stable , the optimization process is considered to have converged and the optimization calculation is terminated. Otherwise, the power allocation optimization needs to be continued. The mathematical expression for convergence judgment is as follows: in, Indicates that the optimization has converged and outputs the final target power allocation strategy, while Indicates that the optimization has not converged yet and the power allocation strategy needs to be adjusted. When the power allocation weight is obtained based on the final optimization Modify the initial power allocation strategy to obtain the target power allocation strategy : in, Represents element-by-element multiplication, indicating the update of the power allocation scheme to satisfy the optimized temperature constraint. Perform decomposition operations to generate the power control instruction set of each energy storage unit. Suppose the power control instruction set is , then its calculation method is as follows: Among them, round Indicates rounding operation to ensure that the power allocation instruction can match the control accuracy of the actual device.

[0031] The above describes the power distribution method of the liquid-cooled string PCS energy storage converter in the embodiment of the present invention. The following describes the power distribution system of the liquid-cooled string PCS energy storage converter in the embodiment of the present invention. Figure 2 , an embodiment of the power distribution system of the liquid-cooled string PCS energy storage converter in the embodiment of the present invention includes: The acquisition module is used to collect the output voltage, output current, coolant inlet temperature, outlet temperature and flow data of each energy storage unit in the liquid-cooled string PCS energy storage converter to construct a multi-dimensional state matrix; Establishing a module for establishing a thermal coupling model between energy storage units according to the temperature parameters in the multidimensional state matrix, and constructing a temperature constraint projection mapping function based on the thermal coupling model; A construction module is used to combine the temperature constraint projection mapping function with the power loss of each energy storage unit to construct a power optimization objective function considering the thermal coupling effect; A calculation module is used to perform random perturbation sampling on the initial power allocation strategy of the energy storage unit, and calculate the temperature change caused by the perturbation through a thermal coupling model to obtain a temperature sensitive gradient vector; The output module is used to dynamically adjust the power allocation weight of each energy storage unit based on the temperature sensitive gradient vector, and output the target power allocation strategy when the iterative change value of the power optimization objective function is less than a preset threshold and the thermal coupling model is stable.

[0032] Through the collaboration of the above components, the thermal management constraints are directly integrated into the optimization process by constructing the temperature constraint projection mapping function, avoiding the convergence problem caused by the use of penalty functions in traditional methods, while ensuring that the energy storage unit always operates within a safe temperature range. The optimization method without model feedback is adopted, and the optimal power allocation can be achieved only by relying on the real-time feedback data of the system, which overcomes the shortcomings of the traditional model-based method that is highly dependent on system parameters and improves the applicability and robustness of the algorithm. The introduction of thermal coupling model and temperature-sensitive gradient vector realizes the coordinated optimization of power allocation and temperature control, effectively solves the thermal interference problem between energy storage units, and significantly improves the overall operation efficiency of the system. The gradient estimation method based on random perturbation sampling is designed to avoid the complex derivative calculation process, reduce the computational complexity of the algorithm, and make it more suitable for real-time control applications. The dynamic weight adjustment mechanism is adopted, combined with the temperature stability criterion, to ensure the convergence of the optimization process and the stability of temperature control, and improve the reliability and safety of the system. Through the iterative optimization of the power optimization objective function, the system loss is significantly reduced.

[0033] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0034] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0035] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A power distribution method for a liquid-cooled string PCS energy storage converter, characterized in that: The method comprises: Collect the output voltage, output current, coolant inlet temperature, outlet temperature and flow data of each energy storage unit in the liquid-cooled string PCS energy storage converter to build a multi-dimensional state matrix; Establishing a thermal coupling model between the energy storage units according to the temperature parameters in the multidimensional state matrix, and constructing a temperature constraint projection mapping function based on the thermal coupling model; The temperature constraint projection mapping function is combined with the power loss of each energy storage unit to construct a power optimization objective function considering the thermal coupling effect; Performing random perturbation sampling on the initial power allocation strategy of the energy storage unit, and calculating the temperature change caused by the perturbation through the thermal coupling model to obtain a temperature sensitive gradient vector; The power allocation weight of each energy storage unit is dynamically adjusted based on the temperature sensitive gradient vector, and when the iterative change value of the power optimization objective function is less than a preset threshold and the thermal coupling model is stable, a target power allocation strategy is output.

2. The power distribution method of the liquid-cooled string PCS energy storage converter according to claim 1 is characterized in that: The output voltage, output current, coolant inlet temperature, outlet temperature and flow rate data of each energy storage unit in the liquid-cooled string PCS energy storage converter are collected to construct a multi-dimensional state matrix, including: The voltage sampling channel and the current sampling channel of each energy storage unit in the liquid-cooled string PCS energy storage converter are synchronously sampled to obtain an output voltage sampling sequence and an output current sampling sequence; Performing digital filtering on the output voltage sampling sequence and the output current sampling sequence to obtain a filtered voltage value sequence and a filtered current value sequence; Collect temperature sensor data on the liquid cooling pipeline of each energy storage unit, perform differential calculation on the coolant inlet temperature and outlet temperature, and obtain a temperature gradient sequence; Perform heat transfer calculation based on the coolant flow data and the temperature gradient sequence to obtain the heat dissipation power of each energy storage unit; The filtered voltage value sequence, the filtered current value sequence and the heat dissipation power are concatenated in a matrix to construct an initial state matrix, and the initial state matrix is ​​normalized and timestamp information is added to obtain a multidimensional state matrix.

3. The power distribution method of the liquid-cooled string PCS energy storage converter according to claim 1 is characterized in that: The step of establishing a thermal coupling model between energy storage units according to the temperature parameters in the multidimensional state matrix, and constructing a temperature constraint projection mapping function based on the thermal coupling model, comprises: Performing time series analysis on the temperature parameters in the multidimensional state matrix to calculate the temperature conduction coefficient and heat capacity coefficient between adjacent energy storage units; Based on the temperature conductivity coefficient and the heat capacity coefficient, a thermal network topology structure of an energy storage unit is established to obtain a thermal coupling relationship matrix; Constructing a temperature state space equation according to the thermal coupling relationship matrix and setting a temperature safety operation range; Converting the temperature state space equation into an inequality constraint group to obtain a temperature constraint boundary condition; The temperature constraint boundary condition is subjected to convex set mapping transformation to construct a projection operator, and a temperature constraint projection mapping function is constructed based on the projection operator and the thermal coupling model.

4. The power distribution method of the liquid-cooled string PCS energy storage converter according to claim 3 is characterized in that: The step of performing a convex set mapping transformation on the temperature constraint boundary condition, constructing a projection operator, and constructing a temperature constraint projection mapping function based on the projection operator and the thermal coupling model includes: Performing a convex set decomposition operation on the temperature constraint boundary condition to divide the temperature constraint space into a plurality of mutually non-intersecting convex subsets; Constructing an objective function of a quadratic programming problem based on the convex subset, and converting the temperature constraint boundary condition into a constraint matrix form; Performing a singular value decomposition operation on the constraint matrix to obtain an orthogonal basis matrix and an eigenvalue matrix; Constructing a projection space conversion matrix based on the orthogonal basis matrix and the eigenvalue matrix, and performing a tensor product operation on the projection space conversion matrix and the thermal coupling model to obtain a tensor product operation result; The tensor product operation result is subjected to singular value threshold processing to obtain a temperature constraint projection kernel matrix, and the temperature constraint projection kernel matrix is ​​subjected to convolution operation with the thermal coupling model to obtain a temperature constraint projection mapping function.

5. The power distribution method of the liquid-cooled string PCS energy storage converter according to claim 2 is characterized in that: The temperature constraint projection mapping function is combined with the power loss of each energy storage unit to construct a power optimization objective function considering the thermal coupling effect, including: Calculating the conduction loss and switching loss of each energy storage unit based on the filtered voltage value sequence and the filtered current value sequence to obtain a total power loss; Performing Taylor expansion on the temperature constraint projection mapping function to obtain a local linearized temperature constraint function, and combining the total power loss with the local linearized temperature constraint function to construct a constrained loss function; Setting a penalty factor according to the constrained loss function to obtain an augmented Lagrangian function; Variation calculation is performed on the augmented Lagrangian function to obtain optimality conditions for power allocation, and a power optimization objective function is constructed based on the optimality conditions.

6. The power distribution method of the liquid-cooled string PCS energy storage converter according to claim 1, characterized in that: The random perturbation sampling is performed on the initial power allocation strategy of the energy storage unit, and the temperature change caused by the perturbation is calculated by the thermal coupling model to obtain the temperature sensitive gradient vector, including: Generate a uniformly distributed random disturbance vector based on the power optimization objective function, perform disturbance sampling on the initial power allocation strategy of the energy storage unit, and obtain the power allocation after disturbance; Substituting the power distribution after the disturbance into the thermal coupling model to calculate the temperature response caused by the disturbance; Performing a differential operation on the temperature response and the initial temperature to obtain a temperature change, and constructing a temperature sensitivity matrix according to the temperature change and the random disturbance vector; A singular value decomposition operation is performed on the temperature sensitivity matrix to extract a main eigenvector, and a temperature sensitive gradient vector is constructed based on the main eigenvector.

7. The power distribution method of the liquid-cooled string PCS energy storage converter according to claim 1, characterized in that: The power allocation weight of each energy storage unit is dynamically adjusted based on the temperature sensitive gradient vector, and when the iterative change value of the power optimization objective function is less than a preset threshold and the thermal coupling model is stable, a target power allocation strategy is output, including: Performing gradient descent calculation on the power allocation weight based on the temperature sensitive gradient vector to obtain an updated power allocation weight; Substituting the updated power allocation weight into the power optimization objective function to calculate an iterative change value of the objective function; Calculating the temperature change rate according to the thermal coupling model, and obtaining a temperature stability mark when the norm of the temperature change rate is less than a temperature stability threshold; Performing a combined judgment on the iterative change value of the objective function and the temperature stability mark to generate a convergence judgment result; The initial power allocation strategy is modified based on the convergence judgment result to obtain a target power allocation strategy, and the target power allocation strategy is decomposed and calculated to obtain a power control instruction set of each energy storage unit.

8. A power distribution system for a liquid-cooled string PCS energy storage converter, characterized in that: A power distribution method for a liquid-cooled string-type PCS energy storage converter according to any one of claims 1 to 7 is used, wherein the power distribution system of the liquid-cooled string-type PCS energy storage converter comprises: The acquisition module is used to collect the output voltage, output current, coolant inlet temperature, outlet temperature and flow data of each energy storage unit in the liquid-cooled string PCS energy storage converter to construct a multi-dimensional state matrix; An establishment module is used to establish a thermal coupling model between each energy storage unit according to the temperature parameters in the multidimensional state matrix, and to construct a temperature constraint projection mapping function based on the thermal coupling model; A construction module is used to combine the temperature constraint projection mapping function with the power loss of each energy storage unit to construct a power optimization objective function considering thermal coupling effect; A calculation module, used for performing random perturbation sampling on the initial power allocation strategy of the energy storage unit, and calculating the temperature change caused by the perturbation through the thermal coupling model to obtain a temperature sensitive gradient vector; An output module is used to dynamically adjust the power allocation weight of each energy storage unit based on the temperature sensitive gradient vector, and output a target power allocation strategy when the iterative change value of the power optimization objective function is less than a preset threshold and the thermal coupling model is stable.

Images