A scene-adaptive distributed photovoltaic response capability evaluation method and device

Abstract

The present application relates to the technical field of distributed power supply application, and specifically provides a scene adaptive distributed photovoltaic response capability evaluation method and device, comprising: obtaining the time sequence input feature vector, performance parameter vector and scene embedding vector of low-voltage distributed photovoltaic equipment within a current time window; taking the time sequence input feature vector, performance parameter vector and scene embedding vector as the input of a pre-trained multi-branch neural network to obtain the low-voltage distributed photovoltaic equipment response capability evaluation result output by the pre-trained multi-branch neural network. The technical scheme provided by the present application fuses the runtime sequence, scene label and performance parameter through a multi-branch neural network to evaluate the low-voltage distributed photovoltaic single-machine power control response capability online, and provides traceable equipment capability judgment basis for dispatching control and stable operation and dynamic support.

Description

Technical Field

[0001] This invention relates to the field of distributed power application technology, specifically to a scenario-adaptive distributed photovoltaic response capability assessment method and apparatus. Background Technology As the scale of distributed photovoltaic (PV) installations in low-voltage distribution networks continues to increase, the low-voltage power supply is gradually shifting from being unobservable and uncontrollable to being observable, measurable, and adjustable. Dispatch control, as well as the stable operation and dynamic support of the distribution network, place higher demands on the ability of individual low-voltage distributed PV devices to adjust their active and reactive power output according to instructions. Given the large number and dispersed distribution of low-voltage distributed PV systems, and the significant differences in equipment models, relying solely on offline sampling or manual experience is insufficient to promptly assess their power control response capabilities. Therefore, it is necessary to utilize high-frequency time-series data collected collaboratively from the cloud and edge to perform online identification and evaluation of PV response capabilities under different operating scenarios.

[0002] Existing methods for assessing the response capability of low-voltage distributed photovoltaic systems still have shortcomings. First, the operating scenarios are mostly preset manually, making it difficult to cover actual changes in operating conditions such as backfeeding from the transformer area, power fluctuations, voltage support, and abnormal execution of control commands. Second, the assessment indicators are often combined with fixed weights, failing to reflect the differences in the importance of adjustable range, continuous operation capability, and fault ride-through capability under different scenarios. Third, some methods rely heavily on single statistical indicators or offline models, failing to simultaneously utilize high-frequency time-series curves, equipment performance parameters, and scenario labels, resulting in unstable online judgments of the power control response capability of individual units. Summary of the Invention

[0003] To overcome the above-mentioned shortcomings, this invention proposes a scenario-adaptive method and apparatus for evaluating the response capability of distributed photovoltaic systems.

[0004] Firstly, a scenario-adaptive distributed photovoltaic (PV) response capability assessment method is provided, the scenario-adaptive distributed PV response capability assessment method comprising: Obtain the time-series input feature vector, performance parameter vector, and scene embedding vector of the low-voltage distributed photovoltaic equipment within the current time window; The time-series input feature vector, performance parameter vector, and scene embedding vector are used as inputs to a pre-trained multi-branch neural network to obtain the low-voltage distributed photovoltaic equipment response capability evaluation results output by the pre-trained multi-branch neural network.

[0005] Preferably, the response capability assessment results of the low-voltage distributed photovoltaic equipment include at least one of the following: active power response value, response capability score, and response capability level.

[0006] Preferably, the training process of the pre-trained multi-branch neural network includes: Training data is constructed using the time-series input feature vectors, performance parameter vectors, scene embedding vectors, and low-voltage distributed photovoltaic (PV) equipment response capability evaluation results within a historical time window of the low-voltage distributed PV equipment. The initial multi-branch neural network is trained using the training data to obtain the pre-trained multi-branch neural network.

[0007] Preferably, the pre-trained multi-branch neural network includes: a temporal feature branch, a performance parameter branch, and a scene feature branch; The mathematical models corresponding to the sequence feature branch, performance parameter branch, and scene feature branch are as follows:

[0008]

[0009]

[0010] In the above formula, For the time-series feature branch output, For performance parameter branch output, Output for the scene feature branch. , , These are time-series feature extraction functions, performance parameter feature extraction functions, and scene feature extraction functions, respectively. , , These are the network parameters for the three branches. , , For low-voltage distributed photovoltaic equipment in the current time window The temporal input feature vector, performance parameter vector, and scene embedding vector are included.

[0011] Furthermore, the low-voltage distributed photovoltaic equipment is in the current time window The temporal input feature vectors within are as follows:

[0012] In the above formula, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The temporal input feature vector within, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The data set inside, For the first Taiwan low-voltage distributed photovoltaic equipment in the time window The data set inside, For the first Taiwan low-voltage distributed photovoltaic equipment in the time window The data set inside, The number of historical windows for the input network, the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The data set inside is as follows:

[0013] In the above formula, For the first Taiwan low-voltage distributed photovoltaic equipment Voltage at grid connection point at all times For the first Taiwan low-voltage distributed photovoltaic equipment Grid-connected current at all times For the first Taiwan low-voltage distributed photovoltaic equipment Active power at all times For the first Taiwan low-voltage distributed photovoltaic equipment Reactive power at all times For the first Taiwan low-voltage distributed photovoltaic equipment Time frequency, For the first Taiwan low-voltage distributed photovoltaic equipment Power control commands received at all times. For the first Taiwan low-voltage distributed photovoltaic equipment Real-time equipment operating status, For the first The total number of moments in each sliding time window.

[0014] Furthermore, the low-voltage distributed photovoltaic equipment is in the current time window The internal performance parameter vector is as follows:

[0015] In the above formula, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention weights for each performance parameter For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner A vector of candidate performance parameters for each performance parameter. For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention weights for each performance parameter For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner A candidate performance parameter vector for the _th performance parameter, wherein the _th ... Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner The candidate performance parameter vectors for each performance parameter are as follows:

[0016] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner The attention weights for each performance parameter are as follows:

[0017] In the above formula, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The proportion of internal power has been increased. For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The proportion of internal power reduction has been lowered. For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal reactive power regulation margin ratio For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal continuous operation qualification rate For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal fault ride-through passability For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal response delay For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal power tracking error, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention score for each performance parameter Total number of performance parameters , for the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention score for each performance parameter.

[0018] Furthermore, the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The percentage adjustments for both active power and active power are as follows:

[0019]

[0020] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window The internal reactive power regulation margin ratio is as follows:

[0021] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window The pass rate for continuous operation is as follows:

[0022] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal fault ride-through pass rate is as follows:

[0023] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window The internal response delay is as follows:

[0024] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window The internal power tracking error is as follows:

[0025] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention scores for each performance parameter are as follows:

[0026] In the above formula, is, For the first The first low-voltage distributed photovoltaic equipment The maximum allowed active power within a time window. For the first The first low-voltage distributed photovoltaic equipment The lower limit of the active power allowed within a time window. For the first The first low-voltage distributed photovoltaic equipment Representative value of active power within a time window For the first Rated active power of low-voltage distributed photovoltaic equipment. For the first The rated apparent power of the low-voltage distributed photovoltaic equipment in Taiwan Used to perform non-negative truncation when sampling error causes the square root term to be less than zero. For the first Taiwan Low-voltage Distributed Photovoltaic Equipment The number of sampling points that exceed voltage limits, frequency limits, or both within a time window. For the first The total number of sampling points within a time window For the first The duration for which a low-voltage distributed photovoltaic system meets grid connection requirements within the fault ride-through assessment range. For the first Total time for low-voltage distributed photovoltaic equipment to overcome faults during the assessment period. To control the timing of command issuance, For the first The actual response of the low-voltage distributed photovoltaic equipment reached the target regulation amount. At that moment, For the first Taiwan low-voltage distributed photovoltaic equipment The target active power is constantly monitored. The first in the candidate performance parameter vector One parameter, , and These are trainable parameters.

[0027] Furthermore, the low-voltage distributed photovoltaic equipment is in the current time window The scene embedding vectors within are as follows:

[0028] In the above formula, Embed functions for the scene. For the first The scene number to which each time window belongs. For the first Class scenario rules, wherein the first The scene numbers to which each time window belongs are as follows:

[0029] In the above formula, This represents the total number of typical scenarios. For the first A scene vector at the station level within a time window. For the first Scene-like center, This is the matrix transpose symbol. This is the feature weight diagonal matrix.

[0030] Furthermore, the first The scene vectors at the station level for each time window are as follows:

[0031] In the above formula, For the first The first low-voltage distributed photovoltaic equipment Normalized features of each time window For the first The number of devices participating in the evaluation within a time window, wherein the number of devices participating in the evaluation within the first time window is... The first low-voltage distributed photovoltaic equipment In the normalized features of the nth time window Features as follows:

[0032] In the above formula, For the first The first low-voltage distributed photovoltaic equipment In the power control scenario characteristics of the first time window One characteristic, The first feature in the power control scenario Minimum value of each feature As a preset positive number, The first feature in the power control scenario The maximum value of each feature.

[0033] Furthermore, the total number of the typical scenarios is as follows:

[0034] In the above formula, For the number of clusters is Profile coefficient at time, For the number of clusters is The Davies-Bouldin index at that time This is the balance coefficient.

[0035] Furthermore, the power control scenario features include: scenario feature vector, average active power within the window, standard deviation of active power within the window, maximum active power ramp rate, maximum voltage deviation, reactive power adjustment margin, control command type code, and operating status code.

[0036] Furthermore, the scenario feature vector, average active power within the window, standard deviation of active power within the window, maximum active power ramp rate, maximum voltage deviation, and reactive power regulation margin are as follows:

[0037]

[0038]

[0039]

[0040]

[0041] In the above formula, , , , , The first The first low-voltage distributed photovoltaic equipment The features of each time window include: scenario feature vector, average active power within the window, standard deviation of active power within the window, maximum active power ramp rate, maximum voltage deviation, and reactive power regulation margin. For the first The total number of sampling points within a time window The time interval between adjacent sampling points. Rated voltage, For the first Active power of a low-voltage distributed photovoltaic system at time t. For the first Taiwan low-voltage distributed photovoltaic equipment Active power at all times For the first Voltage of a low-voltage distributed photovoltaic device at time t.

[0042] Furthermore, the pre-trained multi-branch neural network includes a feature fusion network, the mathematical model of which is as follows:

[0043] In the above formula, Features of fusion; It is a non-linear activation function; For the fusion weight matrix; For fusion bias; This represents vector concatenation. For the first The first low-voltage distributed photovoltaic equipment A comprehensive characterization of scene adaptive performance across time windows. .

[0044] Furthermore, the mathematical model of the output layer of the pre-trained multi-branch neural network is as follows:

[0045]

[0046]

[0047] In the above formula, For the first The equipment in the first Power control response capability score within a time window For the first Active response value at each prediction step size To predict the total number of steps, For the first The equipment in the first The probability of response capability level within a time window , , , , and These are the output layer parameters.

[0048] Furthermore, the loss function used in training the initial multi-branch neural network using the training data is as follows:

[0049] In the above formula, For training loss, For the sample corresponding to the first The equipment in the first A reference score for response capability within a time window. For the first The actual active response value corresponding to each prediction step size. For the sample corresponding to the first The equipment in the first Reference level of response capability within a time window For cross-entropy loss, For regularization terms, , , , This is the loss weight.

[0050] Secondly, a scenario-adaptive distributed photovoltaic response capability assessment device is provided, the scenario-adaptive distributed photovoltaic response capability assessment device comprising: The acquisition module is used to acquire the time-series input feature vector, performance parameter vector, and scene embedding vector of the low-voltage distributed photovoltaic equipment within the current time window; The analysis module is used to take the time-series input feature vector, performance parameter vector and scene embedding vector as input to a pre-trained multi-branch neural network to obtain the low-voltage distributed photovoltaic equipment response capability evaluation result output by the pre-trained multi-branch neural network.

[0051] Thirdly, a computer device is provided, comprising: one or more processors; The processor is used to store one or more programs; When the one or more programs are executed by the one or more processors, the scenario-adaptive distributed photovoltaic response capability assessment method is implemented.

[0052] Fourthly, a computer-readable storage medium is provided, on which a computer program is stored, wherein when the computer program is executed, the scenario-adaptive distributed photovoltaic response capability assessment method is implemented.

[0053] The above-described technical solutions of the present invention have at least one or more of the following beneficial effects: This invention provides a scenario-adaptive distributed photovoltaic (PV) response capability assessment method and apparatus, comprising: acquiring the time-series input feature vector, performance parameter vector, and scenario embedding vector of a low-voltage distributed PV device within the current time window; using the time-series input feature vector, performance parameter vector, and scenario embedding vector as input to a pre-trained multi-branch neural network to obtain the low-voltage distributed PV device response capability assessment result output by the pre-trained multi-branch neural network. The technical solution provided by this invention, by fusing runtime timing, scenario labels, and performance parameters through a multi-branch neural network, online assesses the power control response capability of a single low-voltage distributed PV unit, providing a traceable basis for judging device capability for scheduling control, stable operation, and dynamic support. Specifically: This invention converts operational quantities such as voltage, current, active and reactive power, and frequency collected at the edge into scene vectors, and combines them with control commands and operating status. Then, it forms a scene library through unsupervised clustering. New window data obtains scene labels through scene center matching, reducing the omission of abnormal operating conditions due to fixed manual scene division.

[0054] This invention incorporates the adjustable range, continuous operation, and fault traversal, along with response delay and tracking error, into the same parameter vector, and generates attention weights based on scene labels. The output weighted parameter vector is directly fed into the neural network, avoiding the application of the same set of fixed weights to different scenes.

[0055] This invention employs a multi-branch neural network to process high-frequency time segments, weighted performance parameters, and scene embeddings, and outputs online single-machine response capability scores and levels, while also providing predicted response curves and weak parameters; the scheduling side can use this information to distinguish between devices that can be called immediately, those that need to be called with reduced weight, or those that need to be verified.

[0056] The evaluation results of this invention retain scene labels, parameter weights, response curves, and weak parameters, which can be used for equipment verification and control strategy correction, and for subsequent incremental model training. Attached Figure Description

[0057] Figure 1 This is a schematic diagram of the main steps of the scenario-adaptive distributed photovoltaic response capability assessment method according to an embodiment of the present invention; Figure 2 This is a diagram showing the construction result of a power control scenario library based on high-frequency acquisition data from the cloud and edge, according to an embodiment of the present invention. Figure 3 This is a distribution diagram of attention weights for device performance parameters under different power control scenarios according to an embodiment of the present invention; Figure 4 This is a comparison chart of the power control response curves of a single low-voltage distributed photovoltaic unit according to an embodiment of the present invention; Figure 5 This is a graph showing the online evaluation results of the single-unit power control response capability and the comparison results of the fixed-weight baseline in an embodiment of the present invention. Detailed Implementation

[0058] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0059] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] Example 1 See appendix Figure 1 , Figure 1 This is a schematic flowchart illustrating the main steps of a scenario-adaptive distributed photovoltaic response capability assessment method according to an embodiment of the present invention. Figure 1As shown, the scenario-adaptive distributed photovoltaic response capability assessment method in this embodiment of the invention mainly includes the following steps: Step S101: Obtain the time-series input feature vector, performance parameter vector, and scene embedding vector of the low-voltage distributed photovoltaic equipment within the current time window; Step S102: Use the time-series input feature vector, performance parameter vector and scene embedding vector as input to a pre-trained multi-branch neural network to obtain the low-voltage distributed photovoltaic equipment response capability evaluation result output by the pre-trained multi-branch neural network.

[0061] In this embodiment, the response capability evaluation result of the low-voltage distributed photovoltaic equipment includes at least one of the following: active power response value, response capability score, and response capability level.

[0062] In this embodiment, the training process of the pre-trained multi-branch neural network includes: Training data is constructed using the time-series input feature vectors, performance parameter vectors, scene embedding vectors, and low-voltage distributed photovoltaic (PV) equipment response capability evaluation results within a historical time window of the low-voltage distributed PV equipment. The initial multi-branch neural network is trained using the training data to obtain the pre-trained multi-branch neural network.

[0063] In this embodiment, the pre-trained multi-branch neural network includes: a temporal feature branch, a performance parameter branch, and a scene feature branch; The mathematical models corresponding to the sequence feature branch, performance parameter branch, and scene feature branch are as follows:

[0064]

[0065]

[0066] In the above formula, For the output of the time-series feature branch, For performance parameter branch output, Output for the scene feature branch. , , These are time-series feature extraction functions, performance parameter feature extraction functions, and scene feature extraction functions, respectively. , , These are the network parameters for the three branches. , , For low-voltage distributed photovoltaic equipment in the current time window The temporal input feature vector, performance parameter vector, and scene embedding vector are included.

[0067] In one implementation, the low-voltage distributed photovoltaic device is in the current time window. The temporal input feature vectors within are as follows:

[0068] In the above formula, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The temporal input feature vector within, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The data set inside, For the first Taiwan low-voltage distributed photovoltaic equipment in the time window The data set inside, For the first Taiwan low-voltage distributed photovoltaic equipment in the time window The data set inside, The number of historical windows for the input network, the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The data set inside is as follows:

[0069] In the above formula, For the first Taiwan low-voltage distributed photovoltaic equipment Voltage at grid connection point at all times For the first Taiwan low-voltage distributed photovoltaic equipment Grid-connected current at all times For the first Taiwan low-voltage distributed photovoltaic equipment Active power at all times For the first Taiwan low-voltage distributed photovoltaic equipment Reactive power at all times For the first Taiwan low-voltage distributed photovoltaic equipment Time frequency, For the first Taiwan low-voltage distributed photovoltaic equipment Power control commands received at all times. For the first Taiwan low-voltage distributed photovoltaic equipment Real-time equipment operating status, For the first The total number of moments in each sliding time window.

[0070] In one implementation, the low-voltage distributed photovoltaic device is in the current time window. The internal performance parameter vector is as follows:

[0071] In the above formula, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention weights for each performance parameter For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner A vector of candidate performance parameters for each performance parameter. For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention weights for each performance parameter For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner A candidate performance parameter vector for the _th performance parameter, wherein the _th ... Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner The candidate performance parameter vectors for each performance parameter are as follows:

[0072] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner The attention weights for each performance parameter are as follows:

[0073] In the above formula, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The proportion of internal power has been increased. For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The proportion of internal power reduction was lowered. For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal reactive power regulation margin ratio For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal continuous operation qualification rate For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal fault ride-through passability For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal response delay, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal power tracking error, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention score for each performance parameter This represents the total number of performance parameters. For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention score for each performance parameter.

[0074] In one implementation, the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The percentage adjustments for both active power and active power are as follows:

[0075]

[0076] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window The internal reactive power regulation margin ratio is as follows:

[0077] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window The pass rate for continuous operation is as follows:

[0078] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal fault ride-through pass rate is as follows:

[0079] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window The internal response delay is as follows:

[0080] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window The internal power tracking error is as follows:

[0081] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention scores for each performance parameter are as follows:

[0082] In the above formula, is, For the first The first low-voltage distributed photovoltaic equipment The maximum allowed active power within a time window. For the first The first low-voltage distributed photovoltaic equipment The lower limit of the active power allowed within a time window. For the first The first low-voltage distributed photovoltaic equipment Representative value of active power within a time window For the first Rated active power of low-voltage distributed photovoltaic equipment. For the first The rated apparent power of the low-voltage distributed photovoltaic equipment in Taiwan Used to perform non-negative truncation when sampling error causes the square root term to be less than zero. For the first Taiwan Low-voltage Distributed Photovoltaic Equipment The number of sampling points that exceed voltage limits, frequency limits, or both within a time window. For the first The total number of sampling points within a time window For the first The duration for which a low-voltage distributed photovoltaic system meets grid connection requirements within the fault ride-through assessment range. For the first Total time for low-voltage distributed photovoltaic equipment to overcome faults during the assessment period. To control the timing of command issuance, For the first The actual response of the low-voltage distributed photovoltaic equipment reached the target regulation amount. At that moment, For the first Taiwan low-voltage distributed photovoltaic equipment The target active power is constantly monitored. The first in the candidate performance parameter vector One parameter, , and These are trainable parameters.

[0083] In one implementation, the low-voltage distributed photovoltaic device is in the current time window. The scene embedding vectors within are as follows:

[0084] In the above formula, Embed functions for the scene. For the first The scene number to which each time window belongs. For the first Class scenario rules, wherein the first The scene numbers to which each time window belongs are as follows:

[0085] In the above formula, This represents the total number of typical scenarios. For the first A scene vector at the station level within a time window. For the first Scene-like center, This is the matrix transpose symbol. This is the feature weight diagonal matrix.

[0086] In one implementation, the first The scene vectors at the station level for each time window are as follows:

[0087] In the above formula, For the first The first low-voltage distributed photovoltaic equipment Normalized features of each time window For the first The number of devices participating in the evaluation within a time window, wherein the number of devices participating in the evaluation within the first time window is... The first low-voltage distributed photovoltaic equipment In the normalized features of the nth time window Features as follows:

[0088] In the above formula, For the first The first low-voltage distributed photovoltaic equipment In the power control scenario characteristics of the first time window One characteristic, The first feature in the power control scenario Minimum value of each feature As a preset positive number, The first feature in the power control scenario The maximum value of each feature.

[0089] In one implementation, the total number of typical scenarios is as follows:

[0090] In the above formula, For the number of clusters is Profile coefficient at time, For the number of clusters is The Davies-Bouldin index at that time This is the balance coefficient.

[0091] In one implementation, the power control scenario features include: scenario feature vector, average active power within the window, standard deviation of active power within the window, maximum active power ramp rate, maximum voltage deviation, reactive power adjustment margin, control command type code, and operating status code.

[0092] In one implementation, the scenario feature vector, the average active power within the window, the standard deviation of active power within the window, the maximum active power ramp rate, the maximum voltage deviation, and the reactive power regulation margin are as follows:

[0093]

[0094]

[0095]

[0096]

[0097] In the above formula, , , , , The first The first low-voltage distributed photovoltaic equipment The features of each time window include: scenario feature vector, average active power within the window, standard deviation of active power within the window, maximum active power ramp rate, maximum voltage deviation, and reactive power regulation margin. For the first The total number of sampling points within a time window The time interval between adjacent sampling points. Rated voltage, For the first Active power of a low-voltage distributed photovoltaic system at time t. For the first Taiwan low-voltage distributed photovoltaic equipment Active power at all times For the first Voltage of a low-voltage distributed photovoltaic device at time t.

[0098] In one embodiment, the pre-trained multi-branch neural network includes a feature fusion network, the mathematical model of which is as follows:

[0099] In the above formula, Features of fusion; It is a non-linear activation function; For the fusion weight matrix; For fusion bias; This represents vector concatenation. For the first The first low-voltage distributed photovoltaic equipment A comprehensive characterization of scene adaptive performance across time windows. .

[0100] In one embodiment, the mathematical model of the output layer of the pre-trained multi-branch neural network is as follows:

[0101]

[0102]

[0103] In the above formula, For the first The equipment in the first Power control response capability score within a time window For the first Active response value at each prediction step size To predict the total number of steps, For the first The equipment in the first The probability of response capability level within a time window , , , , and These are the output layer parameters.

[0104] In one implementation, the loss function used in training the initial multi-branch neural network using the training data is as follows:

[0105] In the above formula, For training loss, For the sample corresponding to the first The equipment in the first A reference score for response capability within a time window. For the first The actual active response value corresponding to each prediction step size. For the sample corresponding to the first The equipment in the first Reference level of response capability within a time window For cross-entropy loss, For regularization terms, , , , This is the loss weight.

[0106] In one specific implementation, taking a simulation example of a distribution area containing 60 low-voltage distributed photovoltaic devices, the edge acquisition terminal collects operating quantities such as voltage, current, active power, reactive power, and frequency at 1-second intervals, and records control commands and operating status; the cloud generates device-level feature vectors and distribution area-level scene vectors in a 300-second sliding window.

[0107] Unsupervised clustering of substation-level scenario vectors is performed in the cloud to establish four power control scenarios: active power dispatch control, active voltage support, rapid power fluctuation, and abnormal control response. After new window data enters the system, the scenario number is first determined based on the scenario center, and then the corresponding scenario embedding vector and parameter weights are retrieved.

[0108] During the equipment performance parameter construction phase, the system calculates the active power up-adjustment ratio, active power down-adjustment ratio, reactive power adjustment margin ratio, continuous operation qualification rate, fault ride-through qualification rate, response delay, and power point tracking error for each device. The attention layer generates parameter weights based on the current scenario and forms a weighted performance parameter vector.

[0109] During the online evaluation phase of single-machine response capability, the multi-branch neural network receives high-frequency time-series input, weighted performance parameter input, and scene embedding input, respectively, and outputs the device's response capability score and capability level within the current window, while also outputting the predicted response curve and main weak parameters.

[0110] Simulation results are as follows Figure 2 , 3 As shown in Figures 4 and 5, by Figure 2As can be seen, the scene vectors at the transformer substation level are divided into four categories of power control scenarios; Figure 3 This shows that parameter weights differ in different scenarios; Figure 4 This shows the dynamic trend that the predicted curve of the multi-branch neural network is consistent with the actual response curve; Figure 5 The method of this invention can simultaneously provide equipment scores and ratings, and provides an error comparison with the fixed-weight method. The simulation results above demonstrate that this invention can achieve scenario-adaptive online evaluation of the power control response capability of low-voltage distributed photovoltaic units.

[0111] Example 2 Based on the same inventive concept, the present invention also provides a scenario-adaptive distributed photovoltaic response capability assessment device, the scenario-adaptive distributed photovoltaic response capability assessment device comprising: The acquisition module is used to acquire the time-series input feature vector, performance parameter vector, and scene embedding vector of the low-voltage distributed photovoltaic equipment within the current time window; The analysis module is used to take the time-series input feature vector, performance parameter vector and scene embedding vector as input to a pre-trained multi-branch neural network to obtain the low-voltage distributed photovoltaic equipment response capability evaluation result output by the pre-trained multi-branch neural network.

[0112] Preferably, the response capability assessment results of the low-voltage distributed photovoltaic equipment include at least one of the following: active power response value, response capability score, and response capability level.

[0113] Preferably, the training process of the pre-trained multi-branch neural network includes: Training data is constructed using the time-series input feature vectors, performance parameter vectors, scene embedding vectors, and low-voltage distributed photovoltaic (PV) equipment response capability evaluation results within a historical time window of the low-voltage distributed PV equipment. The initial multi-branch neural network is trained using the training data to obtain the pre-trained multi-branch neural network.

[0114] Preferably, the pre-trained multi-branch neural network includes: a temporal feature branch, a performance parameter branch, and a scene feature branch; The mathematical models corresponding to the sequence feature branch, performance parameter branch, and scene feature branch are as follows:

[0115]

[0116]

[0117] In the above formula, For the output of the time-series feature branch, For performance parameter branch output, Output for the scene feature branch. , , These are time-series feature extraction functions, performance parameter feature extraction functions, and scene feature extraction functions, respectively. , , These are the network parameters for the three branches. , , For low-voltage distributed photovoltaic equipment in the current time window The temporal input feature vector, performance parameter vector, and scene embedding vector are included.

[0118] Furthermore, the low-voltage distributed photovoltaic equipment is in the current time window The temporal input feature vectors within are as follows:

[0119] In the above formula, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The temporal input feature vector within, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The data set inside, For the first Taiwan low-voltage distributed photovoltaic equipment in the time window The data set inside, For the first Taiwan low-voltage distributed photovoltaic equipment in the time window The data set inside, The number of historical windows for the input network, the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The data set inside is as follows:

[0120] In the above formula, For the first Taiwan low-voltage distributed photovoltaic equipment Voltage at grid connection point at all times For the first Taiwan low-voltage distributed photovoltaic equipment Grid-connected current at all times For the first Taiwan low-voltage distributed photovoltaic equipment Active power at all times For the first Taiwan low-voltage distributed photovoltaic equipment Reactive power at all times For the first Taiwan low-voltage distributed photovoltaic equipment Time frequency, For the first Taiwan low-voltage distributed photovoltaic equipment Power control commands received at all times. For the first Taiwan low-voltage distributed photovoltaic equipment Real-time equipment operating status, For the first The total number of moments in each sliding time window.

[0121] Furthermore, the low-voltage distributed photovoltaic equipment is in the current time window The internal performance parameter vector is as follows:

[0122] In the above formula, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention weights for each performance parameter For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner A vector of candidate performance parameters for each performance parameter. For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention weights for each performance parameter For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner A candidate performance parameter vector for the _th performance parameter, wherein the _th ... Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner The candidate performance parameter vectors for each performance parameter are as follows:

[0123] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner The attention weights for each performance parameter are as follows:

[0124] In the above formula, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The proportion of internal power has been increased. For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The proportion of internal power reduction was lowered. For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal reactive power regulation margin ratio For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal continuous operation qualification rate For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal fault ride-through passability For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal response delay, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal power tracking error, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention score for each performance parameter This represents the total number of performance parameters. For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention score for each performance parameter.

[0125] Furthermore, the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The percentage adjustments for both active power and active power are as follows:

[0126]

[0127] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window The internal reactive power regulation margin ratio is as follows:

[0128] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window The pass rate for continuous operation is as follows:

[0129] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal fault ride-through pass rate is as follows:

[0130] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window The internal response delay is as follows:

[0131] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window The internal power tracking error is as follows:

[0132] The first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention scores for each performance parameter are as follows:

[0133] In the above formula, is, For the first The first low-voltage distributed photovoltaic equipment The maximum allowed active power within a time window. For the first The first low-voltage distributed photovoltaic equipment The lower limit of the active power allowed within a time window. For the first The first low-voltage distributed photovoltaic equipment Representative value of active power within a time window For the first Rated active power of low-voltage distributed photovoltaic equipment. For the first The rated apparent power of the low-voltage distributed photovoltaic equipment in Taiwan Used to perform non-negative truncation when sampling error causes the square root term to be less than zero. For the first Taiwan Low-voltage Distributed Photovoltaic Equipment The number of sampling points that exceed voltage limits, frequency limits, or both within a time window. For the first The total number of sampling points within a time window For the first The duration for which a low-voltage distributed photovoltaic system meets grid connection requirements within the fault ride-through assessment range. For the first Total time for low-voltage distributed photovoltaic equipment to overcome faults during the assessment period. To control the timing of command issuance, For the first The actual response of the low-voltage distributed photovoltaic equipment reached the target regulation amount. At that moment, For the first Taiwan low-voltage distributed photovoltaic equipment The target active power is constantly monitored. The first in the candidate performance parameter vector One parameter, , and These are trainable parameters.

[0134] Furthermore, the low-voltage distributed photovoltaic equipment is in the current time window The scene embedding vectors within are as follows:

[0135] In the above formula, Embed functions for the scene. For the first The scene number to which each time window belongs. For the first Class scenario rules, wherein the first The scene numbers to which each time window belongs are as follows:

[0136] In the above formula, This represents the total number of typical scenarios. For the first A scene vector at the station level within a time window. For the first Scene-like center, This is the matrix transpose symbol. This is the feature weight diagonal matrix.

[0137] Furthermore, the first The scene vectors at the station level for each time window are as follows:

[0138] In the above formula, For the first The first low-voltage distributed photovoltaic equipment Normalized features of each time window For the first The number of devices participating in the evaluation within a time window, wherein the number of devices participating in the evaluation within the first time window is... The first low-voltage distributed photovoltaic equipment In the normalized features of the nth time window Features as follows:

[0139] In the above formula, For the first The first low-voltage distributed photovoltaic equipment In the power control scenario characteristics of the first time window One characteristic, The first feature in the power control scenario Minimum value of each feature As a preset positive number, The first feature in the power control scenario The maximum value of each feature.

[0140] Furthermore, the total number of the typical scenarios is as follows:

[0141] In the above formula, For the number of clusters is Profile coefficient at time, For the number of clusters is The Davies-Bouldin index at that time This is the balance coefficient.

[0142] Furthermore, the power control scenario features include: scenario feature vector, average active power within the window, standard deviation of active power within the window, maximum active power ramp rate, maximum voltage deviation, reactive power adjustment margin, control command type code, and operating status code.

[0143] Furthermore, the scenario feature vector, average active power within the window, standard deviation of active power within the window, maximum active power ramp rate, maximum voltage deviation, and reactive power regulation margin are as follows:

[0144]

[0145]

[0146]

[0147]

[0148] In the above formula, , , , , The first The first low-voltage distributed photovoltaic equipment The features of each time window include: scenario feature vector, average active power within the window, standard deviation of active power within the window, maximum active power ramp rate, maximum voltage deviation, and reactive power regulation margin. For the first The total number of sampling points within a time window The time interval between adjacent sampling points. Rated voltage, For the first Active power of a low-voltage distributed photovoltaic system at time t. For the first Taiwan low-voltage distributed photovoltaic equipment Active power at all times For the first Voltage of a low-voltage distributed photovoltaic device at time t.

[0149] Furthermore, the pre-trained multi-branch neural network includes a feature fusion network, the mathematical model of which is as follows:

[0150] In the above formula, Features of fusion; It is a non-linear activation function; For the fusion weight matrix; For fusion bias; This represents vector concatenation. For the first The first low-voltage distributed photovoltaic equipment A comprehensive characterization of scene adaptive performance across time windows. .

[0151] Furthermore, the mathematical model of the output layer of the pre-trained multi-branch neural network is as follows:

[0152]

[0153]

[0154] In the above formula, For the first The equipment in the first Power control response capability score within a time window For the first Active response value at each prediction step size To predict the total number of steps, For the first The equipment in the first The probability of response capability level within a time window , , , , and These are the output layer parameters.

[0155] Furthermore, the loss function used in training the initial multi-branch neural network using the training data is as follows:

[0156] In the above formula, For training loss, For the sample corresponding to the first The equipment in the first A reference score for response capability within a time window. For the first The actual active response value corresponding to each prediction step size. For the sample corresponding to the first The equipment in the first Reference level of response capability within a time window For cross-entropy loss, For regularization terms, , , , This is the loss weight.

[0157] Example 3 Based on the same inventive concept, this invention also provides a computer device, which includes a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to implement corresponding method flows or corresponding functions, thereby realizing the steps of the scenario-adaptive distributed photovoltaic response capability assessment method in the above embodiments.

[0158] Example 4 Based on the same inventive concept, this invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the steps of the scenario-adaptive distributed photovoltaic response capability assessment method in the above embodiments.

[0159] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0160] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0161] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0162] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0163] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A scenario-adaptive method for evaluating the response capability of distributed photovoltaic systems, characterized in that, The method includes: Obtain the time-series input feature vector, performance parameter vector, and scene embedding vector of the low-voltage distributed photovoltaic equipment within the current time window; The time-series input feature vector, performance parameter vector, and scene embedding vector are used as inputs to a pre-trained multi-branch neural network to obtain the low-voltage distributed photovoltaic equipment response capability evaluation results output by the pre-trained multi-branch neural network.

2. The method as described in claim 1, characterized in that, The evaluation results of the response capability of the low-voltage distributed photovoltaic equipment include at least one of the following: active power response value, response capability score, and response capability level.

3. The method as described in claim 1, characterized in that, The training process of the pre-trained multi-branch neural network includes: Training data is constructed using the time-series input feature vectors, performance parameter vectors, scene embedding vectors, and low-voltage distributed photovoltaic (PV) equipment response capability evaluation results within a historical time window of the low-voltage distributed PV equipment. The initial multi-branch neural network is trained using the training data to obtain the pre-trained multi-branch neural network.

4. The method as described in claim 1, characterized in that, The pre-trained multi-branch neural network includes: a temporal feature branch, a performance parameter branch, and a scene feature branch; The mathematical models corresponding to the sequence feature branch, performance parameter branch, and scene feature branch are as follows: In the above formula, For the time-series feature branch output, For performance parameter branch output, Output for the scene feature branch. , , These are time-series feature extraction functions, performance parameter feature extraction functions, and scene feature extraction functions, respectively. , , These are the network parameters for the three branches. , , For low-voltage distributed photovoltaic equipment in the current time window The temporal input feature vector, performance parameter vector, and scene embedding vector are included.

5. The method as described in claim 4, characterized in that, The low-voltage distributed photovoltaic equipment is in the current time window The temporal input feature vectors within are as follows: In the above formula, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The temporal input feature vector within, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The data set inside, For the first Taiwan low-voltage distributed photovoltaic equipment in the time window The data set inside, For the first Taiwan low-voltage distributed photovoltaic equipment in the time window The data set inside, The number of historical windows for the input network, the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The data set inside is as follows: In the above formula, For the first Taiwan low-voltage distributed photovoltaic equipment Voltage at grid connection point at all times For the first Taiwan low-voltage distributed photovoltaic equipment Grid-connected current at all times For the first Taiwan low-voltage distributed photovoltaic equipment Active power at all times For the first Taiwan low-voltage distributed photovoltaic equipment Reactive power at all times For the first Taiwan low-voltage distributed photovoltaic equipment Time frequency, For the first Taiwan low-voltage distributed photovoltaic equipment Power control commands received at all times. For the first Taiwan low-voltage distributed photovoltaic equipment Real-time equipment operating status, For the first The total number of moments in each sliding time window.

6. The method as described in claim 4, characterized in that, The low-voltage distributed photovoltaic equipment is in the current time window The internal performance parameter vector is as follows: In the above formula, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention weights for each performance parameter For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner A vector of candidate performance parameters for each performance parameter. For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention weights for each performance parameter For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner A candidate performance parameter vector for the _th performance parameter, wherein the _th ... Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner The candidate performance parameter vectors for each performance parameter are as follows: The first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner The attention weights for each performance parameter are as follows: In the above formula, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The proportion of internal power has been increased. For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window The proportion of internal power reduction has been lowered. For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal reactive power regulation margin ratio For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal continuous operation qualification rate For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal fault ride-through passability For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal response delay For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal power tracking error, For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention score for each performance parameter This represents the total number of performance parameters. For the first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention score for each performance parameter.

7. The method as described in claim 6, characterized in that, The first Taiwan low-voltage distributed photovoltaic equipment in the current time window The percentage adjustments for both active power and active power are as follows: The first Taiwan low-voltage distributed photovoltaic equipment in the current time window The internal reactive power regulation margin ratio is as follows: The first Taiwan low-voltage distributed photovoltaic equipment in the current time window The pass rate for continuous operation is as follows: The first Taiwan low-voltage distributed photovoltaic equipment in the current time window Internal fault ride-through pass rate is as follows: The first Taiwan low-voltage distributed photovoltaic equipment in the current time window The internal response delay is as follows: The first Taiwan low-voltage distributed photovoltaic equipment in the current time window The internal power tracking error is as follows: The first Taiwan low-voltage distributed photovoltaic equipment in the current time window Inner Attention scores for each performance parameter are as follows: In the above formula, is, For the first The first low-voltage distributed photovoltaic equipment The maximum allowed active power within a time window. For the first The first low-voltage distributed photovoltaic equipment The lower limit of the active power allowed within a time window. For the first The first low-voltage distributed photovoltaic equipment Representative value of active power within a time window For the first Rated active power of low-voltage distributed photovoltaic equipment. For the first The rated apparent power of the low-voltage distributed photovoltaic equipment in Taiwan Used to perform non-negative truncation when sampling error causes the square root term to be less than zero. For the first Taiwan Low-voltage Distributed Photovoltaic Equipment The number of sampling points that exceed voltage limits, frequency limits, or both within a time window. For the first The total number of sampling points within a time window For the first The duration for which a low-voltage distributed photovoltaic system meets grid connection requirements within the fault ride-through assessment range. For the first Total time for low-voltage distributed photovoltaic equipment to overcome faults during the assessment period. To control the timing of command issuance, For the first The actual response of the low-voltage distributed photovoltaic equipment reached the target regulation amount. At that moment, For the first Taiwan low-voltage distributed photovoltaic equipment The target active power is constantly monitored. The first in the candidate performance parameter vector One parameter, , and These are trainable parameters.

8. The method as described in claim 7, characterized in that, The low-voltage distributed photovoltaic equipment is in the current time window The scene embedding vectors within are as follows: In the above formula, Embed functions for the scene. For the first The scene number to which each time window belongs. For the first Class scenario rules, wherein the first The scene numbers to which each time window belongs are as follows: In the above formula, This represents the total number of typical scenarios. For the first A scene vector at the station level within a time window. For the first Scene-like center, This is the matrix transpose symbol. This is the feature weight diagonal matrix.

9. The method as described in claim 8, characterized in that, The first The scene vectors at the station level for each time window are as follows: In the above formula, For the first The first low-voltage distributed photovoltaic equipment Normalized features of each time window For the first The number of devices participating in the evaluation within a time window, wherein the number of devices participating in the evaluation within the first time window is... The first low-voltage distributed photovoltaic equipment In the normalized features of the nth time window Features as follows: In the above formula, For the first The first low-voltage distributed photovoltaic equipment In the power control scenario characteristics of the first time window One characteristic, The first feature in the power control scenario Minimum value of each feature As a preset positive number, The first feature in the power control scenario The maximum value of each feature.

10. The method as described in claim 8, characterized in that, The total number of typical scenarios is as follows: In the above formula, For the number of clusters is Profile coefficient at time, For the number of clusters is The Davies-Bouldin index at that time This is the balance coefficient.

11. The method as described in claim 9, characterized in that, The power control scenario features include: scenario feature vector, average active power within the window, standard deviation of active power within the window, maximum active power ramp rate, maximum voltage deviation, reactive power adjustment margin, control command type code, and operating status code.

12. The method as described in claim 11, characterized in that, The scenario feature vector, average active power within the window, standard deviation of active power within the window, maximum active power ramp rate, maximum voltage deviation, and reactive power regulation margin are as follows: In the above formula, , , , , The first The first low-voltage distributed photovoltaic equipment The features of each time window include: scenario feature vector, average active power within the window, standard deviation of active power within the window, maximum active power ramp rate, maximum voltage deviation, and reactive power regulation margin. For the first The total number of sampling points within a time window The time interval between adjacent sampling points. Rated voltage, For the first Active power of a low-voltage distributed photovoltaic system at time t. For the first Taiwan low-voltage distributed photovoltaic equipment Active power at all times For the first Voltage of a low-voltage distributed photovoltaic device at time t.

13. The method as described in claim 6, characterized in that, The pre-trained multi-branch neural network includes a feature fusion network, whose mathematical model is as follows: In the above formula, Features of fusion; It is a non-linear activation function; For the fusion weight matrix; For fusion bias; This represents vector concatenation. For the first The first low-voltage distributed photovoltaic equipment A comprehensive characterization of scene adaptive performance across time windows. .

14. The method as described in claim 13, characterized in that, The mathematical model for the output layer of the pre-trained multi-branch neural network is as follows: In the above formula, For the first The equipment in the first Power control response capability score within a time window For the first Active response value at each prediction step size To predict the total number of steps, For the first The equipment in the first The probability of response capability level within a time window , , , , and These are the output layer parameters.

15. The method as described in claim 14, characterized in that, The loss function used in training the initial multi-branch neural network using the training data is as follows: In the above formula, For training loss, For the sample corresponding to the first The equipment in the first A reference score for response capability within a time window. For the first The actual active response value corresponding to each prediction step size. For the sample corresponding to the first The equipment in the first Reference level of response capability within a time window For cross-entropy loss, For regularization terms, , , , This is the loss weight.

16. A scenario-adaptive distributed photovoltaic response capability assessment device based on any one of claims 1-15, characterized in that, The device includes: The acquisition module is used to acquire the time-series input feature vector, performance parameter vector, and scene embedding vector of the low-voltage distributed photovoltaic equipment within the current time window; The analysis module is used to take the time-series input feature vector, performance parameter vector and scene embedding vector as input to a pre-trained multi-branch neural network to obtain the low-voltage distributed photovoltaic equipment response capability evaluation result output by the pre-trained multi-branch neural network.

17. A computer device, characterized in that, include: One or more processors; The processor is used to execute one or more programs; When the one or more programs are executed by the one or more processors, the scenario-adaptive distributed photovoltaic response capability assessment method as described in any one of claims 1 to 15 is implemented.

18. A computer-readable storage medium, characterized in that, It contains a computer program, which, when executed, implements the scenario-adaptive distributed photovoltaic response capability assessment method as described in any one of claims 1 to 15.

Images