New energy station resonance risk assessment method and device, medium and equipment

Abstract

The resonance risk assessment method, device, medium, and equipment for new energy power stations provided in this application are based on the operating condition vector of the new energy power station and employ an impedance prediction model that integrates physical laws for impedance prediction. This impedance prediction model is trained using a total loss function that includes data fitting loss and physical residual loss. The physical residual loss can directly embed electromagnetic transient constraints such as Kirchhoff's laws into the training process, ensuring that the model output strictly conforms to physical laws. This effectively overcomes the accuracy errors caused by the simplification of traditional equivalent circuit models, thereby improving the accuracy of resonance risk assessment. Simultaneously, the fully trained model enables second-level impedance characteristic prediction, and the risk frequency is identified in real time by calculating the stability margin, improving analysis efficiency and real-time performance. While ensuring high-precision physical consistency, it can quickly and accurately assess resonance risk, thus meeting the online analysis and rapid early warning needs of new energy power stations under complex and variable operating conditions.

Description

Technical Field

[0001] This application relates to the field of power system technology, and in particular to a method, device, medium and equipment for assessing the resonance risk of new energy power stations. Background Technology

[0002] With the high penetration of power electronic equipment, the impedance characteristics of power stations change dynamically with operating conditions, easily triggering resonance at specific frequencies, posing a serious threat to equipment safety and power grid stability. Currently, the commonly used method in the industry is impedance scanning analysis based on equivalent circuit models, which involves manually modeling and scanning the impedance curve in a simulation environment, and then manually identifying resonance risks in conjunction with stability criteria.

[0003] However, in practical applications, the aforementioned methods suffer from poor accuracy in stability assessment because the equivalent circuit model cannot accurately reflect the transient characteristics of power electronic devices. Furthermore, a single full-band scan is too time-consuming, making it difficult to meet real-time analysis requirements. Therefore, the existing methods fail to achieve the expected accuracy and efficiency in resonant stability assessment. Summary of the Invention

[0004] The purpose of this application is to address at least one of the aforementioned technical deficiencies, particularly the poor accuracy of stability assessments in existing technologies due to the inability of equivalent circuit models to accurately reflect the transient characteristics of power electronic devices. Furthermore, single full-band scans are too time-consuming, making real-time analysis difficult. Therefore, existing methods fail to achieve the expected accuracy and efficiency in resonant stability assessments.

[0005] Firstly, this application provides a method for assessing the resonance risk of new energy power stations, the method comprising:

[0006] Obtain the operating condition vector of the new energy power station;

[0007] The operating condition vector is input into a pre-trained impedance prediction model to obtain the full-band impedance characteristic curve of the new energy power station. The impedance prediction model is trained by a preset total loss function, which includes data fitting loss and physical residual loss. The physical residual loss is used to apply physical constraints to the output of the impedance prediction model.

[0008] Determine the current grid-side impedance characteristic curve, and calculate the stability margin at each frequency point within the full frequency range based on the full-band impedance characteristic curve and the grid-side impedance characteristic curve.

[0009] Frequency points with stability margins less than a preset margin threshold are identified among various frequency points to form a risk frequency set, and the resonance risk level of the new energy power station is determined based on the risk frequency set.

[0010] In one embodiment, the training process of the impedance prediction model includes:

[0011] Obtain working condition samples and their corresponding simulated impedance characteristic curves, and determine a preset pre-trained model, which includes a feature extraction layer, a physical residual layer, and a fusion layer.

[0012] In one iteration, the working condition sample is input into the feature extraction layer and the physical residual layer respectively to obtain the high-dimensional feature vector and physical information of the working condition sample;

[0013] The high-dimensional feature vector and physical information are input into the fusion layer to obtain the predicted impedance characteristic curve. The data fitting loss and physical residual loss are determined based on the working condition sample and its simulated impedance characteristic curve and the predicted impedance characteristic curve.

[0014] A total loss function is constructed based on the data fitting loss and the physical residual loss. The pre-trained model is then trained according to the total loss function until a preset iteration condition is met. The pre-trained model is then determined as the impedance prediction model.

[0015] In one embodiment, obtaining the operating condition sample and its corresponding simulated impedance characteristic curve includes:

[0016] Obtain the pre-constructed electromagnetic transient model of the new energy power station;

[0017] The electromagnetic transient model is used to simulate the operating state of a new energy power station under various working conditions. After frequency domain scanning analysis of the operating state, the simulated impedance characteristic curves are obtained, and samples of various working conditions and their corresponding simulated impedance characteristic curves are recorded.

[0018] In one embodiment, the total loss function includes:

[0019]

[0020] In the formula, Represents the total loss function. Indicates adjustable hyperparameters. This represents the data fitting loss. The physical residual loss is represented by: where the data fitting loss and the physical residual loss are respectively expressed as:

[0021]

[0022]

[0023] In the formula, Indicates the number of training samples. Indicates the first Simulated impedance characteristic curves of training samples, The impedance prediction model represents the first... The full-band impedance characteristic curve predicted from each training sample. This represents the frequency vector for full-band scanning. Indicates the first The working condition vector represented by each training sample. This represents the residual of the nodal admittance equation.

[0024] In one embodiment, the impedance prediction model includes a feature extraction layer, a physical residual layer, and a fusion layer. The step of inputting the operating condition vector into the pre-trained impedance prediction model to obtain the full-band impedance characteristic curve of the new energy power station includes:

[0025] The operating condition vector is input into the feature extraction layer for feature extraction to obtain a high-dimensional feature vector, and the operating condition vector is input into the physical residual layer to obtain physical information;

[0026] The high-dimensional feature vector and the physical information are input into the fusion layer so that the fusion layer outputs a full-band impedance characteristic curve that simultaneously conforms to the data distribution and physical laws.

[0027] In one embodiment, calculating the stability margin at each frequency point within the full frequency band based on the full-band impedance characteristic curve and the grid-side impedance characteristic curve includes:

[0028] For each frequency point within the entire frequency band, obtain the first impedance and the second impedance corresponding to that frequency point from the impedance characteristic curve of the entire frequency band and the impedance characteristic curve of the power grid side, respectively.

[0029] Calculate the quotient of the absolute value of the second impedance and the absolute value of the first impedance, and determine the result as the stability margin at that frequency point.

[0030] When traversing the entire frequency band, determine the stability margin of each frequency point within the entire frequency band.

[0031] In one embodiment, the method further includes:

[0032] When the impedance prediction model is applied to another new energy power station, the physical residual layer in the impedance prediction model is frozen, and the feature extraction layer and fusion layer in the impedance prediction model are fine-tuned to obtain an impedance prediction model suitable for another new energy power station.

[0033] A new impedance prediction model was used to assess the resonance risk of another renewable energy power station.

[0034] Secondly, this application provides a resonant risk assessment device for new energy power stations, the device comprising:

[0035] The operating condition acquisition module is used to acquire the operating condition vector of new energy power plants;

[0036] The impedance prediction module is used to input the operating condition vector into a pre-trained impedance prediction model to obtain the full-band impedance characteristic curve of the new energy power station. The impedance prediction model is trained by a preset total loss function, which includes data fitting loss and physical residual loss. The physical residual loss is used to apply physical constraints to the output of the impedance prediction model.

[0037] The margin calculation module is used to determine the current grid-side impedance characteristic curve and calculate the stability margin at each frequency point in the full frequency range based on the full-band impedance characteristic curve and the grid-side impedance characteristic curve.

[0038] The risk assessment module is used to identify frequency points with a stability margin less than a preset margin threshold among various frequency points to form a risk frequency set, and to determine the resonance risk level of the new energy power station based on the risk frequency set.

[0039] Thirdly, this application provides a storage medium storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the new energy power station resonance risk assessment method as described in any of the above embodiments.

[0040] Fourthly, this application provides a computer device, including: one or more processors, and a memory;

[0041] The memory stores computer-readable instructions, and when the one or more processors execute the computer-readable instructions, they perform the steps of the new energy power station resonance risk assessment method as described in any of the above embodiments.

[0042] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0043] The resonance risk assessment method, device, medium, and equipment for new energy power stations provided in this application are based on the operating condition vector of the new energy power station and employ an impedance prediction model that integrates physical laws for impedance prediction. This impedance prediction model is trained using a total loss function that includes data fitting loss and physical residual loss. The physical residual loss can directly embed electromagnetic transient constraints such as Kirchhoff's laws into the training process, ensuring that the model output strictly conforms to physical laws. This effectively overcomes the accuracy errors caused by the simplification of traditional equivalent circuit models, thereby improving the accuracy and reliability of resonance risk assessment. Simultaneously, the fully trained model enables second-level impedance characteristic prediction, and the risk frequency is identified in real time by calculating the stability margin, improving analysis efficiency and real-time performance. While ensuring high-precision physical consistency, it can quickly and accurately assess resonance risk, thus meeting the online analysis and rapid early warning needs of new energy power stations under complex and variable operating conditions. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of this application 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 this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 A flowchart illustrating a method for assessing resonance risk at a new energy power station, provided as an embodiment of this application;

[0046] Figure 2 A schematic diagram of the structure of a resonant risk assessment device for a new energy power station provided in an embodiment of this application;

[0047] Figure 3 This is an internal structural diagram of a computer device provided in an embodiment of this application. Detailed Implementation

[0048] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0049] In one embodiment, this application provides a method for assessing the resonance risk of new energy power stations. The following embodiments illustrate the application of this method to a server. It is understood that the method for assessing the resonance risk of new energy power stations can be executed by a single server or by a server cluster consisting of multiple servers; this application does not impose any specific limitations on this.

[0050] like Figure 1 As shown, this application provides a method for assessing the resonance risk of new energy power stations, the method comprising:

[0051] S101: Obtain the operating condition vector of the new energy power station.

[0052] Among them, the operating condition vector refers to a set of parameters that can characterize the operating status of a new energy power station, including but not limited to the power output of the power generation equipment, the grid connection status, and environmental conditions.

[0053] In this step, when conducting a resonance risk assessment for a renewable energy power station, various operating parameters can be collected by pre-deploying multiple sensors and monitoring devices within the station. Based on these collected parameters, an operating condition vector for the renewable energy power station can be generated. For example, in a wind farm, by installing power and speed sensors on wind turbines, the power generation and speed of each turbine can be obtained in real time. Simultaneously, voltage and current sensors are installed at grid connection points to monitor the voltage and current status of the grid. Furthermore, environmental parameters can be collected, such as wind speed sensors to measure wind speed and light intensity sensors to measure light intensity (for photovoltaic power stations).

[0054] Specifically, when collecting various operating condition parameters, these data can be preprocessed, such as through data cleaning and normalization. This ultimately generates a standardized operating condition vector, ensuring data accuracy.

[0055] S102: Input the operating condition vector into the pre-trained impedance prediction model to obtain the full-band impedance characteristic curve of the new energy power station.

[0056] The impedance prediction model is used to predict the impedance characteristics of new energy power stations at different frequencies under the current operating condition vector. It includes a feature extraction layer, a physical residual layer, and a fusion layer. Furthermore, the impedance prediction model is trained using a preset total loss function, which includes data fitting loss and physical residual loss. The physical residual loss is used to impose physical constraints on the output of the impedance prediction model. The full-band impedance characteristic curve refers to the continuous curve showing the impedance of a new energy power station as a function of frequency from low to high frequencies (e.g., 1Hz to 3000Hz) under the current operating condition.

[0057] In this step, when the current operating condition vector of the renewable energy power station is determined, this vector is input into a pre-trained impedance prediction model. This model then predicts the impedance characteristics of the power station at different frequencies based on the operating condition vector, thereby outputting the full-frequency impedance characteristic curve of the power station. For example, when the light intensity suddenly changes, by inputting a new operating condition vector, the model can quickly update the impedance characteristic curve, enabling risk assessment based on the latest predicted impedance characteristic curve and improving the accuracy of the risk assessment.

[0058] S103: Determine the current grid-side impedance characteristic curve, and calculate the stability margin at each frequency point within the full frequency range based on the full-band impedance characteristic curve and the grid-side impedance characteristic curve.

[0059] The grid-side impedance characteristic curve refers to the continuous curve showing the impedance of the power grid connecting renewable energy power plants as a function of frequency, from low to high frequencies. Stability margin refers to the degree of difference between the impedance characteristics of the renewable energy power plant and the grid-side impedance characteristics at a specific frequency, used to measure the stability of the power system at that frequency.

[0060] In this step, the grid-side impedance characteristic curve corresponding to the current grid connection point is retrieved from the preset database. Then, the full-band impedance characteristic curve of the new energy power station is compared and analyzed with the grid-side impedance characteristic curve. Specifically, for each frequency point within the full-band range, the difference between the two impedance characteristic curves is calculated. The stability margin can typically be quantified by calculating the difference in impedance magnitude or phase difference. Thus, the stability margin at each frequency point within the full-band range is calculated.

[0061] Furthermore, the grid-side impedance characteristic curves in the pre-set database can be determined through offline simulation, historical measurements, or typical data provided by the grid provider; this application does not impose specific limitations on this.

[0062] S104: Identify frequency points with stability margins less than a preset margin threshold among various frequency points to form a risk frequency set, and determine the resonance risk level of the new energy power station based on the risk frequency set.

[0063] The risk frequency set refers to the set of all frequency points across the entire frequency band where the stability margin is less than a preset margin threshold. The resonance risk level is a quantitative assessment of the resonance risk level of a new energy power station under its current operating conditions. The preset margin threshold is a stability margin threshold predetermined based on engineering experience.

[0064] In this step, each frequency point across the entire frequency band is checked individually, its stability margin is calculated, and compared with a preset margin threshold. When a frequency point's stability margin is found to be below the threshold, that frequency point and its stability margin are recorded, ultimately forming a risk frequency set containing all such frequency points. Next, the resonance risk level is determined based on the characteristics of the risk frequency set. Specifically, the specific resonance risk level can be determined based on characteristics such as the number of frequency points in the risk frequency set, their distribution, and the maximum and minimum values ​​of the stability margin. For example, if the number of frequency points in the risk frequency set is small and their distribution is relatively dispersed, and the minimum stability margin is relatively high, then the resonance risk level can be set as low risk; if the number of frequency points is large and their distribution is relatively concentrated, and the minimum stability margin is low, then the resonance risk level may be set as high risk.

[0065] Furthermore, when determining the specific resonance risk level based on the characteristics of the risk frequency set, the characteristics of the risk frequency set can be matched against pre-set conditional rules, and the level corresponding to the matched rule can be used as the resonance risk level. Alternatively, a corresponding deep learning model can be set to learn the relationship between the characteristics of the risk frequency set and the resonance risk level to determine the specific resonance risk level. The specific determination method can be selected according to the actual situation, and this application does not impose specific restrictions on it.

[0066] In the above embodiments, based on the operating condition vector of the new energy power station, an impedance prediction model integrating physical laws is used for impedance prediction. This impedance prediction model is trained based on a total loss function that includes data fitting loss and physical residual loss. The physical residual loss can directly embed electromagnetic transient constraints such as Kirchhoff's laws into the training process, ensuring that the model output strictly conforms to physical laws. This effectively overcomes the accuracy errors caused by the simplification of traditional equivalent circuit models, thereby improving the accuracy and reliability of resonance risk assessment. Simultaneously, a fully trained model is used to achieve second-level impedance characteristic prediction, and risk frequencies are identified in real time by calculating stability margins, improving analysis efficiency and real-time performance. While ensuring high-precision physical consistency, resonance risk can be quickly and accurately assessed, thus meeting the online analysis and rapid early warning needs of new energy power stations under complex and variable operating conditions.

[0067] In one embodiment, the training process of the impedance prediction model includes:

[0068] S1: Obtain the working condition samples and their corresponding simulated impedance characteristic curves, and determine the preset pre-trained model.

[0069] S2: In one iteration, the working condition sample is input into the feature extraction layer and the physical residual layer respectively to obtain the high-dimensional feature vector and physical information of the working condition sample.

[0070] S3: Input the high-dimensional feature vector and physical information into the fusion layer to obtain the predicted impedance characteristic curve, and determine the data fitting loss and physical residual loss based on the working condition sample and its simulated impedance characteristic curve and predicted impedance characteristic curve.

[0071] S4: Construct a total loss function based on data fitting loss and physical residual loss, and train the pre-trained model according to the total loss function until the preset iteration conditions are met. Then, determine the pre-trained model as the impedance prediction model.

[0072] The pre-trained model includes a feature extraction layer, a physical residual layer, and a fusion layer. The simulated impedance characteristic curve refers to the impedance characteristic curve of a new energy power station under corresponding operating condition samples, generated using an electromagnetic transient simulation tool. The high-dimensional feature vector refers to the feature representation extracted from the operating condition samples through the feature extraction layer, containing both operating condition features and temporal features. Physical information refers to the constraint information related to physical characteristics extracted from the operating condition samples through the physical residual layer. The preset iteration condition refers to the condition for stopping training; it can be set to reaching a preset number of iterations, or to the absolute value of the difference between the loss values ​​of two adjacent iterations being less than a preset threshold. This application does not impose specific restrictions on this.

[0073] In this embodiment, firstly, operating condition samples and their corresponding simulated impedance characteristic curves are acquired as training data, and a preset pre-trained model is determined. This model includes a feature extraction layer, a physical residual layer, and a fusion layer. Then, the pre-trained model is iteratively trained using the training data. In each iteration, the operating condition samples are input into the feature extraction layer and the physical residual layer, respectively, to obtain the high-dimensional feature vector output by the feature extraction layer and the physical information output by the physical residual layer. Next, the high-dimensional feature vector and the physical information are simultaneously input into the fusion layer to obtain the predicted impedance characteristic curve. This allows the fusion layer to calculate the data fitting loss and the physical residual loss based on the operating condition samples, their simulated impedance characteristic curves, and the predicted impedance characteristic curves, respectively. The total loss function is further determined by combining the data fitting loss and the physical residual loss. Finally, backpropagation is performed on the pre-trained model based on this total loss function to update the parameters of the pre-trained model. This completes one iteration of training. Then, it is determined whether the preset iteration conditions are met. If they are met, the pre-trained model is determined as the impedance prediction model; otherwise, the next iteration continues training.

[0074] Specifically, when calculating the data fitting loss and physical residual loss, the data fitting loss can be determined based on the degree of difference between the simulated impedance characteristic curve and the predicted impedance characteristic curve. Simultaneously, the physical residual loss can be calculated by using the residuals of the nodal admittance equation derived from Kirchhoff's laws, combined with the operating condition samples and their simulated and predicted impedance characteristic curves, thereby imposing hard physical constraints on the model's output during the inference process.

[0075] Furthermore, during the forward propagation of the model, the physical residual layer verifies the model's current predicted output in real time based on electromagnetic transient theory. This physical residual layer internally encodes the governing equations describing the system's resonant behavior: the residuals of the nodal admittance equations derived from Kirchhoff's laws. Using automatic differentiation techniques, this physical residual layer can directly calculate the residual between the model's predicted value and the requirements of the physical laws.

[0076] Understandably, training with a total loss function constructed based on data fitting loss and physical residual loss can comprehensively consider data accuracy and physical constraints, allowing the model to continuously optimize during training and ultimately achieve the goal of high-precision impedance characteristic prediction. This training method not only effectively overcomes the problems caused by insufficient model accuracy and low analysis efficiency in traditional methods, but also enables rapid and accurate prediction of impedance characteristics under rapidly changing operating conditions of new energy power plants, providing a reliable basis for resonance risk assessment and system stability analysis.

[0077] In one embodiment, obtaining the operating condition sample and its corresponding simulated impedance characteristic curve includes:

[0078] S1: Obtain the pre-built electromagnetic transient model of the new energy power station.

[0079] S2: Simulate the operating state of new energy power stations under various working conditions using an electromagnetic transient model, and perform frequency domain scanning analysis on the operating state to obtain the simulated impedance characteristic curves. Then, record the samples of various working conditions and their corresponding simulated impedance characteristic curves.

[0080] Among them, the electromagnetic transient model is a mathematical model used to simulate the electromagnetic transient behavior of new energy power plants under different operating conditions. The operating state refers to a series of operating parameters and behaviors of new energy power plants under specific operating conditions, including the power output of the equipment, the grid connection status, the equipment operating mode, and environmental conditions (such as wind speed and light intensity).

[0081] In this embodiment, a pre-constructed electromagnetic transient model of the renewable energy power station is obtained. This model can accurately describe the electromagnetic transient behavior of the power station under different operating conditions. Next, the electromagnetic transient model is used to simulate the operating states of the renewable energy power station under various conditions, including different equipment power outputs, changes in environmental conditions (such as wind speed and solar intensity), and grid connection status. During the simulation, frequency domain scanning analysis is performed on each operating state. By applying small perturbations at different frequencies and measuring the system response, simulated impedance characteristic curves are obtained. Finally, samples of each operating condition and their corresponding simulated impedance characteristic curves are recorded, and the recorded data is used as data for subsequent training of the impedance prediction model.

[0082] Specifically, a site-level electromagnetic transient model, including detailed converter control strategies, cable distribution parameters, and equivalent impedance of transformers and the power grid, can be constructed based on electromagnetic transient simulation tools (such as PSCAD or EMTDC). This is the electromagnetic transient model described in this application. Furthermore, by calling a Python-based automated scheduling interface, seamless bridging and iterative calls between MATLAB control scripts and simulation tools can be achieved. This automated scheduling interface can traverse key operating conditions such as wind speed fluctuations, changes in power grid short-circuit capacity, and switching of reactive power compensation devices according to preset strategies, and perform batch frequency domain scanning analysis. Finally, a large-scale, multi-dimensional database of simulated impedance characteristic curves is generated. Each curve in this database describes the functional relationship between the impedance (including amplitude and phase) of the renewable energy power plant and the scanning frequency under a specific operating condition vector. This database ensures the physical accuracy of the data source and the completeness of the operating condition coverage, laying a solid foundation for training a reliable model that can generalize to unknown operating conditions.

[0083] In one embodiment, the total loss function includes:

[0084]

[0085] In the formula, Represents the total loss function. Indicates adjustable hyperparameters. This represents the data fitting loss. The physical residual loss is represented by: where the data fitting loss and the physical residual loss are respectively expressed as:

[0086]

[0087]

[0088] In the formula, Indicates the number of training samples. Indicates the first Simulated impedance characteristic curves of training samples, The impedance prediction model represents the first... The full-band impedance characteristic curve predicted from each training sample. This represents the frequency vector for full-band scanning. Indicates the first The working condition vector represented by each training sample. This represents the residual of the nodal admittance equation.

[0089] This is understandable; adjustable hyperparameters This is used to balance the weights between data-driven and physical constraints. By training the impedance prediction model with a total loss function that combines data fitting loss and physical residual loss, the model can exhibit excellent generalization ability and physical consistency even with a small number of samples.

[0090] In one embodiment, the impedance prediction model includes a feature extraction layer, a physical residual layer, and a fusion layer. The operating condition vector is input into the pre-trained impedance prediction model to obtain the full-band impedance characteristic curve of the new energy power station, including:

[0091] S1: Input the working condition vector into the feature extraction layer to extract features and obtain a high-dimensional feature vector. Then, input the working condition vector into the physical residual layer to obtain physical information.

[0092] S2: Input high-dimensional feature vectors and physical information into the fusion layer so that the output of the fusion layer is a full-band impedance characteristic curve that conforms to both data distribution and physical laws.

[0093] The fusion layer is a fully connected network that can find the optimal combination between high-dimensional feature vectors and physical information, thereby outputting an impedance characteristic curve that fits the training data and strictly follows physical laws.

[0094] In this embodiment, the operating condition vector is input into the feature extraction layer for feature extraction to obtain a high-dimensional feature vector. Simultaneously, the operating condition vector is input into the physical residual layer to obtain physical information. Subsequently, the high-dimensional feature vector and physical information are input into the fusion layer. The fusion layer uses its internal neural network structure to comprehensively process these data, ultimately outputting a full-band impedance characteristic curve. This curve is not only based on the distribution characteristics of the data itself but is also constrained by the physical information, thus ensuring that it conforms to physical laws.

[0095] Specifically, the feature extraction layer comprises multiple fully connected layers, gated recurrent units, and a multi-head self-attention layer. The multiple fully connected layers are used to perform nonlinear transformations on the operating condition vector to extract preliminary features. Subsequently, considering the inherent temporal correlation of certain operating condition parameters (such as wind speed sequences), gated recurrent units are introduced to effectively capture their dynamic temporal characteristics. Finally, a multi-head self-attention layer is embedded at the end. This layer automatically calculates and quantifies the importance weights of different operating condition parameters on the final impedance characteristics, thus explicitly revealing key influencing factors. This not only improves model performance but also enhances the transparency and interpretability of its decision-making process.

[0096] In one embodiment, the stability margin at each frequency point across the entire frequency band is calculated based on the full-band impedance characteristic curve and the grid-side impedance characteristic curve, including:

[0097] S1: Traverse every frequency point within the entire frequency band and obtain the first impedance and second impedance corresponding to that frequency point from the impedance characteristic curve of the entire frequency band and the impedance characteristic curve of the power grid side.

[0098] S2: Calculate the quotient of the absolute value of the second impedance and the absolute value of the first impedance, and determine the result as the stability margin at that frequency point.

[0099] S3: When traversing the entire frequency band, determine the stability margin of each frequency point within the entire frequency band.

[0100] The full frequency range refers to all frequency intervals considered in impedance characteristic analysis, typically from low frequency (e.g., 1Hz) to high frequency (e.g., 3000Hz).

[0101] In this embodiment, every frequency point within the entire frequency band is traversed, and the corresponding first impedance (impedance of the renewable energy power station) and second impedance (impedance of the grid side) are obtained from the full-band impedance characteristic curve and the grid-side impedance characteristic curve, respectively. For each frequency point, the quotient of the absolute value of the second impedance and the absolute value of the first impedance is calculated, and this calculation result is determined as the stability margin at that frequency point. After all frequency points have been traversed, the stability margin of each frequency point within the entire frequency band can be obtained. For example, at a certain frequency point, a low stability margin indicates that there may be a high risk of resonance between the renewable energy power station and the grid at that frequency.

[0102] It is understandable that by calculating the stability margin point by point, the frequency points in the power system with potential resonance risks can be accurately identified, thus providing a scientific basis for early warning and prevention measures of resonance risks. This method not only effectively overcomes the problem of insufficient stability assessment caused by the lack of full-band analysis in traditional methods, but also can quickly and accurately identify potential resonance risk frequency points under rapidly changing operating conditions of new energy power plants.

[0103] In one example, the stability margin at a certain frequency point can be expressed as:

[0104]

[0105] In the formula, Frequency point stability margin, Frequency points in the grid-side impedance characteristic curve The corresponding impedance, i.e., the second impedance, Frequency points in the full-band impedance characteristic curve The corresponding impedance, i.e., the first impedance.

[0106] In one embodiment, the method for assessing the resonance risk of new energy power stations further includes:

[0107] S1: When the impedance prediction model is transferred to another new energy power station, the physical residual layer in the impedance prediction model is frozen, and the feature extraction layer and fusion layer in the impedance prediction model are fine-tuned to obtain an impedance prediction model suitable for another new energy power station.

[0108] S2: Using a new impedance prediction model to assess the resonance risk of another new energy power station.

[0109] In this embodiment, when the trained impedance prediction model needs to be transferred to another renewable energy power station, the physical residual layer can be frozen because the physical laws and constraints contained in the physical residual layer are generally universal across different power stations. Subsequently, the feature extraction layer and fusion layer in the impedance prediction model are fine-tuned. Fine-tuning the feature extraction layer is to enable the model to adapt to the operating characteristics of the new power station, as different power stations may have differences in equipment type, operating environment, etc. Fine-tuning the feature extraction layer can better capture the operating characteristics of the new power station. Fine-tuning the fusion layer is to optimize the model's output so that it can more accurately reflect the impedance characteristic curve of the new power station.

[0110] Specifically, the fine-tuning process can be carried out by collecting a small number of operating condition samples and their corresponding simulated impedance characteristic curves at the new power station. These data are then used to train the feature extraction layer and fusion layer, adjusting the model parameters until the model achieves satisfactory prediction accuracy on the new power station's data. After fine-tuning, an impedance prediction model suitable for another new energy power station is obtained. This new impedance prediction model is then used to assess the resonance risk of the new power station. It can be understood that when using the fine-tuning strategy, typically only 5-10 sets of fine-tuning are needed to enable the new model to quickly adapt to the new power station with extremely high efficiency while maintaining excellent performance.

[0111] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0112] The following describes the resonant risk assessment device for new energy power stations provided in the embodiments of this application. The resonant risk assessment device for new energy power stations described below can be referred to in correspondence with the resonant risk assessment method for new energy power stations described above.

[0113] like Figure 2 As shown, this application provides a new energy power station resonance risk assessment device 200, the device comprising:

[0114] Operating condition acquisition module 201 is used to acquire the operating condition vector of the new energy power station;

[0115] Impedance prediction module 202 is used to input the operating condition vector into the pre-trained impedance prediction model to obtain the full-band impedance characteristic curve of the new energy power station. The impedance prediction model is trained by a preset total loss function, which includes data fitting loss and physical residual loss. The physical residual loss is used to apply physical constraints to the output of the impedance prediction model.

[0116] The margin calculation module 203 is used to determine the current grid-side impedance characteristic curve and calculate the stability margin at each frequency point in the full frequency range based on the full-band impedance characteristic curve and the grid-side impedance characteristic curve.

[0117] The risk assessment module 204 is used to identify frequency points with a stability margin less than a preset margin threshold in each frequency point to form a risk frequency set, and to determine the resonance risk level of the new energy power station based on the risk frequency set.

[0118] In the above embodiments, based on the operating condition vector of the new energy power station, an impedance prediction model integrating physical laws is used for impedance prediction. This impedance prediction model is trained based on a total loss function that includes data fitting loss and physical residual loss. The physical residual loss can directly embed electromagnetic transient constraints such as Kirchhoff's laws into the training process, ensuring that the model output strictly conforms to physical laws. This effectively overcomes the accuracy errors caused by the simplification of traditional equivalent circuit models, thereby improving the accuracy and reliability of resonance risk assessment. Simultaneously, a fully trained model is used to achieve second-level impedance characteristic prediction, and risk frequencies are identified in real time by calculating stability margins, improving analysis efficiency and real-time performance. While ensuring high-precision physical consistency, resonance risk can be quickly and accurately assessed, thus meeting the online analysis and rapid early warning needs of new energy power stations under complex and variable operating conditions.

[0119] In one embodiment, the impedance prediction module includes:

[0120] The sample acquisition submodule is used to acquire working condition samples and their corresponding simulated impedance characteristic curves, and to determine the preset pre-trained model. The pre-trained model includes a feature extraction layer, a physical residual layer, and a fusion layer.

[0121] The iterative training submodule is used to input the working condition samples into the feature extraction layer and the physical residual layer in one iteration to obtain the high-dimensional feature vector and physical information of the working condition samples.

[0122] The loss determination submodule is used to input high-dimensional feature vectors and physical information into the fusion layer to obtain the predicted impedance characteristic curve, and to determine the data fitting loss and physical residual loss based on the working condition sample and its simulated impedance characteristic curve and the predicted impedance characteristic curve.

[0123] The model training submodule is used to construct a total loss function based on data fitting loss and physical residual loss, and to train the pre-trained model according to the total loss function until the preset iteration conditions are met. Then, the pre-trained model is determined as the impedance prediction model.

[0124] In one embodiment, the sample acquisition submodule includes:

[0125] The model acquisition unit is used to acquire the pre-built electromagnetic transient model of the new energy power station;

[0126] The sample generation unit is used to simulate the operating state of new energy power stations under various working conditions through electromagnetic transient models, and to perform frequency domain scanning analysis on the operating state to obtain the simulated impedance characteristic curves. After that, it records the samples of various working conditions and their corresponding simulated impedance characteristic curves.

[0127] In one embodiment, the impedance prediction model includes a feature extraction layer, a physical residual layer, and a fusion layer, and the impedance prediction module includes:

[0128] The vector input submodule is used to input the working condition vector into the feature extraction layer for feature extraction to obtain a high-dimensional feature vector, and to input the working condition vector into the physical residual layer to obtain physical information;

[0129] The data fusion submodule is used to input high-dimensional feature vectors and physical information into the fusion layer so that the output of the fusion layer is a full-band impedance characteristic curve that conforms to both data distribution and physical laws.

[0130] In one embodiment, the margin calculation module includes:

[0131] The frequency traversal submodule is used to traverse every frequency point in the full frequency band and obtain the first impedance and second impedance corresponding to that frequency point from the full frequency band impedance characteristic curve and the grid-side impedance characteristic curve, respectively.

[0132] The margin calculation submodule is used to calculate the quotient of the absolute value of the second impedance and the absolute value of the first impedance, and to determine the calculation result as the stability margin at that frequency point.

[0133] The margin determination submodule is used to determine the stability margin of each frequency point in the entire frequency band range when traversing the entire frequency band range.

[0134] In one embodiment, the resonant risk assessment device for new energy power stations further includes:

[0135] The model migration module is used to freeze the physical residual layer in the impedance prediction model and fine-tune the feature extraction layer and fusion layer in the impedance prediction model when the impedance prediction model is migrated and applied to another new energy power station, so as to obtain an impedance prediction model suitable for another new energy power station.

[0136] The assessment module is used to conduct a resonance risk assessment of another new energy power station using a new impedance prediction model.

[0137] The division of modules in the aforementioned new energy power station resonance risk assessment device is merely illustrative. In other embodiments, the new energy power station resonance risk assessment device can be divided into different modules as needed to complete all or part of its functions. Each module in the aforementioned new energy power station resonance risk assessment device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0138] In one embodiment, this application also provides a storage medium storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the new energy power station resonance risk assessment method as described in any of the above embodiments.

[0139] In one embodiment, this application also provides a computer device storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the new energy power station resonance risk assessment method as described in any of the above embodiments.

[0140] Indicatively, such as Figure 3 As shown, Figure 3 This is a schematic diagram of the internal structure of a computer device 300 provided in an embodiment of this application. The computer device 300 can be provided as a server. (Refer to...) Figure 3 The computer device 300 includes a processing component 302, which further includes one or more processors, and memory resources represented by memory 301 for storing instructions, such as application programs, that can be executed by the processing component 302. The application programs stored in memory 301 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processing component 302 is configured to execute instructions to perform the resonant risk assessment method for new energy power stations described in any of the above embodiments.

[0141] The computer device 300 may also include a power supply component 303 configured to perform power management of the computer device 300, a wired or wireless network interface 304 configured to connect the computer device 300 to a network, and an input / output (I / O) interface 305. The computer device 300 may operate on an operating system stored in memory 301, such as Windows Server™, Mac OS X™, Unix™, Linux™, Free BSD™, or similar.

[0142] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0143] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, 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 a…" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. In this document, the singular forms "a," "an," and "the" may also include the plural forms unless the context clearly indicates otherwise. It should also be understood that the terms “comprising / including” or “having” specify the presence of the stated features, wholes, steps, operations, components, parts or combinations thereof, but do not exclude the possibility of the presence or addition of one or more other features, wholes, steps, operations, components, parts or combinations thereof. Meanwhile, the term “and / or” as used in this specification includes any and all combinations of the associated listed items.

[0144] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referred to each other.

[0145] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for assessing the resonance risk of new energy power stations, characterized in that, The method includes: Obtain the operating condition vector of the new energy power station; The operating condition vector is input into a pre-trained impedance prediction model to obtain the full-band impedance characteristic curve of the new energy power station. The impedance prediction model is trained by a preset total loss function, which includes data fitting loss and physical residual loss. The physical residual loss is used to apply physical constraints to the output of the impedance prediction model. Determine the current grid-side impedance characteristic curve, and calculate the stability margin at each frequency point within the full frequency range based on the full-band impedance characteristic curve and the grid-side impedance characteristic curve. Frequency points with stability margins less than a preset margin threshold are identified among various frequency points to form a risk frequency set, and the resonance risk level of the new energy power station is determined based on the risk frequency set.

2. The method for assessing the resonance risk of new energy power stations according to claim 1, characterized in that, The training process of the impedance prediction model includes: Obtain working condition samples and their corresponding simulated impedance characteristic curves, and determine a preset pre-trained model, which includes a feature extraction layer, a physical residual layer, and a fusion layer. In one iteration, the working condition sample is input into the feature extraction layer and the physical residual layer respectively to obtain the high-dimensional feature vector and physical information of the working condition sample; The high-dimensional feature vector and physical information are input into the fusion layer to obtain the predicted impedance characteristic curve. The data fitting loss and physical residual loss are determined based on the working condition sample and its simulated impedance characteristic curve and the predicted impedance characteristic curve. A total loss function is constructed based on the data fitting loss and the physical residual loss. The pre-trained model is then trained according to the total loss function until a preset iteration condition is met. The pre-trained model is then determined as the impedance prediction model.

3. The method for assessing the resonance risk of new energy power stations according to claim 2, characterized in that, The acquisition of the operating condition samples and their corresponding simulated impedance characteristic curves includes: Obtain the pre-constructed electromagnetic transient model of the new energy power station; The electromagnetic transient model is used to simulate the operating state of a new energy power station under various working conditions. After frequency domain scanning analysis of the operating state, the simulated impedance characteristic curves are obtained, and samples of various working conditions and their corresponding simulated impedance characteristic curves are recorded.

4. The method for assessing the resonance risk of new energy power stations according to any one of claims 1 to 3, characterized in that, The total loss function includes: In the formula, Represents the total loss function. Indicates adjustable hyperparameters. This represents the data fitting loss. The physical residual loss is represented by: where the data fitting loss and the physical residual loss are respectively expressed as: In the formula, Indicates the number of training samples. Indicates the first Simulated impedance characteristic curves of training samples, The impedance prediction model represents the first... The full-band impedance characteristic curve predicted from each training sample. This represents the frequency vector for full-band scanning. Indicates the first The working condition vector represented by each training sample. This represents the residual of the nodal admittance equation.

5. The method for assessing the resonance risk of new energy power stations according to claim 1, characterized in that, The impedance prediction model includes a feature extraction layer, a physical residual layer, and a fusion layer. The step of inputting the operating condition vector into the pre-trained impedance prediction model to obtain the full-band impedance characteristic curve of the new energy power station includes: The operating condition vector is input into the feature extraction layer for feature extraction to obtain a high-dimensional feature vector, and the operating condition vector is input into the physical residual layer to obtain physical information; The high-dimensional feature vector and the physical information are input into the fusion layer so that the fusion layer outputs a full-band impedance characteristic curve that simultaneously conforms to the data distribution and physical laws.

6. The method for assessing the resonance risk of new energy power stations according to claim 1, characterized in that, The step of calculating the stability margin at each frequency point within the full frequency band based on the full-band impedance characteristic curve and the grid-side impedance characteristic curve includes: For each frequency point within the entire frequency band, obtain the first impedance and the second impedance corresponding to that frequency point from the impedance characteristic curve of the entire frequency band and the impedance characteristic curve of the power grid side, respectively. Calculate the quotient of the absolute value of the second impedance and the absolute value of the first impedance, and determine the result as the stability margin at that frequency point. When traversing the entire frequency band, determine the stability margin of each frequency point within the entire frequency band.

7. The method for assessing the resonance risk of new energy power stations according to claim 1, characterized in that, The method further includes: When the impedance prediction model is applied to another new energy power station, the physical residual layer in the impedance prediction model is frozen, and the feature extraction layer and fusion layer in the impedance prediction model are fine-tuned to obtain an impedance prediction model suitable for another new energy power station. A new impedance prediction model was used to assess the resonance risk of another renewable energy power station.

8. A resonant risk assessment device for new energy power stations, characterized in that, The device includes: The operating condition acquisition module is used to acquire the operating condition vector of new energy power plants; The impedance prediction module is used to input the operating condition vector into a pre-trained impedance prediction model to obtain the full-band impedance characteristic curve of the new energy power station. The impedance prediction model is trained by a preset total loss function, which includes data fitting loss and physical residual loss. The physical residual loss is used to apply physical constraints to the output of the impedance prediction model. The margin calculation module is used to determine the current grid-side impedance characteristic curve and calculate the stability margin at each frequency point in the full frequency range based on the full-band impedance characteristic curve and the grid-side impedance characteristic curve. The risk assessment module is used to identify frequency points with a stability margin less than a preset margin threshold among various frequency points to form a risk frequency set, and to determine the resonance risk level of the new energy power station based on the risk frequency set.

9. A storage medium, characterized in that: The storage medium stores computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the resonant risk assessment method for new energy power stations as described in any one of claims 1 to 7.

10. A computer device, characterized in that, include: One or more processors, and memory; The memory stores computer-readable instructions, which, when executed by the one or more processors, perform the steps of the resonant risk assessment method for new energy power stations as described in any one of claims 1 to 7.

Images