Method, device, equipment and medium for performance evaluation of wind turbine generator

Abstract

The present application relates to the operation and maintenance technical field of wind turbine, discloses a performance evaluation method, device, equipment and medium of wind turbine, the present application is based on meteorological data and carries out the box, so that different operation data belongs to different clusters, the corresponding boundary ideal solution is determined under different clusters, the value range of evaluation index can be accurately indicated, and then, in the process of determining the quality score of each key component, the boundary ideal solution of the corresponding evaluation index in the cluster is taken as the reference, so that the quality score of each key component determined is more accurate, so that the quality score of the target wind turbine finally determined is more accurate.

Description

Technical Field

[0001] This invention relates to the field of wind turbine operation and maintenance technology, specifically to wind turbine performance evaluation methods, devices, equipment, and media. Background Technology

[0002] In the field of wind turbine operation and maintenance, scholars both domestically and internationally have conducted extensive research on the condition assessment of wind turbines using multi-source operating data and multi-attribute decision theory. Current methods tend to select multiple evaluation indicators for wind turbines (e.g., oil temperature, winding temperature, vibration amplitude, etc.). For each indicator, extreme value normalization is first performed to obtain a state vector. Then, Euclidean distance is used to calculate the closeness of the state vector to the boundary vector, resulting in a health score.

[0003] However, evaluation indicators are usually derived directly from the normalization of global data extreme values, which leads to the loss of absolute comparability of the evaluation results of the same wind turbine under different meteorological conditions, resulting in inaccurate evaluation results. Summary of the Invention

[0004] This invention provides a method, apparatus, electronic device, storage medium, and program product for evaluating the performance of wind turbines, in order to solve the problem of inaccurate performance evaluation of wind turbines in related technologies.

[0005] In a first aspect, the present invention provides a method for evaluating the performance of a wind turbine generator set, the method comprising: The system acquires multiple operational data points and weight configuration information of the target wind turbine, as well as meteorological data corresponding to each operational data point. The target wind turbine includes multiple key components, each operational data point includes the collected values ​​corresponding to multiple evaluation indicators, the multiple evaluation indicators include the evaluation indicators associated with multiple key components, and the weight configuration information includes the indicator weights corresponding to multiple evaluation indicators and the component weights of multiple key components. Based on the meteorological data corresponding to each of the multiple operational data points, the multiple operational data points are divided into multiple clusters; In the target cluster, the boundary ideal solutions corresponding to various evaluation indicators are determined respectively; Based on the collected values ​​of evaluation indicators associated with key components of the target in the target operation data, the boundary ideal solutions of the evaluation indicators associated with key components of the target, and the indicator weights, the quality score of key components of the target is determined under the target operation data. Here, the target operation data is any operation data in the target cluster, and the target key component is any one of multiple key components. The quality score of the target wind turbine under the target operating data is determined based on the component weights and quality scores of multiple key components.

[0006] The wind turbine performance evaluation method provided in this embodiment of the invention has the following technical effects: First, multiple operational data points, weight configuration information, and meteorological data of the target wind turbine are acquired. Since the operational status of key components in the wind turbine varies under different meteorological data, this solution bins the operational data based on the meteorological data, resulting in multiple clusters. Then, within each cluster, the corresponding boundary ideal solution can be determined. The boundary ideal solution indicates the boundary values ​​of the evaluation indicators within the cluster, and the indicator weights indicate the degree of correlation between the evaluation indicators and the performance of key components. Taking the target operational data as an example, the quality score of the target key component under the target operational data can be determined based on the collected values ​​of the evaluation indicators associated with the target key component, the indicator weights, and the boundary ideal solution. Finally, based on the component weights and quality scores of multiple key components, the quality score of the target wind turbine under the target operational data is determined.

[0007] Under the aforementioned evaluation method, this scheme divides meteorological data into bins, so that different operational data belong to different clusters. Under each cluster, the corresponding boundary ideal solution is determined, which can accurately indicate the value range of the evaluation index under different meteorological conditions. Furthermore, in the process of determining the quality score of each key component, the boundary ideal solution of the corresponding evaluation index in its respective cluster is used as a reference, so that the quality score of each key component is more accurate, thereby making the final quality score of the target wind turbine more accurate.

[0008] In some optional implementations, the quality score of the target key component under the target operating data is determined based on the collected values ​​of evaluation indicators associated with the target key component in the target operating data, the boundary ideal solutions of the evaluation indicators associated with the target key component, and the indicator weights, including: Based on the collected values ​​of evaluation indicators associated with the key components of the target, construct the target state vector; Based on the boundary ideal solution of the evaluation index associated with the target key component, construct the boundary vector of the target key component; Based on the collected values ​​of evaluation indicators associated with the key components of the target, construct the target covariance matrix; Construct a weight matrix based on the weights of the evaluation indicators associated with the target key components; Based on the target state vector, boundary vector, target covariance matrix, and weight matrix, determine the target Mahalanobis distance corresponding to the target running data; The coefficient of variation is determined based on the Mahalanobis distances corresponding to the operational data included in the target cluster. Based on the target Mahalanobis distance and coefficient of variation, determine the quality score of the target's key components under the target's operating data.

[0009] In this way, for each key component, only evaluation indicators directly related to its physical degradation mechanism are selected, and state vectors, boundary vectors, and covariance matrices are independently constructed. In this process, by introducing regularized weighted Mahalanobis distance, multicollinearity between indicators (e.g., the strong correlation between oil temperature and bearing temperature) can be adaptively eliminated in the component space, which can avoid the same overheating phenomenon being deducted repeatedly, so that the component quality score truly reflects its own health status.

[0010] Furthermore, this scheme calculates the coefficient of variation based on clustered operational data and dynamically adjusts the positive and negative ideal solution preference coefficients, enabling the scoring process to automatically adapt to the data dispersion characteristics under different operating conditions. When a component enters the potential degradation stage and the temporal dispersion of related indicators increases, the objective weight (i.e., entropy weight) can be nonlinearly amplified, significantly improving the sensitivity of Mahalanobis distance to early high-frequency degradation characteristics and overcoming the average masking effect of traditional static weighting on weak faults.

[0011] In some optional implementations, the quality score of the target wind turbine under target operating data is determined based on the component weights and quality scores of multiple key components, including: The overall value is determined based on the component weights and quality scores of multiple key components; Based on the component weights and quality scores of the target key components, as well as the comprehensive value, the target obstacle degree of the target key components to the target wind turbine is determined; The quality score of the target wind turbine under the target operating data is determined based on the obstacle degree corresponding to at least one key component, and the component weight and quality score corresponding to multiple key components.

[0012] In this way, this solution intervenes in the scoring of wind turbine units by using obstacle degree, making the final quality score more accurate.

[0013] In some optional implementations, the quality score of the target wind turbine under the target operating data is determined based on the obstacle degree corresponding to at least one key component, and the component weights and quality scores corresponding to multiple key components, including: Based on at least one obstacle degree and a preset obstacle threshold, determine the penalty factor corresponding to at least one obstacle degree respectively; The initial quality score is determined based on the component weights and quality scores corresponding to multiple key components. The quality score of the target wind turbine under the target operating data is determined based on at least one penalty factor and an initial quality score.

[0014] In this way, by triggering the penalty factor through key components with high obstacle levels, it is ensured that only degraded key components can activate single-point veto, which can break the linear weighted average masking.

[0015] In some optional implementations, before acquiring multiple operational data points of the target wind turbine, weight configuration information, and meteorological data corresponding to each of the multiple operational data points, the method further includes: Obtain the objective weights and subjective weights corresponding to the target evaluation indicators, where the target evaluation indicator is any one of a variety of evaluation indicators; Calculate the total weight based on the objective and subjective weights corresponding to various evaluation indicators; The weights of the target evaluation indicators are determined based on the total weight and the objective and subjective weights corresponding to the target evaluation indicators.

[0016] In this way, using two weights to determine the final weight for any evaluation indicator is more accurate.

[0017] In some optional implementations, multiple evaluation metrics also include the power of the target wind turbine; the method further includes: Based on the power in multiple operational data points, and the meteorological data corresponding to each operational data point, the multiple operational data points are divided into multiple clusters.

[0018] In this way, the scheme combines meteorological data and power data to jointly bin the operational data, constructing multi-condition clusters, fundamentally eliminating the interference of meteorological conditions on evaluation indicators. By independently setting the absolute boundary ideal solution within each cluster, the incomparability problem across operating conditions caused by global extreme value normalization is solved, ensuring the absolute comparability of unit condition evaluations under different wind conditions, and making the quality score truly reflect the degradation of key components themselves rather than environmental differences.

[0019] In some optional implementations, after determining the quality score of the target wind turbine under the target operating data based on the component weights and quality scores of multiple key components, the method further includes: Based on the quality scores of the multiple wind turbines included in the wind turbine group in the target evaluation window, the abnormal wind turbines are identified. Among them, multiple operating data are the operating data collected in the target evaluation window, and the target wind turbine is any one of the multiple wind turbines. Based on the quality scores or obstacle degrees of multiple key components included in the abnormal wind turbine, the abnormal key components are identified in the target evaluation window; Based on the abnormal wind turbine units and abnormal critical components, generate operation and maintenance work orders.

[0020] In this way, this solution can identify abnormal wind turbines by comparing the quality scores of the wind turbine group. Then, based on the quality scores or obstruction levels of each key component inside the wind turbine, it can accurately locate the abnormal key component, realizing step-by-step fault tracing and diagnosis from the group to the individual unit to the component, and automatically generating operation and maintenance work orders. This enables the wind farm operation and maintenance mode to shift from passive maintenance to proactive and precise maintenance, significantly reducing operation and maintenance costs and the risk of unplanned downtime.

[0021] In a second aspect, the present invention provides a performance evaluation device for wind turbine generators, the device comprising: The acquisition module is used to acquire multiple operational data, weight configuration information, and meteorological data corresponding to the multiple operational data of the target wind turbine. The target wind turbine includes multiple key components. Each operational data includes the collected values ​​corresponding to multiple evaluation indicators. The multiple evaluation indicators include evaluation indicators associated with multiple key components. The weight configuration information includes the indicator weights corresponding to multiple evaluation indicators and the component weights of multiple key components. The partitioning module is used to divide multiple operational data into multiple clusters based on the meteorological data corresponding to each of the multiple operational data. The determination module is used to identify the boundary ideal solutions corresponding to various evaluation indicators in the target cluster; based on the collected values ​​of evaluation indicators associated with the target key components in the target operating data, the boundary ideal solutions of the evaluation indicators associated with the target key components, and the indicator weights, the quality score of the target key components under the target operating data is determined, where the target operating data is any operating data in the target cluster, and the target key component is any one of multiple key components; based on the component weights and quality scores of multiple key components, the quality score of the target wind turbine under the target operating data is determined.

[0022] Thirdly, the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the wind turbine performance evaluation method of the first aspect or any corresponding embodiment described above.

[0023] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the wind turbine performance evaluation method of the first aspect or any corresponding embodiment described above.

[0024] Fifthly, the present invention provides a computer program product, including computer instructions for causing a computer to execute the wind turbine performance evaluation method of the first aspect or any corresponding embodiment described above. Attached Figure Description

[0025] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0026] Figure 1 This is a schematic diagram of the first process of a wind turbine performance evaluation method according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the distribution relationship between the evaluation index and the boundary ideal solution according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the distribution of state points under Mahalanobis distance according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the obstacle degree according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the second process of the wind turbine performance evaluation method according to an embodiment of the present invention; Figure 6 This is a comparative diagram of SQI for different performance evaluation methods according to embodiments of the present invention; Figure 7 This is a schematic diagram of a performance evaluation method for a group of wind turbines according to an embodiment of the present invention; Figure 8 This is a structural block diagram of a wind turbine performance evaluation device according to an embodiment of the present invention; Figure 9 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0027] 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.

[0028] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0029] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0030] Currently, scholars both domestically and internationally have conducted extensive research on the state assessment of target wind turbine units using multi-source operational data and multi-attribute decision theory. However, significant shortcomings remain in engineering applications for large-scale wind turbine clusters. Some current methods tend to select multiple evaluation indicators for wind turbine units (e.g., oil temperature, winding temperature, vibration amplitude, etc.). For each evaluation indicator, extreme value normalization is first performed to obtain a state vector. Then, Euclidean distance is used to calculate the closeness of the state vector to the boundary value, i.e., to obtain the health score.

[0031] However, the above method has the following drawbacks: First, it failed to consider the serious multicollinearity problem between fluid thermodynamic variables and electromechanical variables. For example, the rise in oil temperature and the rise in bearing temperature are essentially two manifestations of the same overheating problem. If these are treated as two independent defects and points are deducted separately, the penalty for the same degradation phenomenon is repeatedly amplified, and the evaluation results are distorted.

[0032] Second, traditional static weighting or linear weighting models are prone to average masking effect, that is, a large number of normal evaluation indicators will compensate for and mask the early subtle degradation characteristics of local core key components, and cannot characterize the series barrel effect of wind turbine units.

[0033] Third, the lack of guidance from physical mechanisms means that evaluation indicators are usually derived directly from the normalization of global data extreme values, resulting in the loss of absolute comparability of evaluation results for the same target wind turbine under different meteorological conditions.

[0034] In summary, the current approach leads to inaccurate performance evaluation results for the target wind turbine units and fails to integrate the underlying physical degradation characteristics with the overall collaborative evaluation framework of macro-level cluster economic benefits. Therefore, it cannot be directly used for wind farm optimization scheduling with the goal of minimizing levelized cost of electricity.

[0035] According to an embodiment of the present invention, a performance evaluation method for wind turbine generators is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0036] This embodiment provides a method for evaluating the performance of wind turbine generators, which can be executed by electronic devices (e.g., servers, computers, etc.). Figure 1 This is a flowchart of a wind turbine performance evaluation method according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps: Step S101: Obtain multiple operating data points and weight configuration information of the target wind turbine, as well as meteorological data corresponding to each of the multiple operating data points.

[0037] The target wind turbine can be any one of the target wind turbine groups. The target wind turbine includes several key components, such as the generator and gearbox, which have a high impact on the operation of the wind turbine and can be specified in advance by technical personnel.

[0038] A single data point represents the data collected at one sampling point. Any evaluation time window can include multiple sampling points. Multiple data points can be combined to form a state parameter matrix for the current evaluation time window. Each row in the state parameter matrix indicates the data collected at one sampling point, and each column indicates an evaluation metric. For example, 31 columns represent 31 evaluation dimensions.

[0039] Each operational data point includes collected values ​​corresponding to various evaluation indicators (which can indicate the performance of the wind turbine). These evaluation indicators include those associated with each key component. The evaluation indicators associated with key components can be any of the following: temperature rise characteristics characterizing thermal resistance and frictional degradation, or time-domain vibration characteristics characterizing mechanical wear. For example, temperature rise characteristics could be the temperature rise rate of generator bearings or gearboxes, and time-domain vibration characteristics could be a dimensionless kurtosis index of the high-frequency vibration acceleration of the transmission chain.

[0040] The weighting configuration information includes the weights of various evaluation indicators and the weights of multiple key components included in the target wind turbine.

[0041] Meteorological data can include wind speed, etc.

[0042] Specifically, the Supervisory Control and Data Acquisition (SCADA) system can first acquire the data collected from each wind turbine unit. Then, the electronic equipment can perform preprocessing and feature extraction operations on the acquired data to obtain the operating data.

[0043] For each data collection metric, preprocessing operations may include: First, outliers caused by sensor failure or communication errors are eliminated. For example, for a certain data acquisition indicator, its mean and standard deviation over the entire evaluation time window can be calculated. Then, it can be determined whether the difference between the acquired value and the mean of this indicator is greater than a preset multiple (such as 3 times) of the standard deviation, expressed as... ,in, For collected values, The mean, The standard deviation is the value. If it is, the collected value is considered an outlier and can be removed; otherwise, it can be retained.

[0044] Then, the baseline trajectory of the state parameters can be extracted using the moving average filtering method to filter out transient gusts and high-frequency electrical switching noise. For example, for any acquisition index, the following expression can be used: (1) in, The final collected value is T, where T is the length of the smoothing time window. The original value is collected, t-1 represents the period index, and this formula can be used to indicate that the final collected value of the current sampling point t is the average of the original collected values ​​of the past T sampling points, where i is the index of the sampling point.

[0045] After completing the preprocessing operation, feature extraction can be performed on the preprocessed collected data. For example, indicators required for subsequent evaluation operations can be selected directly as evaluation indicators, and / or, for a certain collection indicator, the collection indicators required to calculate that evaluation indicator can be extracted from it, and the collected value of the evaluation indicator can be obtained after calculation. For specific examples, please refer to the evaluation indicators of the key component layer in Table 2 below.

[0046] Step S102: Divide the multiple operational data into multiple clusters based on the meteorological data corresponding to each of the multiple operational data.

[0047] Specifically, to eliminate the interference of different meteorological conditions on the average results, a binning strategy can be adopted, which divides multiple operational data points into several discrete interval clusters based on the meteorological data corresponding to each operational data point. For example, binning can be based on wind speed.

[0048] Step S103: Determine the boundary ideal solutions corresponding to various evaluation indicators in the target cluster.

[0049] Specifically, for any given cluster, the boundary ideal solution for each evaluation metric can be extracted within that cluster. The boundary ideal solution can include a positive ideal solution (PIS) and a negative ideal solution (NIS). The positive ideal solution indicates the optimal value of the metric corresponding to a brand-new factory condition or the historical best operating envelope, while the negative ideal solution indicates the safety threshold limit when the equipment triggers a SCADA alarm or a serious fault shutdown, also known as the emergency limit.

[0050] For target evaluation indicators, taking target clusters as an example, statistical processing can be performed on each collected value corresponding to the target evaluation indicator in the target cluster to determine the upper and lower limits of the target evaluation indicator (which can be the collected values ​​corresponding to specified quantiles). Furthermore, for benefit-type evaluation indicators (i.e., the larger the better), the upper limit can be used as a positive ideal solution and the lower limit can be used as a negative ideal solution. For cost-type evaluation indicators (i.e., the smaller the better), the upper limit can be used as a negative ideal solution and the lower limit can be used as a positive ideal solution.

[0051] For example, the gearbox oil temperature in the rated wind speed zone will naturally be higher than in the low wind speed zone. If the lowest temperature is taken as the positive ideal solution globally, then a gearbox operating healthily at high wind speeds will be misjudged as deviating too far from the ideal solution, further leading to the judgment of a faulty target wind turbine, resulting in an incorrect judgment. This solution, by dividing the gearboxes, can use different boundary ideal solutions in different clusters, which is more accurate.

[0052] refer to Figure 2 This graph illustrates the distribution of active power of wind turbines between positive and negative ideal solutions, based on wind speed zones. Blue dots represent normal SCADA data, red dots represent degradation or fault points, green lines represent positive ideal solutions, and red dotted lines represent negative ideal solutions. The horizontal axis represents wind speed in m / s, and the vertical axis represents active power in kW. Based on wind speed, the graph can be divided into low-wind-speed zones (wind speed less than the cut-in wind speed), maximum power point tracking (MPPT) / acceleration zones (wind speed between the cut-in and rated wind speeds), and rated power zones (wind speed between the rated and cut-out wind speeds).

[0053] Step S104: Determine the quality score of the target key component under the target operating data based on the collected values ​​of the evaluation indicators associated with the target key component in the target operating data, the boundary ideal solutions of the evaluation indicators associated with the target key component, and the indicator weights.

[0054] Among them, the target running data is any running data in the target cluster, and the target key component is any one of the multiple key components.

[0055] Specifically, for any key component, the collected values ​​of evaluation indicators associated with that key component can be extracted from the target operational data. Since the boundary ideal solution provides the boundary values ​​of the evaluation indicators and the indicator weights can indicate the degree of influence of the evaluation indicators on the performance of the target key component, the quality score of the key component can be determined based on the collected values, boundary values, and weight values ​​of the various evaluation indicators managed by the key component. This score can also be called the Service Quality Index (SQI) or component proximity.

[0056] Step S105: Determine the quality score of the target wind turbine under the target operating data based on the component weights and quality scores of multiple key components.

[0057] Specifically, component weights indicate the degree of influence of a key component on the performance of the target wind turbine, while quality scores indicate the health status of the key components in their current state. Therefore, after determining the quality scores of each key component, a weighted calculation operation can be performed on the quality scores based on the component weights of the key components to obtain the quality score of the target wind turbine under the target operating data. For each piece of operating data, the quality score of the target wind turbine can be determined in the above manner. Finally, the quality scores corresponding to multiple pieces of operating data constitute a quality score column vector corresponding to the target wind turbine.

[0058] In addition, a trend chart of the quality score can be generated based on the quality score of the target wind turbine under multiple operating data points. The horizontal axis represents the data collection time, and the vertical axis represents the quality score. Technicians can use the trend chart to confirm the health status of the target wind turbine.

[0059] The wind turbine performance evaluation method provided in this embodiment first acquires multiple operational data points, weight configuration information, and meteorological data of the target wind turbine. Since the operational status of key components in the wind turbine varies under different meteorological data, this scheme bins the operational data based on the meteorological data to obtain multiple clusters. Then, within each cluster, the corresponding boundary ideal solution can be determined. The boundary ideal solution indicates the boundary value of the evaluation index within the cluster, and the index weight indicates the degree of correlation between the evaluation index and the performance of the key components. Taking the target operational data as an example, the quality score of the target key component under the target operational data can be determined based on the collected values ​​of the evaluation indicators associated with the target key component, the index weights, and the boundary ideal solution. Finally, based on the component weights and quality scores of multiple key components, the quality score of the target wind turbine under the target operational data is determined.

[0060] Under the aforementioned evaluation method, this scheme divides meteorological data into bins, so that different operational data belong to different clusters. Under each cluster, the corresponding boundary ideal solution is determined, which can accurately indicate the value range of the evaluation index under different meteorological conditions. Furthermore, in the process of determining the quality score of each key component, the boundary ideal solution of the corresponding evaluation index in its respective cluster is used as a reference, so that the quality score of each key component is more accurate, thereby making the final quality score of the target wind turbine more accurate.

[0061] In some alternative implementations, each piece of operational data may include the power of the target wind turbine. Accordingly, the multiple pieces of operational data may be divided into multiple clusters based on the power in the multiple pieces of operational data and the meteorological data corresponding to each of the multiple pieces of operational data.

[0062] Thus, since the boundary ideal solutions of the evaluation index will differ under different power levels, in order to eliminate the influence of power, the operational data can be jointly clustered using power and meteorological data, so that a more accurate boundary ideal solution can be determined in subsequent processes, thereby making the final evaluation results more accurate.

[0063] In some optional implementations, prior to step S101 above, the weights of each evaluation index can be determined through the following specific steps: Step 1: Obtain the objective weights and subjective weights corresponding to the target evaluation indicators.

[0064] The target evaluation indicator can be any one of a variety of evaluation indicators.

[0065] Step 2: Calculate the total weight based on the objective and subjective weights corresponding to the various evaluation indicators.

[0066] Step 3: Determine the indicator weights of the target evaluation indicators based on the total weight and the objective and subjective weights corresponding to the target evaluation indicators.

[0067] Among them, subjective weights are pre-set weights, which can generally be obtained through the Analytic Hierarchy Process (AHP) and combined with the prior knowledge of domain experts.

[0068] Objective weights can be calculated based on the real-time dispersion of data through temporal entropy features. For example, for each evaluation indicator, its temporal probability distribution can be calculated based on the corresponding collected values ​​from multiple operational data sets (the sum of multiple collected values ​​can be calculated; for any collected value, the ratio between that value and the sum is determined as the temporal probability, and the temporal probabilities corresponding to each collected value constitute the temporal probability distribution). Then, based on the temporal probability distribution, the entropy value corresponding to the evaluation indicator can be calculated. Next, the entropy value is transformed to the range of 0 to 1 using the number of operational data sets (e.g., 1 - entropy value / lnN, where N is the number of operational data sets). Finally, the sum of the transformed entropy values ​​corresponding to each evaluation indicator is calculated. For any evaluation indicator, the ratio between the entropy value of that indicator and the sum is determined as the objective weight.

[0069] Specifically, for any evaluation indicator, the product of its objective weight and subjective weight can be calculated first. Then, the products of the objective and subjective weights of all evaluation indicators can be summed to obtain the total weight. Finally, for any evaluation indicator, the ratio of the product of its objective and subjective weights to the total weight can be determined as the corresponding indicator weight. For example, the indicator weight can be expressed as follows: (2) Where j is the index of the evaluation indicator, and n is the number of evaluation indicators. This represents the subjective weight of the j-th evaluation index. This represents the objective weight of the j-th evaluation index.

[0070] In this way, subjective weights are derived from the Analytic Hierarchy Process (AHP) combined with prior knowledge from domain experts, ensuring the fundamental importance of key components such as gearboxes and generators in the evaluation system. Objective weights are calculated in real time based on the time-series dispersion of each indicator within a rolling time window, automatically assigning higher weights to indicators with a surge in dispersion in the early stages of degradation. Furthermore, the two are normalized through multiplication, preventing both the rigidity of subjective weight perspectives and the dominance of noise in objective weights.

[0071] When a critical component enters a potential degradation stage and the dispersion of its associated evaluation indicators increases within the rolling time window, the objective weight based on time information entropy increases nonlinearly due to the decrease in entropy value. After being multiplied and synthesized with the subjective weight, the proportion of the evaluation indicator in the combined weight increases rapidly, thereby amplifying its state deviation contribution in the TOPSIS distance calculation. This allows high-frequency degradation characteristics to be captured in the early stages of degradation, avoiding being masked by the linear averaging of a large number of normal indicators.

[0072] Table 1 provides an example of evaluation indicators. The first column lists the names of the evaluation indicators, the second column lists the corresponding key components, the third column lists the subjective weights, the fourth column lists the objective weights, the fifth column lists the indicator weights, and the sixth column lists the ranking of the evaluation indicators from largest to smallest weight.

[0073] Table 1

[0074] In some alternative implementations, step S104 described above may include: Step 1: Construct the target state vector based on the collected values ​​of evaluation indicators associated with the key components of the target.

[0075] Step 2: Construct the boundary vector of the target key component based on the boundary ideal solution of the evaluation index associated with the target key component.

[0076] Step 3: Construct the target covariance matrix based on the collected values ​​of the evaluation indicators associated with the target key components.

[0077] Step 4: Construct a weight matrix based on the weights of the evaluation indicators associated with the target key components.

[0078] Step 5: Determine the target Mahalanobis distance corresponding to the target running data based on the target state vector, boundary vector, target covariance matrix, and weight matrix.

[0079] Step 6: Determine the coefficient of variation based on the Mahalanobis distances corresponding to the running data included in the target cluster.

[0080] Step 7: Determine the quality score of the target key components under the target operating data based on the target Mahalanobis distance and coefficient of variation.

[0081] Specifically, in step one, the collected values ​​of evaluation indicators associated with key components of the target in the target's operational data can be concatenated to obtain the target state vector.

[0082] In step two, the boundary ideal solutions corresponding to the evaluation indicators associated with the target key component can be concatenated to obtain the target boundary vector. For example, the positive ideal solutions of the evaluation indicators associated with the target key component can be combined into a positive boundary vector, and the negative ideal solutions of the evaluation indicators associated with the target key component can be combined into a negative boundary vector. The positive and negative boundary vectors constitute the target boundary vector. For example, the positive boundary vector can be represented as... The negative boundary vector can be represented as k is the cluster index, and m is the number of evaluation indicators corresponding to the target key components.

[0083] In step three, within the target cluster, for any evaluation metric, the variance can be calculated using the collected values ​​corresponding to that evaluation metric from each sub-data set within the target cluster. For any two evaluation metrics, the covariance between these two evaluation metrics can be calculated using the collected values ​​corresponding to that evaluation metric from each sub-data set within the target cluster.

[0084] Furthermore, for various evaluation indicators of key components, an initial covariance matrix of the key components in the target cluster can be constructed based on the variance and covariance corresponding to these evaluation indicators, and this initial covariance matrix can be directly determined as the target covariance matrix.

[0085] Alternatively, to eliminate multicollinearity interference among highly correlated indicators, a regularization (ridge estimation) strategy can be introduced to reconstruct the target covariance matrix. For example, the target covariance matrix can be expressed as follows: (3) in, Let be the target covariance matrix. The initial covariance matrix, This is the shrinkage parameter (typically a value of 0.01). It is an identity matrix.

[0086] In step four, for the various evaluation indicators associated with the target key components, a weight matrix (which can be a diagonal matrix) can be constructed based on the indicator weights corresponding to these evaluation indicators.

[0087] In step five, the target Mahalanobis distance corresponding to the target operating data is determined based on the target state vector, boundary vector, target covariance matrix, and weight matrix. That is, the Mahalanobis distance of the target key components under the target operating data.

[0088] For example, the target Mahalanobis distance can be calculated using the following expression: (4) (5) Where i is the index of the running data (or sampling point), The first Mahalanobis distance, Let be the target state vector. It is a positive boundary vector. This is the weight matrix. Let be the target covariance matrix. This is the second Mahalanobis distance. It is a negative boundary vector.

[0089] For any running data in the target cluster, the Mahalanobis distance of the target's key components under that running data can be calculated in a similar way.

[0090] refer to Figure 3 Blue dots represent normal operating points, while red dots represent abnormal points. If we use traditional Euclidean distance contour lines to determine whether a point is abnormal, hidden abnormal points within the contour lines would be considered normal. Furthermore, this method assumes that all indicator axes are orthogonal and equally weighted, ignoring correlations between indicators. Under this metric, some red abnormal points (e.g., A and C) might fall on the same contour circle as blue normal points and be misjudged as normal.

[0091] In this scheme, the inverse of the covariance matrix is ​​used. Using the weight kernel, adaptive rotation and scaling of the spatial principal axis allows contour lines to align with the true correlation structure of the indicators. Red outliers, deviating from the co-ellipse, are pushed out of the normal contour lines and exposed as anomalies. Furthermore, in the high-dimensional state space of the SCADA system, if the Euclidean distance approximation method (Technique for Order Preference by Similarity to Ideal Solution, TOPSIS) is directly used, the information of strongly correlated indicators will be repeatedly calculated, causing the early subtle degradation of local core critical components to be masked. However, by introducing regularized Mahalanobis distance and adaptively decoupling the covariance structure, the true physical degradation distance of the wind turbine state from the absolute virtual boundary in the multi-dimensional phase space can be accurately reflected.

[0092] In step six, the mean and standard deviation can be calculated based on the Mahalanobis distances of each running data point in the target cluster. The ratio between the standard deviation and the mean is then determined as the coefficient of variation. Specifically, for the first Mahalanobis distance, the corresponding first coefficient of variation is calculated, and for the second Mahalanobis distance, the corresponding second coefficient of variation is calculated.

[0093] In step seven, the dynamic preference coefficients for the positive and negative ideal solutions can be determined based on the first and second coefficients of variation, respectively. Then, the quality score of the target key component under the target operating data can be determined based on the dynamic preference coefficients for the positive and negative ideal solutions, as well as the first and second Mahalanobis distances. For example, the following expression can be used: (6) (7) (8) in, Assess the quality score of the target key components under the target operating data. The first Mahalanobis distance, This is the second Mahalanobis distance. The first coefficient of variation, The second coefficient of variation, For the positive ideal solution, dynamic preference coefficient, The negative ideal solution dynamic preference coefficient is denoted by k, which represents the index of the key component.

[0094] In some alternative implementations, step S105 described above may include: Step 1: Determine the overall value based on the component weights and quality scores of multiple key components.

[0095] Step 2: Determine the target obstacle degree of the target key component to the target wind turbine based on the component weight and quality score of the target key component, as well as the comprehensive value.

[0096] Step 3: Determine the quality score of the target wind turbine under the target operating data based on the obstacle degree corresponding to at least one key component, and the component weight and quality score corresponding to multiple key components.

[0097] Specifically, a comprehensive value can be determined based on the component weights and quality scores of multiple key components. A sub-value for the target key component can be determined based on its component weight and quality score. The ratio between the sub-value and the comprehensive value is then defined as the target obstacle degree of the target key component for the target wind turbine. Similarly, the corresponding obstacle degrees for other key components can be calculated using the same method. Finally, based on the obstacle degree of at least one key component and the component weights and quality scores of the multiple key components, the quality score of the target wind turbine under the target operating data is determined.

[0098] For example, the obstacle degree can be expressed as follows: (9) in, Here, k represents the obstacle degree, k represents the index of the critical component, and K represents the number of critical components included in the target wind turbine. This represents the component weight of the k-th critical component.

[0099] In some optional implementations, step three above, determining the quality score of the target wind turbine under the target operating data based on the obstacle degree corresponding to at least one key component and the component weights and quality scores corresponding to multiple key components, may include: Step 1: Determine the penalty factor corresponding to at least one obstacle degree based on at least one obstacle degree and a preset obstacle degree threshold.

[0100] Step 2: Determine the initial quality score based on the component weights and quality scores corresponding to the multiple key components.

[0101] Step 3: Determine the quality score of the target wind turbine under the target operating data based on at least one penalty factor and the initial quality score.

[0102] Specifically, for any given obstacle level, the obstacle level can be compared with a preset obstacle level threshold. Based on the comparison result, a penalty factor corresponding to the comparison result can be determined. For example, the penalty factor can be expressed as follows: (10) in, Indicates the penalty factor. Indicates the degree of obstacle. To preset the obstacle threshold, This is the preset scaling factor.

[0103] Each key component is assigned a corresponding component weight. For the collected evaluation index values ​​of each key component included in the target operating data, the quality score of each key component under the target operating data can be calculated using the aforementioned method. Then, based on the component weights of each key component, a weighted sum of the quality scores of the key components can be performed to obtain an initial quality score. Finally, the sum of the quality score and the penalty factor is determined as the quality score of the target wind turbine under the target operating data.

[0104] (11) in, Assess the quality score of the target wind turbine under the target operating data. The penalty factor can be the product of the penalty factors of multiple key components, or it can be the largest penalty factor. Let the weight of the k-th key component be . Let K be the quality score of the k-th critical component under the target operating data, where K is the number of critical components included in the target wind turbine.

[0105] Thus, the obstacle level is the sole criterion for determining whether the exponential decay penalty factor is triggered. The penalty is activated only when the obstacle level of a critical component exceeds a preset threshold, and the penalty intensity increases exponentially with the degree of obstacle exceeding the threshold. This ensures that only components that are both important and severely degraded can trigger a single-point veto, while minor fluctuations in normal or minor components will not be over-penalized, thereby achieving a balance between sensitivity and robustness.

[0106] Alternatively, a sub-quality score can be determined based on the product of the component weight, quality score, and penalty factor of each key component. Then, the sub-quality scores corresponding to each key component are summed to obtain the quality score of the target wind turbine.

[0107] refer to Figure 4 This chart illustrates the changing trends in obstacle severity. The horizontal axis represents the SCADA operating time series, and the vertical axis represents the obstacle severity contribution ratio (the ratio of the obstacle severity of a specific component to the obstacle severity of all critical components). Red indicates the yaw brake, with inlet pressure as the primary evaluation indicator; dark orange indicates the gearbox, with coolant temperature as the primary evaluation indicator; yellow indicates the generator, with speed as the primary evaluation indicator; light green indicates the generator, with non-drive end bearing temperature as the primary evaluation indicator; and gray indicates the remaining 27 components. For example, between 08:31:18 and 08:31:20, the yaw brake corresponds to the largest obstacle severity contribution ratio.

[0108] In some alternative implementations, the meteorological data may also include absolute wind direction angles, and correspondingly, at the wind turbine level, the following evaluation indicators may also be calculated: First, the dynamic wind energy capture rate within each evaluation time window can be calculated using the following expression: (12) in, Indicates the dynamic wind energy capture rate. This indicates the length of the evaluation time window, where t represents time, or the number of sampling points. The power of the wind turbine unit. Indicates air density, and A represents the swept area of ​​the wind turbine. This indicates the theoretical maximum wind energy utilization coefficient of the target wind turbine unit of this model under the optimal dimensionless tip speed ratio. Indicates wind speed.

[0109] Thus, a continuous decline in the dynamic wind energy capture rate can accurately characterize the deterioration of blade aerodynamic performance or the increase in mechanical resistance of the transmission chain. Therefore, by analyzing this indicator, it can be determined whether the blades or transmission chain in the target wind turbine are aging and need to be replaced.

[0110] Second, the yaw tracking dynamic error for each evaluation time window can be calculated using the following expression: (13) in, This represents the yaw tracking dynamic error, where N represents the number of sampling points within the evaluation time window. This represents the absolute wind direction angle at the i-th sampling point within the evaluation time window. This represents the actual yaw angle of the cabin at the i-th sampling point within the time window.

[0111] In some optional implementations, a total state vector can be constructed based on the collected values ​​corresponding to all evaluation indicators in the target operating data. A total boundary vector can be constructed based on the boundary ideal solutions corresponding to all evaluation indicators. A total covariance matrix can be constructed based on the collected values ​​of all evaluation indicators in the target cluster. A total weight matrix can be constructed based on the indicator weights of all evaluation indicators. Finally, the total Mahalanobis distance corresponding to the target operating data can be determined based on the total state vector, total boundary vector, total covariance matrix, and total weight matrix. Further, following a similar approach, after determining the total Mahalanobis distance corresponding to all operating data in the target cluster, the total coefficient of variation can be determined based on the Mahalanobis distance corresponding to the operating data included in the target cluster. Based on the total Mahalanobis distance and the total coefficient of variation corresponding to the target operating data, the quality score of the target wind turbine under the target operating data can be determined. The specific calculation process can refer to the processing in steps one to seven of step S104 above, and will not be repeated here.

[0112] In this way, by using the regularized inverse matrix of the total covariance matrix, multicollinearity among all indicators (e.g., the cross-correlation between oil temperature and bearing temperature, vibration and power) is decoupled at once, avoiding the fragmented cross-system collaborative degradation information that might occur with component-by-component modeling. Furthermore, the full Mahalanobis distance preserves the overall measurement of cross-component joint deviation modes such as abnormal gearbox oil temperature and abnormal generator vibration, making the overall system score more realistically reflect the comprehensive effect of multi-component coupling degradation in a series system.

[0113] In some alternative implementations, after step S105 described above, the following steps may also be performed: Step 1: Identify the abnormal wind turbines based on the quality scores of the multiple wind turbines in the target evaluation window.

[0114] Among them, many of the operational data are operational data collected in the target evaluation window, and the target wind turbine is any one of multiple wind turbines.

[0115] Step 2: Identify the critical components of the abnormal wind turbine based on their quality scores or obstacle levels within the target evaluation window.

[0116] Step 3: Generate maintenance work orders based on the abnormal wind turbine units and abnormal critical components.

[0117] Specifically, after determining the quality score corresponding to each wind turbine in the target evaluation window, abnormal wind turbines can be identified based on the quality score (for example, wind turbines with quality scores lower than a preset score threshold are identified as abnormal wind turbines, or wind turbines are sorted from highest to lowest quality score, and those ranked after a preset ranking value are identified as abnormal wind turbines). For abnormal wind turbines, abnormal critical components can be identified based on the quality score or obstacle degree of their constituent key components. For example, critical components with quality scores lower than a preset score threshold are identified as abnormal critical components, or critical components are sorted from highest to lowest quality score, and those ranked after a preset ranking value are identified as abnormal critical components, or critical components with obstacle degrees greater than a preset threshold are identified as abnormal critical components.

[0118] The target evaluation window can include multiple sampling points. Accordingly, during the evaluation process, the quality scores of different wind turbine units at the same sampling point can be sorted, and the quality scores of different key components at the same sampling point can be sorted in abnormal wind turbine units.

[0119] Finally, maintenance work orders can be generated based on the identification information of abnormal wind turbine units and abnormal critical components.

[0120] In this way, technicians can perform preventative operations on abnormal wind turbine units and critical components, shifting from reactive maintenance to proactive and precise maintenance. For example, they can focus on bottleneck components such as the main bearing lubrication system and hydraulic pump station.

[0121] In some alternative implementations, after step S104 described above, the following steps may also be performed: After calculating the quality scores of the target key components under multiple operational data points, the average of these scores can be calculated to obtain the final quality score of the target key component within the target evaluation time window. Furthermore, if an anomaly is determined in the target wind turbine, the abnormal key component can be identified based on the final quality scores of each key component within the target wind turbine.

[0122] In some alternative implementations, after step S105 described above, the following steps may also be performed: Based on the component weights of each key component, the final quality scores of each key component are weighted and summed to obtain the final quality score of the target wind turbine within the target evaluation time window. Alternatively, after determining the quality scores of the target wind turbine under multiple operating data points, the average of the quality scores under each operating data point can be calculated and used as the final quality score of the target wind turbine. Finally, abnormal wind turbines can be identified based on the final quality scores of each wind turbine.

[0123] In some optional implementations, evaluation indicators such as the operation and maintenance cost, overall availability, grid voltage, and frequency fluctuation rate of each wind turbine can also be obtained at the wind turbine group level. Furthermore, taking a target wind turbine as an example, its quality score under the target operating data can also be used as an evaluation indicator. Thus, the operation and maintenance cost, overall availability, grid voltage, frequency fluctuation rate, etc., along with the quality score, can constitute a single operational data point for the target wind turbine. The operational data of each wind turbine constitutes a state parameter matrix for the wind turbine group, where one row represents the operational data of a single wind turbine, and one column represents an evaluation indicator. Then, following a similar approach to the aforementioned embodiments, a quality score can be determined for each row, and the overall quality score of the wind turbine group can be determined based on the quality scores corresponding to all rows (e.g., by averaging). Additionally, it can be determined whether the quality score is less than a preset quality score threshold. If so, an anomaly is identified in the wind turbine group. Then, the individual wind turbines in the wind turbine group can be sorted according to the aforementioned method to identify the abnormal wind turbines, and the abnormal key components can be identified from among the abnormal wind turbines.

[0124] In this way, quality scores can be determined at three levels: wind turbine clusters, wind turbine units, and key components, thus achieving an overall collaborative evaluation framework.

[0125] The following example illustrates the TOPSIS method for comprehensive wind power performance evaluation based on heterogeneous decoupling and improved entropy weight.

[0126] As shown in Table 1, in this application, when evaluating a wind turbine cluster, evaluation operations can be performed at three levels: key component level, single wind turbine level, and target wind turbine cluster level. Correspondingly, the evaluation indicators for these three dimensions can include: At the critical component level, evaluations can be conducted from dimensions such as functional performance, reliability, and safety. For example, functional performance evaluation dimensions can include the assessment of two indicators: generator bearing temperature rise rate and converter IGBT thermal resistance degradation. The generator bearing temperature rise rate is calculated by extracting the drive / non-drive bearing temperatures and using this as the basis for calculating the temperature gradient within a time window. The converter IGBT thermal resistance degradation is calculated by extracting the coolant temperature and module junction temperature and using this as the basis for calculating the steady-state thermal resistance deviation. Reliability evaluation can be based on the assessment of the transmission chain mechanical vibration amplitude, which can be achieved by detecting the high-frequency vibration acceleration of the gearbox / main bearing and calculating the dimensionless kurtosis K. Safety evaluation can be based on the hydraulic and yaw brake pressure deviation, which can be achieved by extracting the difference between the system A2 port pressure and the rated yaw damper inlet pressure. In other words, these evaluation indicators can be calculated in step S101 based on the corresponding evaluation indicators in the operating data.

[0127] At the level of a single wind turbine, evaluation can be conducted from dimensions such as comprehensive performance, reliability, and safety. For example, the evaluation dimension of comprehensive performance can be the dynamic wind energy capture rate, which is the ratio of the actual active power output to the theoretical optimal maximum power at the current wind speed, as shown in expression (12). The evaluation dimension of reliability can be the yaw tracking dynamic error, which can be calculated by extracting anemometer data and the actual yaw angle of the target wind turbine nacelle, and then calculating the root mean square error, as shown in expression (13). The evaluation dimension of safety can be the Mean Time Between Failures (MTBF), which can be assessed by fitting the MTBF distribution based on the SCADA historical alarm logs.

[0128] At the wind turbine cluster level, assessments can be conducted from dimensions such as economics and power quality. For example, economics assessments can include evaluations of operation and maintenance costs and the overall availability of the wind turbine cluster. Operation and maintenance costs can consider spare parts consumption, vessel scheduling, and labor costs, allocated according to power generation. The overall availability of the wind turbine cluster can be measured by the percentage of pure effective operating time after excluding grid curtailment and weather-related outages. Power quality assessments can include grid voltage / frequency fluctuations, which can be statistically analyzed as the variance between the wind turbine-side voltage and the grid frequency.

[0129] Table 2

[0130] The process for determining the quality score of wind turbine units can be as follows: Figure 5 As shown.

[0131] During the data acquisition and preprocessing stage, SCADA data is obtained. After preprocessing and feature extraction of the acquired data, a state parameter matrix is ​​obtained. Then, based on wind speed and power, each row of operational data in the state parameter matrix is ​​binned to obtain multiple clusters.

[0132] In the virtual physical boundary setting phase, the objective weights of each evaluation indicator can be calculated first. Then, based on the objective and subjective weights of each evaluation indicator, the indicator weights of each evaluation indicator are calculated separately, and a weight matrix is ​​constructed. Within each cluster, positive and negative ideal solutions can be set for each evaluation indicator.

[0133] In the multi-indicator fusion calculation stage, a covariance matrix can be constructed within any cluster, and the Mahalanobis distance can be calculated. Then, based on the Mahalanobis distance, the service quality index of key components can be calculated.

[0134] In the nonlinear obstacle degree penalty determination and output stage, the obstacle degree can be calculated based on the service quality index of each key component, and the penalty factor can be determined based on the obstacle degree. Finally, the comprehensive service quality index of the wind turbine can be determined according to the weight of each key component, the service quality score, and the penalty factor.

[0135] refer to Figure 6 The comparison results of calculating the comprehensive service quality index (CSI) of wind turbines using this scheme and other methods are presented. The horizontal axis represents the SCADA operation time series, and the vertical axis represents the comprehensive service quality score. The solid red line represents the CSI of wind turbines determined based on the Entropy Mahalanobis TOPSIS method used in this application; the green dotted line represents the CSI of wind turbines determined based on the Euclidean TOPSIS method; the orange dotted line represents the CSI of wind turbines determined based on the AHP linear evaluation method; the blue dashed line represents the CSI determined based on the traditional Fuzzy evaluation method; and the gray dots represent health warning thresholds. Figure 6 As can be seen from the above, the comprehensive service quality index of wind turbines obtained by the scheme model adopted in this application is more robust to fluctuations in normal operation, and can show a clearer downward trend during abnormal or degradation periods. This avoids the problems of oversensitivity, false fluctuations, and missed detection of hidden anomalies caused by traditional methods that do not consider the correlation of indicators or have fixed weights.

[0136] like Figure 7As shown, a three-tiered evaluation is required in the decision-making process for offshore wind turbine clusters. In the first tier, the critical component tier, evaluation characteristics of each critical component are extracted, such as generator drive-end bearings, tower longitudinal vibration, and gearbox oil temperature. Based on the evaluation indicators of critical components, obstacle degree can be calculated. Furthermore, based on the obstacle degree, the source of the problem can be traced to determine which critical component contributes more to the anomalies of the target wind turbine. Simultaneously, each evaluation indicator is weighted using information entropy, with the indicator weights determined based on objective and subjective weights, as mentioned earlier.

[0137] Furthermore, Mahalanobis distance can be calculated to achieve dimensionless state residuals and feature decoupling. This allows for the determination of the comprehensive service quality index of wind turbines in the second layer. Subsequently, at the wind turbine cluster level, global optimization decisions can be made based on the evaluation results from the three levels.

[0138] This embodiment also provides a wind turbine performance evaluation device, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0139] This embodiment provides a performance evaluation device for wind turbine generators, such as... Figure 8 As shown, it includes: The acquisition module 810 is used to acquire multiple operational data, weight configuration information, and meteorological data corresponding to the multiple operational data of the target wind turbine. The target wind turbine includes multiple key components. Each operational data includes the collected values ​​corresponding to multiple evaluation indicators. The multiple evaluation indicators include evaluation indicators associated with the multiple key components. The weight configuration information includes the indicator weights corresponding to the multiple evaluation indicators and the component weights of the multiple key components. The partitioning module 820 is used to partition the multiple operational data into multiple clusters based on the meteorological data corresponding to each of the multiple operational data. The determination module 830 is used to determine the boundary ideal solutions corresponding to various evaluation indicators in the target cluster; based on the collected values ​​of evaluation indicators associated with the target key components in the target operation data, the boundary ideal solutions of the evaluation indicators associated with the target key components, and the indicator weights, the quality score of the target key components under the target operation data is determined, where the target operation data is any operation data in the target cluster, and the target key component is any one of multiple key components; based on the component weights and quality scores of multiple key components, the quality score of the target wind turbine under the target operation data is determined.

[0140] In some alternative implementations, the determining module 830 is specifically used for: Based on the collected values ​​of evaluation indicators associated with the key components of the target, construct the target state vector; Based on the boundary ideal solution of the evaluation index associated with the target key component, construct the boundary vector of the target key component; Based on the collected values ​​of evaluation indicators associated with the key components of the target, construct the target covariance matrix; Construct a weight matrix based on the weights of the evaluation indicators associated with the target key components; Based on the target state vector, boundary vector, target covariance matrix, and weight matrix, determine the target Mahalanobis distance corresponding to the target running data; The coefficient of variation is determined based on the Mahalanobis distances corresponding to the operational data included in the target cluster. Based on the target Mahalanobis distance and coefficient of variation, determine the quality score of the target's key components under the target's operating data.

[0141] In some alternative implementations, the determining module 830 is specifically used for: A comprehensive value is determined based on the component weights and quality scores of the multiple key components. Based on the component weight and quality score of the target key component, and the comprehensive value, the target obstacle degree of the target key component to the target wind turbine is determined; The quality score of the target wind turbine under the target operating data is determined based on the obstacle degree corresponding to at least one of the key components, and the component weight and quality score corresponding to multiple key components.

[0142] In some alternative implementations, the determining module 830 is specifically used for: Based on at least one of the obstacle degrees and a preset obstacle threshold, a penalty factor corresponding to at least one of the obstacle degrees is determined respectively; The initial quality score is determined based on the component weights and quality scores corresponding to the multiple key components; The quality score of the target wind turbine under the target operating data is determined based on at least one of the penalty factors and the initial quality score.

[0143] In some alternative implementations, the acquisition module 810 is further configured to: Obtain the objective weights and subjective weights corresponding to the target evaluation indicators, where the target evaluation indicator is any one of a variety of evaluation indicators; Calculate the total weight based on the objective and subjective weights corresponding to various evaluation indicators; The weights of the target evaluation indicators are determined based on the total weight and the objective and subjective weights corresponding to the target evaluation indicators.

[0144] In some optional implementations, multiple evaluation metrics also include the power of the target wind turbine; the partitioning module 720 is also used for: Based on the power in multiple operational data points, and the meteorological data corresponding to each operational data point, the multiple operational data points are divided into multiple clusters.

[0145] In some alternative implementations, the apparatus further includes a generation module for: Based on the quality scores of the multiple wind turbines included in the wind turbine group in the target evaluation window, the abnormal wind turbines are identified. Among them, multiple operating data are the operating data collected in the target evaluation window, and the target wind turbine is any one of the multiple wind turbines. Based on the quality scores or obstacle degrees of multiple key components included in the abnormal wind turbine, the abnormal key components are identified in the target evaluation window; Based on the abnormal wind turbine units and abnormal critical components, generate operation and maintenance work orders.

[0146] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0147] The following is a detailed reference. Figure 9 This diagram illustrates a suitable structural schematic for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 901, which can perform various appropriate actions and processes according to a program stored in ROM 902 or a program loaded from memory 908 into RAM 903. RAM 903 also stores various programs and data required for the operation of the electronic device. The processor 901, ROM 902, and RAM 903 are interconnected via bus 904. An input / output (I / O) interface 905 is also connected to bus 904; wherein ROM is a read-only memory and RAM is a random access memory.

[0148] Typically, the following devices can be connected to I / O interface 905: input devices 906 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 907 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 908 including, for example, magnetic tapes, hard disks, etc.; and communication devices 909. Communication device 909 allows electronic devices to exchange data via wireless or wired communication with other devices. Although Figure 8Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0149] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 909, or installed from a memory 908, or installed from a ROM 902. When the computer program is executed by the processor 901, it performs the functions defined in the wind turbine performance evaluation method of the embodiments of the present invention.

[0150] Figure 9 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0151] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the wind turbine performance evaluation method shown in the above embodiments is implemented.

[0152] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0153] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A performance evaluation method for wind turbine generators, characterized in that, The method includes: The system acquires multiple operational data points and weight configuration information of the target wind turbine, as well as meteorological data corresponding to each of the multiple operational data points. The target wind turbine includes multiple key components, each operational data point includes collected values ​​corresponding to multiple evaluation indicators, the multiple evaluation indicators include evaluation indicators associated with the multiple key components, and the weight configuration information includes indicator weights corresponding to the multiple evaluation indicators and component weights of the multiple key components. Based on the meteorological data corresponding to each of the aforementioned operational data points, the multiple operational data points are divided into multiple clusters; In the target cluster, the boundary ideal solutions corresponding to various evaluation indicators are determined respectively; Based on the collected values ​​of evaluation indicators associated with the target key components in the target operation data, the boundary ideal solutions of the evaluation indicators associated with the target key components, and the indicator weights, the quality score of the target key components under the target operation data is determined, wherein the target operation data is any one of the operation data in the target cluster, and the target key component is any one of the multiple key components; The quality score of the target wind turbine under the target operating data is determined based on the component weights and quality scores of the multiple key components.

2. The method according to claim 1, characterized in that, The step of determining the quality score of the target key component under the target operating data based on the collected values ​​of evaluation indicators associated with the target key component in the target operating data, the boundary ideal solutions of the evaluation indicators associated with the target key component, and the indicator weights includes: Based on the collected values ​​of evaluation indicators associated with the target key components, a target state vector is constructed; Based on the boundary ideal solution of the evaluation index associated with the target key component, construct the boundary vector of the target key component; Based on the collected values ​​of evaluation indicators associated with the target key components, a target covariance matrix is ​​constructed; A weight matrix is ​​constructed based on the weights of the evaluation indicators associated with the target key components; Based on the target state vector, the boundary vector, the target covariance matrix, and the weight matrix, determine the target Mahalanobis distance corresponding to the target running data; The coefficient of variation is determined based on the Mahalanobis distances corresponding to the operational data included in the target cluster. Based on the target Mahalanobis distance and the coefficient of variation, the quality score of the target key component under the target operating data is determined.

3. The method according to claim 1, characterized in that, The step of determining the quality score of the target wind turbine under the target operating data based on the component weights and quality scores of multiple key components includes: A comprehensive value is determined based on the component weights and quality scores of the multiple key components. Based on the component weight and quality score of the target key component, and the comprehensive value, the target obstacle degree of the target key component to the target wind turbine is determined; The quality score of the target wind turbine under the target operating data is determined based on the obstacle degree corresponding to at least one of the key components, and the component weight and quality score corresponding to multiple key components.

4. The method according to claim 3, characterized in that, The step of determining the quality score of the target wind turbine under the target operating data based on the obstacle degree corresponding to at least one of the key components, and the component weights and quality scores corresponding to multiple key components, includes: Based on at least one of the obstacle degrees and a preset obstacle threshold, a penalty factor corresponding to at least one of the obstacle degrees is determined respectively; The initial quality score is determined based on the component weights and quality scores corresponding to the multiple key components; The quality score of the target wind turbine under the target operating data is determined based on at least one of the penalty factors and the initial quality score.

5. The method according to any one of claims 1 to 4, characterized in that, Before acquiring multiple operational data points and weight configuration information of the target wind turbine, as well as meteorological data corresponding to each of the multiple operational data points, the method further includes: Obtain the objective weights and subjective weights corresponding to the target evaluation indicators, wherein the target evaluation indicator is any one of the various evaluation indicators; Calculate the total weight based on the objective and subjective weights corresponding to the various evaluation indicators; The indicator weights of the target evaluation indicators are determined based on the total weight, the objective weights, and the subjective weights corresponding to the target evaluation indicators, respectively.

6. The method according to any one of claims 1 to 4, characterized in that, The evaluation indicators all include the power of the target wind turbine; the method also includes: Based on the power in the multiple operational data and the meteorological data corresponding to each of the multiple operational data, the multiple operational data are divided into multiple clusters.

7. The method according to claim 3 or 4, characterized in that, After determining the quality score of the target wind turbine under the target operating data based on the component weights and quality scores of multiple key components, the method further includes: Based on the quality scores corresponding to the multiple wind turbines in the target evaluation window, the abnormal wind turbines are identified. The multiple operational data are operational data collected in the target evaluation window, and the target wind turbine is any one of the multiple wind turbines. Based on the quality score or obstacle degree of multiple key components included in the abnormal wind turbine in the target evaluation window, the abnormal key components are identified; Based on the abnormal wind turbine and the abnormal critical component, an operation and maintenance work order is generated.

8. A performance evaluation device for wind turbine generators, characterized in that, The device includes: The acquisition module is used to acquire multiple operational data points of the target wind turbine, weight configuration information, and meteorological data corresponding to the multiple operational data points respectively. The target wind turbine includes multiple key components, each operational data point includes the collected values ​​corresponding to multiple evaluation indicators, the multiple evaluation indicators include evaluation indicators associated with the multiple key components, and the weight configuration information includes the indicator weights corresponding to the multiple evaluation indicators and the component weights of the multiple key components. The partitioning module is used to divide the multiple operational data into multiple clusters based on the meteorological data corresponding to each of the multiple operational data. The determination module is used to determine the boundary ideal solutions corresponding to various evaluation indicators in the target cluster; determine the quality score of the target key component under the target operating data based on the collected values ​​of the evaluation indicators associated with the target key component in the target operating data, the boundary ideal solutions of the evaluation indicators associated with the target key component, and the indicator weights; wherein the target operating data is any operating data in the target cluster, and the target key component is any one of the multiple key components; and determine the quality score of the target wind turbine under the target operating data based on the component weights and quality scores of the multiple key components.

9. An electronic device, characterized in that, include: A memory and a processor are interconnected, the memory stores computer instructions, and the processor executes the computer instructions to perform the performance evaluation method for the wind turbine as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the performance evaluation method for the wind turbine generator as described in any one of claims 1 to 7.

Images