Health-aware adaptive control method and system for multi-energy storage converter coordination

Abstract

The application discloses a health-aware adaptive control method and system for multiple energy storage converter cooperation, and relates to the technical field of adaptive regulation; the method comprises the following steps: feature extraction is performed based on N energy storage converter operation state data and operation environment data at M time points to obtain a corresponding comprehensive feature set; the comprehensive feature set is input into a state diagnosis model to obtain an energy storage converter state label; a non-directional graph is constructed with the energy storage converter as a node and an electrical connection as an edge, and a spectral clustering algorithm is adopted to divide into G subgraphs based on topological connection tightness and state similarity; topological characteristic parameter sets of the subgraphs are extracted; each subgraph is traversed, and it is judged whether to end regulation according to a state balance threshold value; if the regulation is not ended, an unbalanced scene is identified according to the topological characteristic parameters, and adaptive control is performed for different scenes; the application realizes health-aware adaptive power distribution for multiple energy storage converter cooperation, avoids the coexistence of overload and idling, and suppresses ring current oscillation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of adaptive adjustment, and more particularly, to a health-aware adaptive control method and system for multi-energy storage converter coordination. BACKGROUND

[0002] As a key interface device of the energy storage system, the energy storage converter undertakes the control task of bidirectional energy flow. To meet the power scheduling requirements in multiple scenarios, multiple energy storage converters are often used in parallel to expand the system capacity and flexibility. In such applications, how to ensure the balanced power distribution among multiple energy storage converters through coordinated control, while taking into account the operation safety and life management, has become an important research direction in this field.

[0003] However, most of the existing multi-machine parallel control methods are based on droop control or virtual impedance current sharing, relying only on electrical quantities for distribution, without considering the temperature characteristics, equivalent internal resistance changes, state of charge levels and overall health state differences of each energy storage converter. This control method is prone to power distribution imbalance in actual application, with some energy storage converters running at high load for a long time, leading to high node temperature and accelerated aging, while the other energy storage converters are in low load or even idle state, resulting in insufficient resource utilization. At the same time, uneven power distribution among energy storage converters can also cause circulating current between parallel units, further inducing power oscillation and power quality decline, affecting the safety and stability of the system.

[0004] Therefore, there is an urgent need for a new method that can consider multiple dimensions of operating characteristics in a multi-machine parallel system and achieve adaptive power distribution with health awareness, to avoid the coexistence of overload and idling, and improve the overall stability and life utilization rate of the system. In view of this, the present application proposes a health-aware adaptive control method and system for multi-energy storage converter coordination to solve the above problems. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art and achieve the above-mentioned purposes, the present application provides the following technical solutions: a health-aware adaptive control method for multi-energy storage converter coordination, comprising:

[0006] Based on the operating state data and operating environment data of N energy storage converters at M time points per unit time, feature extraction is performed to obtain a comprehensive feature set corresponding to the N energy storage converters;

[0007] The comprehensive feature set corresponding to the N energy storage converters is input into a state diagnosis model respectively to obtain energy storage converter state labels corresponding to the N energy storage converters; the energy storage converter state labels include hot engine overload state, normal load state and cold engine idle state;

[0008] A power storage converter undirected graph is constructed by taking N power storage converters as nodes and electrical connections between power storage converters as edges; a spectral clustering algorithm is used on the power storage converter undirected graph to divide into G power storage converter subgraphs based on dual criteria of topological connection tightness and state similarity;

[0009] A feature extraction operation is performed on the G power storage converter subgraphs to obtain a topological characteristic parameter set corresponding to the power storage converter subgraphs; the topological characteristic parameter set includes thermal engine density, cold engine utilization rate, and state balance degree;

[0010] Each power storage converter subgraph is traversed to determine whether the state balance degree in the topological characteristic parameter set is greater than or equal to a state balance degree threshold; if the determination result is yes, the adjustment is ended; if the determination result is no, a hot-cold state imbalance scenario including only thermal engine overload, only cold engine idling, and coexistence of thermal engine overload and cold engine idling is identified based on the topological characteristic parameters, and corresponding adaptive control is performed for different hot-cold state imbalance scenarios.

[0011] Further, the method of performing corresponding adaptive control for different hot-cold state imbalance scenarios includes:

[0012] S100: Let the initial value of g be 1, and the value range of g be 1 to G; g is a loop index variable;

[0013] S101: Obtain the topological characteristic parameter set corresponding to the gth power storage converter subgraph; obtain the state balance degree from the topological characteristic parameter set, and if the state balance degree is greater than or equal to a preset state balance degree threshold, execute S104; if the state balance degree is less than the preset state balance degree threshold, execute S102;

[0014] S102: Based on the thermal engine density and cold engine utilization rate in the topological characteristic parameter set, compare and analyze with the preset thermal engine density threshold and cold engine utilization rate threshold to identify a hot-cold state imbalance scenario including only thermal engine overload, only cold engine idling, and coexistence of thermal engine overload and cold engine idling;

[0015] S103: If the hot-cold state imbalance scenario is only thermal engine overload, perform adaptive control for only thermal engine overload; if the hot-cold state imbalance scenario is only cold engine idling, perform adaptive control for only cold engine idling; if the hot-cold state imbalance scenario is coexistence of thermal engine overload and cold engine idling, perform adaptive control for coexistence of thermal engine overload and cold engine idling;

[0016] S104: Let g = g + 1, if g is less than or equal to G, return to S101 for execution, if g is greater than G, end the adaptive control process.

[0017] Further, the identification method of the hot-cold state imbalance scenario includes:

[0018] if the hot engine density ≥ the hot engine density threshold value and the cold engine utilization rate ≥ the cold engine utilization rate threshold value, the hot-cold imbalance scenario is determined to be only hot engine overload;

[0019] if the hot engine density < the hot engine density threshold value and the cold engine utilization rate < the cold engine utilization rate threshold value, the hot-cold imbalance scenario is determined to be only cold engine idling;

[0020] if the hot engine density ≥ the hot engine density threshold value and the cold engine utilization rate < the cold engine utilization rate threshold value, the hot-cold imbalance scenario is determined to be coexistence of hot engine overload and cold engine idling.

[0021] Further, the method of adaptive control for only hot engine overload comprises:

[0022] performing safety threshold calculation on each hot engine node in the energy storage converter subgraph, obtaining the power safety threshold of each hot engine node based on the product of the rated power of the hot engine node and the preset safety proportion threshold value, the hot engine node being the energy storage converter whose state label is identified as the hot engine overload state; accumulating the difference between the real-time output power of all hot engine nodes in the current energy storage converter subgraph and the corresponding power safety threshold value to obtain the first subgraph total overload power; screening the first normal load state node of the energy storage converter whose state label is identified as the normal load state in the energy storage converter subgraph, calculating and accumulating the residual receiving capacity of each first normal load state node to construct a first receiving capacity pool; the residual receiving capacity being the difference between the rated power and the real-time output power of the first normal load state node;

[0023] performing shunt control on the first subgraph total overload power, and distributing the first subgraph total overload power to the corresponding first normal load state node according to the proportion of the residual receiving capacity of each first normal load state node in the first receiving capacity pool.

[0024] Further, the method of adaptive control for only cold engine idling comprises:

[0025] constructing a cold engine adjustment demand degree function based on the cold engine utilization rate and the state balance degree in the set of topological characteristic parameters;

[0026] determining whether the cold engine adjustment demand degree output by the cold engine adjustment demand degree function is greater than the preset cold engine adjustment trigger threshold value;

[0027] if the determination result is no, ending the current adjustment process;

[0028] if the determination result is yes, for each cold engine node, the difference between the rated power and the current output power of the cold engine node is obtained to obtain the idling capacity, and the total idling capacity of the first subgraph is obtained by accumulating all the idling capacities; the cold engine node being the energy storage converter whose state label is identified as the cold engine idling state;

[0029] Extract the total power of the maintenance loss load from the energy storage converter subgraph, denoted as the transfer load power; the maintenance loss load includes electronic device switching loss load, cooling system loss load and control circuit loss load;

[0030] The transfer load power is controlled by shunt, and the transfer load power is distributed to the corresponding cold machine node according to the proportion of the idle capacity of each cold machine node in the total idle capacity of the first subgraph.

[0031] Further, the method for adaptive control of coexistence of hot machine overload and cold machine idling comprises:

[0032] The difference between the real-time output power of all hot machine nodes in the energy storage converter subgraph and the corresponding power safety threshold is accumulated to obtain the total overload power of the second subgraph;

[0033] The difference between the rated power of all cold machine nodes in the energy storage converter subgraph and the current output power is accumulated to obtain the total idle capacity of the second subgraph;

[0034] Determine whether the total idle capacity of the second subgraph can accommodate the total overload power of the second subgraph;

[0035] If the result is sufficient to accommodate, the total overload power of the second subgraph is controlled by shunt, and the total overload power of the second subgraph is distributed to each cold machine node in the total idle capacity of the second subgraph;

[0036] If the result is not sufficient to accommodate, the part of the total overload power of the second subgraph equal to the total idle capacity of the second subgraph is distributed to each cold machine node in the total idle capacity of the second subgraph, and the difference between the total overload power of the second subgraph and the total idle capacity of the second subgraph is obtained. The difference power; the remaining accommodation capacity of the normal load state node in the energy storage converter subgraph is counted and accumulated to form a second accommodation capacity pool;

[0037] The difference power is controlled by shunt, and the difference power is distributed to the corresponding normal load state node according to the proportion of the remaining accommodation capacity of each normal load state node in the second accommodation capacity pool.

[0038] Further, the method for obtaining the set of topology characteristic parameters comprises:

[0039] For each energy storage converter subgraph, the total number of subgraph nodes, the number of hot machines, the number of cold machines, the number of normal states, the total idle power of cold machines and the total rated power of cold machines in the energy storage converter subgraph are counted; the hot machine density is calculated according to the number of hot machines and the total number of subgraph nodes; the cold machine utilization rate is calculated according to the total idle power of cold machines and the total rated power of cold machines; the state balance degree is calculated according to the number of hot machines, the number of cold machines, the number of normal states and the total number of subgraph nodes; the hot machine density, the cold machine utilization rate and the state balance degree are constructed into the corresponding set of topology characteristic parameters.

[0040] Further, the obtaining method of the G energy storage converter subgraphs comprises:

[0041] constructing a topology connection tightness matrix according to line impedance and communication time delay between the energy storage converters;

[0042] converting the energy storage converter state label of each energy storage converter into a corresponding three-dimensional one-hot encoding vector, calculating the cosine similarity between the three-dimensional one-hot encoding vectors of the energy storage converters, and constructing a state similarity matrix;

[0043] fusing the topology connection tightness matrix and the state similarity matrix to form a topology and state coupling matrix according to a preset coupling weight coefficient;

[0044] deducing a comprehensive degree matrix based on the element distribution of the topology and state coupling matrix;

[0045] constructing a Laplacian matrix using the topology and state coupling matrix and the comprehensive degree matrix;

[0046] solving the eigenvalues and corresponding eigenvectors of the Laplacian matrix; and constructing a comprehensive feature matrix from the eigenvectors corresponding to the first G smallest eigenvalues;

[0047] after normalizing the row vectors of the comprehensive feature matrix, dividing them into G energy storage converter subgraphs using a K-means clustering algorithm.

[0048] Further, the operating state data includes DC bus voltage, AC output current, charging power, discharging power and switching frequency; the operating environment data includes ambient temperature, ambient humidity and cooling air speed; and the comprehensive feature set includes node temperature parameter sequence, equivalent internal resistance parameter sequence, state of charge parameter sequence and health state parameter sequence.

[0049] Further, the obtaining method of the node temperature parameter sequence comprises:

[0050] constructing a switching frequency sequence from the switching frequencies at the M time points; and constructing an AC output current sequence from the AC output currents at the M time points;

[0051] constructing a conduction loss sequence and a switching loss sequence from the AC output current sequence, the switching frequency sequence and the conduction loss coefficient and the switching loss coefficient calibrated at the factory; and superimposing the conduction loss sequence and the switching loss sequence to obtain a device heating power sequence;

[0052] according to the cooling air speeds at the M time points, converting the cooling air speeds through the ventilation cross-sectional area to obtain corresponding convective heat transfer capability coefficients, and constructing a heat dissipation capability parameter sequence from the ambient temperature and the ambient humidity at the corresponding time points;

[0053] The node temperature parameter sequence is constructed by time recursion of a thermal model according to a device heating power sequence and a heat dissipation capability parameter sequence.

[0054] Further, the equivalent resistance parameter sequence acquisition method comprises:

[0055] For each energy storage converter, the power direction is determined and merged based on the charging power and discharging power at M time points to obtain a charging and discharging power sequence; the DC side current sequence is constructed according to the conversion relationship between power and voltage based on the charging and discharging power sequence and the DC bus voltage at the M time points; the DC bus voltage change amount sequence is constructed by taking the absolute value of the difference between the DC bus voltages at adjacent time points; the DC bus current change amount sequence is constructed by taking the absolute value of the difference between the DC side currents at adjacent time points; the equivalent resistance estimation value sequence is constructed by time sequence alignment of the DC bus voltage change amount sequence and the DC bus current change amount sequence, and dividing the DC bus voltage change amount by the DC bus current change amount sequence at the same time sequence; and the equivalent resistance parameter sequence is obtained by denoising and median filtering of the equivalent resistance estimation value sequence.

[0056] Further, the state of charge parameter sequence acquisition method comprises:

[0057] For each energy storage converter, the state of charge recursive model is constructed by adopting the ampere-hour integration method based on the DC side current sequence; the state of charge parameters corresponding to the M time points are calculated based on the state of charge recursive model, and the state of charge parameter sequence is constructed;

[0058] Further, the health state parameter sequence acquisition method comprises:

[0059] For each energy storage converter, the node temperature parameter sequence, the equivalent resistance parameter sequence and the state of charge parameter sequence are respectively subjected to dimensionless normalization processing, and the health state parameters corresponding to the M time points are calculated based on the constructed health state function to construct the health state parameter sequence.

[0060] The health-aware adaptive control system for multi-energy storage converter collaboration is used to implement the health-aware adaptive control method for multi-energy storage converter collaboration, and comprises:

[0061] The operation feature modeling module is used to extract features based on the operation state data and operation environment data of the N energy storage converters at the M time points per unit time to obtain a comprehensive feature set corresponding to the N energy storage converters;

[0062] The running state diagnosis module is configured to input the comprehensive feature sets corresponding to the N energy storage converters into a state diagnosis model respectively to obtain energy storage converter state labels corresponding to the N energy storage converters; the energy storage converter state labels include a thermal engine overload state, a normal load state and a cold engine idle state;

[0063] The subgraph clustering construction module constructs an energy storage converter undirected graph with the N energy storage converters as nodes and electrical connections between the energy storage converters as edges; the energy storage converter undirected graph is divided based on a topology connection tightness and a state similarity dual criterion to obtain G energy storage converter subgraphs by using a spectral clustering algorithm;

[0064] The subgraph feature extraction module is configured to perform a feature extraction operation on the G energy storage converter subgraphs to obtain a topology characteristic parameter set corresponding to the energy storage converter subgraphs; the topology characteristic parameter set includes a thermal engine density, a cold engine utilization rate and a state balance degree;

[0065] The adaptive control module iterates through each energy storage converter subgraph to determine whether the state balance degree in the topology characteristic parameter set is greater than or equal to a state balance degree threshold; if the determination result is yes, the adjustment is ended; if the determination result is no, a hot-cold state imbalance scenario including only thermal engine overload, only cold engine idle and coexistence of thermal engine overload and cold engine idle is identified based on the topology characteristic parameters, and corresponding adaptive control is performed for different hot-cold state imbalance scenarios.

[0066] Compared with the prior art, the health perception adaptive control method and system for multiple energy storage converters proposed in the present application have the following technical effects and advantages:

[0067] The present application collects the running state data and the running environment data of multiple energy storage converters, constructs node temperature parameter sequences, equivalent internal resistance parameter sequences, state of charge parameter sequences and health state parameter sequences, and forms comprehensive feature sets. On this basis, the state diagnosis model is used to intelligently identify the running state of each energy storage converter, accurately distinguish the thermal engine overload state, the normal load state and the cold engine idle state, and further combine the topology structure to divide the state distribution into subgraphs, and comprehensively grasp the load imbalance in the local range.

[0068] Compared with the prior art relying only on droop control or power sharing, the application can unify the temperature, internal resistance, state of charge and state of health of the device into the basis for judgment, and realize adaptive power distribution driven by health perception. Through the technical scheme of the application, long-term overloading of part of the energy storage converter device can be effectively avoided to accelerate aging, and at the same time, resource waste caused by long-term idling of part of the energy storage converter device can be reduced; in a multi-machine parallel system, it is helpful to significantly suppress circulating current and power oscillation, and maintain power quality and operation stability. The comprehensive discrimination and collaborative control strategy proposed in the application can significantly improve the energy scheduling efficiency of the energy storage converter, prolong the service life of the energy storage converter device, and reduce the operation and maintenance risk, thereby having outstanding technical value and application prospect in the actual deployment of large-scale energy storage systems. BRIEF DESCRIPTION OF DRAWINGS

[0069] Figure 1 Fig. 1 is a schematic diagram of a health perception adaptive control system for multi-energy storage converter collaboration according to Embodiment 1 of the application;

[0070] Figure 2 Fig. 2 is a flowchart of a health perception adaptive control method for multi-energy storage converter collaboration according to Embodiment 2 of the application;

[0071] Figure 3 Fig. 3 is a flowchart of a method for executing corresponding adaptive control for different hot and cold imbalance scenarios;

[0072] Figure 4 Fig. 4 is a flowchart of a method for obtaining a G-energy storage converter subgraph;

[0073] Figure 5 Fig. 5 is a flowchart of a method for obtaining a node temperature parameter sequence. DETAILED DESCRIPTION

[0074] The technical solutions in the embodiments of the application will be described in detail, clearly and completely below with reference to the drawings in the embodiments of the application. It should be particularly noted that the specific embodiments described below are only used to better illustrate and describe the technical solutions of the application, and are intended to enable those skilled in the art to better understand and implement the application, and should not be understood as limiting the protection scope of the application. Those skilled in the art can modify, adjust or equivalently replace the application according to the content disclosed in the application without departing from the spirit and essence of the application, and these should be regarded as the protection scope of the application.

[0075] Embodiment 1:

[0076] Please refer to Figure 1 As shown in the figure, the embodiment discloses a health perception adaptive control system for multi-energy storage converter collaboration, which comprises that each module is connected through wired and / or wireless connection to realize data transmission.

[0077] The operation feature modeling module is run to perform feature extraction based on operation state data and operation environment data of the N energy storage converters at M time points in a unit of time, to obtain a comprehensive feature set corresponding to the N energy storage converters; the operation state data includes DC bus voltage, AC output current, charging power, discharging power and switching frequency; the operation environment data includes environmental temperature, environmental humidity and cooling wind speed; and the comprehensive feature set includes a node temperature parameter sequence, an equivalent internal resistance parameter sequence, a state of charge parameter sequence and a health state parameter sequence.

[0078] It should be noted that the DC bus voltage is used to represent the voltage level of the DC side bus of the energy storage converter, and the DC bus voltage reflects the stability and margin of the DC side energy supply; the DC bus voltage can be collected in real time by a voltage sensor arranged at both ends of the DC bus, combined with a sampling circuit and an analog-to-digital conversion unit. The AC output current is used to reflect the current output of the energy storage converter to the AC power grid or the load side, and the AC output current can reflect the output power size and current waveform quality; the AC output current can be detected by an AC side current transformer or a Hall sensor, and converted into a current value by a signal conditioning and sampling module.

[0079] The charging power is used to represent the power size of the energy storage converter transferred from the power grid to the energy storage battery under the charging condition, and reflects the battery charging rate and energy input condition. The charging power can be calculated by the product of the DC voltage and the charging current, wherein the charging current is obtained by a DC side current sensor. The discharging power is used to represent the power size of the energy storage converter output from the energy storage battery to the power grid or the load under the discharging condition, and reflects the speed and amplitude of energy release. The discharging power can be calculated by the product of the DC voltage and the discharging current, wherein the discharging current is detected in real time by a DC side sensor. The switching frequency is used to describe the switching action frequency of the internal power device of the energy storage converter, and the switching frequency is closely related to the control strategy and electromagnetic compatibility characteristics of the energy storage converter; the switching frequency can be analyzed by a pulse width modulation signal of the controller.

[0080] The ambient temperature is used to characterize the thermal conditions of the external environment in which the energy storage converter is located, and the ambient temperature is an important factor affecting the heat dissipation efficiency of the power device of the energy storage converter and the chemical reaction rate of the battery. The ambient temperature can be detected in real time by a temperature sensor arranged inside the housing or cabinet of the converter, and the collected data is input to the controller after being processed by an analog-to-digital conversion circuit. The ambient humidity is used to describe the moisture content level of the air around the energy storage converter, and the ambient humidity will affect the surface conductivity of the insulation of the energy storage converter and the condensation risk. High humidity may induce insulation performance degradation or even breakdown failure. The ambient humidity can be measured in real time by a humidity sensor arranged inside the cabinet of the energy storage converter. The cooling air speed is used to reflect the operation effect of the heat dissipation system of the energy storage converter. The cooling air speed can be obtained by a wind speed sensor installed in the heat dissipation channel, or by the speed feedback of the fan driving unit combined with the cross-sectional area of the channel. Through the collection and monitoring of the cooling air speed, it can be judged whether the heat dissipation is sufficient, and support is provided for device temperature rise prediction and thermal safety regulation.

[0081] As shown in Figure 5 The method for obtaining the node temperature parameter sequence comprises:

[0082] The switching frequency at M time points is constructed into a switching frequency sequence, and the alternating current output current at M time points is constructed into an alternating current output current sequence.

[0083] The conduction loss sequence and the switching loss sequence are constructed according to the alternating current output current sequence, the switching frequency sequence, and the conduction loss coefficient and the switching loss coefficient calibrated at the factory; and the device heating power sequence is obtained by superimposing the conduction loss sequence and the switching loss sequence.

[0084] The corresponding convective heat transfer capability coefficient is calculated from the cooling air speed through the ventilation cross-sectional area, and the ambient temperature and the ambient humidity at the corresponding time point are used to construct the heat dissipation capability parameter sequence; the method for calculating the convective heat transfer capability coefficient from the cooling air speed through the ventilation cross-sectional area is specifically that, according to the cooling air speed and the effective cross-sectional area of the ventilation channel, the air volume flow per unit time is calculated; then, according to the air density and the specific heat of air, the air volume flow is converted into mass flow; according to the heat transfer principle, the mass flow, the specific heat of air, and the velocity distribution through the heat dissipation channel are comprehensively calculated to obtain the convective heat transfer capability coefficient proportional to the wind speed.

[0085] The node temperature parameter sequence is constructed by time recursion of the thermal model according to the device heating power sequence and the heat dissipation capability parameter sequence.

[0086] The method for constructing the conduction loss sequence comprises:

[0087] The conduction loss of each of the AC output currents is multiplied by the conduction loss coefficient to obtain a corresponding conduction loss, and the switching loss of each of the AC output currents is multiplied by the switching loss coefficient to obtain a corresponding switching loss.

[0088] The conduction losses are constructed into a conduction loss sequence, and the switching losses are constructed into a switching loss sequence.

[0089] The conduction loss calculation method comprises:

[0090] ;

[0091] wherein, is the conduction loss of the nth energy storage converter at the mth time point, and m is in the range of 1 to M, is the conduction loss coefficient of the nth energy storage converter, and n is in the range of 1 to N, is the AC output current corresponding to the nth energy storage converter at the mth time point.

[0092] The switching loss calculation method comprises:

[0093] ;

[0094] wherein, is the switching loss of the nth energy storage converter at the mth time point, is the switching frequency corresponding to the nth energy storage converter at the mth time point, is the conduction loss coefficient of the nth energy storage converter.

[0095] It should be noted that the conduction loss coefficient is used to represent the energy loss degree of the power device of the energy storage converter in the conduction state per unit current, and the value is determined by the conduction resistance of the device, the device structure and the material characteristics. By applying different amplitudes of current to the power device in the factory test stage and recording the corresponding voltage drop and heating condition, an approximate linear relationship between current and conduction loss can be obtained, and the conduction loss coefficient can be extracted accordingly.

[0096] The switching loss coefficient is used to represent the instantaneous energy loss of the power device of the energy storage converter in the process of turning on and off due to the overlap of voltage and current, and the size is related to the switching speed, parasitic parameters and control strategy of the device. By measuring the energy loss of the power device at different switching frequencies in the factory test, and performing normalization processing, the proportional factor of the switching loss with the change of current and frequency, i.e. the switching loss coefficient, can be obtained.

[0097] For example, in the device calibration process of the energy storage converter, the obtained conduction loss coefficient is 0.2 watt loss per ampere of current, and the switching loss coefficient is 0.01 watt loss per kilohertz frequency and per ampere of current.

[0098] The method for obtaining the device heat power sequence comprises:

[0099] S100: Let the initial value of m be 1, and the value range of m be 1 to M;

[0100] S101: Obtain the conduction loss at the mth time point from the conduction loss sequence, and obtain the switching loss at the mth time point from the switching loss sequence;

[0101] S102: Add the conduction loss at the mth time point to the switching loss to obtain the device heat power at the mth time point; and add the device heat power at the mth time point to the device heat power sequence;

[0102] S103: Let m = m + 1, if m is less than or equal to M, return to S101 for execution, if m is greater than M, end the current process.

[0103] The method for obtaining the heat dissipation capacity parameter sequence comprises:

[0104] S200: Let the initial value of m be 1, and the value range of m be 1 to M;

[0105] S201: Calculate the heat dissipation capacity parameter at the mth time point according to the convection heat transfer capacity coefficient at the mth time point, the ambient temperature, and the ambient humidity;

[0106] S202: Add the heat dissipation capacity parameter at the mth time point to the heat dissipation capacity parameter sequence;

[0107] S203: Let m = m + 1, if m is less than or equal to M, return to S201 for execution, if m is greater than M, end the current process.

[0108] The method for calculating the heat dissipation capacity parameter comprises:

[0109] ;

[0110] wherein, is the heat dissipation capacity parameter corresponding to the nth energy storage converter at the mth time point, is the convection heat transfer capacity coefficient corresponding to the nth energy storage converter at the mth time point, is the ambient temperature corresponding to the mth time point, is the temperature normalization scale, and for example, the temperature normalization scale can be set to 10℃, Let be the ambient humidity at the m-th time point. For example, the humidity normalization scale can be set to 10%. For example, the reference temperature can be set to 25 degrees Celsius, representing the operating environment of the device under normal temperature conditions. The baseline humidity is set to 50%, representing the operating environment of the equipment under moderate humidity conditions. and For dimensionless weights, for example, can be... Set to 0.4, Set it to 0.6.

[0111] It should be noted that, The efficiency reduction factor is 1 when the environment is at the reference point, i.e., the ambient temperature and humidity are the same as the reference temperature and humidity. This means the heat dissipation capacity equals the convective heat transfer coefficient. As the temperature or humidity deviates more significantly, the efficiency reduction factor decreases. and As the value increases, the efficiency reduction factor increases, and the heat dissipation capacity parameter decreases, which conforms to the physical law that the more unfavorable the environment, the lower the equivalent heat dissipation capacity. At the same time, since the dimensionless ratio and constant term 1 are used, the formula does not have a singular point with zero denominator, and the dimension is maintained as watts per degree Celsius, which is convenient for direct coupling with the recursive derivation of the junction temperature.

[0112] The methods for calculating the node temperature parameters in the node temperature parameter sequence include:

[0113] ;

[0114] in, Let be the node temperature parameter corresponding to the nth energy storage converter at the m-th time point. Let be the node temperature parameter corresponding to the nth energy storage converter at the (m-1)th time point. Let be the device heat output power of the nth energy storage converter at the m-th time point. Let be the heat dissipation capacity parameter of the nth energy storage converter at the m-th time point. Let be the ambient temperature at the m-th time point. For time step. The equivalent thermal capacity of the nth energy storage converter is used to characterize the amount of heat that the internal power devices and heat dissipation structures of the energy storage converter can absorb or release under a unit temperature difference. It is a key parameter for measuring the thermal inertia of the energy storage converter. It can be determined experimentally during factory calibration or operation and maintenance. For example, under known power excitation conditions, the rise curve of the node temperature can be recorded, and the equivalent thermal capacity value can be calculated by combining the input energy and the temperature rise amplitude.

[0115] It should be noted that, in order to avoid calculation deviation or confusion of physical meaning caused by inconsistent units and dimensions of physical quantities in the calculation formula of the node temperature parameter, in the calculation process of the node temperature parameter in the present application, the parameters involved in the calculation have been dimensionless and normalized, for example, the numerical range of all parameters in the node temperature parameter formula is limited to [0, 1] to ensure the scientificity of the formula structure and the comparability of the results.

[0116] The method for obtaining the equivalent resistance parameter sequence comprises:

[0117] For each energy storage converter, the charging and discharging power at M time points is determined and combined based on the power direction to obtain a charging and discharging power sequence; based on the charging and discharging power sequence and the DC bus voltage at M time points, a DC side current sequence is constructed according to the conversion relationship between power and voltage; the DC bus voltage at M time points is subtracted at adjacent time points and the absolute value is taken to construct a DC bus voltage change sequence; the DC side current sequence is subtracted at adjacent time points and the absolute value is taken to construct a DC bus current change sequence; the DC bus voltage change sequence and the DC bus current change sequence are time-aligned, and the DC bus voltage change is divided by the DC bus current change sequence at the same time sequence to construct an equivalent resistance estimation value sequence; the equivalent resistance estimation value sequence is denoised and median filtered to obtain an equivalent resistance parameter sequence.

[0118] The method for obtaining the charging and discharging power sequence specifically includes: when it is detected that the charging power is greater than zero and the discharging power is zero, it is determined that the energy storage converter is in a charging condition, the charging power is assigned a positive sign and used as the numerical value of the charging and discharging power; when it is detected that the discharging power is greater than zero and the charging power is zero, it is determined that the energy storage converter is in a discharging condition, the discharging power is assigned a negative sign and used as the numerical value of the charging and discharging power; if both are zero, it is considered to be in a stationary condition, and the charging and discharging power is zero.

[0119] The method for calculating the DC side current in the DC side current sequence comprises:

[0120] ;

[0121] wherein, is the DC side current corresponding to the nth energy storage converter at the mth time point, is the charging and discharging power corresponding to the nth energy storage converter at the mth time point, is the DC bus voltage corresponding to the nth energy storage converter at the mth time point.

[0122] The method for obtaining the state of charge parameter sequence comprises:

[0123] For each energy storage converter, a recursive state-of-charge model is constructed using the ampere-hour integration method based on the DC-side current sequence. Based on the recursive state-of-charge model, the state-of-charge parameters corresponding to M time points are calculated and constructed into a state-of-charge parameter sequence.

[0124] The recursive model for the state of charge is as follows:

[0125] ;

[0126] in, Let be the state-of-charge parameters of the nth energy storage converter at the m-th time point. For the nth energy storage converter at the nth time The state of charge parameters corresponding to each time point. Let n be the rated capacity of the energy storage battery corresponding to the nth energy storage converter. Let be the DC-side current of the nth energy storage converter at the m-th time point. This represents the sampling time corresponding to the m-th time point. Indicates the first The sampling time corresponding to the time point. For example, the charge / discharge efficiency coefficient can be set to 0.95 during charging and 0.98 during discharging. It represents a time infinitesimal element in the continuous time domain.

[0127] It should be noted that the integral term The physical meaning is to calculate from arrive The cumulative value of DC-side current over time within a time period, combined with the charge / discharge efficiency coefficient, is used to characterize the charge inflow / outflow of the energy storage battery within that time period. Then, the change in state of charge is recursively derived using the ampere-hour integral method.

[0128] The method for obtaining the health status parameter sequence includes:

[0129] For each energy storage converter, the node temperature parameter sequence, equivalent internal resistance parameter sequence, and state of charge parameter sequence are respectively subjected to dimensionless normalization. Based on the constructed health state function, the health state parameters corresponding to M time points are calculated, and a health state parameter sequence is constructed. For example, the value ranges corresponding to the node temperature parameter sequence, equivalent internal resistance parameter sequence, and state of charge parameter sequence can all be normalized to [0, 1].

[0130] The calculation method for the health status parameters includes:

[0131] ;

[0132] in, Let be the health status parameters of the nth energy storage converter at the m-th time point. Let be the equivalent internal resistance parameter of the nth energy storage converter at the m-th time point. The reference state of charge parameter can be set to 0.6, for example. , and For the corresponding weighting coefficients, For example, as determined through accelerated aging tests at the manufacturing stage, the equivalent internal resistance parameter has a dominant influence on the health status parameter. Set it to 0.5, Set it to 0.3, Set it to 0.2.

[0133] It should be noted that the junction temperature parameter sequence reflects the thermal stress level of the power devices in the energy storage converter at various operating times. Continuous monitoring of the junction temperature allows for assessment of the heat dissipation and thermal safety margin of the power devices under different load and environmental conditions. The equivalent internal resistance parameter sequence characterizes the changes in internal resistance of the battery or energy storage unit in the energy storage converter during operation. Internal resistance is a crucial indicator of battery health and aging, dynamically changing with the number of charge-discharge cycles, temperature, and health status. By obtaining the time-series evolution of the equivalent internal resistance, battery performance degradation trends can be detected promptly, providing support for maintenance and fault early warning.

[0134] The state of charge (SOC) parameter sequence describes the energy storage level of an energy storage system. SOC reflects the remaining usable capacity of the battery and is a crucial basis for power dispatch, energy management, and grid support. Accurate estimation of SOC ensures the energy storage system operates within safe limits, preventing overcharging and over-discharging, and improving energy utilization efficiency. The health state parameter sequence, as a comprehensive index fused from multiple sources, quantifies the overall health level of each energy storage converter. The health state parameters comprehensively consider multiple factors such as node temperature, equivalent internal resistance, and SOC, and calculate a unified health value using weighted calculations, thus providing a unified criterion for system state assessment, coordinated dispatch, and dynamic optimization.

[0135] In summary, this invention constructs a comprehensive feature set and sequentially models node temperature, equivalent internal resistance, state of charge, and health status, achieving a comprehensive characterization of the multi-dimensional operational characteristics and trends of energy storage converters. This not only helps improve the safety and reliability of single-unit energy storage converter operation but also provides data support and decision-making basis for collaborative control and health sensing allocation among multiple energy storage converters.

[0136] The running state diagnosis module is configured to input the comprehensive feature sets corresponding to the N energy storage converters into the state diagnosis model respectively to obtain energy storage converter state labels corresponding to the N energy storage converters. The energy storage converter state labels include a thermal engine overload state, a normal load state, and a cold engine idle state. The energy storage converter state labels are used to represent the running load state of each energy storage converter.

[0137] It should be noted that the thermal engine overload state refers to an energy storage converter that bears a large power output and has a high running load in multiple energy storage converters. The cold engine idle state refers to an energy storage converter that has a low power output and insufficient resource utilization in multiple energy storage converters.

[0138] The training method of the state diagnosis model includes:

[0139] A state diagnosis dataset is constructed in advance. The state diagnosis dataset includes ZT sets of state diagnosis data and energy storage converter state labels corresponding to the ZT sets of state diagnosis data, where ZT is a positive integer. The state diagnosis data includes a comprehensive feature set. The state diagnosis dataset is divided into a training set and a validation set. The training set is used for parameter learning of the state diagnosis model, and the validation set is used for real-time monitoring of the generalization performance and overfitting degree of the state diagnosis model.

[0140] A deep neural network with a multilayer perceptron as the core is used as the state diagnosis model. The state diagnosis data is input into the deep neural network after standardization and vectorization processing. The deep neural network is composed of an input layer, hidden layers, and an output layer. Each hidden layer uses a nonlinear activation function to extract features, and the output layer uses a Softmax activation function to obtain a probability distribution corresponding to each energy storage converter state label. Finally, the energy storage converter state label corresponding to the maximum probability is taken as the prediction result of the state diagnosis model. In the training process, a cross-entropy loss function is used as the optimization objective, a gradient descent type optimization algorithm is used to update the network weights, and an early stopping strategy is set: when the prediction accuracy on the validation set reaches or exceeds a prediction accuracy threshold, it is determined that the state diagnosis model has converged and the training is terminated. For example, the prediction accuracy threshold can be set to 95% in this application.

[0141] It should be noted that the junction temperature parameter sequence directly reflects the thermal stress level of the power device. In the hot machine overload state, long-term large current and frequent switching operation can cause the junction temperature to rise significantly and remain in a high range. In the cold machine idle state, the device temperature rise is small and changes gently. In the normal load state, the junction temperature is within the safe range and presents regular fluctuations with the load. Therefore, the junction temperature parameter sequence can be used as an important basis for distinguishing different load states. The equivalent internal resistance parameter sequence reflects the impedance characteristics of the battery or energy storage unit of the energy storage converter in operation. The hot machine overload state is often accompanied by larger current fluctuations and faster internal resistance growth trend. In the cold machine idle state, the internal resistance changes slowly and the value is low. In the normal load state, the internal resistance presents regular fluctuations with the charging and discharging process. The equivalent internal resistance parameter sequence can reveal the load degree and health consumption difference of the energy storage converter device, and provide support for state label classification.

[0142] The state of charge parameter sequence is used to characterize the energy storage level of the energy storage converter. In the hot machine overload state, the state of charge decreases significantly faster, and the charge and discharge amplitude is larger. In the cold machine idle state, the state of charge curve is relatively flat, and the energy exchange is not active. In the normal state, the state of charge changes in accordance with the scheduling demand. Through the dynamic change characteristics of the state of charge parameter, it can be judged whether the energy storage converter device is in the hot machine overload or the cold machine idle. The health status parameter sequence, as a comprehensive index integrating temperature, internal resistance and state of charge, can overall depict the operation health level of the energy storage converter device. When the energy storage converter device is in the hot machine overload state, the health status index presents a downward trend or is at a low level. In the cold machine idle state, the index changes less. In the normal state, the index remains stable. This feature further improves the accuracy of state label classification.

[0143] In summary, the four types of parameter sequences in the comprehensive feature set reflect the significant differences of the energy storage converter in different load states from four dimensions of thermal characteristics, electrical characteristics, energy characteristics and health characteristics. By inputting these data into a deep neural network for training, the state diagnosis model can automatically extract the nonlinear correlation between multiple dimensions, thereby accurately identifying the high load, normal load and cold machine idle state labels corresponding to each energy storage converter, thereby supporting the technical solution of the present application.

[0144] The subgraph clustering construction module constructs an energy storage converter undirected graph with N energy storage converters as nodes and electrical connections between energy storage converters as edges. The spectral clustering algorithm is used on the energy storage converter undirected graph to obtain G energy storage converter subgraphs based on the dual criteria of topological connection density and state similarity.

[0145] As shown in Figure 4 The method for obtaining the G energy storage converter subgraphs includes:

[0146] constructing a topology connection tightness matrix according to line impedance and communication time delay between energy storage converters;

[0147] convert the energy storage converter state label of each energy storage converter into a corresponding three-dimensional one-hot encoding vector, calculate the cosine similarity between the three-dimensional one-hot encoding vectors of each energy storage converter, and construct a state similarity matrix;

[0148] For example, in this application, the three-dimensional one-hot encoding vector of the thermal overload state is [1, 0, 0]; the three-dimensional one-hot encoding vector of the normal load state is [0, 1, 0]; and the three-dimensional one-hot encoding vector of the cold machine idle state is [0, 0, 1].

[0149] According to the preset coupling weight coefficient, the topology connection tightness matrix and the state similarity matrix are fused to form a topology and state coupling matrix;

[0150] Based on the element distribution of the topology and state coupling matrix, a comprehensive degree matrix is derived. The elements outside the diagonal of the comprehensive degree matrix are all 0.

[0151] A Laplacian matrix is constructed using the topology and state coupling matrix and the comprehensive degree matrix.

[0152] The eigenvalues and corresponding eigenvectors of the Laplacian matrix are solved; and the eigenvectors corresponding to the first G small eigenvalues are constructed into a comprehensive eigenvector matrix.

[0153] After normalizing the row vectors of the comprehensive eigenvector matrix, the K-means clustering algorithm is used to divide it into G energy storage converter subgraphs.

[0154] It should be noted that extracting G eigenvalues is to prepare data for the next step of clustering into G energy storage converter subgraphs. If the values are inconsistent, such as taking 3 eigenvectors and clustering to obtain 4 energy storage converter subgraphs, it will cause the feature dimension and the number of target energy storage converter subgraphs to be inconsistent, and the precise energy storage converter subgraph division cannot be realized. Therefore, in the technical solution of the present application, the two G values must be the same.

[0155] The calculation method of the elements in the topology connection tightness matrix includes:

[0156]

[0157] wherein, is the matrix element corresponding to the position (i, j) in the topology connection tightness matrix, the value range of i is 1 to N, and the value range of j is 1 to N, represents an edge set composed of connected edges between energy storage converters. represents that there is a connected edge between energy storage converter i and energy storage converter j, ​indicates that there is no communication edge between energy storage converter i and energy storage converter j. is the line impedance between energy storage converter i and energy storage converter j, is the communication delay between energy storage converter i and energy storage converter j, is the communication delay weight coefficient. In an example, the communication delay weight coefficient can be set to 0.1 in the present application. The communication delay weight coefficient reflects the degree of physical connection closeness. The lower the impedance and the smaller the delay, the greater the weight.

[0158] It should be noted that for the line impedance, an offline calibration and online identification method can be used. Specifically, in the offline stage, the high-precision bridge is used to measure the segmented impedance of the connection cable between the energy storage converters, and the resistance and inductance parameters of each segment are recorded. The initial impedance model is constructed in combination with the physical length and material of the cable. In online operation, the current excitation signal is injected, the corresponding voltage response signal is collected, and the least square method is used to identify the equivalent impedance value of the line. At the same time, the impedance is thermally corrected in combination with the temperature of the connection cable monitored by the temperature sensor in real time. For example, the resistance of the copper cable increases by about 4% for every 10℃ increase in temperature. For the communication delay, a bidirectional time marking method is used for real-time measurement. Specifically, the first energy storage converter sends a probe data packet containing a time stamp to the second energy storage converter. The second energy storage converter records the receiving time stamp after receiving the probe data packet and immediately returns a response packet containing the sending time stamp and the receiving time stamp. The first energy storage converter records the time stamp when receiving the response packet, and the communication delay is the difference between the two time stamps. .

[0159] The method for constructing the topology and state coupling matrix comprises:

[0160] ;

[0161] wherein, is the topology and state coupling matrix, is the topology connection closeness matrix, is the state similarity matrix, is the preset coupling weight coefficient. In an example, the coupling weight coefficient can be set to 0.6 in the present application.

[0162] The method for calculating the matrix elements in the comprehensive degree matrix comprises:

[0163] ;

[0164] wherein, ​​​​​is the matrix element corresponding to the position (p, p) in the integrated degree matrix, and p is in the range of 1 to N. is the matrix element corresponding to the position (p, q) in the topological connection density matrix.

[0165] The method for constructing the Laplacian matrix comprises:

[0166] ;

[0167] wherein, is the Laplacian matrix, is the unit matrix, is the integrated degree matrix.

[0168] The subgraph feature extraction module is configured to perform a feature extraction operation on the G energy storage converter subgraphs to obtain a topological characteristic parameter set corresponding to the energy storage converter subgraphs; the topological characteristic parameter set comprises a heat engine density, a cold engine utilization rate and a state balance degree.

[0169] The method for obtaining the topological characteristic parameter set comprises:

[0170] For each energy storage converter subgraph, the total number of subgraph nodes, the number of heat engines, the number of cold engines, the number of normal states, the total idle power of cold engines and the total rated power of cold engines in the energy storage converter subgraph are counted; the heat engine density is calculated according to the number of heat engines and the total number of subgraph nodes; the cold engine utilization rate is calculated according to the total idle power of cold engines and the total rated power of cold engines; the state balance degree is calculated according to the number of heat engines, the number of cold engines, the number of normal states and the total number of subgraph nodes; and the heat engine density, the cold engine utilization rate and the state balance degree are constructed into a corresponding topological characteristic parameter set.

[0171] The method for calculating the heat engine density comprises:

[0172] ;

[0173] wherein, is the heat engine density corresponding to the gth energy storage converter subgraph, is the number of heat engines corresponding to the gth energy storage converter subgraph, is the total number of subgraph nodes corresponding to the gth energy storage converter subgraph.

[0174] The method for calculating the cold engine utilization rate comprises:

[0175] ;

[0176] wherein, is the cold engine utilization rate corresponding to the gth energy storage converter subgraph, is the total idle power of cold engines corresponding to the gth energy storage converter subgraph, is the cold machine rated total power corresponding to the gth energy storage converter subgraph.

[0177] The method for calculating the state balance degree comprises:

[0178]

[0179] wherein, is the state balance degree corresponding to the gth energy storage converter subgraph, d is an index variable in the summation formula, is the proportion of the number of energy storage converters of the dth state in the gth energy storage converter subgraph to the total number of nodes in the subgraph.

[0180] when d = 1, , is the number of hot machines corresponding to the gth energy storage converter subgraph; when d = 2, , is the number of cold machines corresponding to the gth energy storage converter subgraph; when d = 3, , is the number of normal states corresponding to the gth energy storage converter subgraph, is the total number of nodes corresponding to the gth energy storage converter subgraph. denotes the Shannon entropy of the three types of states, i.e., hot machines, cold machines and normal states.

[0181] The adaptive control module iterates through each energy storage converter subgraph to determine whether the state balance degree in the set of topological characteristic parameters is greater than or equal to the state balance degree threshold value. If the determination result is yes, the adjustment is ended. If the determination result is no, the hot-cold state imbalance scenarios including only hot machine overload, only cold machine idling and coexistence of hot machine overload and cold machine idling are identified based on the set of topological characteristic parameters, and corresponding adaptive control is performed for different hot-cold state imbalance scenarios.

[0182] As shown in Figure 3 , the method for performing corresponding adaptive control for different hot-cold state imbalance scenarios comprises:

[0183] S100: let the initial value of g be 1, and the value range of g be 1 to G; g is a loop index variable;

[0184] S101: obtain the set of topological characteristic parameters corresponding to the gth energy storage converter subgraph; obtain the state balance degree from the set of topological characteristic parameters, if the state balance degree is greater than or equal to the preset state balance degree threshold value, execute S104; if the state balance degree is less than the preset state balance degree threshold value, execute S102;

[0185] ​It should be noted that, for example, in the present application, the state balance degree threshold value can be set to 0.7, and the state balance degree greater than or equal to the preset state balance degree threshold value indicates that the three state distributions of hot machine, cold machine and normal state have reached balance, and the risks of hot machine overload and cold machine idling are low, so the adjustment process of the current energy storage converter subgraph can be directly ended.

[0186] S102: Based on the hot machine density and the cold machine utilization rate in the set of topological characteristic parameters, by comparing and analyzing with the preset hot machine density threshold value and the cold machine utilization rate threshold value, the hot-cold state imbalance scenario including only hot machine overload, only cold machine idling and coexistence of hot machine overload and cold machine idling is identified.

[0187] S103: If the hot-cold state imbalance scenario is only hot machine overload, adaptive control is performed for only hot machine overload; if the hot-cold state imbalance scenario is only cold machine idling, adaptive control is performed for only cold machine idling; if the hot-cold state imbalance scenario is coexistence of hot machine overload and cold machine idling, adaptive control is performed for coexistence of hot machine overload and cold machine idling.

[0188] S104: Let g=g+1, if g is less than or equal to G, return to S101 for execution, if g is greater than G, end the adaptive control process.

[0189] The identification method of the hot-cold state imbalance scenario includes:

[0190] If the hot machine density is greater than or equal to the hot machine density threshold value and the cold machine utilization rate is greater than or equal to the cold machine utilization rate threshold value, the hot-cold state imbalance scenario is determined to be only hot machine overload.

[0191] If the hot machine density is less than the hot machine density threshold value and the cold machine utilization rate is less than the cold machine utilization rate threshold value, the hot-cold state imbalance scenario is determined to be only cold machine idling.

[0192] If the hot machine density is greater than or equal to the hot machine density threshold value and the cold machine utilization rate is less than the cold machine utilization rate threshold value, the hot-cold state imbalance scenario is determined to be coexistence of hot machine overload and cold machine idling.

[0193] It should be noted that, if the hot machine density is less than the hot machine density threshold value and the cold machine utilization rate is greater than or equal to the cold machine utilization rate threshold value, it indicates that the overall operation is normal and no adjustment is needed, so it is not classified as a hot-cold state imbalance scenario. For example, in the present application, the hot machine density threshold value can be set to 0.5, and the cold machine utilization rate threshold value can be set to 0.3.

[0194] The method of adaptive control for only hot machine overload includes:

[0195] The safety threshold calculation is performed on each thermal engine node in the energy storage converter subgraph, and the power safety threshold of each thermal engine node is obtained based on the product of the rated power of the thermal engine node and the preset safety proportion threshold, the thermal engine node being the energy storage converter whose state label is identified as a thermal engine overload state; the difference between the real-time output power of all thermal engine nodes in the current energy storage converter subgraph and the corresponding power safety threshold is accumulated to obtain the first subgraph total overload power; the first normal load state node of the energy storage converter subgraph whose state label is in a normal load state is screened, the residual receiving capacity of each first normal load state node is calculated and accumulated to construct a first receiving capacity pool; the residual receiving capacity is the difference between the rated power and the real-time output power of the first normal load state node.

[0196] For example, in the present application, the safety proportion threshold can be set to 80%, that is, 80% of the rated power of the thermal engine node is the corresponding safety threshold.

[0197] The first subgraph total overload power is controlled, and the first subgraph total overload power is distributed to the corresponding first normal load state node according to the proportion of the residual receiving capacity of each first normal load state node in the first receiving capacity pool.

[0198] The method for adaptive control of only cold engine idling includes:

[0199] A cold engine adjustment demand degree function is constructed based on the cold engine utilization rate and the state balance degree in the set of topological characteristic parameters.

[0200] It is judged whether the cold engine adjustment demand degree output by the cold engine adjustment demand degree function is greater than the preset cold engine adjustment trigger threshold.

[0201] The construction method of the cold engine adjustment demand degree function includes:

[0202] ;

[0203] Wherein, is the cold engine adjustment demand degree of the gth energy storage converter subgraph, and is the corresponding weight coefficient, . For example, in the present application, the value of can be set to 0.6, and the value of can be set to 0.4.

[0204] If the judgment result is no, the current adjustment process is ended.

[0205] If the judgment result is yes, for each cold machine node, the rated power of the cold machine node is subtracted from the current output power to obtain an idle capacity, and all idle capacities are accumulated to obtain a first subgraph total idle capacity; the cold machine node is an energy storage converter whose state label is identified as a cold machine idle state;

[0206] The total power of the maintenance loss load is extracted from the energy storage converter subgraph, and is recorded as a transferred load power;

[0207] It should be noted that the maintenance loss load includes electronic device switching loss load, cooling system loss load, and control circuit loss load. Among them, the electronic device switching loss load is the loss power generated by the IGBT device in the switching process; the cooling system loss load is the operating power consumption of the cooling fan; and the control circuit loss load is the real-time power consumption of the converter control board.

[0208] The transferred load power is controlled in shunt, and the transferred load power is distributed to the corresponding cold machine node according to the proportion of the idle capacity of each cold machine node to the first subgraph total idle capacity.

[0209] For example, there are three cold machine nodes, the idle capacity of the first cold machine node is , the idle capacity of the second cold machine node is , and the idle capacity of the third cold machine node is . The first subgraph total idle capacity is , and the transferred load power is . That is, the power allocated to the first cold machine node is , the power allocated to the second cold machine node is , and the power allocated to the first cold machine node is .

[0210] The method for adaptive control of coexistence of hot machine overload and cold machine idling includes:

[0211] The difference between the real-time output power of all hot machine nodes in the energy storage converter subgraph and the corresponding power safety threshold is accumulated to obtain a second subgraph total overload power;

[0212] The difference between the rated power of all cold machine nodes in the energy storage converter subgraph and the current output power is accumulated to obtain a second subgraph total idle capacity;

[0213] It is judged whether the second subgraph total idle capacity can accommodate the second subgraph total overload power;

[0214] If the judgment result is sufficient to accommodate, i.e., the second subgraph total idle capacity is greater than or equal to the second subgraph total overload power, the second subgraph total overload power is controlled in shunt, and the second subgraph total overload power is distributed to each cold machine node in the second subgraph total idle capacity.

[0215] If the result of the judgment is not enough to be admitted, that is, the total idle capacity of the second subgraph is less than the total overload power of the second subgraph, then the part of the total overload power of the second subgraph equal to the total idle capacity of the second subgraph is allocated to each cold machine node in the total idle capacity of the second subgraph, and the difference between the total overload power of the second subgraph and the total idle capacity of the second subgraph is obtained. The remaining receiving capacity of the normal load state node in the energy storage converter subgraph is counted and accumulated to form a second receiving capacity pool.

[0216] The excess power is controlled to be distributed, and the excess power is distributed to the corresponding normal load state node according to the proportion of the remaining receiving capacity of each normal load state node to the second receiving capacity pool.

[0217] Embodiment 2:

[0218] Please refer to Figure 2 The embodiment provides a health-aware adaptive control method for multi-energy storage converter cooperation, which comprises:

[0219] Based on the operation state data and operation environment data of N energy storage converters at M time points per unit time, feature extraction is performed to obtain a comprehensive feature set corresponding to the N energy storage converters;

[0220] The comprehensive feature set corresponding to the N energy storage converters is input into a state diagnosis model respectively to obtain energy storage converter state labels corresponding to the N energy storage converters; the energy storage converter state labels comprise a hot machine overload state, a normal load state and a cold machine idle state;

[0221] A non-directed graph of energy storage converters is constructed with the N energy storage converters as nodes and electrical connections between the energy storage converters as edges; a spectral clustering algorithm is used on the non-directed graph of energy storage converters to obtain G energy storage converter subgraphs based on a double criterion of topological connection density and state similarity;

[0222] Feature extraction is performed on the G energy storage converter subgraphs to obtain a topological characteristic parameter set corresponding to the energy storage converter subgraphs; the topological characteristic parameter set comprises a hot machine density, a cold machine utilization rate and a state balance degree;

[0223] Each energy storage converter subgraph is traversed to determine whether the state balance degree in the topological characteristic parameter set is greater than or equal to a state balance degree threshold value; if the result of the determination is yes, the adjustment is ended; if the result of the determination is no, a hot-cold state imbalance scenario including only hot machine overload, only cold machine idle and coexistence of hot machine overload and cold machine idle is identified based on the topological characteristic parameters, and corresponding adaptive control is performed for different hot-cold state imbalance scenarios.

[0224] The above merely describes specific embodiments of the present application, but the protection scope of the present application is not limited thereto, any modification or substitution within the technical range disclosed by the present application can be easily thought of by those skilled in the art, and should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0225] Finally: the above merely describes preferred embodiments of the present application, and is not used to limit the present application, any modification, equivalent substitution, improvement, etc. within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A health-aware adaptive control method for multi-energy storage converter coordination, characterized in that, The method comprises the following steps: Based on the operation state data and operation environment data of N energy storage converters at M time points per unit time, feature extraction is performed to obtain a comprehensive feature set corresponding to the N energy storage converters; the comprehensive feature set comprises a node temperature parameter sequence, an equivalent internal resistance parameter sequence, a state of charge parameter sequence, and a health state parameter sequence; The comprehensive feature set corresponding to the N energy storage converters is input into a state diagnosis model respectively to obtain an energy storage converter state label corresponding to the N energy storage converters; the energy storage converter state label comprises a thermal engine overload state, a normal load state, and a cold engine idle state; A non-directed graph of energy storage converters is constructed with the N energy storage converters as nodes and electrical connections between the energy storage converters as edges; A spectral clustering algorithm is used on the non-directed graph of energy storage converters to obtain G energy storage converter subgraphs based on the dual criteria of topological connection density and state similarity; Feature extraction is performed on the G energy storage converter subgraphs to obtain a topological characteristic parameter set corresponding to the energy storage converter subgraphs; the topological characteristic parameter set comprises a thermal engine density, a cold engine utilization rate, and a state balance degree; The method for obtaining the topological characteristic parameter set comprises: For each energy storage converter subgraph, the total number of subgraph nodes, the number of thermal engines, the number of cold engines, the number of normal states, the total idle power of cold engines, and the total rated power of cold engines in the energy storage converter subgraph are counted; the thermal engine density is calculated based on the number of thermal engines and the total number of subgraph nodes; the cold engine utilization rate is calculated based on the total idle power of cold engines and the total rated power of cold engines; the state balance degree is calculated based on the number of thermal engines, the number of cold engines, the number of normal states, and the total number of subgraph nodes; and the thermal engine density, the cold engine utilization rate, and the state balance degree are constructed into a corresponding topological characteristic parameter set; Each energy storage converter subgraph is traversed to determine whether the state balance degree in the topological characteristic parameter set is greater than or equal to a state balance degree threshold value; if the determination result is yes, the adjustment is ended; if the determination result is no, a hot-cold state imbalance scenario including only thermal engine overload, only cold engine idle, and coexistence of thermal engine overload and cold engine idle is identified based on the topological characteristic parameters, and corresponding adaptive control is performed for different hot-cold state imbalance scenarios.

2. The health-aware adaptive control method for multi-energy storage converter coordination according to claim 1, wherein, The method for performing corresponding adaptive control for different hot-cold state imbalance scenarios comprises: S100: let the initial value of g be 1, and the value range of g be 1 to G; g is a loop index variable; S101: obtain the topological characteristic parameter set corresponding to the gth energy storage converter subgraph; from the topological characteristic parameter set, obtain the state balance degree; if the state balance degree is greater than or equal to a preset state balance degree threshold value, execute S104; if the state balance degree is less than the preset state balance degree threshold value, execute S102; S102: based on the thermal engine density and the cold engine utilization rate in the topological characteristic parameter set, identify a hot-cold state imbalance scenario including only thermal engine overload, only cold engine idle, and coexistence of thermal engine overload and cold engine idle by comparing and analyzing with a preset thermal engine density threshold value and a cold engine utilization rate threshold value; S103: If the hot-cold imbalance scenario is only hot machine overload, adaptive control is performed for only hot machine overload; if the hot-cold imbalance scenario is only cold machine idle, adaptive control is performed for only cold machine idle; if the hot-cold imbalance scenario is coexistence of hot machine overload and cold machine idle, adaptive control is performed for coexistence of hot machine overload and cold machine idle; S104: Let g = g + 1, if g is less than or equal to G, return to S101 for execution, if g is greater than G, end the adaptive control process.

3. The health-aware adaptive control method for multi-energy storage converter coordination according to claim 2, wherein, The identification method of the hot-cold imbalance scenario comprises: If the hot machine density is greater than or equal to the hot machine density threshold value and the cold machine utilization rate is greater than or equal to the cold machine utilization rate threshold value, the hot-cold imbalance scenario is determined to be only hot machine overload; If the hot machine density is less than the hot machine density threshold value and the cold machine utilization rate is less than the cold machine utilization rate threshold value, the hot-cold imbalance scenario is determined to be only cold machine idle; If the hot machine density is greater than or equal to the hot machine density threshold value and the cold machine utilization rate is less than the cold machine utilization rate threshold value, the hot-cold imbalance scenario is determined to be coexistence of hot machine overload and cold machine idle.

4. The health-aware adaptive control method for multi-energy storage converter coordination according to claim 2, wherein, The method for adaptive control for only hot machine overload comprises: Performing safety threshold calculation on each hot machine node in the energy storage converter subgraph, obtaining the power safety threshold of each hot machine node based on the product of the rated power of the hot machine node and the preset safety proportion threshold value, the hot machine node being the energy storage converter whose state label is identified as the hot machine overload state; accumulating the difference between the real-time output power of all hot machine nodes in the current energy storage converter subgraph and the corresponding power safety threshold value to obtain the first subgraph total overload power; screening the first normal load state node of the energy storage converter whose state label is in the normal load state in the energy storage converter subgraph, calculating and accumulating the residual receiving capacity of each first normal load state node to construct a first receiving capacity pool; the residual receiving capacity being the difference between the rated power and the real-time output power of the first normal load state node; Performing shunt control on the first subgraph total overload power, and distributing the first subgraph total overload power to the corresponding first normal load state node according to the proportion of the residual receiving capacity of each first normal load state node in the first receiving capacity pool.

5. The health-aware adaptive control method for multi-energy storage converter coordination according to claim 2, wherein, The method for adaptive control for only cold machine idle comprises: Constructing a cold machine adjustment demand degree function based on the cold machine utilization rate and the state balance degree in the set of topological characteristic parameters; Judging whether the cold machine adjustment demand degree output by the cold machine adjustment demand degree function is greater than the preset cold machine adjustment trigger threshold value; If the judgment result is no, ending the current adjustment process; If the judgment result is yes, for each cold machine node, the difference between the rated power and the current output power of the cold machine node is obtained to obtain the idle capacity, and the total idle capacity of the first subgraph is obtained by accumulating all idle capacities; the cold machine node being the energy storage converter whose state label is identified as the cold machine idle state; Extracting the total power of the maintenance loss load from the energy storage converter subgraph, which is recorded as the transferred load power; the maintenance loss load including electronic device switching loss load, cooling system loss load and control circuit loss load; Performing shunt control on the transferred load power, and distributing the transferred load power to the corresponding cold machine node according to the proportion of the idle capacity of each cold machine node in the first subgraph total idle capacity.

6. The health-aware adaptive control method for multi-energy storage converter coordination according to claim 2, wherein, The method for adaptive control of coexistence of thermal engine overload and cold engine idling comprises: accumulating the difference between the real-time output power of all thermal engine nodes in the energy storage converter subgraph and the corresponding power safety threshold to obtain the second subgraph total overload power; accumulating the difference between the rated power of all cold engine nodes in the energy storage converter subgraph and the current output power to obtain the second subgraph total idle capacity; determining whether the second subgraph total idle capacity can accommodate the second subgraph total overload power; if the result is sufficient to accommodate, then the second subgraph total overload power is controlled to be distributed to each cold engine node in the second subgraph total idle capacity; if the result is insufficient to accommodate, then the second subgraph total overload power is distributed to each cold engine node in the second subgraph total idle capacity, and the difference between the second subgraph total overload power and the second subgraph total idle capacity is obtained; the residual accommodation capacity of the normal load state nodes in the energy storage converter subgraph is counted and accumulated to form a second accommodation capacity pool; the difference power is controlled to be distributed to the corresponding normal load state nodes according to the proportion of the residual accommodation capacity of each normal load state node in the second accommodation capacity pool.

7. The health-aware adaptive control method for multi-energy storage inverter coordination according to claim 1, wherein, The method for obtaining G energy storage converter subgraphs comprises: constructing a topology connection tightness matrix according to the line impedance and communication time delay between the energy storage converters; converting the energy storage converter state label of each energy storage converter into a corresponding three-dimensional one-hot encoding vector, calculating the cosine similarity between the three-dimensional one-hot encoding vectors of the energy storage converters, and constructing a state similarity matrix; fusing the topology connection tightness matrix and the state similarity matrix according to the preset coupling weight coefficient to form a topology and state coupling matrix; deducing a comprehensive degree matrix based on the element distribution of the topology and state coupling matrix; constructing a Laplacian matrix using the topology and state coupling matrix and the comprehensive degree matrix; solving the eigenvalues and corresponding eigenvectors of the Laplacian matrix; constructing a comprehensive feature matrix from the eigenvectors corresponding to the first G smallest eigenvalues; after normalizing the row vectors of the comprehensive feature matrix, using the K-means clustering algorithm to divide them into G energy storage converter subgraphs.

8. The health-aware adaptive control method for multi-energy storage inverter coordination according to claim 1, wherein, The operating state data includes DC bus voltage, AC output current, charging power, discharging power and switching frequency; the operating environment data includes ambient temperature, ambient humidity and cooling air speed.

9. The health-aware adaptive control method for multi-energy storage inverter coordination according to claim 1, wherein, The method for obtaining the node temperature parameter sequence comprises: constructing a switching frequency sequence from the switching frequencies at M time points; constructing an AC output current sequence from the AC output currents at M time points; constructing a conduction loss sequence and a switching loss sequence from the AC output current sequence, the switching frequency sequence and the conduction loss coefficient and the switching loss coefficient calibrated at the factory; superimposing the conduction loss sequence and the switching loss sequence to obtain a device heating power sequence; according to the cooling air speed at M time points, the corresponding convective heat transfer capacity coefficient is obtained through the ventilation cross-sectional area conversion, and the corresponding time point ambient temperature and ambient humidity are constructed to form a heat dissipation capacity parameter sequence; The node temperature parameter sequence is constructed by time recursion of a thermal model according to a device heat power sequence and a heat dissipation capability parameter sequence.

10. The health-aware adaptive control method for multi-energy storage inverter coordination according to claim 9, wherein, The equivalent resistance parameter sequence is obtained by: For each energy storage converter, the power direction is determined and merged based on the charging power and discharging power at the M time points to obtain a charging and discharging power sequence; the DC side current sequence is constructed according to the conversion relationship between power and voltage based on the charging and discharging power sequence and the DC bus voltage at the M time points; the DC bus voltage change sequence is constructed by taking the absolute value of the difference between the DC bus voltages at adjacent time points; the DC bus current change sequence is constructed by taking the absolute value of the difference between the DC side currents at adjacent time points; the equivalent resistance estimation value sequence is constructed by aligning the DC bus voltage change sequence and the DC bus current change sequence in time sequence, and dividing the DC bus voltage change by the DC bus current change sequence at the same time sequence; and the equivalent resistance parameter sequence is obtained by denoising and median filtering the equivalent resistance estimation value sequence.

11. The health-aware adaptive control method for multi-energy storage inverter coordination according to claim 10, wherein, The state of charge parameter sequence is obtained by: For each energy storage converter, the state of charge recursive model is constructed by adopting the ampere-hour integration method based on the DC side current sequence; and the state of charge parameters corresponding to the M time points are calculated based on the state of charge recursive model to construct the state of charge parameter sequence. The health state parameter sequence is obtained by: For each energy storage converter, the node temperature parameter sequence, the equivalent resistance parameter sequence and the state of charge parameter sequence are respectively dimensionless normalized, and the health state parameters corresponding to the M time points are calculated based on the constructed health state function to construct the health state parameter sequence.

12. A multi-energy storage converter coordination oriented health-aware adaptive control system for implementing the multi-energy storage converter coordination oriented health-aware adaptive control method of any one of claims 1-11, characterized in that, It comprises: The operation feature modeling module extracts features based on the operation state data and operation environment data of the N energy storage converters at the M time points per unit time to obtain a comprehensive feature set corresponding to the N energy storage converters; The operation state diagnosis module inputs the comprehensive feature set corresponding to the N energy storage converters into the state diagnosis model to obtain energy storage converter state labels corresponding to the N energy storage converters; the energy storage converter state labels include thermal engine overload state, normal load state and cold engine idle state; The subgraph clustering construction module constructs an energy storage converter undirected graph with the N energy storage converters as nodes and the electrical connections between the energy storage converters as edges; The spectral clustering algorithm is used on the energy storage converter undirected graph to divide it into G energy storage converter subgraphs based on the topological connection density and state similarity; The subgraph feature extraction module extracts features from the G energy storage converter subgraphs to obtain a topological characteristic parameter set corresponding to the energy storage converter subgraphs; the topological characteristic parameter set includes thermal engine density, cold engine utilization rate and state balance degree. The adaptive control module traverses each energy storage converter subgraph, judges whether the state balance degree in the set of topological characteristic parameters is greater than or equal to the state balance degree threshold value, if the judgment result is yes, the adjustment is ended, if the judgment result is no, the topological characteristic parameters are identified to include a hot-cold imbalance scenario of only thermal machine overload, only cold machine idling and coexistence of thermal machine overload and cold machine idling, and corresponding adaptive control is performed for different hot-cold imbalance scenarios.

Images