Method and system for improving overload capability of network configuration type converter based on DPWM

Abstract

The application belongs to the technical field of power electronic intelligent control, and provides a network type converter overload capacity improvement method and system based on DPWM, which comprises the following steps: obtaining a converter d / q axis current instruction and synthesizing a current instruction included angle, constructing a data set of the current instruction included angle, a clamping offset angle and a junction temperature, and establishing a direct nonlinear mapping of the current instruction included angle to the optimal clamping offset angle through an artificial neural network model; under an overload condition, the real-time current instruction included angle is used to quickly output the optimal clamping offset angle from the model; the optimal clamping offset angle is injected into an electric angle for sector selection, combined results are used to generate a zero sequence component injection voltage instruction, and a clamping mode is dynamically switched to align the clamping interval with a current peak value area, so that the insulated gate bipolar transistor peak value section zero switching action is realized to suppress the junction temperature. The application simplifies calculation, improves response speed, accurately suppresses the junction temperature, and significantly improves the converter overload capacity.

Description

Technical Field

[0001] This invention belongs to the field of power electronics intelligent control technology, and provides a method and system for improving the overload capacity of grid-type converters based on DPWM. Background Technology

[0002] With the rapid development of high-proportion renewable energy power systems, grid-connected converters, with their active voltage and frequency support capabilities, have become core equipment for renewable energy grid integration, microgrids, and distribution network stability control. When the grid experiences voltage dips, short-circuit faults, or severe load fluctuations, grid-connected converters need to operate under short-term overload conditions to ensure grid stability. At this time, the converter output current increases sharply, and the thermal stress on power devices rises significantly, making them highly susceptible to failure due to excessive junction temperatures. Therefore, improving the thermal safety of devices and the overload tolerance of the converter under overload conditions has become a key technical requirement for the engineering application of grid-connected converters.

[0003] Discontinuous pulse width modulation (DPWM) is widely used to suppress junction temperature in power devices due to its ability to reduce switching actions and losses. Its core lies in the rational setting of the DPWM clamp offset angle to match the clamping interval with the loss hotspot for optimal thermal control. Existing DPWM clamp angle optimization methods are mostly based on analytical calculations using the switching loss function (SLF). This involves acquiring voltage and current signals to solve for the power factor angle, and then determining the optimal clamp angle based on a loss model. However, these methods have significant drawbacks: First, traditional analytical models only consider switching losses and cannot fully characterize the mapping relationship between total device losses and junction temperature, leading to the optimal clamp angle deviating from the actual minimum. Second, the control process requires the introduction of multi-dimensional variables such as voltage phase angle, current phase angle, and power factor angle, resulting in complex calculations and a large real-time computational load. This leads to significant lag during transient processes with millisecond-level surges in overload current, hindering rapid response. Third, traditional methods rely on multi-physical quantity sampling and complex model solving, resulting in high system hardware and software complexity. Furthermore, in network-based control architectures, they are susceptible to parameter drift, making it difficult to meet the thermal control accuracy and robustness requirements under overload conditions. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method and system for improving the overload capacity of a grid-type converter based on DPWM, which simplifies calculations, improves response speed, accurately suppresses junction temperature, and significantly enhances the overload capacity and operational reliability of the converter.

[0005] The technical solution of the present invention includes:

[0006] Obtain the d-axis current command and q-axis current command of the grid-type converter, and combine the d-axis current command and q-axis current command into a current command angle.

[0007] A dataset of current command angle and clamp offset angle with junction temperature is constructed based on multiphysics thermal simulation. Then, a direct nonlinear mapping from current command angle to optimal clamp offset angle is established through an artificial neural network model.

[0008] Under converter overload conditions, the optimal clamping offset angle is quickly output through an artificial neural network model based on the real-time current command angle.

[0009] The three-phase voltage command is subjected to extreme value extraction, and the three-phase voltage command is subjected to coordinate transformation to obtain the fundamental phase angle. The optimal clamping offset angle is superimposed with the fundamental phase angle to select the sector. The zero-sequence component is injected into the voltage command by combining the sector selection and extreme value extraction results. The clamping mode is dynamically switched so that the clamping interval is aligned with the current peak region. Then, the insulated gate bipolar transistor operates with zero switching in the current peak segment to suppress junction temperature, thereby improving the overload capacity of the grid converter.

[0010] Furthermore, the formula for the angle between the combined current command of the d-axis current command and the q-axis current command is as follows:

[0011] ;

[0012] In the formula, This is the d-axis current command. This is a q-axis current command. The angle between the current command and the ...

[0013] Furthermore, the dataset is represented as:

[0014] ;

[0015] The mapping relationship between junction temperature and input variables can be expressed as:

[0016] ;

[0017] In the formula, This is the clamping offset angle. For the junction temperature, For the dataset, This represents the minimum value of the current command angle scan. This represents the maximum value of the current command angle scan. This is the minimum value of the clamping offset angle. This represents the maximum value of the clamping offset angle.

[0018] Furthermore, the direct nonlinear mapping is:

[0019] ;

[0020] In the formula, The optimal clamping offset angle is... It is an artificial neural network model.

[0021] Furthermore, aligning the clamping interval with the current peak region specifically includes extreme value selection, coordinate transformation, sector selection, and zero-sequence component injection.

[0022] Furthermore, the extreme values ​​are selected as follows:

[0023] ;

[0024] In the formula, This is the three-phase reference voltage input.

[0025] Furthermore, the fundamental phase angle is obtained through coordinate transformation, using the following formula:

[0026] ;

[0027] In the formula, The fundamental phase angle, For the α-axis voltage component, This represents the β-axis voltage component.

[0028] Furthermore, the formula for sector selection is:

[0029] ;

[0030] In the formula, To control the angle, For logical sectors.

[0031] Furthermore, the formula for zero-order component injection is:

[0032] ;

[0033] In the formula, For the final modulated wave, .

[0034] This invention also provides a DPWM-based overload capacity enhancement system for grid-connected converters, implementing the aforementioned DPWM-based overload capacity enhancement method for grid-connected converters, comprising:

[0035] The feature extraction module is used to obtain the d-axis current command and q-axis current command of the grid-type converter and synthesize them to obtain the current command angle.

[0036] The electrothermal mapping module is used to construct a dataset of current command angle, clamp offset angle and junction temperature based on multiphysics thermal simulation data, and to establish a direct nonlinear mapping from the current command angle to the optimal clamp offset angle through an artificial neural network model.

[0037] The real-time decision module is used to output the optimal clamping offset angle based on the real-time current command angle through an artificial neural network model under converter overload conditions.

[0038] The dynamic modulation execution module is used to inject the optimal clamp offset angle into the modulation signal generation path. By dynamically switching the clamp mode through sector logic, the clamp interval is aligned with the current peak region, so that the insulated gate bipolar transistor can achieve zero switching operation in the current peak segment to suppress junction temperature.

[0039] The technical solution provided by this invention has the following advantages compared with the prior art:

[0040] 1. By extracting the d / q axis current commands of the grid-type converter and synthesizing a single current command angle, feature dimensionality reduction is achieved. This method abandons the complex calculations required by traditional methods that introduce multi-dimensional variables such as voltage phase angle and current phase angle. It combines a dataset constructed by multiphysics thermal simulation with a direct nonlinear mapping from the current command angle to the optimal clamping offset angle established by an artificial neural network, which fully characterizes the mapping relationship between the total device loss and junction temperature. This solves the problem that the optimal clamping angle deviates from the actual minimum point due to the traditional analytical model only considering switching losses.

[0041] 2. Under overload conditions, the optimal clamping offset angle can be quickly output from the model. Then, the extreme values ​​of the three-phase voltage command are extracted, coordinates are transformed, and the optimal clamping offset angle is injected into the electrical angle for sector selection. The results are combined to generate a zero-sequence component and inject voltage command to dynamically switch the clamping mode. This allows the clamping interval to be precisely aligned with the current peak area, achieving zero-switching action of the IGBT peak segment. This eliminates the problem of overload transient response lag caused by the complexity of the calculation process and the large amount of real-time computation in traditional methods. At the same time, it simplifies the sampling of multiple physical quantities and the solution of complex models, reduces the complexity of system software and hardware, reduces the impact of parameter drift under the grid-type control architecture, and improves the accuracy and robustness of thermal control. Finally, by accurately suppressing junction temperature, the overload capacity and operational reliability of the grid-type converter are significantly improved.

[0042] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 The waveforms are shown for different traditional DPWM modulation methods.

[0045] Figure 2 This is a flowchart of the overload capacity improvement method and system control for grid-type converters based on DPWM proposed in this invention.

[0046] Figure 3 This is a block diagram of the overload capacity improvement method for grid-type converters based on DPWM according to the present invention.

[0047] Figure 4 This invention presents a method for improving the overload capacity of grid-type converters based on DPWM and an optimal control model for the system.

[0048] Figure 5 This is a control block diagram of the DPWM modulation strategy based on variable clamping angle proposed in this invention.

[0049] Figure 6 The scatter plot shows the effect of selecting the optimal αopt in the ANN model proposed in this invention.

[0050] Figure 7 The theoretical value of SLF proposed in this invention is compared with the optimal αopt of the proposed scheme under the same working conditions.

[0051] Figure 8 This invention proposes a model to investigate the IGBT junction temperature fluctuation under different DPWM offset angles.

[0052] Figure 9 The model proposed in this invention has loss waveforms in αopt and non-optimal αopt. (a) shows the relationship between power loss and time in the on-state loss, (b) shows the relationship between power loss and time in the switching loss, (c) shows the relationship between power loss and time in the on-state loss, and (d) shows the relationship between power loss and time in the switching loss. Detailed Implementation

[0053] The following detailed description of a specific embodiment of the present invention is provided in conjunction with the accompanying drawings. However, it should be understood that the scope of protection of the present invention is not limited to the specific embodiment.

[0054] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the technical solution of this 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, they should not be construed as limitations on this invention.

[0055] In the description of the embodiments of the present invention, unless otherwise stated, "a plurality of" means two or more.

[0056] In grid-connected (GFM) converters, IGBTs (Insulated Gate Bipolar Transistors) serve as the core switching devices, and their current-carrying capacity directly determines the upper limit of the equipment's overload capacity in the face of grid faults or severe disturbances. Due to the voltage source characteristics of GFMs, severe overcurrent phenomena are easily triggered under faults such as grid voltage dips, leading to increased internal electrothermal stress in the power devices. If the junction temperature rise of the power devices caused by the current surge cannot be effectively suppressed, the devices will fail due to thermal breakdown. Therefore, optimizing the modulation strategy to reduce the junction temperature of the devices during overload periods is crucial to improving the transient overload tolerance of the converter.

[0057] The reason why there is a close relationship between the IGBT device level and the network system level is that this relationship is reflected in the fact that the power loss of the IGBT directly affects the junction temperature. T j This limitation restricts the overall performance and reliability of the entire network-type system. This invention, by optimizing the modulation strategy, compared to the one-sidedness of traditional strategies that rely solely on switching loss prediction, fully considers the impact of conduction losses and complex thermal dynamics. It directly uses minimizing junction temperature as the global optimization objective, achieving high-precision, real-time thermal management across the entire power factor range. This effectively reduces junction temperature, thus having a beneficial impact on the network-type system. The following section provides a detailed analysis and explanation of losses at the device level.

[0058] IGBTs, as high-efficiency power switching devices, are widely used in frequency converters, electric vehicle drive systems, and renewable energy inverters. However, IGBTs generate significant power losses during operation, leading to thermal stress and device aging. According to literature, the total power loss of an IGBT is mainly divided into two parts: switching losses and conduction losses. These losses directly affect the junction temperature of the device, thus limiting the overall performance of the system.

[0059] Traditional methods manage heat through passive cooling (such as air cooling or water cooling), but cannot dynamically respond to load changes. This invention proposes an active thermal control strategy that optimizes switching losses and conduction losses in real time by adjusting the switching frequency and gate drive voltage, respectively, to achieve more precise thermal management.

[0060] Total power loss of IGBT ( P total It can be decomposed into switching losses ( P sw ) and on-state loss ( Pcond The two most important components are switching loss and conduction loss. Switching loss is the transient loss generated by the IGBT during turn-on and turn-off. It mainly originates from the charge storage effect inside the device and the charging / discharging process of parasitic capacitance. Switching loss is closely related to the switching frequency, load current, and DC bus voltage. In high-frequency applications, switching loss often accounts for a large proportion of the total loss. Conduction loss, on the other hand, is the steady-state loss generated by the IGBT in the on state. It is mainly composed of saturation voltage drop (…). V ce ) and conduction current ( I c The ohmic losses caused by the switching voltage (IGBT) constitute the total power loss. On-state losses are related to the duty cycle and gate drive voltage, and are often more significant in low-frequency or high-load applications. In addition, the total IGBT losses may also include smaller reverse recovery losses and drive losses, but these are usually included in the switching losses. Therefore, the total power loss can be expressed as:

[0061] ;

[0062] Switching losses include turn-on losses ( E on ) and shutdown loss ( E off Energy values ​​are typically obtained directly from datasheets. Average switching losses can be calculated as follows:

[0063] ;

[0064] In the formula, E on This represents the energy loss during a single activation. E off This represents the energy loss during a single shutdown. f sw This represents the switching frequency.

[0065] On-state loss P cond The power loss generated by the IGBT during conduction (i.e., when the gate drive signal is high) mainly originates from the saturation voltage drop between the collector and emitter. V ce With the current through the IGBT power device I c and the duty cycle of PWM D The product of P cond It can be calculated as follows:

[0066] ;

[0067] In the formula, V ceIt is not a constant value; it varies with different gate drive voltages. I c It exhibits a certain non-linear relationship.

[0068] Based on the above, the total loss of the IGBT can be calculated as follows:

[0069] ;

[0070] Traditional methods manage heat through passive cooling (such as air cooling or water cooling), but cannot dynamically respond to load changes. In existing technologies, Discontinuous Pulse Width Modulation (DPWM) is a modulation technique that significantly reduces total system losses by stopping switching operations within a specific sector of each fundamental cycle. This invention deeply integrates the DPWM strategy into the control architecture. Its core thermal control mechanism lies in the fact that in three-phase current, each phase current has two regions with the largest amplitude (peak regions) within the fundamental cycle. DPWM modulation technology can clamp the switching operation of this phase near the current peak region through modulation logic, that is, keeping the switching transistor in the on or off state within the clamping region.

[0071] like Figure 1 The diagram illustrates four commonly used DPWM modulation waveforms: 0, 1, 2, and 3. The blue dashed line represents the original modulation waveform, the green curve represents the calculated injected zero-sequence component, and the red curve represents the DPWM modulation waveform after injecting the zero-sequence component into the original modulation waveform. It can be seen that the DPWM waveform is clamped to the positive or negative terminal of the DC bus in certain intervals. Within the clamped interval, the switching devices do not operate, resulting in reduced switching losses for that phase during that time period. P sw The current will drop to zero. Under overload fault conditions of grid-connected converters, due to the surge in reactive current, the output current vector deviates drastically from the voltage vector. The clamping interval calculated based on the fixed power factor angle in the traditional method often deviates from the current peak value, which cannot minimize losses. At this time, the traditional fixed clamping mode will fail.

[0072] like Figures 2 to 9 As shown, this invention provides a method for improving the overload capacity of a grid-type converter based on DPWM, comprising:

[0073] Obtain the d-axis current command and q-axis current command of the grid-type converter, and combine the d-axis current command and q-axis current command into a current command angle.

[0074] A dataset of current command angle and clamp offset angle with junction temperature is constructed based on multiphysics thermal simulation. Then, a direct nonlinear mapping from current command angle to optimal clamp offset angle is established through an artificial neural network model.

[0075] Under converter overload conditions, the optimal clamping offset angle is quickly output through an artificial neural network model based on the real-time current command angle.

[0076] The three-phase voltage command is subjected to extreme value extraction, and the three-phase voltage command is subjected to coordinate transformation to obtain the fundamental phase angle. The optimal clamping offset angle is superimposed with the fundamental phase angle to select the sector. The zero-sequence component is injected into the voltage command by combining the sector selection and extreme value extraction results. The clamping mode is dynamically switched so that the clamping interval is aligned with the current peak region. Then, the insulated gate bipolar transistor operates with zero switching in the current peak segment to suppress junction temperature, thereby improving the overload capacity of the grid converter.

[0077] In the embodiments provided by this invention, the dataset is represented as follows:

[0078] ;

[0079] The mapping relationship between junction temperature and input variables can be expressed as:

[0080] ;

[0081] In the formula, This is the clamping offset angle. .

[0082] This invention essentially aims to reduce total loss by optimizing the modulation strategy. P total Active intervention to reduce junction temperature T j The artificial neural network (ANN) used in this invention is a mathematical model designed to simulate the ability of biological nervous systems to process complex nonlinear mappings. The global nonlinear mapping relationship from the input layer to the output layer of an ANN can be expressed by the following general nested matrix formula:

[0083] ;

[0084] In the formula, x is the input vector of the network; y is the output vector of the network; L W represents the total number of layers in the network, including hidden and output layers. k For the first k The layer's weight matrix represents the connection strength between neurons. k For the first k The layer's bias vector is used to adjust the activation threshold; f k For the first k The activation function of the layer.

[0085] After the ANN model is trained, the above formula is solidified into the online execution logic of the controller. Only simple matrix multiplication and addition operations are needed to complete the mapping from input to output in milliseconds, without the need to solve complex physical equations in real time.

[0086] In the DPWM control system proposed in this invention, the ANN is used as the core inference engine. Its main task is to capture the feature vector of the converter output current in real time and output the optimal pulse width modulation offset parameters. The core input variable of the ANN is the current command angle, and the output variable is the optimal DPWM offset angle. α opt The specific mapping relationship can be represented as:

[0087] ;

[0088] In the formula, The optimal clamping offset angle is... It is an artificial neural network model.

[0089] like Figure 2 As shown, firstly, the three-phase AC current signal is acquired in real time using a sensor. i abc Then, through coordinate transformation (Clark & ​​Park transformation), the time-varying AC signal is converted into the DC component d-axis in a synchronous rotating coordinate system. i d and the q-axis i q Unlike traditional methods that require complex power calculations, this solution directly extracts the command value of the current loop, specifically the d-axis current command. and q-axis current command This is a core feature. The advantage of doing so is that it allows for the prediction of current change trends and eliminates the influence of sensor noise on the input of the artificial neural network (ANN).

[0090] exist Figure 2 As can be seen, under the traditional modulation strategy, it is necessary to calculate the current command angle at the converter output side. i and the angle of internal potential u Generally by extraction , , u dref , u qref The corresponding included angle is then calculated to obtain the power factor angle. The calculation method is as follows:

[0091] ;

[0092] ;

[0093] ;

[0094] In power systems with renewable energy integration, grid-connected converters, by simulating the external characteristics of synchronous generators, possess the active capability to provide voltage and frequency support. However, when the converter operates under overload or a short-circuit fault occurs in the grid, the reactive current output by the grid-connected converter surges in order to maintain grid voltage stability. This invention addresses the characteristics of a decreasing internal potential angle and an increasing current command angle under this operating condition, pointing out that the change in the current command angle plays a decisive role in DPWM heat loss analysis. Based on this dominant factor, this invention ignores the internal potential angle variable and uses the current command angle as the dominant variable, achieving variable-dimensionality reduction control and improving the speed and efficiency of control response.

[0095] This invention, based on multiphysics thermal simulation experimental data, constructs an electrothermal coupling simulation model of a converter, covering a wide range of current command angles. and clamp offset angle α Lower junction temperature Scanning and data acquisition are performed to build an initial dataset. The dataset is then learned and simulated using an ANN, and real-time output of the junction temperature data is generated. The minimum optimal clamping offset angle improves the real-time response speed of the system and enables the lowest junction temperature control of the device.

[0096] The control block diagram of the modulation strategy for adjusting the minimum peak junction temperature by changing the DPWM clamp angle mentioned in this invention is as follows: Figure 5 As shown.

[0097] First, the three-phase reference voltage is obtained through Clark transformation. Extracting the fundamental phase angle θ ,definition α This refers to the clamping offset angle. In traditional SVPWM, the device switches frequently throughout the entire cycle; however, this strategy adjusts the clamping offset angle. α This causes the device to generate a continuous 60Hz frequency within each half-fundamental cycle. 0 The non-operational range (clamping region) and the phase shift angle are adjusted. α Using synthetic control angle Then, the parity of the current sector is determined through sector logic. In the zero-sequence signal generation path, the system detects the maximum and minimum values ​​of the three-phase voltage in real time and calculates the upper clamping in parallel. With lower clamp The required zero-sequence component. Finally, the zero-sequence component selection logic module dynamically switches the clamping mode based on the sector parity, injects the selected zero-sequence voltage into the original reference wave, and generates the corrected modulation wave. Thus achieving the clamping range αA DPWM strategy with flexibly adjustable phase shift angle is used to optimize switching losses at a specific power factor. The first step in this process is extreme value extraction, as shown in the following equation:

[0098] ;

[0099] In the formula These are the three-phase reference voltage inputs. After extreme value extraction, the three-phase coordinate system is mapped to the stationary coordinate system using Clark transformation, and the fundamental phase angle is calculated. θ , can be represented as:

[0100] ;

[0101] In the formula, For the α-axis voltage component, This represents the β-axis voltage component.

[0102] Introducing an external control variable: clamp offset angle α Calculate the control angle used for logical judgment. ctrl :

[0103] ;

[0104] The space is divided into 6 sectors, each sector is 60... 0 Centered on, determine the current ctrl The logic sector that fell in k , k The range is 0~5, with a total of 6 sectors. k It can be represented as:

[0105] ;

[0106] Even sector ( k =0,2,4) corresponds to the peak region of the positive half-cycle. Odd-numbered sectors ( k =1,3,5) corresponds to the negative half-cycle peak region. The injected zero-sequence component... u z It can be represented as:

[0107] ;

[0108] Final modulated wave u * mod It can be calculated as follows:

[0109] ;

[0110] In the formula, This is the original, unclamped modulated wave. This refers to the injected zero-order component.

[0111] This embodiment introduces the theoretically optimal trajectory of the traditional switching loss function (SLF) as a benchmark reference. SLF is a metric used to evaluate the switching loss performance of different DPWM (Discontinuous Pulse Width Modulation) modes. Its physical essence is: measuring the proportion of switching losses that DPWM can reduce relative to traditional SVPWM (Continuous Pulse Width Modulation) at a specific power factor angle. Its mathematical expression is:

[0112] ;

[0113] In the formula, SLF represents the loss ratio of the DPWM strategy and the traditional SVPWM strategy at a certain power factor; is the system power factor angle. Since the traditional SLF theory is based on the power factor angle, while this invention uses a current-dominated approach... i Coordinate system. To achieve comparison within the same dimension, this invention maps the theoretically optimal trajectory of SLF to... i Space, for each specific current command angle i Operating point, combined with the system's filter inductance L By analyzing the output impedance parameters, the corresponding actual power factor angle is calculated, and then substituted into the SLF analytical function to obtain its theoretical optimal clamping offset angle.

[0114] By comparing and analyzing the predicted output values ​​of the ANN model with the optimal measured data obtained through scanning, the results are as follows: Figure 6 As shown, simulation results indicate that the ANN model's predicted output agrees well with the measured optimal data. Within the full power factor angle range (-90°), 0 Up to 90 0 The model's prediction error remained consistently at 1. 0 Within this range. This extremely low error performance fully demonstrates the excellent nonlinear fitting capability of the ANN model. Even when the grid-type converter enters extreme overload and current vector fluctuates drastically, the model can still maintain high-precision decision output, meeting the real-time and high-precision requirements of optimal control for the decision core.

[0115] To verify the proposed method of using an artificial neural network (ANN) to select the optimal clamping offset angle, this invention is presented. opt To demonstrate its scientific validity, under uniform operating conditions, the theoretical optimal trajectory of the traditional switching loss function (SLF) was compared with the output results of the proposed scheme (see [reference]). Figure 7 The comparison results show that the trajectory output by the ANN model is highly consistent with the traditional SLF baseline trajectory. This strongly confirms that the neural network model constructed in this invention can accurately capture and characterize the optimal thermal management law of the grid-type converter under different overload conditions.

[0116] Unlike traditional SLF strategies that rely on complex power factor angle calculations, this invention demonstrates through an ANN model that it is only necessary to extract the current command angle from the inner loop output of the control system. i As the dominant variable, feature dimensionality reduction can be achieved, bypassing traditional complex switching loss function models, and accurately obtaining the optimal clamping offset angle that minimizes the power device loss. opt This current-driven mapping mechanism reduces the lag in traditional computing links and significantly improves the system's response speed under conditions of current surges.

[0117] It is worth noting that, although Figure 7 The ANN output and the SLF theoretical value show slight differences in local values, but this invention proves through thermal analysis that these differences not only do not degrade the control effect, but also demonstrate the robustness of the algorithm: thermal insensitivity: based on the thermodynamic characteristics of power devices, their peak junction temperature T j There exists a significant "flat region" near the optimal operating point. Within this region, the junction temperature exhibits extremely low sensitivity to small fluctuations in the offset angle, demonstrating significant thermal insensitivity.

[0118] Engineering optimization logic: Based on this characteristic, the ANN model in this invention, while ensuring the peak junction temperature is kept at the lowest level, does not pursue an absolute analytical solution in a mathematical sense, but provides a more direct and robust mapping relationship. Compared with the tedious calculation of strictly following the SLF theoretical values, the ANN model avoids the risk of theoretical model failure due to system parameter drift (such as filter inductor saturation) by learning the optimal distribution in the flat region.

[0119] This invention further verifies, through simulation experiments, the decisive influence of selecting the optimal modulation strategy on the thermal safety of the grid-type converter. Figure 8 This demonstrates the optimal clamping offset angle achieved using the output of a neural network. opt Non-ideal offset angle bad Dynamic comparison waveforms showing the effect of suppressing peak junction temperature in the device. Analysis of the experimental results indicates that using the optimal clamping offset angle can significantly suppress the junction temperature. Figure 8 It can be seen that, under the same overload conditions, setting the optimal clamping offset angle is crucial. opt The peak junction temperature of the device at that time was 138.9°C, compared to selecting a non-ideal offset angle. bad At 162.1°C, the junction temperature decreased significantly by 23.2°C. Simulation data shows that only at... opt Under this strategy, the device junction temperature can be effectively locked below 150°C (the thermal safety upper limit specified in the datasheet). However, if the offset angle is not properly selected ( badThe temperature can only be reduced by an additional 5.4°C, still far exceeding the safety threshold, making it highly susceptible to thermal breakdown. This comparative result strongly demonstrates that junction temperature is highly sensitive to clamp offset angle. This verifies the core logic of the strategy proposed in this invention: the converter must be combined with a DPWM clamping strategy based on accurate prediction by the machine-side ANN to truly achieve decisive suppression of temperature rise. This invention ensures that the core power devices of the grid-type converter always operate within a safe thermal stress range when subjected to large current surges, significantly improving the system's overload ride-through limit and operational robustness.

[0120] The core advantage of the strategy proposed in this invention lies in the active reconstruction of the distribution characteristics of the conduction loss and switching loss of power devices through online optimization using neural networks. Figure 9 A detailed comparison was made between using a non-optimal clamping offset angle and the optimal clamping offset angle under the same load conditions. opt The loss evolution of IGBT devices within one fundamental frequency cycle. For example... Figure 9 As shown in (a) and 9(c), when the offset angle is not set properly, the clamping interval (i.e., the inactive interval) of the modulation wave deviates from the peak region of the phase current. At this time, the IGBT remains in a high-frequency switching state even at the moment of maximum current, generating huge switching loss pulses, resulting in extremely high instantaneous thermal stress. Through the real-time correction of the ANN model in this invention, as... Figure 9 As shown in (b) and (d), the DPWM strategy can precisely shift the clamping interval to the current peak region. This mechanism ensures that the IGBT remains in a constant on or off state when the phase current magnitude is at its maximum, achieving zero switching loss at that moment and fundamentally reducing the energy input at the most severe heat-generating point. Simulation data demonstrates that, based on optimal... opt The loss reconstruction mechanism plays a decisive role in suppressing device temperature rise. Although the on-state loss is mainly limited by the load current amplitude, by intervening with the optimal clamping offset angle, this invention successfully avoids the overlap of "high on-state loss" and "high switching loss" on the time axis, reducing the peak value of the equivalent total loss. (Comparison) Figure 9 As can be seen from the energy distribution curves in (c) and (d), the switching energy pulse distribution under the optimal strategy is more uniform and effectively avoids the high current region. This benign reconstruction of the loss distribution ensures a significant reduction in the overall temperature rise of the device under overload conditions, verifying the significant progress of this scheme in protecting core power devices.

[0121] The beneficial effects of this DPWM-based grid converter overload capacity enhancement method and system are reflected in multiple dimensions. First, the optimal clamping offset angle is accurately predicted by ANN. optThis significantly suppresses the junction temperature of power devices, allowing the converter to briefly output a higher proportion of reactive power during current surges or fault conditions caused by renewable energy integration. This enhances the system's active support capability, ensures voltage and frequency stability during grid transients, and effectively prevents the risk of cascading grid disconnection. Secondly, it utilizes the current command angle... i The feedforward control logic, acting as the dominant variable, improves the response speed of the thermal management system during dynamic processes and reduces the risk of overload shutdown caused by instantaneous junction temperature spikes. The strategy proposed in this invention ensures deep synergy between device-level loss optimization and grid-based system-level dynamic response. By utilizing the thermally insensitive flat region of the junction temperature near the optimal operating point, the system exhibits strong robustness against parameter disturbances, achieving seamless integration from microscopic "precise loss suppression" to macroscopic "robust grid support." In summary, the intelligent optimization of the DPWM offset angle using artificial neural networks not only fundamentally solves the limitations of traditional SLF analytical methods in terms of calculation lag and insufficient accuracy but also substantially improves the instantaneous overload limit and operational reliability of the converter, providing a solid technical guarantee for the stable operation of high-penetration power electronic systems.

[0122] This invention also provides a DPWM-based overload capacity enhancement system for grid-connected converters, implementing the aforementioned DPWM-based overload capacity enhancement method for grid-connected converters, comprising:

[0123] The feature extraction module is used to obtain the d-axis current command and q-axis current command of the grid-type converter and synthesize them to obtain the current command angle.

[0124] The electrothermal mapping module is used to construct a dataset of current command angle, clamp offset angle and junction temperature based on multiphysics thermal simulation data, and to establish a direct nonlinear mapping from the current command angle to the optimal clamp offset angle through an artificial neural network model.

[0125] The real-time decision module is used to output the optimal clamping offset angle based on the real-time current command angle through an artificial neural network model under converter overload conditions.

[0126] The dynamic modulation execution module is used to inject the optimal clamp offset angle into the modulation signal generation path. By dynamically switching the clamp mode through sector logic, the clamp interval is aligned with the current peak region, so that the insulated gate bipolar transistor can achieve zero switching operation in the current peak segment to suppress junction temperature.

[0127] It should be noted that any parts not disclosed or specifically described in this invention are existing technology or conventional configurations, and their specific structures and working principles will not be elaborated further. In this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0128] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. It can be applied to various fields suitable for the present invention. Other modifications can be readily implemented by those skilled in the art. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and examples shown and described herein.

Claims

1. A method for improving the overload capacity of a grid-type converter based on DPWM, characterized in that, include: Obtain the d-axis current command and q-axis current command of the grid-type converter, and combine the d-axis current command and q-axis current command into the included angle of the current command; A dataset of current command angle and clamp offset angle versus junction temperature is constructed based on multiphysics thermal simulation. Then, a direct nonlinear mapping from current command angle to optimal clamp offset angle is established through an artificial neural network model. Under converter overload conditions, the optimal clamping offset angle is quickly output through an artificial neural network model based on the real-time current command angle. The extreme values ​​of the three-phase voltage commands are extracted, and the coordinate transformation of the three-phase voltage commands is performed to obtain the fundamental phase angle. The optimal clamping offset angle is superimposed with the fundamental phase angle to select the sector. The zero-sequence component is injected into the voltage command by combining the sector selection and extreme value extraction results. The clamping mode is dynamically switched so that the clamping interval is aligned with the current peak region. Then, the insulated gate bipolar transistor operates with zero switching in the current peak segment to suppress the junction temperature, thereby improving the overload capacity of the grid converter. The formula for the angle between the combined current command of the d-axis current command and the q-axis current command is: ； In the formula, This is the d-axis current command. This is a q-axis current command. The included angle is the current command angle; The dataset is represented as follows: ； The mapping relationship between junction temperature and input variables can be expressed as: ； In the formula, This is the clamping offset angle. For the junction temperature, For the dataset, This represents the minimum value of the current command angle scan. This represents the maximum value of the current command angle scan. This is the minimum value of the clamping offset angle. This represents the maximum value of the clamping offset angle; The direct nonlinear mapping is: ； In the formula, The optimal clamping offset angle is... It is an artificial neural network model; Based on multiphysics thermal simulation experimental data, an electrothermal coupling simulation model of the converter was built, covering a wide range of current command angles. and the junction temperature at clamp offset angle α Scanning and data acquisition are performed to build an initial dataset. The dataset is then learned and simulated using an ANN, and real-time output of the junction temperature data is generated. The minimum optimal clamping offset angle.

2. The method for improving the overload capacity of a grid-type converter based on DPWM according to claim 1, characterized in that, Aligning the clamping interval with the current peak region specifically includes extreme value selection, coordinate transformation, sector selection, and zero-sequence component injection.

3. The method for improving the overload capacity of a grid-type converter based on DPWM according to claim 2, characterized in that, The extreme values ​​are selected as follows: ； In the formula, This is the three-phase reference voltage input.

4. The method for improving the overload capacity of a grid-type converter based on DPWM according to claim 3, characterized in that, The coordinate transformation is used to obtain the fundamental phase angle, and the formula is as follows: ； In the formula, The fundamental phase angle, For the α-axis voltage component, This represents the β-axis voltage component.

5. The method for improving the overload capacity of a grid-type converter based on DPWM according to claim 4, characterized in that, The formula for sector selection is: ； In the formula, To control the angle, For logical sectors.

6. The method for improving the overload capacity of a grid-type converter based on DPWM according to claim 5, characterized in that, The formula for zero-order component injection is: ； In the formula, For the final modulated wave, This is the original, unclamped modulated wave. This refers to the injected zero-order component.

7. A DPWM-based overload capacity enhancement system for grid-type converters, implementing the DPWM-based overload capacity enhancement method for grid-type converters according to any one of claims 1 to 6, characterized in that, include: The feature extraction module is used to obtain the d-axis current command and q-axis current command of the grid-type converter and synthesize them to obtain the included angle of the current command. The electrothermal mapping module is used to construct a dataset of current command angle, clamp offset angle and junction temperature based on multiphysics thermal simulation data, and to establish a direct nonlinear mapping from the current command angle to the optimal clamp offset angle through an artificial neural network model. The real-time decision-making module is used to output the optimal clamping offset angle based on the real-time current command angle through an artificial neural network model under converter overload conditions. The dynamic modulation execution module is used to inject the optimal clamp offset angle into the modulation signal generation path. By dynamically switching the clamp mode through sector logic, the clamp interval is aligned with the current peak region, so that the insulated gate bipolar transistor can achieve zero switching operation in the current peak segment to suppress junction temperature.

Images