A network-constructing energy storage converter voltage support method based on thermal security boundary expansion

Abstract

This invention provides a voltage support method for grid-connected energy storage converters based on thermal safety boundary extension, belonging to the field of new energy grid connection and power electronics control technology. The method includes: establishing an electrothermal coupling model of the grid-connected energy storage converter; defining a transient thermal safety boundary based on junction temperature constraints and power quality constraints; reducing switching losses and extending the global thermal safety boundary through transient frequency reduction control; performing adaptive modulation switching based on junction temperature feedback and thermal sensitivity matrix to achieve local hotspot redistribution; constructing a collaborative framework of upper-level active thermal management and lower-level adaptive virtual impedance control, and using the DDPG algorithm to adjust the virtual impedance online, enabling the output current to dynamically track the extended transient thermal safety boundary. This invention effectively improves the transient overload capacity and grid connection point voltage support effect of grid-connected energy storage converters while ensuring that the maximum junction temperature of power devices does not exceed the thermal safety threshold, and is suitable for transient stability control scenarios in new energy high-penetration power systems.

Description

Technical Field

[0001] This invention belongs to the field of new energy grid connection and power electronic control technology, and relates to a voltage support method for grid-connected energy storage converters based on thermal safety boundary extension. Background Technology

[0002] With the advancement of the "dual carbon" target, the penetration rate of new energy sources in the power system continues to increase. The characteristics of low system inertia and weak damping are becoming increasingly prominent. Therefore, grid-forming (GFM) energy storage converters, possessing voltage source characteristics and active support capabilities, have become crucial equipment for supporting the stable operation of new power systems. Especially under large disturbances such as short circuits, the ability of GFMs to quickly provide sufficient reactive power support directly affects the voltage recovery level at the grid connection point and the transient stability of the system.

[0003] However, the transient voltage support capability of GFM is strictly limited by the physical thermal limits of power devices. During faults, to suppress voltage drops, the converter often needs to output reactive current several times its rated value, causing the junction temperature of the power devices to rise sharply in a short period of time. Once the junction temperature exceeds the thermal safety threshold, it will trigger the risk of thermal runaway or even damage the devices. Therefore, in essence, the transient voltage support problem of GFM is a problem of releasing support capability under thermal safety constraints. How to improve the transient support capability of the converter as much as possible while ensuring the thermal safety of the power devices has become an important problem that urgently needs to be solved in the field of grid control.

[0004] Existing research on the transient support problem of power flow faults (GFFs) mainly focuses on electrical control, with representative methods including current limiting control and virtual impedance control. Traditional current limiting control introduces an instantaneous current limiting module between the current and voltage loops to limit the fault current to a constant value; virtual impedance control achieves flexible current limiting by adjusting the equivalent output impedance of the converter and maintains the voltage source characteristics of the GFM during faults. Building on this, previous research has systematically explored fixed virtual impedance parameter tuning, adaptive impedance regulation based on transient current characteristics, two-stage virtual impedance control that balances current limiting and power angle stability, minimum virtual impedance design based on impedance phasor diagrams, and quantitative parameter tuning methods for low-voltage ride-through. However, optimizing current limiting and support characteristics solely from an electrical control perspective is insufficient to fundamentally overcome the limitations imposed by the thermal constraints of power devices on the transient support capability of GFMs.

[0005] Therefore, Active Thermal Management (ATM) offers a new approach to improving the thermal safety of converters. Existing ATM research largely focuses on issues such as junction temperature suppression, thermal stress equalization, lifetime extension, and derating control under steady-state operating conditions; some studies also consider control optimization under junction temperature constraints and active thermal control of grid-connected inverters during transient faults. However, these works primarily focus on improving device thermal safety and reliability, and essentially lack a control framework that aligns with the transient support requirements of GFM. Therefore, it is necessary to propose a voltage support method for grid-connected energy storage converters based on thermal safety boundary extension to improve the system's voltage support capability during voltage sags. Summary of the Invention

[0006] In view of this, the purpose of this invention is to provide a voltage support method for grid-connected energy storage converters based on thermal safety boundary extension. This method defines the transient thermal safety boundary of the grid-connected energy storage converter by establishing a power device loss model and a thermal network model. Furthermore, it extends the transient thermal safety boundary of the converter by reducing power device losses during faults and achieving local hotspot redistribution through transient frequency reduction and adaptive modulation switching based on junction temperature feedback. On this basis, adaptive virtual impedance control enables the converter output current to dynamically track the extended transient thermal safety boundary. Thus, while ensuring that the maximum junction temperature of the power devices does not exceed the thermal safety threshold, the transient overload capacity and grid connection point voltage support effect of the grid-connected energy storage converter are improved.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A voltage support method for grid-connected energy storage converters based on thermal safety boundary extension, the method specifically includes the following steps: S1. Establish an electrothermal coupling model for a grid-type energy storage converter, wherein the electrothermal coupling model includes a power device loss model and a thermal network model; S2. Based on the junction temperature constraints of the power devices and the grid-connected power quality constraints, define the transient thermal safety boundary of the grid-type energy storage converter; S3. Based on the relationship between the switching frequency and the power device loss, the global thermal safety boundary is extended through transient frequency reduction control; S4. Based on the junction temperature feedback and thermal sensitivity matrix of the power device, adaptive modulation switching is performed to realize the redistribution of local hot spots, so as to further extend the transient thermal safety boundary. S5. Construct a collaborative control framework for thermal safety boundary extension and boundary tracking, the framework including an upper-layer active thermal management module and a lower-layer adaptive virtual impedance controller; S6. The upper-layer active thermal management module outputs an extended transient thermal safety boundary based on the thermal state of the power device, and the lower-layer adaptive virtual impedance controller adjusts the virtual impedance parameters online so that the output current of the grid-type energy storage converter dynamically tracks the extended transient thermal safety boundary.

[0008] Furthermore, in step S1, the loss model includes a conduction loss model and a switching loss model; the conduction loss is calculated by averaging the product of the device current and its saturation voltage drop over one fundamental frequency period; the switching loss is calculated by averaging the sum of the turn-on loss and the turn-off loss over one fundamental frequency period, and the switching loss is proportional to the switching frequency. The thermal network model adopts the Foster equivalent thermal network model, which equates the internal heat conduction process of the power module to a dynamic network composed of series and parallel thermal resistance and thermal capacity. The loss of the power device is injected into the thermal network as a heat source and is conducted step by step through the heat transfer path to form the dynamic junction temperature response of the power device.

[0009] Furthermore, in step S2, the transient thermal safety boundary is defined as: the maximum current value that the grid-type energy storage converter can safely output under the condition that the maximum dynamic junction temperature of the power device does not exceed the maximum allowable junction temperature and the harmonic distortion rate of the grid connection point voltage does not exceed the set upper limit.

[0010] Furthermore, in step S3, the transient frequency reduction control specifically involves reducing the switching frequency of the grid-type energy storage converter from the rated frequency to a preset lower limit frequency when a fault occurs, in order to reduce the overall switching loss during the transient process and raise the global thermal safety boundary; the preset lower limit frequency is jointly determined by the switching speed of the power device and the harmonic distortion rate constraint of the grid connection point voltage.

[0011] Furthermore, in step S4, the adaptive modulation switching specifically includes: S41. The thermal sensitivity matrix is ​​pre-constructed, where the rows of the matrix correspond to different power devices, the columns correspond to different modulation methods, and the matrix elements represent the junction temperature reduction benefit of the corresponding device after switching from the reference modulation method to the target modulation method. S42. Monitor the junction temperature of each power device in real time. When the junction temperature of any power device exceeds a preset warning threshold, identify the device as the current hot spot device. S43. Based on the thermal sensitivity matrix, select the modulation method that has the greatest cooling benefit for the current hot spot device as the optimal modulation method, and switch to the optimal modulation method to realize the dynamic migration of local hot spots.

[0012] Furthermore, in step S5, virtual impedance control achieves flexible adjustment of fault current by introducing adjustable equivalent resistance and equivalent reactance on the output side of the grid-type energy storage converter.

[0013] Furthermore, in step S6, the lower-level adaptive virtual impedance controller is constructed using a deep deterministic policy gradient algorithm. The agent of the deep deterministic policy gradient algorithm uses the voltage drop depth, the output current amplitude, the rate of change of the output current, and the tracking error of the output current to the transient thermal safety boundary as the state space, the reference value of the virtual impedance as the action space, and designs a reward function with the goal of minimizing the current tracking error, suppressing the current overshoot, and smoothing the virtual impedance adjustment.

[0014] Furthermore, the transient frequency reduction control and adaptive modulation switching of the upper active thermal management module work together to increase the maximum safe output current of the grid-type energy storage converter from the traditional current limiting value to the preset target overload value.

[0015] Furthermore, the main circuit topology of the grid-type energy storage converter is a three-level active midpoint clamping topology, and its control method is virtual synchronous generator control.

[0016] The beneficial effects of this invention are as follows: 1) This invention defines the transient thermal safety boundary of GFM based on power device loss models and thermal network models, and proposes an active thermal management method consisting of transient frequency reduction and adaptive modulation switching. This method achieves active expansion of the thermal safety boundary through global switching loss suppression and local thermal stress redistribution; 2) This invention constructs a DDPG adaptive virtual impedance control method oriented towards thermal safety boundaries. By adjusting the virtual impedance online, the output current is dynamically tracked to the thermal safety boundary, thereby taking into account the initial impact suppression of faults, the release of support capacity during the fault duration, and the smooth recovery after fault clearance. 3) The proposed method was verified under typical fault conditions and multiple operating conditions based on electrothermal co-simulation and a semi-physical experimental platform. The results show that, under the premise of ensuring that the dynamic junction temperature of the device does not exceed the thermal safety threshold, the proposed method improves the transient voltage support effect at the grid connection point by about 35%, and also shows good adaptability under different fault depths and continuous fault scenarios.

[0017] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0018] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 The main circuit topology of the VSG and its control system structure; Figure 2 A thermal network model for the power module; Figure 3 The junction temperature characteristic curves and thermal safety boundaries of the device at different switching frequencies; Figure 4 The thermal stress distribution characteristics of the A-phase bridge arm device under different modulation strategies; Figure 5 A collaborative control framework for thermal safety boundary extension and boundary tracking; Figure 6 This is a schematic diagram of the DDPG agent training and interaction process; Figure 7 Training the reward convergence curve for the DDPG agent; Figure 8 This section describes the active thermal management actions and the junction temperature response of power devices under different methods. Figure 9 Output three-phase current using different methods; Figure 10 Comparison of reactive power output and grid connection point voltage under different methods; Figure 11 The results are from a semi-physical experiment at different voltage drop depths. Figure 12 The results are from a semi-physical experiment under continuous failure conditions. Detailed Implementation

[0019] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0020] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures, and should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0021] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0022] This invention provides a voltage support method for grid-type energy storage converters based on thermal safety boundary extension, comprising the following steps: S1, Establish the electrothermal coupling model of the grid-type energy storage converter: This invention focuses on a grid-connected system of a virtual synchronous generator (VSG) controlled energy storage converter. The main circuit employs a three-level active neutral point clamped (ANPC) topology. Its main circuit and control structure are as follows: Figure 1 As shown in the figure. L 1. C 2 represents the filter's inductance and capacitance parameters. U dc DC side voltage C 1 represents the parameters of the parallel capacitor on the DC side. T a1~6 , T b1~6 , T c1~6 These correspond to the power devices for phases a, b, and c, respectively. L 2 represents the equivalent inductance parameter of the power grid side line.

[0023] VSG control provides inertia, damping, and autonomous voltage regulation support to the system by simulating the rotor motion equations and excitation regulation characteristics of a synchronous generator. A virtual inertia is introduced into the active-frequency loop. J With damping coefficient D To support the stability of the power grid frequency, its control equation is shown in equation (1).

[0024] (1) In the formula, P m For mechanical power, P e Electromagnetic power, P refThis is the active power command value; ω 0 represents the grid synchronization angular velocity; k p This is the frequency modulation coefficient.

[0025] The reactive-voltage control loop maintains the terminal voltage stability by adjusting the internal potential, and the control equation is shown in equation (2).

[0026] (2) In the formula, E ref Internal potential command, E m This is the output value of the internal potential; k qp This is the reactive power proportionality coefficient. k qi This is the reactive voltage integral coefficient; Q ref This is the reactive power command value. Q This represents the actual value of reactive power.

[0027] VSG output current i abc The modulation signal directly determines the switching state and on-state current of the power device, thereby generating switching losses and on-state losses, and causing the junction temperature to rise. Therefore, it is necessary to establish a loss model and a thermal network model for the IGBT to accurately assess the thermal dynamics during transient overcurrent.

[0028] The junction temperature of IGBTs is mainly caused by the accumulation of internal losses within the power module, which includes both conduction losses and switching losses. P con Depends on device current i (t) and its saturation pressure drop v ce ( i , T j The product of ) in one fundamental frequency period T The average on-state loss within 0 can be expressed as: (3) In the formula, d (t) is the switching function (duty cycle), which is directly determined by the modulation strategy. Therefore, active control of loss can be achieved by changing the modulation strategy.

[0029] Switching losses P sw Due to turn-on losses E on With shutdown loss E off The composition, whose periodic average value can be expressed as: (4) In the formula, N The number of switching actions within a cycle, and the switching frequency. f sw Proportional. Therefore, during transient processes, the decrease... f sw It can directly suppress switching losses.

[0030] After obtaining the power module losses, the device junction temperature can be further calculated based on the Foster equivalent thermal network, the structure of which is as follows: Figure 2 As shown.

[0031] This model equates the internal heat conduction process of the power module to a dynamic network composed of series and parallel thermal resistances and capacitances. The power losses of the IGBT and diodes are also considered. P T , P D As a heat source injected into the thermal network, the heat is conducted step by step through the heat transfer path, ultimately forming the dynamic response of the device junction temperature. T JT , T JD Its calculation expression is shown in equation (5).

[0032] (5) In the formula, R thJC , R thCH , R thHA These are the thermal resistances from the device junction to the case, from the case to the heat sink, and from the heat sink to the external environment, respectively, used to characterize the steady-state thermal resistance characteristics of each heat transfer link. C th The heat capacity of the corresponding branch is used to reflect the heat accumulation effect; T c and T a These are the module housing temperature and the ambient temperature, respectively. τ The time constants corresponding to each branch of the thermal network determine the response speed during the dynamic change of junction temperature.

[0033] S2, defining transient thermal safety boundaries based on power device junction temperature constraints and grid-connected power quality constraints: The overcurrent safety capability of GFM is not a fixed constant, but is determined by the thermal state of the power devices and the control actions. To characterize this overcurrent capability, this invention defines the maximum current that the converter can safely output under junction temperature constraints as the transient thermal safety boundary, i.e.: (6) In the formula, This is the transient thermal safety boundary. T j,k ( t ) is the first k Each device at time t Dynamic junction temperature; T j,max THD is the maximum allowable junction temperature of the device; THD is the harmonic distortion rate of the grid-connected voltage. lim Max is the upper limit of harmonic distortion rate; where max is the upper limit of harmonic distortion rate. k,t T j,k ( t This represents the maximum dynamic junction temperature of all power devices throughout the entire fault process.

[0034] S3, based on the relationship between switching frequency and device losses, extends the global thermal safety boundary through transient frequency reduction: Switching frequency affects switching losses, thereby altering the dynamic evolution of device junction temperature. To analyze the impact of switching frequency on device junction temperature and transient thermal safety margin, Figure 3 The junction temperature characteristic curves of the device at different switching frequencies are presented.

[0035] Depend on Figure 3 It can be seen that as the switching frequency decreases, the junction temperature characteristic curve of the device shifts downward as a whole, and its intersection with the upper limit of junction temperature of 150℃ shifts to the right, indicating that reducing the switching frequency can effectively expand the global transient thermal safety boundary of the converter.

[0036] The adjustable range of the switching frequency is constrained by both hardware capabilities and output performance. Upper limit. f max Limited by the device's own switching speed and maximum allowable junction temperature; lower limit f min It is mainly determined by the output voltage harmonic distortion rate (THD), which can be expressed as: (7) In the formula, t on , t off THD refers to the device's turn-on and turn-off times. lim This is the upper limit of harmonic distortion allowed by the grid connection standard.

[0037] According to GB / T14549, the harmonic distortion rate (THD) of the grid connection point voltage should be controlled within 5%. Simulation results based on the system parameters of this invention show that when the switching frequency is below 3kHz, the THD of the grid connection point voltage will exceed the standard limit. Therefore, this invention selects 3kHz as the lower limit for fault transient frequency reduction control to balance improving thermal safety boundaries and maintaining power quality.

[0038] S4, based on device junction temperature feedback and thermal sensitivity matrix, performs adaptive modulation switching to achieve local hotspot redistribution: Although transient frequency reduction can effectively reduce overall system switching losses and raise the system's transient thermal safety margin, for ANPC topologies, the thermal stress distribution among power devices is not uniform. Local hotspot devices often reach their junction temperature limits before other devices, thus becoming a bottleneck restricting the converter's transient overcurrent capability. Therefore, it is necessary to redistribute local hotspots through modulation strategy switching.

[0039] This invention introduces a thermal sensitivity matrix to describe the cooling benefits of different modulation schemes for various power devices. Let the reference modulation scheme be... 0, Switch to modulation mode Afterwards, the device The cooling benefit is defined as: (8) If Δ T k,m >0 indicates the modulation method m For devices k It has a cooling effect; Δ T k,m The larger the value, the more significant the thermal mitigation effect of the modulation method on the device. Furthermore, a thermal sensitivity matrix can be constructed: (9) The rows correspond to the devices, and the columns correspond to the modulation methods. If the current hotspot device is... k The optimal modulation scheme is: m * =argmaxΔ T k*,m This means selecting the modulation method that maximizes the cooling benefit for the currently hot-spotted device in order to achieve hotspot migration.

[0040] To quantitatively analyze the influence of different modulation strategies on the thermal stress distribution characteristics of power devices, this invention compares and analyzes the junction temperature distribution of each IGBT in phase A under rated operating conditions using SPWM, SVPWM, and various discontinuous pulse width modulation strategies (DPWM). The results are as follows: Figure 4 As shown.

[0041] Depend on Figure 4 It is evident that the thermal stress distribution of each device in the A-phase bridge arm differs significantly under different modulation strategies, indicating that the modulation method significantly affects the location and temperature rise level of hotspot devices. By appropriately switching the modulation strategy, local hotspots can be migrated from high-temperature devices to devices with larger thermal margins, thereby achieving thermal stress redistribution and providing support for expanding the transient thermal safety boundary of the converter. Subsequent simulation results show that, under junction temperature constraints...T j,max At ≤150℃, the combined effect of transient frequency reduction and adaptive modulation switching can increase the maximum safe output current of the converter from 1.5pu to 2.3pu, which is traditionally conservatively current-limited.

[0042] S5, Constructing a collaborative control framework for thermal safety boundary extension and boundary tracking: This invention proposes, as follows Figure 5 The thermal boundary extension and boundary tracking collaborative control framework shown in the figure has an upper-layer active thermal management for thermal boundary extension and a lower-layer DDPG adaptive virtual impedance control for boundary tracking. The two work together to release transient support capabilities and achieve smooth recovery throughout the fault process.

[0043] In terms of control implementation, the upper-level thermal boundary extension module outputs transient thermal safety boundaries based on the device's thermal state and active thermal management actions. This information is then sent to the lower-level controller as a boundary tracking reference. Upper-level active thermal management comprises two parts: transient frequency reduction and adaptive modulation switching based on junction temperature feedback. Transient frequency reduction is used to quickly reduce overall switching losses in the early stages of a fault, raising the transient thermal safety boundary globally. Furthermore, an adaptive modulation switching mechanism based on junction temperature feedback is introduced. When the device junction temperature exceeds a warning threshold... T warn At the same time, the controller identifies the current hot spot device and calls its corresponding optimal modulation mode, while setting a minimum hold time to avoid frequent switching, realizing the dynamic migration of local hot spots between the outer tube, inner tube and clamping tube, thereby further expanding the transient thermal safety boundary of the converter.

[0044] The lower-level control decision-making layer uses transient thermal safety boundaries. To achieve the desired result, the virtual impedance parameters are adjusted online based on the voltage drop depth at the grid connection point, the output current, and its changing trend. Virtual impedance control, by introducing an adjustable equivalent impedance on the GFM output side, achieves flexible regulation of the fault current. Its specific form can be expressed as: (10) In the formula, u sdref , u sqref Voltage loops d , q Shaft voltage reference value; E d and E q These are the virtual internal potentials. d , q Axial components; i d , i q They are respectivelyd , q Axis current components, R v and X v These are virtual resistance and virtual reactance, respectively. X v= ωL v .

[0045] From the perspective of the entire fault process, the lower-level boundary tracking mechanism can be divided into three stages: the initial stage of the fault mainly focuses on suppressing the current surge, which requires appropriately increasing the virtual impedance to reduce the impact; the continuous stage of the fault mainly focuses on releasing the support capability, which requires gradually adjusting the virtual impedance under the thermal boundary constraint to make the output current approach the threshold. During the fault clearing phase, the virtual impedance needs to be readjusted to suppress current fluctuations during recovery and achieve a smooth transition. Thus, upper boundary expansion and lower boundary tracking work in coordination to ensure the safe release of transient support capabilities throughout the fault process.

[0046] S6 employs an adaptive virtual impedance controller to adjust the virtual impedance parameters online, enabling the output current to dynamically track the extended transient thermal safety boundary. Within the aforementioned collaborative framework, to achieve online continuous optimization and adjustment of the lower-level virtual impedance parameters, this invention further employs the DDPG algorithm to construct an adaptive virtual impedance controller.

[0047] The DDPG algorithm uses an Actor-Critic architecture to handle the continuous action space problem. The key design of the agent is as follows: (1) State Space Design s t ) To reflect the severity of the fault, the output current state, and the boundary tracking requirements, the agent is defined at time [time value missing]. t The state variables are s t : (11) In the formula, ΔU This refers to the voltage drop depth. I ( t () represents the amplitude of the GFM output current; I’ ( t () represents the rate of change of current; e I ( t The tracking error represents the tracking performance of the lower-level controller on the transient thermal safety boundary.

[0048] (12) (2) Action Space Design a t ) This invention uses virtual impedance as the action output of the intelligent agent, that is: (13) in, v , v The agent at time 1 t The output provides reference values ​​for virtual resistance and virtual inductance. The output is first normalized to the [0,1] interval, and then linearly mapped to obtain the actual virtual impedance parameters. This method enables continuous and adjustable control of the virtual impedance, avoiding control values ​​exceeding the engineering implementation range.

[0049] (3) Reward Function Design r t ) reward function r t It is the core of guiding the agent's learning strategy. Designed as follows: (14) in, w 0 is the base reward coefficient. w 1. w 2 and w 3 is the weighting coefficient, Δ a t This represents the change in action between adjacent control moments. The first term is the basic reward; the second term is used to penalize current tracking errors; the third term applies a stronger penalty when the current crosses the thermal boundary reference to prevent the output current from exceeding the maximum allowable range; and the fourth term is used to suppress drastic changes in the virtual impedance parameter, improving the smoothness and feasibility of the control process.

[0050] (4) Training and online deployment The training and interaction process of DDPG agents is as follows: Figure 6 As shown in the figure. The left side of the figure illustrates the interaction process between the intelligent agent and the simulation environment, at time... t The Actor network is based on the current state. s t Output Action a t The environment transitions to the next state under the action. s t+1 And return the reward r tThe right side illustrates the network parameter update process. The Critic network estimates action value using interactive samples and constructs a loss function in conjunction with the target network. It continuously optimizes network parameters through gradient updates and soft updates. After multiple rounds of iterative training, the agent gradually learns the mapping relationship between voltage drop depth, output current change trends, and virtual impedance adjustment, thus forming a virtual impedance adjustment strategy for transient thermal safety boundary tracking.

[0051] During the online deployment phase, only the trained Actor network is invoked, and output is quickly based on real-time measured system status. v , v This enables online adaptive tuning of the virtual impedance parameter.

[0052] (5) Training Result Analysis To verify the training effect of the DDPG agent, this invention records the changes in the cumulative reward per round and the average cumulative reward during the training process. The results are as follows: Figure 7 As shown.

[0053] Depend on Figure 7 It is evident that in the early stages of training, the agent had not yet established an effective virtual impedance regulation pattern, resulting in significant fluctuations in the cumulative reward per round and a low average cumulative reward. As the number of training rounds increased, the average cumulative reward rose rapidly and gradually entered a stable fluctuation range, indicating that the agent had established a mapping relationship between fault states and virtual impedance regulation.

[0054] The cumulative reward per round still fluctuates somewhat in the later stages of training, mainly due to the exploration mechanism and differences in the system's dynamic response under different fault states. Overall, the average cumulative reward tends to stabilize, indicating that the trained Actor network can form a relatively stable control strategy among boundary tracking, current over-limit suppression, and virtual impedance smoothing.

[0055] To verify the effectiveness of the proposed method in extending the thermal safety boundary, this invention uses an electrothermal co-simulation and hardware-in-the-loop experimental platform to verify the method under typical fault conditions and multiple operating conditions. The thermal parameters of the power devices are shown in Table 1, and the network control parameters are shown in Table 2.

[0056] Table 1 Thermal parameters of power module

[0057] Table 2 Key parameters of grid-connected energy storage converters

[0058] Example 1: Simulation analysis under typical fault conditions: System settings tA three-phase short-circuit fault occurred at 1.5s, and the grid connection voltage dropped to 0.5pu for 1s. To verify the effectiveness of the proposed method, the following three sets of comparative experiments were designed. Method A is a traditional conservative method, using fixed-parameter virtual impedance control without introducing active thermal management, and the thermal safety boundary is set at 1.5pu; Method B is a high overload method without thermal management, also using fixed-parameter virtual impedance control, but increasing the current limit value to 2.3pu, used to test the extreme thermal withstand capability of the converter under conditions without active thermal management, and to verify the risk of thermal runaway; Method C is the method proposed in this invention, which introduces DDPG online adjustment of virtual impedance and combines it with an active thermal management mechanism.

[0059] During a fault, the real-time action of active thermal management and the dynamic junction temperature response waveforms of power devices under different control methods are shown below. Figure 8 As shown.

[0060] Depend on Figure 8 (a) As can be seen, after the fault occurs, method C rapidly reduces the switching frequency from 12kHz to 3kHz to reduce the overall switching loss during the transient process and suppress the global temperature rise of the device. Subsequently, based on the junction temperature feedback, it adaptively switches between SVM, DPWM2, and DPWM0 to dynamically transfer the heat load between the outer layer, inner layer, and clamping device, avoiding the continuous accumulation of local hot spots. After the fault is cleared, the switching frequency and modulation method return to normal operation.

[0061] Junction temperature response of different methods, such as Figure 8 As shown in (b), Method A employs conservative current limiting, ensuring the device junction temperature remains below the thermal safety threshold, but its transient overcurrent support capability is not fully released. Method B increases the current limit to 2.3 pu, but due to the lack of active thermal management, the junction temperature of hot-spot devices rapidly exceeds the thermal safety threshold, posing a significant risk of thermal runaway. In contrast, Method C, under a 2.3 pu overload output, maintains the highest dynamic junction temperature of the device within the thermal safety threshold through active thermal management. The results demonstrate that the active thermal management method proposed in this invention can effectively release the transient overload support capability of the converter while ensuring device thermal safety. The dynamic response waveform of the three-phase current output of the converter during the fault is shown in Figure [Figure number missing]. Figure 9 As shown.

[0062] Depend on Figure 9It is evident that both Method A and Method B, employing a fixed virtual impedance, exhibited significant transient inrush currents at the moment of fault occurrence. Method A exceeded the current limit at 1.5 pu, while Method B, although increasing the current limit to 2.3 pu, still experienced a brief surge current exceeding the set boundary. This indicates that fixed-parameter virtual impedance struggles to simultaneously suppress current surges in the initial fault phase and release subsequent support capabilities. In contrast, Method C did not exhibit current overruns in the early stages of the fault, effectively confining the output current within the thermal boundary. This demonstrates that the DDPG agent can rapidly increase the virtual impedance based on the fault state, thereby suppressing current surges and improving system transient stability.

[0063] The dynamic response waveforms of the converter's output reactive power and grid connection point voltage during the fault period are as follows: Figure 10 As shown.

[0064] Depend on Figure 10 It can be seen that all three methods can provide reactive power support to the grid after a fault occurs, but their support capabilities differ significantly. Compared to method A, method C, due to its higher transient overcurrent carrying capacity, can output more reactive power, thus achieving a better voltage support effect. Quantitative results show that method C can generate an additional 0.69 pu of reactive power compared to method A, and improve the voltage support effect at the grid connection point by about 35%. The proposed method can effectively improve the reactive power support capability during converter faults under thermal safety constraints, thereby improving the voltage recovery level at the grid connection point.

[0065] Example 2, Hardware-in-the-loop experimental verification: To further verify the feasibility and effectiveness of the proposed method under real-time operating conditions, based on the aforementioned mechanism analysis and simulation verification, this invention builds a hardware-in-the-loop experimental platform based on the MT8020. The main circuit is deployed in the FPGA module of the MT8020 to realize fast electromagnetic transient simulation, and the control algorithm is compiled and deployed in the CPU module of the MT8020 to complete the control decision calculation. The host computer is only used for waveform monitoring and data recording during the experiment.

[0066] To verify the real-time adaptability of the proposed method to different fault severity levels, the grid connection point voltage was set to drop to 0.6 pu, 0.4 pu, and 0.2 pu while maintaining consistent system parameters, control parameters, and fault duration. The output current and grid connection point voltage responses under different fault depths were compared and analyzed. The semi-physical experimental results are as follows: Figure 11 As shown.

[0067] Depend on Figure 11 (a) It is evident that the proposed method effectively supports the grid connection point voltage under various voltage dip conditions. Figure 11(b) As can be seen, after the fault occurred, the output current rose rapidly and stabilized near the thermal safety boundary. Under the 0.2 pu and 0.4 pu drop conditions, the current did not exceed the 2.3 pu limit. Under the 0.6 pu drop condition, due to the relatively low voltage support requirement, the output current did not reach the upper limit. Further... Figure 11 (c) As can be seen, the virtual impedance can adaptively adjust according to the severity of the fault. The deeper the voltage drop, the larger the value of the virtual impedance, in order to enhance the current suppression capability in the early stage of the fault. During the fault duration, the virtual impedance is gradually adjusted and maintained at the corresponding level, so that the output current does not exceed the thermal safety boundary. The above results show that the proposed method can achieve adaptive adjustment of virtual impedance under semi-physical real-time closed-loop conditions, while taking into account both fault current constraints and grid connection point voltage support requirements.

[0068] To verify the boundary tracking capability of the proposed method under continuous time-varying fault conditions, the following continuous fault condition was set: First, the grid voltage drops to 0.4 pu for 0.5 s; then the fault deepens to 0.2 pu for 0.5 s; subsequently, the fault eases to 0.6 pu for 0.5 s; finally, the fault ends and the system returns to normal operation. This condition can simultaneously cover the dynamic process of fault deepening, fault easing, and fault clearing, and better reflects the response capability of the control method to time-varying faults under real-time operating conditions. Experimental results are as follows: Figure 12 As shown.

[0069] Depend on Figure 12 As can be seen, under continuous voltage dip conditions, the proposed method can adaptively adjust the virtual impedance parameter according to the grid connection point voltage change, ensuring that the output current remains within the transient thermal safety boundary. When the fault depth increases from 0.4 pu to 0.2 pu, the virtual impedance increases accordingly to suppress current overshoot; when the fault depth decreases to 0.6 pu and the fault is cleared, the virtual impedance gradually decreases, and the output current recovers smoothly. The results show that the proposed method has good dynamic adaptability and current constraint capability under continuous fault conditions.

[0070] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A voltage support method for grid-connected energy storage converters based on thermal safety boundary extension, characterized in that, The method specifically includes the following steps: S1. Establish an electrothermal coupling model for a grid-type energy storage converter, which includes a power device loss model and a thermal network model. S2. Based on the junction temperature constraints of the power devices and the grid-connected power quality constraints, define the transient thermal safety boundary of the grid-type energy storage converter; S3. Based on the relationship between the switching frequency and the power device loss, the global thermal safety boundary is extended through transient frequency reduction control; S4. Based on the junction temperature feedback and thermal sensitivity matrix of the power device, adaptive modulation switching is performed to realize the redistribution of local hot spots, so as to further extend the transient thermal safety boundary. S5. Construct a collaborative control framework for thermal safety boundary extension and boundary tracking, the framework including an upper-layer active thermal management module and a lower-layer adaptive virtual impedance controller; S6. The upper-layer active thermal management module outputs an extended transient thermal safety boundary based on the thermal state of the power device, and the lower-layer adaptive virtual impedance controller adjusts the virtual impedance parameters online so that the output current of the grid-type energy storage converter dynamically tracks the extended transient thermal safety boundary.

2. The voltage support method for grid-connected energy storage converters based on thermal safety boundary extension according to claim 1, characterized in that, In step S1, the loss model includes a conduction loss model and a switching loss model; the conduction loss is calculated by averaging the product of the device current and its saturation voltage drop over one fundamental frequency period; the switching loss is calculated by averaging the sum of the turn-on loss and the turn-off loss over one fundamental frequency period, and the switching loss is proportional to the switching frequency. The thermal network model adopts the Foster equivalent thermal network model, which equates the internal heat conduction process of the power module to a dynamic network composed of series and parallel thermal resistance and thermal capacity. The loss of the power device is injected into the thermal network as a heat source and is conducted step by step through the heat transfer path to form the dynamic junction temperature response of the power device.

3. The voltage support method for grid-connected energy storage converters based on thermal safety boundary extension according to claim 2, characterized in that, In step S2, the transient thermal safety boundary is defined as: the maximum current value that the grid-type energy storage converter can safely output under the condition that the maximum dynamic junction temperature of the power device does not exceed the maximum allowable junction temperature and the harmonic distortion rate of the grid connection point voltage does not exceed the set upper limit.

4. The voltage support method for grid-connected energy storage converters based on thermal safety boundary extension according to claim 3, characterized in that, In step S3, the transient frequency reduction control specifically involves reducing the switching frequency of the grid-type energy storage converter from the rated frequency to a preset lower limit frequency when a fault occurs, in order to reduce the overall switching loss during the transient process and raise the global thermal safety boundary; the preset lower limit frequency is jointly determined by the switching speed of the power device and the harmonic distortion rate constraint of the grid connection point voltage.

5. The voltage support method for grid-connected energy storage converters based on thermal safety boundary extension according to claim 4, characterized in that, In step S4, the adaptive modulation switching specifically includes: S41. The thermal sensitivity matrix is ​​pre-constructed, where the rows of the matrix correspond to different power devices, the columns correspond to different modulation methods, and the matrix elements represent the junction temperature reduction benefit of the corresponding device after switching from the reference modulation method to the target modulation method. S42. Monitor the junction temperature of each power device in real time. When the junction temperature of any power device exceeds a preset warning threshold, identify the device as the current hot spot device. S43. Based on the thermal sensitivity matrix, select the modulation method that has the greatest cooling benefit for the current hot spot device as the optimal modulation method, and switch to the optimal modulation method to realize the dynamic migration of local hot spots.

6. The voltage support method for grid-connected energy storage converters based on thermal safety boundary extension according to claim 5, characterized in that, In step S5, virtual impedance control achieves flexible regulation of fault current by introducing adjustable equivalent resistance and equivalent reactance on the output side of the grid-type energy storage converter.

7. The voltage support method for grid-connected energy storage converters based on thermal safety boundary extension according to claim 6, characterized in that, In step S6, the lower-level adaptive virtual impedance controller is constructed using a deep deterministic policy gradient algorithm. The agent of the deep deterministic policy gradient algorithm uses the voltage drop depth, the output current amplitude, the rate of change of the output current, and the tracking error of the output current to the transient thermal safety boundary as the state space, the reference value of the virtual impedance as the action space, and designs a reward function with the goal of minimizing the current tracking error, suppressing the current over-limit, and smoothing the virtual impedance adjustment.

8. The voltage support method for grid-connected energy storage converters based on thermal safety boundary extension according to claim 7, characterized in that, The transient frequency reduction control and adaptive modulation switching of the upper active thermal management module work together to increase the maximum safe output current of the grid-type energy storage converter from the traditional current limiting value to the preset target overload value.

9. A method for voltage support of a grid-connected energy storage converter based on thermal safety boundary extension as described in claim 8, characterized in that, The main circuit topology of the grid-type energy storage converter is a three-level active midpoint clamping topology, and its control method is virtual synchronous generator control.

Images