Offshore wind farm wind turbine complex maintenance strategy optimization method, device and equipment

Abstract

The present invention discloses a composite maintenance strategy optimization method, device and equipment for wind turbines in offshore wind farms. The present invention predicts the operation status by considering the monitoring data of each wind turbine in the offshore wind farm, optimizes the maintenance strategy according to the predicted operation status of each wind turbine, obtains the optimal preventive maintenance strategy for each wind turbine, and considers the current operation status of each wind turbine and the actual maintenance resource constraints, selects the target unit combination to be maintained and optimizes its maintenance strategy, obtains the optimal actual maintenance strategy for the target unit to be maintained, thereby achieving preventive maintenance and actual maintenance of each wind turbine in the offshore wind farm.

Description

Technical Field

[0001] The present invention relates to the field of offshore wind power technology, and in particular to a method, device and equipment for optimizing a composite maintenance strategy for wind turbines in an offshore wind farm. Background Art

[0002] The existing operation and maintenance solutions for wind turbines in offshore wind farms usually adopt a combination of an onshore monitoring center and an offshore operation and maintenance base. The onshore monitoring center first monitors the operation of each wind turbine in the offshore wind farm in real time throughout the day, especially the rotation and friction of key components of the wind turbine. The monitoring results are analyzed centrally to determine the operating status of each wind turbine. Then, targeted inspection and maintenance are carried out on each wind turbine during the low wind period. At the same time, regular maintenance and fault handling of each wind turbine are completed through the offshore operation and maintenance base. The operation and maintenance personnel maintain each wind turbine according to the annual plan, and go to the site to handle equipment failures that cannot be reset according to the maintenance instructions of the onshore monitoring center.

[0003] The existing technology does not consider online data such as real-time operation data and equipment failure information when maintaining each wind turbine, but only considers offline data such as maintenance records and spare parts. Massive data of different time scales and different data sources are not effectively analyzed and integrated, making it difficult to predict the reliability and operation status of each wind turbine to optimize the maintenance strategy, and preventive maintenance cannot be achieved. Moreover, after the maintenance strategy is formed, the execution effect of the maintenance strategy is still limited by the actual maintenance resources. It is difficult to balance the performance of each wind turbine and the benefits of the offshore wind farm to optimize the operation and maintenance strategy, and actual operation and maintenance cannot be achieved. Summary of the invention

[0004] In order to overcome the defects of the prior art, the present invention provides a method, device and equipment for optimizing the composite maintenance strategy of wind turbines in an offshore wind farm, which can realize preventive maintenance and actual maintenance of each wind turbine in the offshore wind farm.

[0005] In order to solve the above technical problems, in a first aspect, the present invention provides a method for optimizing a composite maintenance strategy for wind turbines in an offshore wind farm, comprising:

[0006] For each wind turbine set, based on the monitoring data of the wind turbine set, the current operating state of the wind turbine set is evaluated, and the operating state is predicted in combination with the historical operating state of the wind turbine set to obtain the predicted operating state of the wind turbine set;

[0007] The predicted operating state of each wind turbine is used as an input parameter of a pre-established preventive maintenance strategy optimization model, so that the preventive maintenance strategy optimization model takes the lowest discount cost as the optimization goal, and correspondingly optimizes the maintenance method and monitoring cycle of each wind turbine to obtain the optimal preventive maintenance strategy for the wind turbine;

[0008] According to the current operating status of all the wind turbines, all the wind turbines are classified to obtain a plurality of combinations of turbines to be maintained, and according to the predefined maintenance resource restriction condition, all the combinations of turbines to be maintained are clustered to obtain a plurality of target combinations of turbines to be maintained;

[0009] For each target combination of units to be maintained, an objective function is constructed with the minimum ratio of total maintenance cost to total power generation as the optimization target, constraints are constructed according to the maintenance resource restriction conditions, and an actual maintenance strategy optimization model is established by combining the objective function and the constraints to obtain the optimal actual maintenance strategy for the target units to be maintained.

[0010] Further, the current operating state of the wind turbine generator set is evaluated according to the monitoring data of the wind turbine generator set, specifically:

[0011] According to the monitoring data of the wind turbine generator set, a fuzzy comprehensive evaluation method is adopted to evaluate the current operating status of the wind turbine generator set.

[0012] Furthermore, before performing operating state prediction in combination with the historical operating state of the wind turbine generator set to obtain the predicted operating state of the wind turbine generator set, the method further includes:

[0013] The historical operating state of the wind turbine generator set is determined according to the historical monitoring data of the wind turbine generator set or the n-state left-right non-spanning model.

[0014] Furthermore, the operation state prediction is performed in combination with the historical operation state of the wind turbine generator set to obtain the predicted operation state of the wind turbine generator set, specifically:

[0015] Combining the current operating state and the historical operating state of the wind turbine generator set, obtaining the transition probability of the wind turbine generator set between different operating states, and sorting the transition probabilities of the wind turbine generator set between different operating states in descending order to obtain a predicted operating state sequence of the wind turbine generator set;

[0016] For each operating state change pair in the predicted operating state sequence of the wind turbine generator set, the target monitoring data associated with the operating state change pair is used as a multi-input variable of a pre-established particle swarm multi-objective optimization prediction model, so as to evaluate the predicted operating state of the wind turbine generator set according to the adjustment amount of the target monitoring data output by the particle swarm multi-objective optimization prediction model.

[0017] Furthermore, according to the current operating status of all the wind turbines, all the wind turbines are classified to obtain a plurality of wind turbine combinations to be maintained, specifically:

[0018] The generator sets whose current operating statuses satisfy the predefined generator set maintenance conditions among all the wind turbine sets are taken as generator sets to be maintained, thereby obtaining a plurality of generator sets to be maintained;

[0019] All the units to be maintained belonging to the same area are divided into the same combination of units to be maintained, so as to obtain a plurality of combinations of units to be maintained.

[0020] Furthermore, according to the predefined maintenance resource restriction condition, all the combinations of units to be maintained are clustered to obtain several target combinations of units to be maintained, specifically:

[0021] According to the maintenance vessel quantity restriction condition in the maintenance resource restriction condition, determining the available quantity of maintenance vessels as the partition quantity;

[0022] Based on the number of partitions, all the combinations of units to be maintained are clustered to obtain a plurality of target combinations of units to be maintained.

[0023] Furthermore, the objective function is:

[0024] minZ = (CC + CL + CD) / Wf;

[0025] Among them, CC is the total cost of maintenance personnel scheduling, maintenance ship operation, and maintenance path traffic, CL is the maintenance downtime loss, CD is the delayed maintenance penalty fee, and Wf is the total power generation.

[0026] Furthermore, the constraints include at least one of the constraints on the number of maintenance vessels, the load capacity of maintenance vessels, the number of maintenance personnel, the timing constraints on unit maintenance, the effective working time constraints on the maintenance vessels per day, the maintenance downtime constraints, the delayed maintenance time constraints, the number of maintenance vessel unit maintenance times per day, and the maintenance route constraints.

[0027] In a second aspect, an embodiment of the present invention provides a device for optimizing a composite maintenance strategy for wind turbines in an offshore wind farm, comprising:

[0028] A unit operation state prediction module is used for evaluating the current operation state of each wind turbine unit according to the monitoring data of the wind turbine unit, and performing operation state prediction in combination with the historical operation state of the wind turbine unit to obtain the predicted operation state of the wind turbine unit;

[0029] A preventive maintenance strategy optimization module is used to use the predicted operating state of each wind turbine as an input parameter of a pre-established preventive maintenance strategy optimization model, so that the preventive maintenance strategy optimization model takes the lowest discount cost as the optimization goal, and correspondingly optimizes the maintenance method and monitoring cycle of each wind turbine to obtain the optimal preventive maintenance strategy for the wind turbine;

[0030] A unit classification module for maintenance, used to classify all the wind turbines according to their current operating states to obtain a plurality of combinations of units to be maintained, and to cluster all the combinations of units to be maintained according to predefined maintenance resource constraints to obtain a plurality of target combinations of units to be maintained;

[0031] The actual maintenance strategy optimization module is used to construct an objective function for each target combination of units to be maintained, with the ratio of total maintenance cost to total power generation being minimized as the optimization objective, construct constraints based on the maintenance resource restriction conditions, and establish an actual maintenance strategy optimization model in combination with the objective function and the constraints to solve and obtain the optimal actual maintenance strategy for the target units to be maintained.

[0032] In the third aspect, an embodiment of the present invention provides a device for optimizing a composite maintenance strategy for wind turbines in an offshore wind farm, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the memory is coupled to the processor, and when the processor executes the computer program, the method for optimizing a composite maintenance strategy for wind turbines in an offshore wind farm as described above is implemented.

[0033] The embodiments of the present invention have the following beneficial effects:

[0034] By evaluating the current operating status of each wind turbine according to the monitoring data of the wind turbine, and predicting the operating status in combination with the historical operating status of the wind turbine, the predicted operating status of the wind turbine is obtained; the predicted operating status of each wind turbine is used as the input parameter of the pre-established preventive maintenance strategy optimization model, so that the preventive maintenance strategy optimization model takes the lowest discount cost as the optimization goal, and correspondingly optimizes the maintenance method and monitoring cycle of each wind turbine to obtain the optimal preventive maintenance strategy for the wind turbine; according to the current operating status of all wind turbines, all wind turbines are classified to obtain a number of combinations of units to be maintained. According to the predefined maintenance resource constraint conditions, all the combinations of units to be maintained are clustered to obtain several target combinations of units to be maintained; for each target combination of units to be maintained, the objective function is constructed with the minimum ratio of total maintenance cost to total power generation as the optimization target, the constraint conditions are constructed according to the maintenance resource constraint conditions, and the actual maintenance strategy optimization model is established in combination with the objective function and the constraint conditions, and the optimal actual maintenance strategy of the target unit to be maintained is solved to complete the optimized maintenance strategy, so that the optimal preventive maintenance strategy of each wind turbine unit and the optimal actual maintenance strategy of the target unit to be maintained can be combined in the future to perform composite operation and maintenance on the corresponding wind turbine unit. Compared with the prior art, the embodiment of the present invention predicts the operating status by considering the monitoring data of each wind turbine unit in the offshore wind farm, optimizes the maintenance strategy according to the predicted operating status of each wind turbine unit, obtains the optimal preventive maintenance strategy of each wind turbine unit, and considers the current operating status of each wind turbine unit and the actual maintenance resource constraint conditions, selects the target combination of units to be maintained and optimizes its maintenance strategy, and obtains the optimal actual maintenance strategy of the target unit to be maintained, thereby realizing preventive maintenance and actual maintenance of each wind turbine unit in the offshore wind farm. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A schematic flow chart of a method for optimizing a composite maintenance strategy for wind turbines in an offshore wind farm in the first embodiment of the present invention;

[0036] Figure 2 A data flow diagram of a particle swarm multi-objective optimization prediction algorithm exemplified in the first embodiment of the present invention;

[0037] Figure 3 It is a structural schematic diagram of a composite maintenance strategy optimization device for wind turbines in an offshore wind farm in the second embodiment of the present invention. DETAILED DESCRIPTION

[0038] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0039] It should be noted that the step numbers in the text are only for the convenience of explaining the specific embodiment and do not limit the order of execution of the steps. The method provided in this embodiment can be executed by a related terminal device, and the following description is taken as an example of a processor as the execution subject.

[0040] like Figure 1 As shown, the first embodiment provides a method for optimizing a composite maintenance strategy for wind turbines in an offshore wind farm, including steps S1 to S3:

[0041] S1. For each wind turbine, the current operating state of the wind turbine is evaluated according to the monitoring data of the wind turbine, and the operating state is predicted in combination with the historical operating state of the wind turbine to obtain the predicted operating state of the wind turbine;

[0042] S2, taking the predicted operating status of each wind turbine as the input parameter of the pre-established preventive maintenance strategy optimization model, making the preventive maintenance strategy optimization model take the lowest discount cost as the optimization goal, correspondingly optimizing the maintenance method and monitoring cycle of each wind turbine, and obtaining the optimal preventive maintenance strategy for the wind turbine;

[0043] S3. Classify all wind turbines according to their current operating status to obtain a number of combinations of turbines to be maintained, and cluster all combinations of turbines to be maintained according to predefined maintenance resource constraints to obtain a number of target combinations of turbines to be maintained;

[0044] S4. For each target combination of units to be maintained, an objective function is constructed with the minimum ratio of total maintenance cost to total power generation as the optimization goal, constraints are constructed according to maintenance resource restrictions, and an actual maintenance strategy optimization model is established by combining the objective function and constraints to obtain the optimal actual maintenance strategy for the target units to be maintained.

[0045] As an example, in step S1, for each wind turbine, the wind turbine is divided into several key components, generally 5 to 6, and the monitoring data of the wind turbine is obtained, including gearbox oil temperature, vibration signal, generator temperature, blade noise, blade stress, control panel temperature, pitch angle, etc. According to the monitoring data of the wind turbine, the function of each component is evaluated to evaluate the current operating state of the wind turbine. The operating state is predicted by combining the current operating state and the historical operating state of the wind turbine to obtain the predicted operating state of the wind turbine, thereby obtaining the predicted operating state of each wind turbine.

[0046] In step S2, a preventive maintenance strategy optimization model is established in advance, and the predicted operating status of each wind turbine is used as an input parameter of the preventive maintenance strategy optimization model. The preventive maintenance strategy optimization model takes the lowest discount cost as the optimization goal, and correspondingly optimizes the maintenance method and monitoring cycle of each wind turbine to obtain the optimal preventive maintenance strategy for the wind turbine, thereby obtaining the optimal preventive maintenance strategy for each wind turbine.

[0047] In step S3, according to the current operating status of all wind turbines, it is determined whether each wind turbine needs to be maintained, and according to the maintenance requirements of all wind turbines, all wind turbines are classified to obtain a number of combinations of units to be maintained. Based on the actual limited maintenance resources, maintenance resource constraints are predefined, and according to the predefined maintenance resource constraints, all combinations of units to be maintained are clustered to obtain a number of target combinations of units to be maintained, so as to reasonably use limited maintenance resources to maintain the generators with maintenance requirements.

[0048] In step S4, for each target combination of units to be maintained, an objective function is constructed with the minimum ratio of total maintenance cost to total power generation as the optimization target, constraints are constructed according to maintenance resource restrictions, and an actual maintenance strategy optimization model is established in combination with the objective function and constraints. The actual maintenance strategy optimization model is solved to obtain the optimal actual maintenance strategy for the target units to be maintained, thereby obtaining the optimal actual maintenance strategy for each target unit to be maintained.

[0049] This embodiment predicts the operating status by considering the monitoring data of each wind turbine unit in the offshore wind farm, optimizes the maintenance strategy according to the predicted operating status of each wind turbine unit, and obtains the optimal preventive maintenance strategy for each wind turbine unit. In addition, considering the current operating status of each wind turbine unit and the actual maintenance resource constraints, a target combination of units to be maintained is selected and its maintenance strategy is optimized to obtain the optimal actual maintenance strategy for the target unit to be maintained, thereby achieving preventive maintenance and actual maintenance of each wind turbine unit in the offshore wind farm.

[0050] In a preferred embodiment, the current operating state of the wind turbine set is evaluated based on the monitoring data of the wind turbine set, specifically: the current operating state of the wind turbine set is evaluated using a fuzzy comprehensive evaluation method based on the monitoring data of the wind turbine set.

[0051] As an example, in the process of evaluating the status of each component function based on the monitoring data of the wind turbine, considering that there are many evaluation factors and they are fuzzy and random, a fuzzy comprehensive evaluation method can be used to evaluate the current operating status of the wind turbine.

[0052] This embodiment uses a fuzzy comprehensive evaluation method to evaluate the current operating state of the wind turbine generator set, which is conducive to quickly and accurately evaluating the current operating state of the wind turbine generator set.

[0053] In a preferred embodiment, before predicting the operating status in combination with the historical operating status of the wind turbine to obtain the predicted operating status of the wind turbine, it also includes: determining the historical operating status of the wind turbine according to the historical monitoring data of the wind turbine or the n-state left-right no-span model.

[0054] In a preferred embodiment, the operating state prediction is performed in combination with the historical operating state of the wind turbine to obtain the predicted operating state of the wind turbine, specifically: the current operating state and the historical operating state of the wind turbine are combined to obtain the transition probability of the wind turbine between different operating states, and the transition probabilities of the wind turbine between different operating states are sorted in descending order to obtain the predicted operating state sequence of the wind turbine; for each operating state change pair in the predicted operating state sequence of the wind turbine, the target monitoring data associated with the operating state change pair is used as the multi-input variable of the pre-established particle swarm multi-objective optimization prediction model, so as to evaluate the predicted operating state of the wind turbine according to the adjustment amount of the target monitoring data output by the particle swarm multi-objective optimization prediction model.

[0055] As an example, the historical operating state of the wind turbine is determined based on the historical monitoring data of the wind turbine or the n-state left-right no-span model. The transition probability of the wind turbine between different operating states due to performance degradation is obtained by combining the current operating state and the historical operating state of the wind turbine.

[0056] The transition probabilities of the wind turbines between different operating states are sorted in descending order of the transition probabilities to obtain a predicted operating state sequence of the wind turbines, wherein the predicted operating sequence of the wind turbines includes a plurality of operating state change pairs.

[0057] For each operating state change pair in the predicted operating state sequence of the wind turbine, the target monitoring data associated with the operating state change pair, that is, the monitoring data that causes the change in the operating state of the wind turbine, is used as the multi-input variable of the pre-established particle swarm multi-objective optimization prediction model, so as to evaluate the predicted operating state of the wind turbine according to the adjustment amount of the target monitoring data output by the particle swarm multi-objective optimization prediction model.

[0058] The data flow diagram of the particle swarm multi-objective optimization prediction algorithm is as follows: Figure 2 As shown in FIG. 1 , the process of the particle swarm multi-objective optimization prediction model obtaining the adjustment amount of the target monitoring data according to the target monitoring data is as follows:

[0059] 1. Input the target monitoring data as multiple input variables into the multi-objective optimization module, and the multi-objective optimization module randomly generates multiple multi-dimensional particles, each of which corresponds to the adjustment amount of a set of target monitoring data of the wind turbine in the current operating state;

[0060] 2. Perform an iterative search on all multidimensional particles to obtain the value of each multidimensional particle;

[0061] 3. According to the objective function of multi-objective optimization prediction, the pros and cons of the values ​​of each multi-dimensional particle are judged, and the values ​​of each multi-dimensional particle are taken as their own historical optimal values, and all historical optimal values ​​are combined into a non-inferior solution set;

[0062] 4. After each iteration, each multidimensional particle automatically updates its own parameters according to its own historical optimal value and the global optimal value of the entire particle swarm;

[0063] 5. Repeat operations 2-3 for each multidimensional particle after the updated parameters until all multidimensional particles converge to the optimal position, thereby obtaining the adjustment amount of the target monitoring data;

[0064] 6. Based on the adjustment amount of target monitoring data, a preventive maintenance strategy optimization model is established for the key components of wind turbines.

[0065] Among them, operation 2 specifically includes:

[0066] 21. Input each multidimensional particle into the wind turbine power generation process model, which is established by offline learning of the historical operation data of the wind turbine in the full sample space;

[0067] 22. Through Bayesian reasoning, the adjustment probability distribution of the target monitoring data is obtained. The target monitoring data includes adjustable monitoring data such as gearbox oil temperature, generator temperature, control panel temperature, and pitch angle.

[0068] The objective function of the multi-objective optimization prediction in operation 3 is:

[0069]

[0070] f:X→R k ,f max (x) = η(x), f min (x) = (gearbox oil temperature, generator temperature, control panel temperature, pitch angle) T (1);

[0071] In formula (1), η represents the power generation efficiency of the wind turbine corresponding to the multidimensional particle, and the optimal value is a set of target monitoring data that minimizes the objective function value within the search range.

[0072] The objective function of the multi-objective optimization prediction also includes seeking the optimum between reducing the gearbox oil temperature, generator temperature, control panel temperature and pitch angle and improving the power generation efficiency, so as to achieve the highest power generation efficiency and the maximum reduction of the gearbox oil temperature, generator temperature, control panel temperature and pitch angle.

[0073] The objective function of the multi-objective optimization prediction in operation 3 is a wind turbine power generation process model based on a Bayesian network. Through Bayesian reasoning, the corresponding wind turbine operating state can be derived from the adjustment amount of each set of target monitoring data, so as to select the adjustment amount of the target monitoring data corresponding to the optimal operating state of the wind turbine.

[0074] Operation 6 specifically includes: judging whether the current wind turbine is operating in a steady-state load condition or a variable load condition, and switching the control loop according to the load judgment result, that is, dynamically changing the weights and adjustment modes of multiple optimization objectives in the non-inferior solution set, so that multiple optimization objectives have variable load adaptive optimization characteristics. Under stable load conditions, the preventive maintenance logic focuses on improving the power generation efficiency of the wind turbine, while under variable load conditions, the preventive maintenance logic focuses on reducing the temperature of key components of the wind turbine, and quantified expert knowledge is embedded in the preventive maintenance logic as an optimization constraint.

[0075] This embodiment uses a particle swarm multi-objective optimization prediction algorithm to predict the operating state, which is conducive to quickly and accurately obtaining the predicted operating state of the wind turbine.

[0076] In a preferred embodiment, all wind turbines are classified according to their current operating status to obtain several combinations of turbines to be maintained, specifically: the turbines among all wind turbines whose current operating status meets predefined turbine maintenance conditions are classified as turbines to be maintained to obtain several combinations of turbines to be maintained; all turbines to be maintained in the same area are classified into the same combination of turbines to be maintained to obtain several combinations of turbines to be maintained.

[0077] In a preferred embodiment, all the combinations of units to be maintained are clustered according to the predefined maintenance resource constraint conditions to obtain a number of target combinations of units to be maintained, specifically: according to the maintenance vessel quantity constraint conditions in the maintenance resource constraint conditions, the available number of maintenance vessels is determined as the number of partitions; based on the number of partitions, all the combinations of units to be maintained are clustered to obtain a number of target combinations of units to be maintained.

[0078] As an example, according to the actual wind turbine operation standard, the wind turbine maintenance conditions are predefined, and it is determined whether the current operation state of each wind turbine meets the generator maintenance condition. The generators whose current operation states meet the predefined generator maintenance conditions among all wind turbines are taken as generators to be maintained, and several generators to be maintained are obtained. The region to which each generator to be maintained belongs is obtained, and all generators to be maintained belonging to the same region are divided into the same generator combination to be maintained, and several generator combinations to be maintained are obtained.

[0079] Based on the actual limited maintenance resources, maintenance resource constraints are defined in advance, and it is determined whether all combinations of units to be maintained can be maintained at the same time based on the existing maintenance resources. If it is determined that all combinations of units to be maintained can be maintained at the same time, then each combination of units to be maintained is directly used as a target combination of units to be maintained. Otherwise, all combinations of units to be maintained need to be clustered according to the maintenance resource constraints to obtain several target combinations of units to be maintained.

[0080] It is understandable that for an offshore wind farm of a certain size, the allocation of maintenance resources is often limited. Considering the limited maintenance resources of the offshore wind farm, clustering partitioning is performed on the basis of the initial partitioning of all wind turbines.

[0081] Assuming that the number of maintenance vessels that can simultaneously carry out maintenance on an offshore wind farm is m, and the number of maintenance personnel p ≥ m, consider the following situations:

[0082] 1. If the number of generators in alarm or delayed maintenance is equal to the number of available maintenance vessels, that is, k = m, maintenance work can be carried out simultaneously, and the generators in alarm or delayed maintenance are given priority.

[0083] 2. If the number of alarmed or delayed maintenance units is greater than the number of available maintenance vessels, that is, k>m, due to the limitation of maintenance resources, the k partitions initially divided in the first stage cannot be maintained at the same time. Therefore, the hierarchical clustering method is adopted from top to bottom, with the number of available maintenance vessels m as the number of partitions and the distance between partitions as the class distance, to merge the k partitions; the class distance adopts the WARD distance, that is, the sum of squared deviations. This method adopts the idea of ​​variance analysis. During classification, the distance within the class can be made as small as possible, while the distance between classes can be made as large as possible, so as to realize the reasonable partitioning of the units to be maintained under the constraint of maintenance resources.

[0084] Specifically, N i Represents the i-th group G i The number of wind turbines in, i = 1, 2, ..., k, X i,rs For the i-th group G i The distance from the rth wind turbine group to the sth wind turbine group, r, s = 1, 2, ..., N i , when r = s, X i,rs =0,X i is the average distance within the i-th group, then the sum of squares of deviations within the i-th group is S i =0.5ΣΣ(X i,rs -X i ) 2 , r, s = 1, 2, ..., N i If group G i and Group G j Merge into new class G r , then G i and G j The sum of squares of the differences between 2 ij =S r -S i -S j .

[0085] Therefore, the implementation process of the second stage cluster partitioning is as follows: based on the k partitions in the first stage, according to D 2 ij Select the adjacent subclass G p and G q , so that the WARD distance D between classes 2 pq =min(D 2 ij ); subclass G p and G q Merge into new class G r , calculate the intra-class WARD distance S r ; Calculate the new class G r The inter-class WARD distance with other subclasses is updated; the number of classes is reduced to k-1, until the number of partitions k=m+1; sorting is performed according to the inter-class WARD distance, merging new classes, and forming a candidate set {O 1 , O 2 , …, O h}, the number of elements in the set h = m(m+1) / 2, the number of elements in the set O h In order to sort by the WARD distance between classes, n units to be maintained are divided into different combinations of k = m classes O h = {G h1 , G h2, …, G hm}; Reduce the classification tree to m - 1, calculate the within-class distance, update the between-class distance, sort according to the between-class WARD distance, merge new classes again, divide the wind turbines, divide the n maintenance units into different combinations of k = m - 1 classes to form a candidate set, and loop through the above operations until m = 1.

[0086] 3. If the number of alarmed or deferred maintenance units is less than the number of available maintenance vessels, i.e., k < m, then use k units as the condensation points of the initial classes, combine with the number of available maintenance vessels m, re-cluster to form a candidate set {O m , O m-1 , …, O k}. The element O m in the set is the different combination O m = {G m1 , G m2 , …, G mm} obtained by dividing the nth unit to be maintained into k = m classes.

[0087] In this embodiment, by performing two-stage partitioning, namely initial partitioning and clustering partitioning, on all wind turbines, it is beneficial to ensure that the actual implementation of the maintenance strategy is not restricted by the actual maintenance resources, and improve the implementation effect of the actual maintenance strategy.

[0088] In a preferred embodiment, the objective function is:

[0089] minZ = (CC + CL + CD) / Wf (2);

[0090] where CC is the total cost of maintenance personnel scheduling, maintenance vessel operation, and maintenance path traffic, CL is the maintenance downtime loss, CD is the deferred maintenance penalty cost, and Wf is the total power generation.

[0091] In a preferred embodiment, the constraint conditions include at least one of the constraint conditions of the number of maintenance vessels, the load capacity constraint of maintenance vessels, the number of maintenance personnel constraint, the maintenance time sequence constraint of units, the daily effective working hours constraint of maintenance vessels, the maintenance downtime constraint, the deferred maintenance time constraint, the number of units maintained by a maintenance vessel per day constraint, and the maintenance route constraint.

[0092] As an example, for each combination of target units to be maintained, an objective function is constructed with the minimum ratio of the total maintenance cost to the total power generation as the optimization goal. The objective function is:

[0093] minZ = (CC + CL + CD) / Wf (2);

[0094] In formula (2), CC is the total cost of maintenance personnel scheduling, maintenance vessel operation, and maintenance path transportation, CL is the maintenance downtime loss, CD is the delayed maintenance penalty cost, and Wf is the total power generation.

[0095] Constraints are constructed according to maintenance resource constraints, including at least one of a maintenance vessel quantity constraint, a maintenance vessel load constraint, a maintenance personnel quantity constraint, a unit maintenance sequence constraint, a maintenance vessel daily effective working time constraint, a maintenance downtime constraint, a delayed maintenance time constraint, a maintenance vessel daily unit maintenance times constraint, and a maintenance route constraint.

[0096] The actual maintenance strategy optimization model is established by combining the objective function and constraint conditions, and the actual maintenance strategy optimization model is solved to obtain the optimal actual maintenance strategy for the target unit to be maintained.

[0097] Specifically, considering the current operating status of wind turbines and the maintenance resources of offshore wind farms, combined with the results of cluster analysis, the initial group of offshore wind farm zone maintenance units {G n1 , G n2 , …, G nm}, G nm Indicates that the mth maintenance partition contains a cluster of units to be maintained, G nm = {N m1 , N m2 ,…N mb , N m(b+1) ,…,N ma},N ma Indicates the unit number of the ath unit to be maintained in the cluster. The cluster contains b (b<=a) alarm units and (ab) planned maintenance units.

[0098] In order to reduce downtime losses, priority is given to maintaining alarm units, and then considering planned maintenance units, forming an initial maintenance plan for the unit cluster with the shortest maintenance path.

[0099] Among them, the solution algorithm includes two categories: exact algorithm and approximate algorithm. The computational cost of the exact algorithm increases exponentially with the expansion of the problem scale, while the approximate algorithm can achieve higher solution speed and stability. In this embodiment, the approximate algorithm adopts a saving algorithm, which has the advantage of fast operation speed and can solve large-scale optimization combination problems. The basic idea is to merge the two loops in the transportation problem into one loop in turn under the constraint conditions, so that the total distance after each merger is reduced to the maximum extent until all requirements are met. The loops include two categories. One is an initialization line of OiO formed by connecting the port with each unit to be maintained separately. The formal distance of the i-th line is S i =S 0i +Si0 The second type is to connect the unit i to be maintained and the unit j together to form a line OijO, and the path saving is S(i,j)=S 0i +S 0j -S ij , the larger the value, the more distance is saved when connecting unit i and unit j together. The saved distance of each unit is calculated to form the data S(i,j); the initial actual maintenance strategy set P of the cluster of units to be maintained is formed by repeated solution, where the set P is represented by a matrix, and the order of each row is the initial maintenance order; the total load, number of people, and time of each initial maintenance strategy p in the set are calculated according to the constraints, and the effective actual maintenance strategy set P' is screened out; the effective actual maintenance strategy set P' is further screened by the objective function Z to obtain the optimal actual maintenance strategy P".

[0100] This embodiment constructs and solves an actual maintenance strategy optimization model, comprehensively considers the current operating status of each wind turbine and the actual maintenance resource constraints, optimizes the maintenance strategy of the target combination of units to be maintained, and can achieve actual maintenance of each wind turbine in an offshore wind farm.

[0101] Based on the same inventive concept as the first embodiment, the second embodiment provides Figure 3 The device for optimizing the composite maintenance strategy of wind turbines in an offshore wind farm shown in the figure comprises: a unit operation state prediction module 21, which is used for evaluating the current operation state of each wind turbine according to the monitoring data of the wind turbine, and predicting the operation state in combination with the historical operation state of the wind turbine to obtain the predicted operation state of the wind turbine; a preventive maintenance strategy optimization module 22, which is used for taking the predicted operation state of each wind turbine as the input parameter of a pre-established preventive maintenance strategy optimization model, so that the preventive maintenance strategy optimization model takes the lowest discount cost as the optimization goal, and correspondingly optimizes the maintenance method and monitoring cycle of each wind turbine to obtain the optimal preventive maintenance of the wind turbine. Maintenance strategy; a module for classifying wind turbines to be maintained 23, which is used to classify all wind turbines according to their current operating status to obtain a number of combinations of wind turbines to be maintained, and to cluster all combinations of wind turbines to be maintained according to predefined maintenance resource constraints to obtain a number of target combinations of wind turbines to be maintained; an actual maintenance strategy optimization module 24, which is used to construct an objective function for each target combination of wind turbines to be maintained with the ratio of total maintenance cost to total power generation being minimized as the optimization objective, to construct constraints according to maintenance resource constraints, to establish an actual maintenance strategy optimization model in combination with the objective function and constraints, and to solve and obtain the optimal actual maintenance strategy for the target combination of wind turbines to be maintained.

[0102] In a preferred embodiment, the current operating state of the wind turbine set is evaluated based on the monitoring data of the wind turbine set, specifically: the current operating state of the wind turbine set is evaluated using a fuzzy comprehensive evaluation method based on the monitoring data of the wind turbine set.

[0103] In a preferred embodiment, before predicting the operating status in combination with the historical operating status of the wind turbine to obtain the predicted operating status of the wind turbine, it also includes: determining the historical operating status of the wind turbine according to the historical monitoring data of the wind turbine or the n-state left-right no-span model.

[0104] In a preferred embodiment, the operating state prediction is performed in combination with the historical operating state of the wind turbine to obtain the predicted operating state of the wind turbine, specifically: the current operating state and the historical operating state of the wind turbine are combined to obtain the transition probability of the wind turbine between different operating states, and the transition probabilities of the wind turbine between different operating states are sorted in descending order to obtain the predicted operating state sequence of the wind turbine; for each operating state change pair in the predicted operating state sequence of the wind turbine, the target monitoring data associated with the operating state change pair is used as the multi-input variable of the pre-established particle swarm multi-objective optimization prediction model, so as to evaluate the predicted operating state of the wind turbine according to the adjustment amount of the target monitoring data output by the particle swarm multi-objective optimization prediction model.

[0105] In a preferred embodiment, all wind turbines are classified according to their current operating status to obtain several combinations of turbines to be maintained, specifically: the turbines among all wind turbines whose current operating status meets predefined turbine maintenance conditions are classified as turbines to be maintained to obtain several combinations of turbines to be maintained; all turbines to be maintained in the same area are classified into the same combination of turbines to be maintained to obtain several combinations of turbines to be maintained.

[0106] In a preferred embodiment, all the combinations of units to be maintained are clustered according to the predefined maintenance resource constraint conditions to obtain a number of target combinations of units to be maintained, specifically: according to the maintenance vessel quantity constraint conditions in the maintenance resource constraint conditions, the available number of maintenance vessels is determined as the number of partitions; based on the number of partitions, all the combinations of units to be maintained are clustered to obtain a number of target combinations of units to be maintained.

[0107] In a preferred embodiment, the objective function is:

[0108] minZ = (CC + CL + CD) / Wf (3);

[0109] Among them, CC is the total cost of maintenance personnel scheduling, maintenance ship operation, and maintenance path transportation, CL is the maintenance downtime loss, CD is the delayed maintenance penalty cost, and Wf is the total power generation.

[0110] In a preferred embodiment, the constraints include at least one of a maintenance vessel quantity constraint, a maintenance vessel load constraint, a maintenance personnel quantity constraint, a unit maintenance sequence constraint, a maintenance vessel daily effective working time constraint, a maintenance downtime constraint, a delayed maintenance time constraint, a maintenance vessel daily unit maintenance number constraint, and a maintenance route constraint.

[0111] Based on the same inventive concept as the first embodiment, the third embodiment provides a device for optimizing a composite maintenance strategy for wind turbines in an offshore wind farm, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. The memory is coupled to the processor, and when the processor executes the computer program, it implements the method for optimizing a composite maintenance strategy for wind turbines in an offshore wind farm as described in the first embodiment, and can achieve the same beneficial effects as the first embodiment.

[0112] In summary, the implementation of the embodiments of the present invention has the following beneficial effects:

[0113] By evaluating the current operating status of each wind turbine according to the monitoring data of the wind turbine, and predicting the operating status in combination with the historical operating status of the wind turbine, the predicted operating status of the wind turbine is obtained; the predicted operating status of each wind turbine is used as the input parameter of the pre-established preventive maintenance strategy optimization model, so that the preventive maintenance strategy optimization model takes the lowest discount cost as the optimization goal, and correspondingly optimizes the maintenance method and monitoring cycle of each wind turbine to obtain the optimal preventive maintenance strategy for the wind turbine; according to the current operating status of all wind turbines, all wind turbines are classified to obtain a number of combinations of units to be maintained. And according to the predefined maintenance resource constraint conditions, all combinations of units to be maintained are clustered to obtain several target combinations of units to be maintained; for each target combination of units to be maintained, the objective function is constructed with the minimum ratio of total maintenance cost to total power generation as the optimization target, the constraint conditions are constructed according to the maintenance resource constraint conditions, and the actual maintenance strategy optimization model is established in combination with the objective function and the constraint conditions, and the optimal actual maintenance strategy of the target unit to be maintained is solved to complete the optimized maintenance strategy, so that the optimal preventive maintenance strategy of each wind turbine unit and the optimal actual maintenance strategy of the target unit to be maintained can be combined in the future to perform composite operation and maintenance on the corresponding wind turbine unit. The embodiment of the present invention predicts the operating status by considering the monitoring data of each wind turbine unit in the offshore wind farm, optimizes the maintenance strategy according to the predicted operating status of each wind turbine unit, obtains the optimal preventive maintenance strategy of each wind turbine unit, and considers the current operating status of each wind turbine unit and the actual maintenance resource constraint conditions, selects the target combination of units to be maintained and optimizes its maintenance strategy, and obtains the optimal actual maintenance strategy of the target unit to be maintained, thereby realizing preventive maintenance and actual maintenance of each wind turbine unit in the offshore wind farm.

[0114] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

[0115] Those skilled in the art can understand that all or part of the processes in the above embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes in the above embodiments. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

Claims

1. A composite maintenance strategy optimization method for wind turbines in offshore wind farms. It is characterized in that include: For each wind turbine set, based on the monitoring data of the wind turbine set, the current operating state of the wind turbine set is evaluated, and the operating state is predicted in combination with the historical operating state of the wind turbine set to obtain the predicted operating state of the wind turbine set; The predicted operating state of each wind turbine is used as an input parameter of a pre-established preventive maintenance strategy optimization model, so that the preventive maintenance strategy optimization model takes the lowest discount cost as the optimization goal, and correspondingly optimizes the maintenance method and monitoring cycle of each wind turbine to obtain the optimal preventive maintenance strategy for the wind turbine; According to the current operating status of all the wind turbines, all the wind turbines are classified to obtain a plurality of combinations of turbines to be maintained, and according to the predefined maintenance resource restriction condition, all the combinations of turbines to be maintained are clustered to obtain a plurality of target combinations of turbines to be maintained; For each target combination of units to be maintained, an objective function is constructed with the minimum ratio of total maintenance cost to total power generation as the optimization target, constraints are constructed according to the maintenance resource restriction conditions, and an actual maintenance strategy optimization model is established by combining the objective function and the constraints to obtain the optimal actual maintenance strategy for the target units to be maintained; The operation state prediction is performed in combination with the historical operation state of the wind turbine generator set to obtain the predicted operation state of the wind turbine generator set, specifically: Combining the current operating state and the historical operating state of the wind turbine generator set, obtaining the transition probability of the wind turbine generator set between different operating states, and sorting the transition probabilities of the wind turbine generator set between different operating states in descending order to obtain a predicted operating state sequence of the wind turbine generator set; For each operating state change pair in the predicted operating state sequence of the wind turbine generator set, the target monitoring data associated with the operating state change pair is used as a multi-input variable of a pre-established particle swarm multi-objective optimization prediction model, so as to evaluate the predicted operating state of the wind turbine generator set according to the adjustment amount of the target monitoring data output by the particle swarm multi-objective optimization prediction model.

2. The offshore wind farm wind turbine complex maintenance strategy optimization method according to claim 1, It is characterized in that The evaluating the current operating state of the wind turbine generator set according to the monitoring data of the wind turbine generator set is specifically: According to the monitoring data of the wind turbine generator set, a fuzzy comprehensive evaluation method is adopted to evaluate the current operating status of the wind turbine generator set.

3. The composite maintenance strategy optimization method for wind turbines in an offshore wind farm according to claim 1, It is characterized in that Before performing operating state prediction in combination with the historical operating state of the wind turbine generator set to obtain the predicted operating state of the wind turbine generator set, the method further includes: The historical operating status of the wind turbine generator set is determined according to the historical monitoring data of the wind turbine generator set.

4. The offshore wind farm wind turbine complex maintenance strategy optimization method according to claim 1, It is characterized in that According to the current operating status of all the wind turbines, all the wind turbines are classified to obtain a plurality of wind turbine combinations to be maintained, specifically: The generator sets whose current operating statuses satisfy the predefined generator set maintenance conditions among all the wind turbine sets are taken as generator sets to be maintained, thereby obtaining a plurality of generator sets to be maintained; All the units to be maintained belonging to the same area are divided into the same combination of units to be maintained, so as to obtain a plurality of combinations of units to be maintained.

5. The offshore wind farm wind turbine complex maintenance strategy optimization method according to claim 1, It is characterized in that According to the predefined maintenance resource restriction condition, all the combinations of units to be maintained are clustered to obtain several target combinations of units to be maintained, specifically: According to the maintenance vessel quantity restriction condition in the maintenance resource restriction condition, determining the available quantity of maintenance vessels as the partition quantity; Based on the number of partitions, all the combinations of units to be maintained are clustered to obtain a plurality of target combinations of units to be maintained.

6. The offshore wind farm wind turbine complex maintenance strategy optimization method according to claim 1, It is characterized in that The objective function is: minZ = (CC + CL + CD) / Wf; Among them, CC is the total cost of maintenance personnel scheduling, maintenance ship operation, and maintenance path traffic, CL is the maintenance downtime loss, CD is the delayed maintenance penalty fee, and Wf is the total power generation.

7. The offshore wind farm wind turbine complex maintenance strategy optimization method according to claim 1, It is characterized in that The constraints include at least one of a maintenance vessel quantity constraint, a maintenance vessel load constraint, a maintenance personnel quantity constraint, a unit maintenance sequence constraint, a maintenance vessel daily effective working time constraint, a maintenance downtime constraint, a delayed maintenance time constraint, a maintenance vessel daily unit maintenance times constraint, and a maintenance route constraint.

8. A composite maintenance strategy optimization device for wind turbines in offshore wind farms, It is characterized in that include: A unit operation state prediction module is used for evaluating the current operation state of each wind turbine unit according to the monitoring data of the wind turbine unit, and performing operation state prediction in combination with the historical operation state of the wind turbine unit to obtain the predicted operation state of the wind turbine unit; A preventive maintenance strategy optimization module is used to use the predicted operating state of each wind turbine as an input parameter of a pre-established preventive maintenance strategy optimization model, so that the preventive maintenance strategy optimization model takes the lowest discount cost as the optimization goal, and correspondingly optimizes the maintenance method and monitoring cycle of each wind turbine to obtain the optimal preventive maintenance strategy for the wind turbine; A unit classification module for maintenance, used to classify all the wind turbines according to their current operating states to obtain a plurality of combinations of units to be maintained, and to cluster all the combinations of units to be maintained according to predefined maintenance resource constraints to obtain a plurality of target combinations of units to be maintained; An actual maintenance strategy optimization module is used to construct an objective function for each target combination of units to be maintained, taking the minimum ratio of total maintenance cost to total power generation as the optimization objective, construct constraint conditions according to the maintenance resource restriction conditions, and establish an actual maintenance strategy optimization model in combination with the objective function and the constraint conditions to solve and obtain the optimal actual maintenance strategy for the target units to be maintained; The operation state prediction is performed in combination with the historical operation state of the wind turbine generator set to obtain the predicted operation state of the wind turbine generator set, specifically: Combining the current operating state and the historical operating state of the wind turbine generator set, obtaining the transition probability of the wind turbine generator set between different operating states, and sorting the transition probabilities of the wind turbine generator set between different operating states in descending order to obtain a predicted operating state sequence of the wind turbine generator set; For each operating state change pair in the predicted operating state sequence of the wind turbine generator set, the target monitoring data associated with the operating state change pair is used as a multi-input variable of a pre-established particle swarm multi-objective optimization prediction model, so as to evaluate the predicted operating state of the wind turbine generator set according to the adjustment amount of the target monitoring data output by the particle swarm multi-objective optimization prediction model.

9. A composite maintenance strategy optimization device for wind turbines in offshore wind farms, It is characterized in that It comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, the memory is coupled to the processor, and when the processor executes the computer program, it implements the method for optimizing the composite maintenance strategy of wind turbines in an offshore wind farm as described in any one of claims 1 to 7.

Images