Intelligent new energy power generation equipment monitoring method and system

Abstract

The invention discloses an intelligent new energy power generation equipment monitoring method and system, and belongs to the technical field of power generation monitoring, and the method comprises the steps: power generation equipment data integration, power generation equipment nearest neighbor number searching, power generation equipment neighborhood guiding clustering and power generation equipment monitoring. According to the scheme, the step length and the nearest neighbor number are dynamically adjusted according to the intensity degree, the temporary optimal nearest neighbor number is obtained, the nearest neighbor stability test is carried out based on the nearest neighbor overlapping degree and the nearest neighbor effective rate difference value between adjacent iterations, the optimal nearest neighbor number is determined, and the accuracy of equipment state recognition is improved; screening neighborhood optimization points according to the local density, performing two-dimensional sorting by combining the correlation degree and the fluctuation index, determining an affiliation guide point, setting a threshold value based on the local density and a neighbor representative proportion, determining a condensation center point, constructing an initial cluster according to the condensation center point and the correlation degree, calculating cluster correlation based on the affinity and the correlation degree, and obtaining a final clustering result. And a reliable basis is provided for state judgment of the new energy power generation equipment.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power generation monitoring, and particularly refers to an intelligent new energy power generation equipment monitoring method and system. BACKGROUND

[0002] The new energy power generation equipment monitoring method is a technical scheme that fuses big data analysis and artificial intelligence algorithm, identifies abnormal features and potential fault risks in equipment operation by analyzing multi-dimensional data of new energy power generation equipment in real time, and provides scientific decision basis for operation and maintenance personnel, power generation enterprises and energy supervision departments, so as to guarantee stable and safe operation of new energy power generation equipment.

[0003] However, in the existing new energy power generation equipment monitoring method, the equipment operation data is affected by environmental factors, the data density distribution fluctuates significantly, it is difficult to accurately capture the dynamic change structure of data distribution, resulting in low equipment state recognition accuracy and poor monitoring result consistency; in the existing new energy power generation equipment monitoring method, the equipment operation data is non-uniformly distributed due to environmental changes, traditional clustering is easy to misclassify similar data under different environments as the same class, and the clustering center selection lacks clear guidance, which cannot dynamically adapt to the data distribution characteristics under different equipment states, resulting in low equipment state recognition accuracy and slow monitoring response. SUMMARY

[0004] In view of the above, in order to overcome the defects of the prior art, the application provides a kind of intelligent new energy power generation equipment monitoring method and system, in view of the existing new energy power generation equipment monitoring method Equipment operating data is influenced by environmental factors, data density distribution fluctuation is significant, it is difficult to accurately capture the dynamic change structure of data distribution, leading to low equipment state recognition accuracy, monitoring result consistency is poor, the initial local density is used for high-density effective data screening in the scheme, the step size and the nearest neighbor number are dynamically adjusted according to the density, the temporarily preferred nearest neighbor number is obtained, the nearest neighbor stability test is carried out based on the nearest neighbor overlap degree and the near neighbor efficiency difference between adjacent iterations, and the optimal nearest neighbor number is determined, which provides reliable parameters for subsequent clustering, improves the accuracy of equipment state recognition and the consistency of monitoring results, and avoids the influence of monitoring jump on operation and maintenance decision;In view of the existing new energy power generation equipment monitoring method, the equipment operating data is non-uniformly distributed due to environmental changes, the traditional clustering is easy to misclassify similar data under different environments as the same class, and the clustering center selection lacks clear guidance, and cannot dynamically adapt to the data distribution characteristics under different states of the equipment, leading to low equipment state recognition accuracy and slow monitoring response, the scheme selects neighborhood optimization point according to local density, determines attribution guide point by combining correlation degree and fluctuation index two-dimensional sorting, sets local aggregation threshold and near neighbor attribution threshold based on local density and near neighbor representative ratio, determines condensation center point, constructs initial cluster according to condensation center and correlation degree, calculates cluster correlation based on affinity and correlation degree, iteratively merges the cluster with the largest correlation, and obtains the final clustering result, improves the accuracy of equipment state classification and monitoring timeliness, provides reliable basis for state judgment of new energy power generation equipment, and ensures that operation and maintenance personnel respond to equipment abnormities quickly.

[0005] The technical scheme adopted by the application is as follows: the application provides an intelligent new energy power generation equipment monitoring method, which comprises the following steps:

[0006] Step S1: integrate power generation equipment data;

[0007] Step S2: search for the nearest neighbor number of the power generation equipment;

[0008] Step S3: neighborhood guide clustering of power generation equipment;

[0009] Step S4: power generation equipment monitoring.

[0010] Further, in step S1, the power generation equipment data integration is to collect historical new energy power generation equipment data and real-time new energy power generation equipment data, wherein the historical new energy power generation equipment data and the real-time new energy power generation equipment data both include photovoltaic module related data, environmental data and inverter data, the historical new energy power generation equipment data is labeled with a label, the label includes a normal state, an abnormal early warning state and a fault state, and the collected historical new energy power generation equipment data and real-time new energy power generation equipment data are subjected to data cleaning and data normalization.

[0011] Further, in step S2, the power generation equipment nearest neighbor number search specifically includes the following steps:

[0012] Step S21: high-density effective data screening; for each historical new energy power generation equipment data, an initial local density of the data is calculated based on an initial nearest neighbor number, and the number n of data whose initial local density is not lower than the local density of the data itself in the nearest neighbors is counted, the historical new energy power generation equipment data with the number n greater than or equal to 2 is marked as effective data, all effective data and real-time new energy power generation equipment data are integrated to construct a power generation equipment monitoring data set;

[0013] Step S22: adaptive step neighbor number optimization; an adaptive density threshold is set according to the initial local density distribution of all data in the power generation equipment monitoring data set, the nearest neighbor set of each data in the power generation equipment monitoring data set is determined based on the current nearest neighbor number, the density of the power generation equipment monitoring data set is obtained based on the nearest neighbor set, an initial step is set, the step is dynamically adjusted according to the density, and the nearest neighbor number is adjusted again according to the adjusted step, the step and the nearest neighbor number are repeatedly iterated and adjusted until the proportion of data without reverse nearest neighbors in the power generation equipment monitoring data set is not greater than 5%, and the iteration is stopped, and the nearest neighbor number at this time is taken as a temporary optimal nearest neighbor number;

[0014] Step S23: nearest neighbor stability test; taking the temporary optimal nearest neighbor number as a starting point, the iteration is continued twice to obtain three consecutive iteration corresponding nearest neighbor sets, the nearest neighbor overlap degree and the nearest neighbor efficiency difference value between the two adjacent iterations are calculated, if the nearest neighbor overlap degree of the first group is not less than that of the second group, and the nearest neighbor efficiency difference value of the first group is not greater than that of the second group, it is determined that the nearest neighbor set has reached a stable state, and the nearest neighbor number corresponding to the last iteration in the first group of adjacent iterations is taken as the optimal nearest neighbor number; otherwise, taking the nearest neighbor number of the latest iteration as a new starting point, the iteration is performed twice again until the nearest neighbor set reaches a stable state and the iteration is stopped.

[0015] Further, in step S3, the power generation equipment neighborhood guided clustering specifically includes the following steps:

[0016] Step S31: neighborhood guide point optimization; for each data in the power plant monitoring data set, in its optimal nearest neighbor set, the data whose local density is not lower than the local density of the data itself is screened out as a neighborhood preferred point, the correlation between the data and the neighborhood preferred point is calculated, and the fluctuation index of the neighborhood preferred point is calculated, the neighborhood preferred points are sorted in descending order of correlation and in ascending order of fluctuation index, only the neighborhood preferred points that are simultaneously in the top 20% in terms of correlation and in the top 20% in terms of fluctuation index are retained, if there is only one neighborhood preferred point retained, the neighborhood preferred point is selected as the attribution guide point of the data; if there are multiple neighborhood preferred points retained, the neighborhood preferred point with the highest correlation is selected as the attribution guide point of the data; if there is no neighborhood preferred point retained, the data itself is defined as the attribution guide point;

[0017] Step S32: condensation center adaptive extraction; the data whose attribution guide point is itself and whose local density is not less than the local aggregation threshold value is screened out from the power plant monitoring data set as a condensation candidate point, for each condensation candidate point, the proportion of data whose attribution guide point is the condensation candidate point in its optimal nearest neighbor set is calculated, and the near neighbor representative proportion of each condensation candidate point is obtained, only the condensation candidate points whose proportion is not less than the near neighbor attribution threshold value are retained as condensation center points;

[0018] Step S33: correlation assignment builds initial cluster; an initial cluster is created for each condensation center point, and the condensation center point itself is included in the cluster, for all non-condensation center points in the power plant monitoring data set, the correlation between the non-condensation center points and all condensation center points is calculated, each non-condensation center point is assigned to the cluster corresponding to the condensation center point with the largest correlation, and the construction of all initial clusters is completed;

[0019] Step S34: cluster correlation adaptive merging; based on the affinity between data and clusters, and the correlation between condensation center points, the cluster correlation between each two clusters is obtained, the two clusters with the largest cluster correlation are merged into a new cluster, and the mean of the two cluster condensation center points is taken as the condensation center point of the new cluster, the cluster merging operation is repeatedly executed until the cluster correlation between all clusters is less than the merging threshold value, the final clustering result is obtained, and the label with the most number of historical new energy power plant data in each cluster is selected as the cluster label.

[0020] Further, in step S4, the power plant monitoring is based on the final clustering result, the cluster label to which the real-time new energy power plant data belongs is taken as the monitoring result, and the state corresponding to the real-time new energy power plant data is obtained.

[0021] The intelligent new energy power plant monitoring system provided by the application comprises a power plant data integration module, a power plant nearest neighbor number search module, a power plant neighborhood guide clustering module and a power plant monitoring module.

[0022] The power generation equipment data integration module collects historical new energy power generation equipment data and real-time new energy power generation equipment data, performs data cleaning and data normalization, and sends the data to a power generation equipment nearest neighbor number search module;

[0023] The power generation equipment nearest neighbor number search module performs high-density effective data screening based on an initial local density, dynamically adjusts a step size and a nearest neighbor number according to a density degree, obtains a temporarily preferred nearest neighbor number, performs a nearest neighbor stability test based on a nearest neighbor overlap degree and a nearest neighbor efficiency difference between adjacent iterations, determines an optimal nearest neighbor number, and sends the data to a power generation equipment neighborhood guided clustering module;

[0024] The power generation equipment neighborhood guided clustering module screens neighborhood preferred points according to a local density, determines an attribution guide point in combination with a correlation degree and a fluctuation index two-dimensional sorting, sets a threshold based on a local density and a nearest neighbor representative ratio, determines a condensation center point, constructs an initial cluster according to the condensation center point and the correlation degree, calculates cluster relevance based on an affinity and the correlation degree, iteratively merges clusters with the greatest relevance, obtains a final clustering result, and sends the data to a power generation equipment monitoring module;

[0025] The power generation equipment monitoring module takes a cluster label to which real-time new energy power generation equipment data belongs as a monitoring result based on the final clustering result.

[0026] The above scheme has the following beneficial effects:

[0027] (1) In view of the problem in the prior new energy power generation equipment monitoring method that equipment operation data is affected by environmental factors, data density distribution fluctuates significantly, and it is difficult to accurately capture the dynamic change structure of data distribution, resulting in low equipment state recognition accuracy and poor monitoring result consistency, the scheme performs high-density effective data screening based on an initial local density, effectively eliminates redundant data interference, and improves the data quality basis for equipment state recognition; a step size and a nearest neighbor number are dynamically adjusted according to a density degree, a temporarily preferred nearest neighbor number is obtained, the nearest neighbor number is dynamically adjusted with the equipment data density, environmental sensitivity data characteristics are adapted, real-time data distribution structure is accurately matched, a nearest neighbor stability test is performed based on a nearest neighbor overlap degree and a nearest neighbor efficiency difference between adjacent iterations, an optimal nearest neighbor number is determined, reliable parameters are provided for subsequent clustering, the accuracy of equipment state recognition and the consistency of monitoring results are improved, and the operation and maintenance decision is prevented from being affected by monitoring jumps.

[0028] (2) In view of the problems that in the existing new energy power generation equipment monitoring method, the equipment operation data is non-uniformly distributed due to environmental changes, the traditional clustering is easy to misclassify similar data under different environments as the same class, and the clustering center selection lacks clear guidance and cannot dynamically adapt to the data distribution characteristics of different states of the equipment, resulting in low equipment state recognition accuracy and slow monitoring response, the scheme selects the optimal point according to the local density, combines the correlation and fluctuation index double-dimensional sorting to determine the attribution guide point, provides a clear clustering direction for each data, reduces the state misjudgment under non-uniform data distribution; based on the local density and the proportion of the nearest neighbor representative, the local aggregation threshold and the nearest neighbor attribution threshold are set to determine the condensation center point, ensure that the center matches the data distribution characteristics of each state of the equipment, provide a reliable core for subsequent clustering, and avoid state recognition errors caused by center deviation; according to the condensation center and the correlation, the initial cluster is constructed, the cluster correlation is calculated based on the affinity and the correlation, the clusters with the largest correlation are iteratively merged, and the final clustering result is obtained, which improves the accuracy and monitoring timeliness of the equipment state classification, provides a reliable basis for the state judgment of the new energy power generation equipment, and ensures that the operation and maintenance personnel can quickly respond to equipment abnormalities. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 A flowchart of an intelligent new energy power generation equipment monitoring method provided by the present application is shown in the figure.

[0030] Figure 2 A schematic diagram of an intelligent new energy power generation equipment monitoring system provided by the present application is shown in the figure.

[0031] Figure 3 A flowchart of step S2 is shown in the figure.

[0032] Figure 4 A flowchart of step S3 is shown in the figure.

[0033] The accompanying drawings are used to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, to explain the present application, and do not constitute a limitation of the present application. DETAILED DESCRIPTION

[0034] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0035] In the description of the present application, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.

[0036] Embodiment one, refer to Figure 1 The present application provides a kind of intelligent new energy power generation equipment monitoring method, the method comprises the following steps:

[0037] Step S1: power generation equipment data integration;Collect historical new energy power generation equipment data and real-time new energy power generation equipment data, and carry out data cleaning and data normalization;

[0038] Step S2: power generation equipment nearest neighbor search;Based on initial local density, high-density effective data screening is carried out, and according to the density degree dynamic adjustment step and nearest neighbor number, the temporary optimal nearest neighbor number is obtained, based on the nearest neighbor overlap degree and near neighbor efficiency difference between adjacent iterations, near neighbor stability test is carried out, and the optimal nearest neighbor number is determined;

[0039] Step S3: power generation equipment neighborhood guided clustering;According to local density, neighborhood optimization point is screened, and combined with correlation degree and fluctuation index two-dimensional sorting, attribution guide point is determined, based on local density and near neighbor representative ratio, threshold value is set, and condensation center point is determined, according to condensation center point and correlation degree, initial cluster is constructed, based on affinity and correlation degree, cluster correlation is calculated, the cluster with maximum correlation is iteratively merged, and the final clustering result is obtained;

[0040] Step S4: power generation equipment monitoring;Based on the final clustering result, the cluster label to which real-time new energy power generation equipment data belongs is taken as monitoring result.

[0041] Embodiment two, refer to Figure 1The embodiment is based on the above embodiment, in step S1, the power generation equipment data integration is to collect historical new energy power generation equipment data and real-time new energy power generation equipment data, the new energy power generation equipment refers to a solar power generation equipment, the historical new energy power generation equipment data and the real-time new energy power generation equipment data all include photovoltaic module related data, environmental data and inverter data, the photovoltaic module related data, the environmental data and the inverter data are all collected every 5 minutes; the photovoltaic module related data includes voltage, current, power and surface temperature; the environmental data includes light intensity, environmental temperature, wind speed and wind direction; the inverter data includes direct current input voltage, direct current input current, alternating current output voltage, alternating current output current, alternating current output frequency and conversion efficiency; the historical new energy power generation equipment data is labeled with a label, the label includes a normal state, an abnormal early warning state and a fault state, and is only used for cluster result evaluation; the label division standard is that the conversion efficiency is not less than 90% for the normal state, the conversion efficiency is less than 90% and not less than 75% for the abnormal early warning state, and the conversion efficiency is less than 75% or the alternating current output frequency is 0 for the fault state; and the collected historical new energy power generation equipment data and real-time new energy power generation equipment data are subjected to data cleaning and data normalization; the data cleaning is to delete error values in the data by using the 3σ principle; and the data normalization is to use the maximum-minimum scaling method to unify the numerical data to the range of [0, 1].

[0042] Embodiment three, refer to Figure 1 and Figure 3 The embodiment is based on the above embodiment, in step S2, the power generation equipment nearest neighbor number search specifically includes the following contents:

[0043] Step S21: high-density effective data screening; there are low-density redundant data in the historical new energy power generation equipment data, these data are mostly abnormal data in the device start-stop transition state or extreme environment, if directly included in the monitoring data set, it will cause deviation in subsequent clustering and reduce the device state recognition accuracy; for each historical new energy power generation equipment data, based on the initial nearest neighbor number The initial local density of the data is calculated, and the number n of data whose initial local density in the nearest neighbor is not lower than the local density of itself is counted, the historical new energy power generation equipment data with the number n greater than or equal to 2 is marked as effective data, all effective data and real-time new energy power generation equipment data are integrated to construct a power generation equipment monitoring data set; the redundant low-density data are removed, high-quality data support is provided for subsequent nearest neighbor number optimization and clustering, and the reliability of the device state monitoring is improved; the formula used is as follows:

[0044] ;

[0045] In the formula, x i and x jare the ith and jth historical new energy power generation equipment data respectively, i and j are data indexes, is x i is the initial local density, k0 is the minimum nearest neighbor number, is the ceiling function, N is the feature dimension of the historical new energy power generation equipment data, is x i is the initial nearest neighbor set of x is x i is the Euclidean distance between x j and x i , is the average Euclidean distance of all nearest neighbor data in x , is the minimum value function, is the natural exponential function;

[0046] Step S22: adaptive step neighbor number optimization; the operation data of the new energy power generation equipment dynamically changes with environmental factors such as light and temperature, and a fixed nearest neighbor number cannot adapt to the fluctuation of data density, which is easy to cause the neighborhood division to be too coarse or too fine when clustering, affecting the equipment state distinguishing effect; an adaptive density threshold is set according to the initial local density distribution of all data in the power generation equipment monitoring data set, the nearest neighbor set of each data in the power generation equipment monitoring data set is determined based on the current nearest neighbor number, the density of the power generation equipment monitoring data set is obtained based on the nearest neighbor set, and an initial step is set, the step is dynamically adjusted according to the density, and the nearest neighbor number is adjusted again according to the adjusted step, the step and the nearest neighbor number are repeatedly iterated and adjusted, and the iteration is stopped when the proportion of data without anti-nearest neighbor in the power generation equipment monitoring data set is not greater than 5%, and the nearest neighbor number at this time is taken as a temporary preferred nearest neighbor number; the nearest neighbor number can accurately reflect the data distribution structure, provide reasonable neighborhood division basis for subsequent clustering, and improve the accuracy and adaptability of equipment state clustering; the used formula is as follows:

[0047] ;

[0048] ;

[0049] ;

[0050] In the formula, is the density of the power generation equipment monitoring data set at the tth iteration, t is the iteration number index, A eff is the power generation equipment monitoring data set, is the number of data in A eff , x a and x b are the ath and bth data in the power generation equipment monitoring data set respectively, a and b are data indexes, is x b initial local density of k t and k t+1 are the nearest neighbor number at the tth and (t+1)th iteration, respectively, is the nearest neighbor set of x a at the tth iteration, and are the step size at the tth and (t+1)th iteration, respectively, and are the mean and standard deviation of the initial local density of all data in the power plant monitoring data set, respectively, x a the anti-nearest neighbor of x a refers to the data in the power plant monitoring data set that takes x a as the nearest neighbor, is the maximum value function;

[0051] Step S23: nearest neighbor stability test; the temporary nearest neighbor number obtained by adaptive step size optimization may fluctuate and cannot ensure stability in the dynamic change scenario of new energy equipment data. If directly used for clustering, it may cause frequent jumps in equipment condition monitoring results, affecting operation and maintenance decisions. Taking the temporarily preferred nearest neighbor number as the starting point, continue to perform two iterations to obtain the nearest neighbor sets corresponding to t temp , t temp+1 and t temp+2 three consecutive iterations, calculate the nearest neighbor overlap and nearest neighbor efficiency difference between the two adjacent iterations, the first group is (t temp , t temp+1 ), and the second group is (t temp+1 , t temp+2 ). If the nearest neighbor overlap of the first group is not less than the second group, and the nearest neighbor efficiency difference of the first group is not greater than the second group, it is determined that the nearest neighbor set has reached a stable state, and the nearest neighbor number corresponding to the last iteration t temp+1 in the first group of adjacent iterations is taken as the optimal nearest neighbor number; otherwise, taking the nearest neighbor number of the latest iteration as the new starting point, performing two iterations again until the nearest neighbor set reaches a stable state to stop iteration; providing reliable parameters for subsequent neighborhood-guided clustering, improving the consistency of equipment condition monitoring, and ensuring that operation and maintenance personnel can develop operation and maintenance strategies based on stable results; the formula used is as follows:

[0052] ;

[0053] ;

[0054] where k temp is the temporarily preferred nearest neighbor number, t temp is the iteration number corresponding to the temporarily preferred nearest neighbor number, t temp+1 and t temp+2is t temp the number of iterations after, and are the nearest neighbor overlap and the nearest neighbor efficiency difference of the power plant monitoring data set in the t temp and t temp+1 th iteration, and are the nearest neighbor set of x temp and x temp+1 in the t a th iteration, is and the number of data in the intersection, is and the number of data in the union, is and the number of data in the intersection, is and the number of data in the intersection, and are the number of data in and .

[0055] By performing the above operations, in view of the problems in the existing new energy power plant monitoring method that the device operation data is affected by environmental factors, the data density distribution fluctuates significantly, it is difficult to accurately capture the dynamic change structure of the data distribution, resulting in low device state recognition accuracy and poor monitoring result consistency, the present scheme performs high-density effective data screening based on the initial local density, effectively eliminates redundant data interference, and improves the data quality basis of device state recognition. According to the density degree, the step size and the number of nearest neighbors are dynamically adjusted to obtain a temporarily preferred number of nearest neighbors, so that the number of nearest neighbors is dynamically adjusted with the device data density, adapts to the environmental sensitive data characteristics, and accurately matches the real-time data distribution structure. Based on the nearest neighbor overlap and the nearest neighbor efficiency difference between adjacent iterations, the nearest neighbor stability is tested, the optimal number of nearest neighbors is determined, reliable parameters are provided for subsequent clustering, the accuracy of device state recognition and the consistency of monitoring results are improved, and the influence of monitoring jump on operation and maintenance decision is avoided.

[0056] Embodiment four, refer to Figure 1 and Figure 4 , this embodiment is based on the above embodiment, in step S3, the power plant neighborhood guided clustering specifically includes the following steps:

[0057] Step S31: neighborhood guide point optimization; the data distribution of new energy power generation equipment is non-uniformly affected by environmental factors, and traditional clustering is easy to misclassify similar data in different environments as the same class, leading to misjudgment of equipment state, and lacking clear attribution guide basis, and the randomness of clustering results is strong; for each data in the power generation equipment monitoring data set, in its optimal nearest neighbor set, the data with local density not lower than the local density of the data itself is selected as a neighborhood preferred point, the correlation between the data and the neighborhood preferred point is calculated, and the fluctuation index of the neighborhood preferred point is calculated, the neighborhood preferred points are sorted in descending order of correlation and in ascending order of fluctuation index, only the neighborhood preferred points with correlation ranking in the top 20% and fluctuation index ranking in the top 20% are reserved, if the reserved neighborhood preferred points are only one, the neighborhood preferred point is selected as the attribution guide point of the data; if there are multiple reserved neighborhood preferred points, the neighborhood preferred point with the highest correlation is selected as the attribution guide point of the data; if there is no reserved neighborhood preferred point, the data itself is defined as the attribution guide point; the attribution guide point is determined according to the rule, the clear clustering direction is provided for each data, the misclustering under non-uniform data distribution is reduced, the accuracy of equipment state classification is improved, and it is especially suitable for the data distribution fluctuation scenario caused by solar equipment due to light change; the used formula is as follows:

[0058] ;

[0059] ;

[0060] In the formula, x c is the cth neighborhood preferred point, c is the neighborhood preferred point index, and are the Euclidean distance and correlation between x a and x c , d max is the maximum Euclidean distance between all data pairs in the power generation equipment monitoring data set, is the fluctuation index of x c , is the optimal nearest neighbor set of x c , is the number of data in , x f is the fth data in the optimal nearest neighbor set, f is the data index, is the local density of x f based on the optimal nearest neighbor number, is the average local density of all data in the optimal nearest neighbor set of x c ;

[0061] Step S32: Adaptive extraction of agglomeration centers; Traditional clustering agglomeration centers are mostly manually set or randomly selected, which cannot adapt to the data distribution characteristics of new energy equipment under different states. This can easily lead to the cluster centers deviating from the core data of the actual equipment state, affecting the accuracy of state differentiation. Data with its own guiding point and local density not less than the local clustering threshold are selected from the power generation equipment monitoring dataset as agglomeration candidate points. The local clustering threshold is the 25th percentile of the local density of all data labeled as normal state in the power generation equipment monitoring dataset. For each agglomeration candidate point, the proportion of data with its own guiding point in its best nearest neighbor set is calculated to obtain the nearest neighbor representative proportion of each agglomeration candidate point. Only agglomeration candidate points with a proportion not less than the nearest neighbor belonging threshold are retained as agglomeration center points. The nearest neighbor belonging threshold is the 30th percentile of the nearest neighbor representative proportion of all data labeled as normal state in the power generation equipment monitoring dataset. Adaptive extraction of agglomeration centers requires no manual intervention, can dynamically adapt to the data distribution under different equipment states, improve the accuracy of clustering centers, provide a reliable core for subsequent initial cluster construction, and ensure the rationality of equipment state classification. The formula used is as follows:

[0062] ;

[0063] In the formula, x g It is the g-th agglomeration candidate point, where g is the index of the agglomeration candidate point. It is x g The proportion of neighboring representatives, It is x g The optimal nearest neighbor set, yes The amount of data in Is Central Order equals x g Data x h Quantity, x h It is the h-th data in the best nearest neighbor set, where h is the data index. It is x h The guiding point of belonging;

[0064] Step S33: association distribution builds initial clusters; new energy equipment data is large in amount, if directly clustering full amount data, calculation complexity is high, time-consuming is long, and initial cluster division is easy to be chaotic due to scattered data distribution, affecting subsequent cluster merging efficiency and state monitoring speed; an initial cluster is created for each condensation center point, and the condensation center point itself is included in the cluster, the association degree of all non-condensation center points in the power generation equipment monitoring data set with all condensation center points is calculated, each non-condensation center point is distributed to the cluster corresponding to the condensation center point with the maximum association degree, and the construction of all initial clusters is completed; data grouping is quickly completed, ensuring that the data in the initial cluster has high similarity, reducing the workload of subsequent cluster merging, while ensuring that the initial cluster can preliminarily reflect the different states of the equipment, laying a foundation for the final clustering result;

[0065] Step S34: cluster related adaptive merging; the initial cluster may be over-subdivided, similar data under the same device state is divided into multiple clusters, leading to misjudgment of the device state, and lacking a unified cluster merging standard, which cannot adapt to the dynamic change characteristics of new energy equipment data; based on the affinity between data and clusters, and the association degree between condensation center points, the cluster correlation between each two clusters is obtained, the two clusters with the largest cluster correlation are merged into a new cluster, and the mean of the two cluster condensation center points is taken as the condensation center point of the new cluster, the cluster merging operation is repeatedly executed until the cluster correlation between all clusters is less than the merging threshold , the final clustering result is obtained, and the label with the most number of historical new energy power generation equipment data in each cluster is selected as the cluster label; the similarity between clusters is comprehensively measured, avoiding mismerging caused by single dimension judgment, ensuring that the merging standard adapts to data distribution, and improving the accuracy of device state monitoring; the formula used is as follows:

[0066] ;

[0067] ;

[0068] In the formula, x s and x u are the s-th and u-th data in the cluster, s and u are data indexes, is the association degree between x a and x s , C p and C q are the p-th and q-th clusters, and are the data quantities in C p and C q , p and q are cluster indexes, and are x a and x u and C paffinity between x and x, is x s affinity between x and x, q affinity between x and x, H is a merging threshold, μ H and σ H are respectively the mean and standard deviation of cluster correlation between all clusters, x lp and x lq are respectively the condensation center points in C p and C q , is x lp and x lq , is the cluster correlation between C p and C q .

[0069] By performing the above operations, for the existing new energy power generation equipment monitoring method, the equipment operation data is non-uniformly distributed due to environmental changes, the traditional clustering is easy to misclassify similar data under different environments as the same class, and the selection of clustering center lacks clear guidance and cannot dynamically adapt to the data distribution characteristics of different states of the equipment, resulting in low equipment state recognition accuracy and slow monitoring response. The present scheme selects the neighborhood optimization point according to the local density, combines the correlation degree and the volatility index for two-dimensional sorting, determines the attribution guide point, provides a clear clustering direction for each data, and reduces the state misjudgment under non-uniform data distribution; based on the local density and the proportion of near neighbors, the local aggregation threshold and the near neighbor attribution threshold are set, the condensation center point is determined, the center and the equipment state data distribution characteristics are matched, a reliable core is provided for subsequent clustering, and the state recognition error caused by center deviation is avoided; the initial cluster is constructed according to the condensation center and the correlation degree, the cluster correlation is calculated based on the affinity and the correlation degree, the clusters with the largest correlation are iteratively merged, and the final clustering result is obtained, the accuracy and monitoring timeliness of equipment state classification are improved, a reliable basis is provided for state judgment of new energy power generation equipment, and the operation and maintenance personnel are ensured to quickly respond to equipment abnormalities.

[0070] Embodiment five, refer to Figure 1 , this embodiment is based on the above embodiment, in step S4, the power generation equipment monitoring is based on the final clustering result, the cluster label to which the real-time new energy power generation equipment data belongs is taken as the monitoring result, and the state corresponding to the real-time new energy power generation equipment data is obtained; if the monitoring result is a normal state, the real-time power generation equipment data and the corresponding state are automatically recorded to the equipment monitoring database; if the monitoring result is an abnormal early warning state, an early warning prompt is generated and stored to the equipment operation and maintenance terminal, and the operation and maintenance personnel are prompted to pay attention to the equipment operation trend; if the monitoring result is a fault state, the equipment fault early warning mechanism is immediately triggered, and the equipment fault information is synchronously pushed to the equipment operation and maintenance terminal.

[0071] Embodiment six, refer toFigure 2 The embodiment is based on the above-mentioned embodiment, and the application provides an intelligent new energy power generation equipment monitoring system, which comprises a power generation equipment data integration module, a power generation equipment nearest neighbor number searching module, a power generation equipment neighborhood guided clustering module and a power generation equipment monitoring module.

[0072] The power generation equipment data integration module collects historical new energy power generation equipment data and real-time new energy power generation equipment data, performs data cleaning and data normalization, and sends the data to the power generation equipment nearest neighbor number searching module.

[0073] The power generation equipment nearest neighbor number searching module performs high-density effective data screening based on an initial local density, dynamically adjusts a step length and a nearest neighbor number according to a dense degree, obtains a temporarily optimal nearest neighbor number, performs a nearest neighbor stability test based on a nearest neighbor overlap degree and a nearest neighbor efficiency difference between adjacent iterations, determines an optimal nearest neighbor number, and sends the data to the power generation equipment neighborhood guided clustering module.

[0074] The power generation equipment neighborhood guided clustering module screens neighborhood optimal points according to a local density, determines an attribution guide point by combining a correlation degree and a fluctuation index two-dimensional sorting, sets a threshold based on a local density and a nearest neighbor representative ratio, determines a condensation center point, constructs an initial cluster according to the condensation center point and the correlation degree, calculates cluster correlation based on an affinity and the correlation degree, iteratively merges a cluster with the largest correlation, obtains a final clustering result, and sends the data to the power generation equipment monitoring module.

[0075] The power generation equipment monitoring module takes a cluster label to which real-time new energy power generation equipment data belongs as a monitoring result based on the final clustering result.

[0076] It should be noted that, in this document, relationship terms such as first and second are used merely to distinguish one entity or action from another, without necessarily requiring or implying that these entities or actions are in any way actually related or ordered. Moreover, the terms "comprises", "comprising", or any other variations thereof are intended to cover non-exclusive inclusions, so that a process, method, article, or apparatus that comprises a list of elements does not only include those elements, but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0077] Although the embodiments of the application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the application.

[0078] The above describes the present application and its embodiments, which are not limited, and the drawings only show one of the embodiments of the present application, and the actual structure is not limited thereto. In general, if a person skilled in the art is inspired thereby, without departing from the purpose of the present application, without creative design, similar structure and embodiments of the technical solution are not creative, and should belong to the protection scope of the present application.

Claims

1. A method for monitoring intelligent new energy power generation equipment, characterized in that: The method comprises the following steps: Step S1: power generation equipment data integration; collecting historical new energy power generation equipment data and real-time new energy power generation equipment data, and performing data cleaning and data normalization; Step S2: power generation equipment nearest neighbor number search; high-density effective data screening is performed based on initial local density, the step size and the nearest neighbor number are dynamically adjusted according to the density degree, a temporary optimal nearest neighbor number is obtained, near neighbor stability test is performed based on the nearest neighbor overlap degree and the near neighbor efficiency difference between adjacent iterations, and the optimal nearest neighbor number is determined; Step S3: power generation equipment neighborhood guided clustering; the neighborhood optimal point is screened according to the local density, the attribution guide point is determined by combining the correlation degree and the fluctuation index two-dimensional sorting, the threshold is set based on the local density and the near neighbor representative ratio, the condensation center point is determined, the initial cluster is constructed according to the condensation center point and the correlation degree, the cluster correlation is calculated based on the affinity and the correlation degree, the clusters with the largest correlation are iteratively merged, and the final clustering result is obtained; Step S4: power generation equipment monitoring; the cluster label to which the real-time new energy power generation equipment data belongs is taken as the monitoring result. 2.The intelligent new energy power generation equipment monitoring method of claim 1, wherein: In step S2, the power generation equipment nearest neighbor number search specifically comprises the following steps: Step S21: high-density effective data screening; for each historical new energy power generation equipment data, the initial local density of the data is calculated based on the initial nearest neighbor number, and the number n of data whose initial local density in the nearest neighbor is not lower than the local density of itself is counted, the historical new energy power generation equipment data with the number n greater than or equal to 2 is marked as effective data, all effective data and real-time new energy power generation equipment data are integrated to construct a power generation equipment monitoring data set; Step S22: adaptive step size neighbor number optimization; an adaptive density threshold is set according to the initial local density distribution of all data in the power generation equipment monitoring data set, the nearest neighbor set of each data in the power generation equipment monitoring data set is determined based on the current nearest neighbor number, the density degree of the power generation equipment monitoring data set is obtained based on the nearest neighbor set, the initial step size is set, the step size is dynamically adjusted according to the density degree, and the nearest neighbor number is adjusted again according to the adjusted step size, the step size and the nearest neighbor number are repeatedly iterated and adjusted until the proportion of data without anti-nearest neighbor in the power generation equipment monitoring data set is not greater than 5%, and then the iteration is stopped, and the nearest neighbor number at this time is taken as the temporary optimal nearest neighbor number; Step S23: near neighbor stability test. 3.The intelligent new energy power generation equipment monitoring method of claim 2, wherein: In step S23, the near neighbor stability test is performed from the temporary optimal nearest neighbor number as the starting point, and two iterations are continuously performed to obtain three consecutive iteration corresponding nearest neighbor sets, the nearest neighbor overlap degree and the near neighbor efficiency difference between the two adjacent iterations are calculated, if the nearest neighbor overlap degree of the first group is not less than that of the second group, and the near neighbor efficiency difference of the first group is not greater than that of the second group, it is determined that the nearest neighbor set has reached a stable state, and the nearest neighbor number corresponding to the last iteration in the first group of adjacent iterations is taken as the optimal nearest neighbor number; otherwise, the nearest neighbor number of the latest iteration is taken as the new starting point, and two iterations are performed again until the nearest neighbor set reaches a stable state, and then the iteration is stopped. 4.The intelligent new energy power generation device monitoring method of claim 3, wherein: In step S3, the power generation equipment neighborhood guided clustering specifically comprises the following steps: Step S31: neighborhood guide point optimization; for each data in the power plant monitoring data set, in its optimal nearest neighbor set, the data with local density not lower than the local density of the data itself is selected as a neighborhood preferred point, the correlation between the data and the neighborhood preferred point is calculated, and the fluctuation index of the neighborhood preferred point is calculated, the neighborhood preferred points are sorted in descending order of correlation and in ascending order of fluctuation index, only the neighborhood preferred points with correlation in the top 20% and fluctuation index in the top 20% are retained, if there is only one neighborhood preferred point, it is selected as the attribution guide point of the data; if there are multiple neighborhood preferred points, the neighborhood preferred point with the highest correlation is selected as the attribution guide point of the data; if there is no neighborhood preferred point, the data itself is set as the attribution guide point; Step S32: adaptive extraction of condensation center; Step S33: initial cluster construction by correlation assignment; an initial cluster is created for each condensation center point, and the condensation center point itself is included in the cluster, the correlation between all non-condensation center points in the power plant monitoring data set and all condensation center points is calculated, each non-condensation center point is assigned to the cluster corresponding to the condensation center point with the maximum correlation, and the construction of all initial clusters is completed; Step S34: adaptive merging of cluster correlation.

5. The intelligent new energy power generation equipment monitoring method according to claim 4, characterized in that: In step S32, the adaptive extraction of condensation center is to select data with attribution guide point being itself and local density not less than local aggregation threshold from the power plant monitoring data set as condensation candidate points, for each condensation candidate point, the proportion of data with attribution guide point being the condensation candidate point in its optimal nearest neighbor set is calculated to obtain the neighbor representative proportion of each condensation candidate point, only the condensation candidate points with proportion not less than neighbor attribution threshold are retained as condensation center points. 6.The intelligent new energy power generation device monitoring method of claim 5, wherein: In step S34, the adaptive merging of cluster correlation is based on the affinity between data and clusters, and the correlation between condensation center points, to obtain the cluster correlation between each two clusters, the two clusters with the maximum cluster correlation are merged into a new cluster, and the mean of the two cluster condensation center points is taken as the condensation center point of the new cluster, the cluster merging operation is repeatedly performed until the cluster correlation between all clusters is less than the merging threshold, the final clustering result is obtained, and the label with the most number of historical new energy power plant data in each cluster is selected as the cluster label. 7.The intelligent new energy power generation device monitoring method of claim 6, wherein: In step S1, the power plant data integration is to collect historical new energy power plant data and real-time new energy power plant data, the historical new energy power plant data and real-time new energy power plant data both include photovoltaic module related data, environmental data and inverter data, the historical new energy power plant data is labeled with labels, the labels include normal state, abnormal warning state and fault state, and the collected historical new energy power plant data and real-time new energy power plant data are subjected to data cleaning and data normalization. 8.The intelligent new energy power generation device monitoring method of claim 7, wherein: In step S4, the power plant monitoring is based on the final clustering result, the cluster label to which the real-time new energy power plant data belongs is taken as the monitoring result, and the state corresponding to the real-time new energy power plant data is obtained.

9. An intelligent new energy power generation equipment monitoring system for implementing the intelligent new energy power generation equipment monitoring method according to any one of claims 1-8, characterized in that: The power generation equipment data integration module, the power generation equipment nearest neighbor number search module, the power generation equipment neighborhood guide clustering module and the power generation equipment monitoring module are included. The power generation equipment data integration module collects historical new energy power generation equipment data and real-time new energy power generation equipment data, performs data cleaning and data normalization, and sends the data to the power generation equipment nearest neighbor number search module; The power generation equipment nearest neighbor number search module performs high-density effective data screening based on initial local density, dynamically adjusts the step size and the nearest neighbor number according to the density level, obtains a temporarily preferred nearest neighbor number, performs nearest neighbor stability inspection based on the nearest neighbor overlap degree and the nearest neighbor efficiency difference between adjacent iterations, determines the optimal nearest neighbor number, and sends the data to the power generation equipment neighborhood guide clustering module; The power generation equipment neighborhood guide clustering module screens neighborhood preferred points according to local density, determines the attribution guide point by combining the correlation degree and the fluctuation index two-dimensional sorting, sets the threshold based on the local density and the near neighbor representative ratio, determines the condensation center point, constructs the initial cluster according to the condensation center point and the correlation degree, calculates the cluster correlation based on the affinity and the correlation degree, iteratively merges the cluster with the largest correlation, obtains the final clustering result, and sends the data to the power generation equipment monitoring module; The power generation equipment monitoring module takes the cluster label to which the real-time new energy power generation equipment data belongs as the monitoring result based on the final clustering result.

Images