A steady-state line loss determination method and system for primary frequency modulation of a wind farm

Abstract

The application discloses a kind of steady-state line loss determination method and system for wind farm primary frequency modulation, the method comprises: obtaining wind farm current operating data and determining current working condition and its working condition characteristics, and the working condition node in adaptive working condition space line loss model is matched with current working condition;When matching succeeds, the representative steady-state line loss of matching node is used to determine the current steady-state line loss;When matching fails, the current steady-state line loss is determined according to the number of adjacent working condition nodes, using adjacent node weighting or steady-state line loss prediction model, and a complete working condition node is established to join the model.The working condition node is established based on the steady-state working condition and corresponding representative steady-state line loss identified.The influence of data transmission delay and different operating conditions on line loss determination accuracy can be reduced, and the accuracy and adaptability of wind farm primary frequency modulation line loss compensation can be improved.

Description

Technical Field

[0001] This invention belongs to the field of active power control technology in wind farms, and particularly relates to a method and system for determining steady-state line loss for primary frequency regulation in wind farms. Background Technology

[0002] With the continuous increase in the installed capacity of new energy sources, wind farms participating in primary frequency regulation has become an important means of power grid frequency regulation. During the primary frequency regulation process of wind farms, in order to ensure that the actual active power at the grid connection point can accurately respond to the power regulation requirements of primary frequency regulation, it is necessary to consider the active power loss between the wind turbine side and the grid connection point, and make corresponding corrections to the active power regulation amount on the wind turbine side based on the line loss.

[0003] In existing technologies, line losses are typically calculated based on the total active power of the wind turbine and the active power at the grid connection point, or a fixed line loss function is pre-established, and the corresponding line loss is calculated based on the current active power. However, the total active power of the wind turbine is usually aggregated by the wind turbine energy management system and transmitted to the primary frequency regulation system, and this data transmission has a certain delay. In contrast, the active power at the grid connection point can usually be directly obtained from the grid connection point measurement equipment. Therefore, when the active power of the wind farm changes dynamically, the total active power of the wind turbine and the active power at the grid connection point may correspond to different operating times, and directly using these two figures to calculate line losses can easily lead to significant deviations.

[0004] Furthermore, wind farm line losses are not only related to active power but are also affected by various operating factors such as reactive power, grid connection voltage, and wind speed. Under different operating conditions, even with similar active power levels, actual line losses may differ. Using a fixed line loss function or a fixed loss coefficient is insufficient to fully characterize the line loss characteristics under different operating conditions, which can easily lead to deviations in line loss compensation during primary frequency regulation.

[0005] Meanwhile, the operating conditions of wind farms continuously change with seasons, wind conditions, equipment status, and operating modes, and existing line loss data is usually insufficient to cover all operating conditions. When the current operating condition is not covered by existing line loss data, existing methods cannot accurately obtain the corresponding line loss, and existing line loss models usually lack the ability to continuously supplement and update operating condition information based on subsequent operating data, which can easily lead to a decline in the accuracy and adaptability of the model after long-term operation.

[0006] Therefore, how to accurately determine the steady-state line loss for primary frequency regulation under different operating conditions of wind farms, and how to continuously supplement and update the existing line loss information as the operating conditions change, are technical problems that need to be solved. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for determining steady-state line loss in primary frequency regulation of wind farms, so as to improve the accuracy of steady-state line loss determination under different operating conditions and improve the adaptability of the line loss determination process to uncovered operating conditions and changes in the operating environment.

[0008] To achieve the above objectives, the present invention employs the following technical solution:

[0009] In a first aspect, the present invention provides a method for determining steady-state line losses for primary frequency regulation in wind farms, comprising:

[0010] Obtain the current operating data of the wind farm, and determine the current operating condition and its corresponding characteristics based on the current operating data;

[0011] The operating conditions are matched based on the operating condition characteristics corresponding to the current operating condition and the operating condition characteristics corresponding to each operating condition node in the adaptive operating condition space line loss model.

[0012] If the match is successful, the steady-state line loss corresponding to the current operating condition is determined based on the representative steady-state line loss corresponding to the matched operating condition node. If the match fails, the neighboring operating condition nodes that are similar to the current operating condition are determined from each operating condition node. When the number of neighboring operating condition nodes reaches the preset number of nodes, the steady-state line loss corresponding to the current operating condition is determined based on the representative steady-state line loss corresponding to the neighboring operating condition nodes. When the number of neighboring operating condition nodes does not reach the preset number of nodes, the steady-state line loss corresponding to the current operating condition is determined based on the operating condition characteristics and steady-state line loss prediction model corresponding to the current operating condition.

[0013] The adaptive operating condition spatial line loss model is pre-constructed based on multiple operating condition nodes. Each operating condition node corresponds to a steady-state operating condition identified from wind farm operation data, and includes the operating condition characteristics of the corresponding steady-state operating condition and a representative steady-state line loss determined based on the steady-state operation sample corresponding to the steady-state operating condition.

[0014] When matching fails, the operating condition features corresponding to the current operating condition are used as the operating condition features to complete the operating condition node, and the steady-state line loss corresponding to the current operating condition is used as the representative steady-state line loss corresponding to the complete operating condition node. The complete operating condition node is then established and added to the adaptive operating condition space line loss model.

[0015] Optionally, the operating condition characteristics include the active power at the grid connection point, the total active power of the wind turbine, the reactive power at the grid connection point, and the normalized characteristic values ​​corresponding to the grid connection point voltage and wind speed.

[0016] Optionally, the steady-state operating conditions and operating characteristics corresponding to each operating condition node used to pre-construct the adaptive operating condition space line loss model are determined in the following ways:

[0017] Multiple operation samples are obtained based on wind farm operation data within a preset time window, and the operating condition characteristics corresponding to each operation sample are determined.

[0018] Density clustering is performed based on the operating condition characteristics corresponding to each of the operating samples to identify steady-state sample clusters, eliminate disturbed samples and isolated samples, and determine the operating samples in the steady-state sample clusters as steady-state operating samples to obtain the set of steady-state operating samples corresponding to the steady-state operating conditions.

[0019] The median of each feature is determined based on the operating condition characteristics of each steady-state operating sample in the steady-state operating sample set, and the operating condition characteristics corresponding to the steady-state operating condition are determined based on the median of each feature.

[0020] Optionally, representative steady-state line losses can be determined based on steady-state operating samples corresponding to steady-state conditions, including:

[0021] For each steady-state operating sample in the set of steady-state operating samples, obtain the total active power of the wind turbine and the active power at the grid connection point corresponding to each steady-state operating sample, and determine the line loss sample corresponding to the steady-state operating sample based on the difference between the total active power of the wind turbine and the active power at the grid connection point.

[0022] By collecting the line loss samples corresponding to each of the steady-state operating samples, a set of line loss samples corresponding to the steady-state operating condition is obtained;

[0023] Abnormal line loss samples are removed from the line loss sample set, and the representative steady-state line loss corresponding to the steady-state operating condition is determined based on the remaining line loss samples after removing abnormal line loss samples.

[0024] Optionally, abnormal line loss samples are removed from the line loss sample set, and representative steady-state line loss corresponding to the steady-state operating condition is determined based on the remaining line loss samples after removing abnormal line loss samples, including:

[0025] Determine the median of the line loss sample set, and determine the absolute deviation of each line loss sample from the median;

[0026] The median of each absolute deviation is determined as the median absolute deviation, and abnormal line loss samples are identified based on the median absolute deviation.

[0027] The abnormal line loss samples are removed, and the median of the remaining line loss samples is determined as the representative steady-state line loss corresponding to the steady-state operating condition.

[0028] Optionally, the step of matching the operating conditions based on the operating condition characteristics corresponding to the current operating condition and the operating condition characteristics corresponding to each operating condition node in the adaptive operating condition space line loss model includes:

[0029] Calculate the Euclidean distance between the working condition feature corresponding to the current working condition and the working condition feature corresponding to each working condition node, and determine the minimum Euclidean distance among the Euclidean distances.

[0030] When the minimum Euclidean distance is not greater than the first distance threshold, the working condition is determined to be successfully matched, and the working condition node corresponding to the minimum Euclidean distance is determined as the matched working condition node.

[0031] When the minimum Euclidean distance is greater than the first distance threshold, the working condition matching is determined to be unsuccessful.

[0032] Optionally, determining the steady-state line loss corresponding to the current operating condition when matching fails includes:

[0033] The working condition node whose Euclidean distance from the current working condition is not greater than the second distance threshold is determined as the neighboring working condition node, where the second distance threshold is greater than the first distance threshold.

[0034] Determine whether the number of adjacent working condition nodes has reached the preset number of nodes;

[0035] When the number of neighboring operating condition nodes reaches the preset number of nodes, select the preset number of neighboring operating condition nodes with the smallest Euclidean distance to the current operating condition from the neighboring operating condition nodes. Determine the corresponding weights based on the Euclidean distances of each selected neighboring operating condition node. Determine the steady-state line loss corresponding to the current operating condition based on the representative steady-state line loss and corresponding weights of each neighboring operating condition node.

[0036] When the number of adjacent working condition nodes does not reach the preset number of nodes, the working condition characteristics corresponding to the current working condition are input into the steady-state line loss prediction model for prediction to obtain the steady-state line loss corresponding to the current working condition.

[0037] Optionally, the training process of the steady-state line loss prediction model includes:

[0038] Multiple steady-state operating conditions identified from wind farm operation data are obtained. The operating condition characteristics corresponding to each steady-state operating condition are used as model inputs, and the representative steady-state line loss corresponding to each steady-state operating condition is used as model outputs. Based on the model inputs and model outputs, the extreme gradient boosting algorithm is used for training to obtain the steady-state line loss prediction model.

[0039] In the subsequent operation of the wind farm, new operating data is acquired, and steady-state operating conditions are identified based on the new operating data. The operating condition characteristics corresponding to the steady-state operating conditions are used as incremental training inputs, and the corresponding representative steady-state line loss is used as the incremental training output to incrementally train the steady-state line loss prediction model.

[0040] The operating condition characteristics and representative steady-state line loss corresponding to the completed operating condition nodes are not used as incremental training data for the steady-state line loss prediction model.

[0041] Optionally, each of the aforementioned working condition nodes further includes a timestamp for recording the most recent creation or update time of the corresponding working condition node, and the method further includes:

[0042] During the subsequent operation of the wind farm, new operating data is acquired, and steady-state operating conditions are identified based on the new operating data. New operating condition nodes are established according to the operating condition characteristics and representative steady-state line loss corresponding to the steady-state operating conditions.

[0043] The working condition characteristics corresponding to the newly added working condition node are matched with the working condition characteristics corresponding to each working condition node in the adaptive working condition space line loss model to determine whether there is a working condition node in the adaptive working condition space line loss model that corresponds to the newly added working condition node.

[0044] If it does not exist, the newly added working condition node is added to the adaptive working condition spatial line loss model; if it exists, the working condition node that is closest to the newly added working condition node is determined from the matching working condition nodes, the representative steady-state line loss corresponding to the closest working condition node is updated according to the representative steady-state line loss corresponding to the newly added working condition node, and its timestamp is updated.

[0045] Based on the timestamps corresponding to each working condition node, delete working condition nodes that have not been updated for a preset retention period.

[0046] Secondly, the present invention provides a steady-state line loss determination system for primary frequency regulation in wind farms, comprising:

[0047] The data processing module is used to acquire the current operating data of the wind farm and determine the current operating condition and the corresponding operating condition characteristics based on the current operating data.

[0048] The working condition matching module is used to match the working condition characteristics corresponding to the current working condition with the working condition characteristics corresponding to each working condition node in the adaptive working condition space line loss model.

[0049] The line loss determination module is used to determine the steady-state line loss corresponding to the current operating condition based on the representative steady-state line loss corresponding to the matched operating condition node if the matching is successful; if the matching fails, it determines the neighboring operating condition nodes that are similar to the current operating condition from each operating condition node. When the number of neighboring operating condition nodes reaches the preset number of nodes, it determines the steady-state line loss corresponding to the current operating condition based on the representative steady-state line loss corresponding to the neighboring operating condition nodes. When the number of neighboring operating condition nodes does not reach the preset number of nodes, it determines the steady-state line loss corresponding to the current operating condition based on the operating condition characteristics and steady-state line loss prediction model corresponding to the current operating condition.

[0050] The adaptive operating condition spatial line loss model is pre-constructed based on multiple operating condition nodes. Each operating condition node corresponds to a steady-state operating condition identified from wind farm operation data, and includes the operating condition characteristics of the corresponding steady-state operating condition and a representative steady-state line loss determined based on the steady-state operation sample corresponding to the steady-state operating condition.

[0051] The model completion module is used to, when matching fails, use the operating condition features corresponding to the current operating condition as the operating condition features for completing the operating condition node, and use the steady-state line loss corresponding to the determined current operating condition as the representative steady-state line loss corresponding to the completed operating condition node, to establish the completed operating condition node, and add the completed operating condition node to the adaptive operating condition space line loss model. Compared with the prior art, the beneficial effects achieved by this invention are:

[0052] This invention provides a method and system for determining steady-state line losses in primary frequency regulation of wind farms. It identifies steady-state operating conditions from actual wind farm operation data and forms corresponding operating condition characteristics and representative steady-state line losses based on steady-state operation samples. The line loss information used to construct operating condition nodes can reduce the time asynchrony between the total active power of the wind turbine and the active power at the grid connection point caused by data transmission delays. An adaptive operating condition spatial line loss model is constructed by building operating condition nodes corresponding to multiple steady-state operating conditions, enabling the current operating condition to be matched with existing operating conditions based on multi-dimensional operating condition characteristics, thereby targeting… The system obtains the corresponding steady-state line loss for different operating states. When the current operating condition is not covered by existing operating condition nodes, it further determines the corresponding steady-state line loss based on neighboring operating condition nodes or the steady-state line loss prediction model, and supplements the current operating condition into the adaptive operating condition space line loss model to achieve continuous completion of the uncovered operating conditions. At the same time, the operating condition nodes can be updated using the subsequently identified steady-state operating conditions, and the steady-state line loss prediction model can be incrementally trained, so that the line loss determination results can adapt to changes in wind conditions, equipment status, and operating modes, thereby improving the accuracy and long-term adaptability of line loss compensation during the primary frequency regulation process of wind farms. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of the current operating condition steady-state line loss determination process provided in an embodiment of the present invention;

[0054] Figure 2 This is a schematic diagram of the adaptive working condition spatial line loss model matching process provided in an embodiment of the present invention;

[0055] Figure 3 The flowchart for updating the adaptive working condition spatial line loss model is provided in the embodiments of the present invention. Detailed Implementation

[0056] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0057] Example 1

[0058] This embodiment provides a method for determining steady-state line loss in primary frequency regulation of a wind farm. The method can be executed by a data processing device deployed in the wind farm's primary frequency regulation control system, the power control system of the wind farm, or connected to it in communication. During the primary frequency regulation control process of the wind farm, the active power adjustment on the wind turbine side can be compensated for line loss based on the steady-state line loss determined in this embodiment, so that the active power at the grid connection point after transmission through the in-farm lines is closer to the target power corresponding to the primary frequency regulation.

[0059] In this embodiment, the wind farm operating data includes at least the active power at the grid connection point, the total active power of the wind turbines, the reactive power at the grid connection point, the voltage at the grid connection point, and the wind speed. The active power, reactive power, and voltage at the grid connection point can be obtained from the grid connection point measurement device. The total active power of the wind turbines can be obtained from the wind turbine energy management system or the site monitoring system. The wind speed can be obtained from the wind farm's wind measurement device or the wind speed measurement device corresponding to the wind turbine. Since the dimensions and numerical ranges of different operating parameters are different, the above five types of operating parameters are normalized before being used for steady-state condition identification and condition matching to obtain the corresponding normalized feature values. The same normalization rules and parameters are used in the model building stage and the subsequent online condition matching stage, thereby ensuring the comparability of the condition features obtained at different stages.

[0060] It should be noted that the normalization process is used to form the operating condition characteristics involved in operating condition identification and operating condition distance calculation, but it does not change the actual values ​​of the total active power of the wind turbine and the active power at the grid connection point used to calculate the actual line loss. In other words, the line loss sample is determined based on the original power data, while operating condition identification and operating condition matching are performed based on the corresponding normalized operating condition characteristics.

[0061] In execution Figure 1 Before determining the steady-state line loss under the current operating condition, an adaptive operating condition spatial line loss model can be established in advance based on the historical operating data of the wind farm. The adaptive operating condition spatial line loss model includes multiple operating condition nodes, each of which represents a steady-state operating condition of the wind farm and includes at least the operating condition characteristics and representative steady-state line loss corresponding to that steady-state operating condition.

[0062] Specifically, a sliding window can be used to continuously capture wind farm operation data. In one specific implementation, the preset time window can be 5 seconds, and multiple operation samples are acquired within each preset time window. An operation sample includes the active power at the grid connection point, the total active power of the wind turbine, the reactive power at the grid connection point, the voltage at the grid connection point, and the wind speed at the same sampling time. The operating condition feature vector corresponding to the above five types of operation data is formed based on the normalized feature values ​​corresponding to the above five types of operation data.

[0063] For multiple operating samples within a preset time window, the density-based spatial clustering algorithm DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is used for density clustering. DBSCAN can aggregate operating samples with similar operating conditions and sufficient sample density in the feature space to form steady-state sample clusters based on a preset neighborhood radius and the minimum number of samples required to form core samples. Operating samples deviating from these steady-state sample clusters are identified as disturbed or isolated samples. After removing disturbed and isolated samples, the operating samples in the steady-state sample clusters are determined as steady-state operating samples, and each steady-state operating sample constitutes a steady-state operating sample set corresponding to a steady-state operating condition.

[0064] Therefore, line loss is not determined directly based on the total active power of the wind turbine and the active power at the grid connection point at a single moment. Instead, a relatively stable operating sample is first determined over a period of time, and then the steady-state line loss is determined based on this steady-state operating sample. Through this process, even when there is a certain transmission delay in the total active power data of the wind turbine, the impact of the time asynchrony between the two types of power data on the line loss calculation results can be reduced because the changes in relevant operating quantities within the steady-state condition are relatively small.

[0065] For the first in the steady-state operation sample set For each steady-state operating sample, obtain its corresponding total active power of the wind turbine. Active power at grid connection point Determine the line loss sample corresponding to the steady-state operating sample. :

[0066]

[0067] The above processing is performed on each steady-state operating sample in the steady-state operating sample set to obtain the line loss sample set corresponding to the steady-state operating condition.

[0068] Considering that outliers may exist in the line loss sample due to communication anomalies, measurement errors, or short-term disturbances, outlier removal is performed on the line loss sample set based on the absolute deviation of the median.

[0069] Specifically, the median of the line loss sample set is first determined. And calculate the absolute deviation between each line loss sample and the median:

[0070]

[0071] in, Indicates the first The absolute deviation between the line loss sample corresponding to each steady-state operating sample and the median;

[0072] Further determine the median of each absolute deviation to obtain the median absolute deviation. :

[0073]

[0074] in, This represents the number of steady-state samples in the steady-state sample set. This indicates a median-taking operation; abnormal line loss samples are identified based on the MAD and preset anomaly detection parameters. For example, line loss samples whose deviation from the median of the line loss samples exceeds the anomaly range determined based on MAD can be identified as abnormal line loss samples. After removing abnormal line loss samples, the median of the remaining line loss samples is determined as the representative steady-state line loss corresponding to this steady-state operating condition. .

[0075] While determining the representative steady-state line loss, the median is determined for each dimension of the operating condition characteristics in the steady-state operating sample set. For example, the median of the normalized active power characteristic at the grid connection point, the normalized active power characteristic of the wind turbine, the normalized reactive power characteristic at the grid connection point, the normalized voltage characteristic at the grid connection point, and the normalized wind speed characteristic are determined in the steady-state operating sample set, and the operating condition characteristics corresponding to this steady-state operating condition are formed by the median of these five characteristics.

[0076] Therefore, a working condition node can be represented as:

[0077]

[0078] in, Indicates the first The five-dimensional working condition features corresponding to each working condition node. Indicates the first Representative steady-state line loss corresponding to each operating condition node This represents the timestamp of the corresponding working condition node.

[0079] In one specific implementation, a timestamp can record the establishment time of the corresponding operating condition node, and be updated to the most recent update time when the node is subsequently updated based on new actual steady-state operating condition data. By repeating the above process for different historical steady-state operating conditions, multiple operating condition nodes are obtained, and an adaptive operating condition spatial line loss model is constructed from these multiple operating condition nodes.

[0080] Furthermore, a steady-state line loss prediction model can be pre-trained before determining online line losses. Specifically, multiple steady-state operating conditions identified from actual wind farm operation data are acquired. The five-dimensional operating condition features corresponding to each steady-state operating condition are used as model input, and the representative steady-state line loss corresponding to each steady-state operating condition is used as model output. The model is trained using the Extreme Gradient Boosting (XGBoost) algorithm to obtain the steady-state line loss prediction model. This steady-state line loss prediction model is used to predict the corresponding steady-state line loss based on the operating condition characteristics of the current operating condition when there are insufficient operating condition nodes near the current operating condition.

[0081] During model training, the prediction error between the predicted steady-state line loss output by the steady-state line loss prediction model and the representative steady-state line loss of the corresponding training samples can be used as the model optimization objective. The prediction error is reduced through iterative training. The loss function used to characterize the prediction error can be the squared error loss function, or other loss functions suitable for regression tasks. Training parameters such as the model's learning rate, number of decision trees, and decision tree depth can be set according to the number of training samples and the required model prediction accuracy.

[0082] It should be noted that the extreme gradient boosting algorithm itself can be implemented using the XGBoost algorithm commonly used in this field. Its specific training processes, such as decision tree construction, gradient iteration, and model parameter optimization, can adopt the conventional implementation methods of the corresponding algorithm, and this embodiment does not impose specific limitations on them. This embodiment mainly utilizes the extreme gradient boosting algorithm to establish a mapping relationship between operating condition characteristics and representative steady-state line loss, so as to predict the steady-state line loss corresponding to the current operating condition when the adaptive operating condition spatial line loss model cannot provide enough neighboring operating condition nodes.

[0083] After completing the pre-construction of the adaptive operating condition spatial line loss model and the steady-state line loss prediction model, execute... Figure 1 The process for determining the steady-state line loss under the current operating conditions, as shown, includes:

[0084] S1: Obtain the current operating data of the wind farm, and determine the current operating condition and the corresponding operating condition characteristics based on the current operating data;

[0085] S2: Perform condition matching based on the condition characteristics corresponding to the current condition and the condition characteristics corresponding to each condition node in the adaptive condition space line loss model;

[0086] S3: If the matching is successful, determine the steady-state line loss corresponding to the current working condition based on the representative steady-state line loss corresponding to the matched working condition node; if the matching fails, determine the neighboring working condition node that is similar to the current working condition from each working condition node. When the number of neighboring working condition nodes reaches the preset number of nodes, determine the steady-state line loss corresponding to the current working condition based on the representative steady-state line loss corresponding to the neighboring working condition node. When the number of neighboring working condition nodes does not reach the preset number of nodes, determine the steady-state line loss corresponding to the current working condition based on the working condition characteristics and steady-state line loss prediction model corresponding to the current working condition.

[0087] In this embodiment, step S1 includes:

[0088] Obtain the active power at the grid connection point, the total active power of the wind turbines, the reactive power at the grid connection point, the voltage at the grid connection point, and the wind speed corresponding to the current operating cycle of the wind farm, as the current operating data.

[0089] The active power at the grid connection point, total active power of the wind turbine, reactive power at the grid connection point, grid connection point voltage, and wind speed in the current operating data are normalized to obtain the normalized characteristic values ​​corresponding to each operating parameter. The normalization rules and parameters used for the current operating data are consistent with those used in the pre-constructed adaptive operating condition space line loss model, so that the operating condition characteristics corresponding to the current operating condition are in the same feature space as the operating condition characteristics corresponding to each operating condition node, and can be used for subsequent distance calculation and operating condition matching.

[0090] Specifically, the operating condition characteristics corresponding to the current operating condition can be represented as a five-dimensional operating condition feature vector. :

[0091]

[0092] in, This represents the normalized characteristic value corresponding to the active power at the current grid connection point. This represents the normalized characteristic value corresponding to the total active power of the current wind turbine. This represents the normalized characteristic value corresponding to the reactive power at the current grid connection point. This represents the normalized characteristic value corresponding to the current grid connection point voltage. This represents the normalized eigenvalue corresponding to the current wind speed.

[0093] In this embodiment, the current operating condition refers to the wind farm's operating status as reflected by current operating data, and the five-dimensional operating condition feature vector is used to characterize the current operating condition. Since wind farm line losses are not only related to active power levels but also affected by operating factors such as reactive power, voltage, and wind speed, characterizing the current operating condition using these multiple operating parameters can comprehensively reflect the similarity between the current operating state and historical steady-state operating conditions during subsequent operating condition matching.

[0094] It should be noted that the normalization process in this step is only used to form the operating condition characteristics participating in the operating condition matching, and does not mean that the normalized grid-connected active power and the total active power of the wind turbine will be used for actual line loss calculation. The representative steady-state line loss used when pre-constructing the operating condition nodes is still determined based on the actual total active power of the wind turbine and the actual active power of the grid-connected point in the corresponding steady-state operation sample.

[0095] In this embodiment, steps S2-S3 perform operating condition matching based on the operating condition characteristics corresponding to the current operating condition and the operating condition characteristics corresponding to each operating condition node in the adaptive operating condition space line loss model, and determine the steady-state line loss corresponding to the current operating condition based on the operating condition matching result. Wherein, Figure 2 The specific process of matching the operating conditions and determining the steady-state line loss in the event of matching failure is further illustrated.

[0096] like Figure 2 As shown, in step S21, the Euclidean distance between the working condition feature corresponding to the current working condition and the working condition feature corresponding to each working condition node is calculated, and the minimum Euclidean distance among all Euclidean distances is determined.

[0097] As mentioned earlier, the five-dimensional working condition characteristics corresponding to the current working condition are represented as follows:

[0098]

[0099] For the adaptive working condition spatial line loss model, the first... Each working condition node, and its corresponding five-dimensional working condition characteristics, can be represented as follows:

[0100]

[0101] in, , , , and They respectively represent the formation of the first In the steady-state operation sample set of each operating node, the median of the normalized characteristics of the active power at the grid connection point, the normalized characteristics of the total active power of the wind turbine, the normalized characteristics of the reactive power at the grid connection point, the normalized characteristics of the voltage at the grid connection point, and the normalized characteristics of the wind speed are taken. The median of these five characteristics collectively represents the... The steady-state operating conditions corresponding to each operating condition node.

[0102] Based on the five-dimensional working condition characteristics corresponding to the current working condition and the Five-dimensional working condition features corresponding to each working condition node Current operating conditions and the first Euclidean distance between each working node It can be determined according to the following formula:

[0103]

[0104] Since the operating condition characteristics corresponding to the current operating condition and the operating condition characteristics corresponding to each operating condition node are obtained using the same normalization rules and normalization parameters, different types of operating parameters have a unified numerical scale when calculating distances. The smaller the Euclidean distance, the closer the current operating condition is to the steady-state operating condition represented by the corresponding operating condition node.

[0105] By traversing each working condition node in the adaptive working condition spatial line loss model, the corresponding Euclidean distance set is obtained, and the minimum Euclidean distance is determined from it. , where K represents the number of working condition nodes in the current adaptive working condition spatial line loss model;

[0106] Determine whether the minimum Euclidean distance is not greater than a first distance threshold. If the distance is not greater than the first distance threshold, proceed to step S22; when If the distance exceeds the first distance threshold, it is determined that the current working condition is not directly covered by the existing working condition node, and the process proceeds to step S23.

[0107] Step S22: Confirm that the working condition matching is successful, and determine the working condition node corresponding to the minimum Euclidean distance as the matched working condition node.

[0108] If a match is successful, it indicates that the current operating condition is sufficiently close to the historical steady-state operating condition corresponding to that operating condition node in the five-dimensional operating condition feature space. Therefore, the representative steady-state line loss corresponding to that operating condition node can be determined as the steady-state line loss corresponding to the current operating condition. Using the representative steady-state line loss of an existing steady-state operating condition for determination can avoid directly using the real-time total active power of the wind turbine and the active power at the grid connection point, which may have time asynchrony, to recalculate the line loss during a single frequency regulation process.

[0109] Step S23: In the event of failure to match working conditions, the working condition node whose Euclidean distance from the current working condition is not greater than the second distance threshold is determined as the neighboring working condition node, wherein the second distance threshold is greater than the first distance threshold.

[0110] The first distance threshold is used to determine whether existing operating condition nodes can directly represent the current operating condition. The second distance threshold is used to expand the search range after a direct match fails, in order to obtain similar operating condition nodes that can be used to estimate the steady-state line loss of the current operating condition. Therefore, the operating condition range corresponding to the second distance threshold is larger than the direct matching range.

[0111] Further determine whether the number of adjacent working condition nodes has reached the preset number of nodes. In one specific embodiment, the preset number of nodes is 3; in other embodiments, the preset number of nodes can also be preset according to actual application needs.

[0112] When the number of adjacent working condition nodes reaches the preset number of nodes, proceed to step S24.

[0113] Step S24: Select a preset number of neighboring working condition nodes from the neighboring working condition nodes that have the smallest Euclidean distance from the current working condition.

[0114] For example, when there are 5 neighboring operating condition nodes within the second distance threshold range and the preset number of nodes is 3, the 3 neighboring operating condition nodes with the smallest Euclidean distance are selected from the 5 neighboring operating condition nodes, instead of using all 5 nodes for subsequent calculations. This allows priority to be given to historical nodes whose operating condition characteristics are closest to the current operating condition, reducing the impact of line loss corresponding to more distant operating conditions on the current line loss determination result.

[0115] Step S25: Determine the corresponding weights based on the Euclidean distances of the selected neighboring operating condition nodes, and determine the steady-state line loss corresponding to the current operating condition based on the representative steady-state line loss and corresponding weights of each neighboring operating condition node.

[0116] Specifically, for the preset number of neighboring operating condition nodes selected in step S24, the weights corresponding to each neighboring operating condition node are determined according to the Euclidean distance between each neighboring operating condition node and the current operating condition. The representative steady-state line loss of the corresponding neighboring operating condition node is weighted using each weight to obtain the steady-state line loss corresponding to the current operating condition.

[0117] In one specific implementation, the preset number of nodes is 3. Let the three adjacent working condition nodes selected in step S24 be the nth node, ... The, the The and the first There are three working condition nodes, and the Euclidean distances between the three neighboring working condition nodes and the current working condition are respectively... , and ,in, .

[0118] The corresponding weights are determined based on the Euclidean distance between each neighboring working condition node and the current working condition. For the ... Each neighboring working condition node has its corresponding weight. Determined according to the following formula:

[0119]

[0120] in, Indicates the first The Euclidean distance between each neighboring node and the current node. Indicates the first The weights are assigned to neighboring operating condition nodes. As can be seen from the above weight determination method, the smaller the Euclidean distance between a neighboring operating condition node and the current operating condition, the greater its corresponding weight. This means that operating condition nodes whose operating characteristics are closer to the current operating condition have a greater impact on the determination of the current steady-state line loss.

[0121] Let the representative steady-state line losses corresponding to the three adjacent operating condition nodes be as follows: , and Then, a weighted average is performed based on the representative steady-state line loss corresponding to each neighboring operating condition node and its corresponding weight to obtain the steady-state line loss corresponding to the current operating condition. Specifically, it is determined according to the following formula:

[0122]

[0123] in, Indicates the first The representative steady-state line loss corresponding to each neighboring operating node.

[0124] In a specific form, the above formula can be expressed as:

[0125]

[0126] Therefore, when the current operating condition cannot be directly matched with existing operating condition nodes, but there are a sufficient number of neighboring operating condition nodes near the current operating condition, the steady-state line loss of the current operating condition can be determined by comprehensively utilizing the representative steady-state line loss corresponding to multiple similar steady-state operating conditions, and the neighboring operating condition nodes that are closer to the current operating condition have a greater weight in determining the line loss.

[0127] When the number of adjacent working condition nodes does not reach the preset number of nodes, proceed to step S26.

[0128] Step S26: Input the operating condition characteristics corresponding to the current operating condition into the steady-state line loss prediction model for prediction, and obtain the steady-state line loss corresponding to the current operating condition.

[0129] Specifically, the five-dimensional normalized working condition characteristics corresponding to the current operating condition are used as input to the XGBoost steady-state line loss prediction model. The steady-state line loss prediction model outputs the corresponding steady-state line loss prediction value, which is then used as the steady-state line loss corresponding to the current operating condition. This processing is suitable for situations where the current operating condition is located in a sparse region of the existing operating condition space, i.e., where it cannot be directly matched with existing operating condition nodes, nor is there a sufficient number of neighboring operating condition nodes for reliable local weighted estimation.

[0130] The steady-state line loss prediction model is trained using the aforementioned Extreme Gradient Boosting (XGBoost) algorithm. Its input consists of five-dimensional operating condition features identified from actual wind farm operation data, and its output is the representative steady-state line loss for that operating condition. This method allows for the determination of the steady-state line loss for the current operating condition even when sufficient historical steady-state operating condition nodes are lacking in the vicinity. This is based on the mapping relationship between the learned operating condition features and the representative steady-state line loss.

[0131] Therefore, after the direct matching of operating conditions fails, the local line loss is first estimated using the adjacent operating condition nodes; the steady-state line loss prediction model is only called when the number of adjacent operating condition nodes is insufficient, thus forming a hierarchical line loss determination method of "adjacent operating condition weighting - prediction model supplementation".

[0132] Step S27: Output the steady-state line loss corresponding to the current operating condition.

[0133] Specifically, based on the different processing results of steps S2 and S3, the steady-state line loss corresponding to the current operating condition can be determined in any of the following ways:

[0134] When the minimum Euclidean distance is not greater than the first distance threshold, the working condition matching is successful, and the representative steady-state line loss corresponding to the matched working condition node is determined as the steady-state line loss corresponding to the current working condition.

[0135] When the minimum Euclidean distance is greater than the first distance threshold and the number of neighboring working condition nodes within the range of the second distance threshold reaches the preset number of nodes, the steady-state line loss determined in step S25 based on the weighted average of the selected neighboring working condition nodes is taken as the steady-state line loss corresponding to the current working condition.

[0136] When the minimum Euclidean distance is greater than the first distance threshold and the number of neighboring working condition nodes within the range of the second distance threshold does not reach the preset number of nodes, the steady-state line loss prediction value output by the steady-state line loss prediction model in step S26 is taken as the steady-state line loss corresponding to the current working condition.

[0137] The steady-state line loss obtained through the corresponding path is used as the output steady-state line loss corresponding to the current operating condition, so as to provide line loss compensation for the primary frequency regulation process of the wind farm.

[0138] Furthermore, when step S2 determines that the working condition matching fails, after determining the steady-state line loss corresponding to the current working condition through step S25 or step S26, the adaptive working condition spatial line loss model is also completed based on the current working condition and the corresponding steady-state line loss.

[0139] Specifically, based on the five-dimensional working condition characteristics corresponding to the current working condition. As the operating condition feature corresponding to the completed operating condition node, the steady-state line loss corresponding to the current operating condition determined in step S25 or step S26. As a representative steady-state line loss corresponding to the completed operating condition node, the timestamp corresponding to the establishment of the completed operating condition node is recorded. This establishes complete working condition nodes:

[0140]

[0141] The completed operating condition nodes are added to the adaptive operating condition space line loss model, allowing operating conditions not directly covered by existing operating condition nodes to enter the adaptive operating condition space. When the same or similar operating conditions reappear, the completed operating condition nodes can participate in operating condition matching or neighboring operating condition node search, thereby gradually expanding the range of operating conditions covered by the adaptive operating condition space line loss model.

[0142] It should be noted that the representative steady-state line loss corresponding to the completed operating condition node is derived from the weighted calculation results of neighboring operating condition nodes or the prediction results of the steady-state line loss prediction model, rather than being calculated based on the subsequently identified steady-state operating samples. Therefore, in this embodiment, the completed operating condition node can be added to the adaptive operating condition space line loss model to participate in subsequent operating condition matching, but its corresponding operating condition characteristics and representative steady-state line loss are not used as incremental training data for the steady-state line loss prediction model.

[0143] As the wind farm continues to operate, the adaptive operating condition spatial line loss model can be updated based on newly collected actual operating data, specifically as follows: Figure 3 As shown.

[0144] Step S41: Acquire new operational data during the subsequent operation of the wind farm.

[0145] The newly added operational data refers to the actual wind farm operational data collected after the adaptive operating condition spatial line loss model is put into use, and its data type can be the same as the data type used in the initial model construction stage.

[0146] Step S42: Identify steady-state operating conditions based on the newly added operating data, and establish new operating condition nodes according to the operating condition characteristics and representative steady-state line loss corresponding to the steady-state operating conditions.

[0147] Specifically, for the newly added operating data, the aforementioned sliding window and DBSCAN density clustering method are used to identify steady-state sample clusters, and disturbed samples and isolated samples are removed to obtain the set of steady-state operating samples corresponding to the newly added steady-state operating conditions.

[0148] The operating characteristics corresponding to the newly added steady-state operating condition are determined based on the median of each operating condition characteristic in the steady-state operating sample set. A line loss sample set is formed based on the difference between the actual total active power of the wind turbine and the active power at the grid connection point corresponding to each steady-state operating sample. Abnormal line loss samples are removed using the MAD algorithm, and the median of the remaining line loss samples is determined as the representative steady-state line loss corresponding to the newly added steady-state operating condition.

[0149] The new operating condition node is established based on the operating condition characteristics, representative steady-state line loss, and corresponding timestamp of the newly added steady-state operating condition.

[0150] It should be noted that the newly added operating condition node in step S42 is established based on the steady-state operating conditions identified from the subsequent actual operation data of the wind farm, which is different from the supplementary operating condition node formed by the aforementioned weighted neighboring node or steady-state line loss prediction model.

[0151] Step S43: Match the operating condition characteristics corresponding to the newly added operating condition node with the operating condition characteristics corresponding to each operating condition node in the adaptive operating condition space line loss model to determine whether there is an operating condition node in the adaptive operating condition space line loss model that corresponds to the newly added operating condition node.

[0152] In one specific implementation, the same Euclidean distance calculation method as in step S21 can be used for comparison. When the Euclidean distance between the existing working condition node and the newly added working condition node is not greater than the first distance threshold, it is determined that a matching working condition node exists; otherwise, it is determined that no matching working condition node exists.

[0153] When no matching operating condition node exists, step S44 is executed to add the new operating condition node to the adaptive operating condition spatial line loss model. In this case, the new operating condition node is used to supplement the operating condition regions that were not previously covered by actual steady-state operating data.

[0154] When a matching working condition node exists, proceed to step S45 to determine the working condition node that is closest to the newly added working condition node from among the matching working condition nodes. Specifically, the Euclidean distance between each matching working condition node and the newly added working condition node can be compared, and the working condition node with the smallest Euclidean distance can be determined as the closest working condition node.

[0155] Step S46: Update the representative steady-state line loss corresponding to the closest operating condition node according to the representative steady-state line loss corresponding to the newly added operating condition node, and update the timestamp corresponding to the operating condition node.

[0156] In this embodiment, the representative steady-state line loss corresponding to the existing operating condition node is not directly replaced by the representative steady-state line loss corresponding to the newly added operating condition node. Instead, a weighted update is performed by combining the representative steady-state line loss before the update and the representative steady-state line loss corresponding to the newly added operating condition node.

[0157] The updated representative steady-state line loss is determined by the following formula:

[0158]

[0159] in, This represents the updated representative steady-state line loss. This represents the representative steady-state line loss of the closest existing operating condition node before the update. This represents the representative steady-state line loss corresponding to the newly added operating condition node. This represents the update coefficient.

[0160] In one specific implementation:

[0161]

[0162] That is, based on the existing representative steady-state line loss, the line loss is gradually updated using the representative steady-state line loss corresponding to the newly added actual steady-state operating conditions.

[0163] By adopting the above method, on the one hand, existing operating condition nodes can be continuously corrected based on subsequent actual operating data, and on the other hand, it can be avoided that a single addition of steady-state operating condition data will cause excessively large changes in the line loss value corresponding to the existing operating condition nodes.

[0164] After completing the representative steady-state line loss update, the timestamp corresponding to the operating condition node is updated to the time corresponding to this update, so that the timestamp can represent the time when the operating condition node last received the actual steady-state operating condition data update.

[0165] Step S47: Delete the working condition nodes that have not been updated for a preset retention period based on the timestamps corresponding to each working condition node.

[0166] Specifically, all operating condition nodes in the adaptive operating condition spatial line loss model can be traversed, and the current time can be compared with the most recent establishment or update time of each operating condition node. When the difference between the two exceeds a preset retention period, the corresponding operating condition node is deleted from the adaptive operating condition spatial line loss model. In one specific implementation, the preset retention period can be set to 15 days.

[0167] By adopting the above-mentioned node elimination mechanism, the problem of model obsolescence caused by the continuous retention of historical line loss nodes that have not appeared for a long time can be reduced, and the current adaptive operating condition spatial line loss model can focus more on reflecting the line loss relationship of the wind farm in the recent equipment status and operating environment.

[0168] remove Figure 3 In addition to updating the adaptive working condition spatial line loss model, the actual steady-state working conditions identified by subsequent newly added operating data can be used to incrementally train the steady-state line loss prediction model.

[0169] Specifically, the operating characteristics corresponding to the steady-state operating conditions identified by the newly added operating data are used as the incremental training input, and the representative steady-state line loss determined based on the actual steady-state operating samples of the steady-state operating condition is used as the incremental training output. On the basis of the existing XGBoost steady-state line loss prediction model, incremental training is continued to be carried out so that the model parameters are continuously updated with the newly added actual steady-state operating data.

[0170] Specifically, supplementary operating condition nodes established through weighted summation of neighboring operating condition nodes or steady-state line loss prediction models due to direct matching failures are not used as incremental training data. This avoids the model using its own prediction results or data based on those prediction results as supervised targets for iterative learning, reducing the accumulation of prediction errors in subsequent training processes. Only steady-state operating conditions and corresponding representative steady-state line losses re-identified based on subsequent actual operating data participate in incremental training.

[0171] Through the above process, the adaptive operating condition spatial line loss model and the steady-state line loss prediction model form a complementary online update mechanism with distinct data sources: the adaptive operating condition spatial line loss model can quickly expand the currently queried operating condition coverage by supplementing operating condition nodes, and further utilize subsequent actual steady-state data for correction and updates; the steady-state line loss prediction model continuously optimizes itself using the actually identified steady-state operating condition data as incremental training data. Therefore, the steady-state line loss determination process can adapt to long-term operating environment changes such as equipment aging, seasonal variations, and changes in operating modes.

[0172] This embodiment extracts representative steady-state line losses after identifying steady-state operating conditions, reducing the impact of time asynchrony between the total active power of the wind turbine and the active power at the grid connection point on line loss calculation during dynamic processes. An adaptive operating condition space is constructed through multi-dimensional operating condition features, and a hierarchical determination mechanism using direct matching, neighboring node weighting, and sequential fallback of the prediction model is adopted to improve the adaptability of steady-state line loss acquisition under different operating conditions. Simultaneously, through operating condition completion, actual steady-state operating condition updates, expired node elimination, and incremental training of the prediction model, the model can be continuously updated according to the long-term operating status changes of the wind farm, thereby improving the accuracy of line loss compensation and the long-term adaptability of the model during primary frequency regulation.

[0173] Example 2

[0174] This embodiment provides a steady-state line loss determination system for primary frequency regulation of wind farms, which implements the steady-state line loss determination method for primary frequency regulation of wind farms in Embodiment 1. The system can be integrated into the primary frequency regulation control device of the wind farm, or it can be used as an independent data processing device to communicate with the primary frequency regulation control device, the wind turbine energy management system and the wind farm measurement system.

[0175] The steady-state line loss determination system includes a data processing module, an operating condition matching module, a line loss determination module, and a model completion module.

[0176] The data processing module is used to acquire the current operating data of the wind farm and determine the current operating condition and the corresponding operating condition characteristics based on the current operating data.

[0177] The working condition matching module is used to match the working condition characteristics corresponding to the current working condition with the working condition characteristics corresponding to each working condition node in the adaptive working condition space line loss model.

[0178] The line loss determination module is used to determine the steady-state line loss corresponding to the current operating condition based on the representative steady-state line loss corresponding to the matched operating condition node if the matching is successful; if the matching fails, it determines the neighboring operating condition nodes that are similar to the current operating condition from each operating condition node. When the number of neighboring operating condition nodes reaches the preset number of nodes, it determines the steady-state line loss corresponding to the current operating condition based on the representative steady-state line loss corresponding to the neighboring operating condition nodes. When the number of neighboring operating condition nodes does not reach the preset number of nodes, it determines the steady-state line loss corresponding to the current operating condition based on the operating condition characteristics and steady-state line loss prediction model corresponding to the current operating condition.

[0179] The adaptive operating condition spatial line loss model is pre-constructed based on multiple operating condition nodes. Each operating condition node corresponds to a steady-state operating condition identified from wind farm operation data, and includes the operating condition characteristics of the corresponding steady-state operating condition and a representative steady-state line loss determined based on the steady-state operation sample corresponding to the steady-state operating condition.

[0180] The model completion module is used to, when matching fails, use the operating condition features corresponding to the current operating condition as the operating condition features to complete the operating condition node, use the steady-state line loss corresponding to the current operating condition as the representative steady-state line loss corresponding to the completed operating condition node, establish the completed operating condition node, and add the completed operating condition node to the adaptive operating condition space line loss model.

[0181] The specific functions of each module described above are explained in the relevant content of the method in Embodiment 1, and will not be repeated here.

[0182] Example 3

[0183] This embodiment provides a computer-readable storage medium storing a computer program that, when executed, implements the steady-state line loss determination method for primary frequency regulation of wind farms as described in Embodiment 1.

[0184] 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.

[0185] 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.

[0186] 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.

[0187] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for determining steady-state line loss for primary frequency regulation in wind farms, characterized in that, include: Obtain the current operating data of the wind farm, and determine the current operating condition and its corresponding characteristics based on the current operating data; The operating conditions are matched based on the operating condition characteristics corresponding to the current operating condition and the operating condition characteristics corresponding to each operating condition node in the adaptive operating condition space line loss model. If the match is successful, the steady-state line loss corresponding to the current operating condition is determined based on the representative steady-state line loss corresponding to the matched operating condition node. If the match fails, the neighboring operating condition nodes that are similar to the current operating condition are determined from each operating condition node. When the number of neighboring operating condition nodes reaches the preset number of nodes, the steady-state line loss corresponding to the current operating condition is determined based on the representative steady-state line loss corresponding to the neighboring operating condition nodes. When the number of neighboring operating condition nodes does not reach the preset number of nodes, the steady-state line loss corresponding to the current operating condition is determined based on the operating condition characteristics and steady-state line loss prediction model corresponding to the current operating condition. The adaptive operating condition spatial line loss model is pre-constructed based on multiple operating condition nodes. Each operating condition node corresponds to a steady-state operating condition identified from wind farm operation data, and includes the operating condition characteristics of the corresponding steady-state operating condition and a representative steady-state line loss determined based on the steady-state operation sample corresponding to the steady-state operating condition. When matching fails, the operating condition features corresponding to the current operating condition are used as the operating condition features to complete the operating condition node, and the steady-state line loss corresponding to the current operating condition is used as the representative steady-state line loss corresponding to the complete operating condition node. The complete operating condition node is then established and added to the adaptive operating condition space line loss model.

2. The method for determining steady-state line loss for primary frequency regulation in wind farms according to claim 1, characterized in that, The operating characteristics include the active power at the grid connection point, the total active power of the wind turbine, the reactive power at the grid connection point, and the normalized characteristic values ​​corresponding to the grid connection point voltage and wind speed.

3. The method for determining steady-state line loss for primary frequency regulation in wind farms according to claim 2, characterized in that, The steady-state operating conditions and operating characteristics corresponding to each operating condition node used to pre-build the adaptive operating condition spatial line loss model are determined in the following way: Multiple operation samples are obtained based on wind farm operation data within a preset time window, and the operating condition characteristics corresponding to each operation sample are determined. Density clustering is performed based on the operating condition characteristics corresponding to each of the operating samples to identify steady-state sample clusters, eliminate disturbed samples and isolated samples, and determine the operating samples in the steady-state sample clusters as steady-state operating samples to obtain the set of steady-state operating samples corresponding to the steady-state operating conditions. The median of each feature is determined based on the operating condition characteristics of each steady-state operating sample in the steady-state operating sample set, and the operating condition characteristics corresponding to the steady-state operating condition are determined based on the median of each feature.

4. The method for determining steady-state line loss for primary frequency regulation in wind farms according to claim 3, characterized in that, Representative steady-state line losses are determined based on steady-state operating samples corresponding to steady-state conditions, including: For each steady-state operating sample in the set of steady-state operating samples, obtain the total active power of the wind turbine and the active power at the grid connection point corresponding to each steady-state operating sample, and determine the line loss sample corresponding to the steady-state operating sample based on the difference between the total active power of the wind turbine and the active power at the grid connection point. By collecting the line loss samples corresponding to each of the steady-state operating samples, a set of line loss samples corresponding to the steady-state operating condition is obtained; Abnormal line loss samples are removed from the line loss sample set, and the representative steady-state line loss corresponding to the steady-state operating condition is determined based on the remaining line loss samples after removing abnormal line loss samples.

5. The method for determining steady-state line loss for primary frequency regulation in wind farms according to claim 4, characterized in that, Abnormal line loss samples are removed from the line loss sample set, and representative steady-state line loss corresponding to the steady-state operating condition is determined based on the remaining line loss samples after removing abnormal line loss samples, including: Determine the median of the line loss sample set, and determine the absolute deviation of each line loss sample from the median; The median of each absolute deviation is determined as the median absolute deviation, and abnormal line loss samples are identified based on the median absolute deviation. The abnormal line loss samples are removed, and the median of the remaining line loss samples is determined as the representative steady-state line loss corresponding to the steady-state operating condition.

6. The method for determining steady-state line loss for primary frequency regulation in wind farms according to claim 2, characterized in that, The step of matching operating conditions based on the operating condition characteristics corresponding to the current operating condition and the operating condition characteristics corresponding to each operating condition node in the adaptive operating condition space line loss model includes: Calculate the Euclidean distance between the working condition feature corresponding to the current working condition and the working condition feature corresponding to each working condition node, and determine the minimum Euclidean distance among the Euclidean distances. When the minimum Euclidean distance is not greater than the first distance threshold, the working condition is determined to be successfully matched, and the working condition node corresponding to the minimum Euclidean distance is determined as the matched working condition node. When the minimum Euclidean distance is greater than the first distance threshold, the working condition matching is determined to be unsuccessful.

7. The method for determining steady-state line loss for primary frequency regulation in wind farms according to claim 6, characterized in that, When a match fails, determining the steady-state line loss corresponding to the current operating condition includes: The working condition node whose Euclidean distance from the current working condition is not greater than the second distance threshold is determined as the neighboring working condition node, where the second distance threshold is greater than the first distance threshold. Determine whether the number of adjacent working condition nodes has reached the preset number of nodes; When the number of neighboring operating condition nodes reaches the preset number of nodes, select the preset number of neighboring operating condition nodes with the smallest Euclidean distance to the current operating condition from the neighboring operating condition nodes. Determine the corresponding weights based on the Euclidean distances of each selected neighboring operating condition node. Determine the steady-state line loss corresponding to the current operating condition based on the representative steady-state line loss and corresponding weights of each neighboring operating condition node. When the number of adjacent working condition nodes does not reach the preset number of nodes, the working condition characteristics corresponding to the current working condition are input into the steady-state line loss prediction model for prediction to obtain the steady-state line loss corresponding to the current working condition.

8. The method for determining steady-state line loss for primary frequency regulation in wind farms according to claim 1, characterized in that, The training process of the steady-state line loss prediction model includes: Multiple steady-state operating conditions identified from wind farm operation data are obtained. The operating condition characteristics corresponding to each steady-state operating condition are used as model inputs, and the representative steady-state line loss corresponding to each steady-state operating condition is used as model outputs. Based on the model inputs and model outputs, the extreme gradient boosting algorithm is used for training to obtain the steady-state line loss prediction model. In the subsequent operation of the wind farm, new operating data is acquired, and steady-state operating conditions are identified based on the new operating data. The operating condition characteristics corresponding to the steady-state operating conditions are used as incremental training inputs, and the corresponding representative steady-state line loss is used as the incremental training output to incrementally train the steady-state line loss prediction model. The operating condition characteristics and representative steady-state line loss corresponding to the completed operating condition nodes are not used as incremental training data for the steady-state line loss prediction model.

9. The method for determining steady-state line loss for primary frequency regulation in wind farms according to claim 1, characterized in that, Each of the aforementioned working condition nodes also includes a timestamp for recording the most recent creation or update time of the corresponding working condition node, and the method further includes: During the subsequent operation of the wind farm, new operating data is acquired, and steady-state operating conditions are identified based on the new operating data. New operating condition nodes are established according to the operating condition characteristics and representative steady-state line loss corresponding to the steady-state operating conditions. The working condition characteristics corresponding to the newly added working condition node are matched with the working condition characteristics corresponding to each working condition node in the adaptive working condition space line loss model to determine whether there is a working condition node in the adaptive working condition space line loss model that corresponds to the newly added working condition node. If it does not exist, the newly added working condition node is added to the adaptive working condition spatial line loss model; if it exists, the working condition node that is closest to the newly added working condition node is determined from the matching working condition nodes, the representative steady-state line loss corresponding to the closest working condition node is updated according to the representative steady-state line loss corresponding to the newly added working condition node, and its timestamp is updated. Based on the timestamps corresponding to each working condition node, delete working condition nodes that have not been updated for a preset retention period.

10. A steady-state line loss determination system for primary frequency regulation in wind farms, characterized in that, include: The data processing module is used to acquire the current operating data of the wind farm and determine the current operating condition and the corresponding operating condition characteristics based on the current operating data. The working condition matching module is used to match the working condition characteristics corresponding to the current working condition with the working condition characteristics corresponding to each working condition node in the adaptive working condition space line loss model. The line loss determination module is used to determine the steady-state line loss corresponding to the current operating condition based on the representative steady-state line loss corresponding to the matched operating condition node if the matching is successful; if the matching fails, it determines the neighboring operating condition nodes that are similar to the current operating condition from each operating condition node. When the number of neighboring operating condition nodes reaches the preset number of nodes, it determines the steady-state line loss corresponding to the current operating condition based on the representative steady-state line loss corresponding to the neighboring operating condition nodes. When the number of neighboring operating condition nodes does not reach the preset number of nodes, it determines the steady-state line loss corresponding to the current operating condition based on the operating condition characteristics and steady-state line loss prediction model corresponding to the current operating condition. The adaptive operating condition spatial line loss model is pre-constructed based on multiple operating condition nodes. Each operating condition node corresponds to a steady-state operating condition identified from wind farm operation data, and includes the operating condition characteristics of the corresponding steady-state operating condition and a representative steady-state line loss determined based on the steady-state operation sample corresponding to the steady-state operating condition. The model completion module is used to, when matching fails, use the operating condition features corresponding to the current operating condition as the operating condition features to complete the operating condition node, use the steady-state line loss corresponding to the current operating condition as the representative steady-state line loss corresponding to the completed operating condition node, establish the completed operating condition node, and add the completed operating condition node to the adaptive operating condition space line loss model.

Images