An offshore wind farm opportunity maintenance and path coordination optimization method for extreme weather

Abstract

The application discloses an offshore wind farm opportunity maintenance and path collaborative optimization method for extreme weather, comprising: environment perception and physical state extrapolation, updating the degradation state of unit components based on weather forecast and physical evolution model; weather capacity pre-play and preemptive evaluation, evaluating future maintenance capacity through rolling simulation, identifying weather lockout risk and calculating postponement penalty cost; hierarchical scheduling planning of air-sea collaboration, using a heuristic optimization algorithm to preferentially construct the skeleton route of the water surface operation and maintenance carrier for emergency tasks, and dynamically optimizing the opportunity maintenance task based on marginal cost and inserting it into the route; active perception and closed-loop dynamic re-planning, scheduling the air patrol carrier to patrol the high-uncertainty unit, using the state posterior update rule to correct the cognitive bias, and triggering local re-planning when encountering sudden risks. The application collaboratively optimizes air-sea multi-resources and maintenance opportunity, significantly improves the operation and maintenance robustness under extreme weather, and reduces the total system cost.

Description

Technical Field

[0001] This invention belongs to the field of offshore wind farm operation and maintenance technology, specifically relating to an opportunistic maintenance and path collaborative optimization method for offshore wind farms in extreme weather conditions. Background Technology

[0002] As offshore wind power expands into deeper waters, the operating environment for wind turbines is becoming increasingly harsh. Unlike onshore wind power, offshore wind farms face far more severe challenges in operation and maintenance scheduling:

[0003] First, there is the issue of accessibility limitations due to extreme weather. Maritime weather is non-stationary and time-varying, with frequent extreme weather events such as typhoons and cold waves. Maintenance vessels are subject to strict wave resistance thresholds, resulting in the complete closure of maintenance windows during storms. Existing maintenance strategies often lack accurate forecasting of weather windows, easily leading to a supply-demand mismatch where there are no maintenance plans when sea conditions are good, but when equipment malfunctions requiring urgent repairs, they cannot go to sea due to weather closures.

[0004] Second, there is the issue of "cognitive-physical" state bias. The operations and maintenance center cannot monitor the actual physical state of the internal components of offshore wind turbines in real time, relying primarily on SCADA data and physical models for state estimation. However, due to individual component differences and random damage at the microscopic level, the deviation between model estimates and the actual physical state tends to widen as the unobserved time increases. This "cognitive bias" can easily lead to misjudgments of maintenance timing, resulting in missed maintenance windows or premature maintenance waste.

[0005] Third, there is a disconnect between resource scheduling and path planning. Existing O&M research often separates maintenance decisions from logistics scheduling. Furthermore, traditional corrective maintenance or condition-based maintenance based on fixed thresholds lacks system-level coordination, making it difficult to achieve optimal long-term O&M costs in dynamically changing marine environments.

[0006] Therefore, there is an urgent need for an operation and maintenance method that can effectively integrate physical models and active perception in extreme weather conditions, and collaboratively optimize maintenance timing and navigation routes. Summary of the Invention

[0007] To address the problems existing in the prior art, this invention proposes a method for opportunistic maintenance and path coordination optimization for offshore wind farms in extreme weather conditions. This method aims to solve the problems of divergent cognitive biases in component degradation status, complex failure modes affected by extreme weather, and lack of coordination in the scheduling of sea, land, and air resources under dynamic weather windows in the existing offshore wind power operation and maintenance.

[0008] To achieve the above-mentioned technical objectives, the present invention provides the following technical solution:

[0009] A method for coordinated optimization of opportunistic maintenance and path selection in offshore wind farms under extreme weather conditions includes the following steps:

[0010] Based on the physical evolution model of the turbine components, combined with real-time monitoring data and meteorological forecast data obtained from offshore wind farms, the estimated degradation status of each turbine component in the offshore wind farm is updated, and it is determined whether each turbine component has experienced functional failure.

[0011] Based on rolling time-domain simulation, the maintenance needs of each unit component within a future set time window are predicted. Combined with meteorological forecast data, the maximum operational capacity of the water surface maintenance vehicle is determined. Based on supply and demand matching analysis, preemptive scheduling is carried out to pre-position emergency repair tasks. Based on the game rule of the remaining life wastage value and the expected failure risk cost, the delay penalty cost of each potential maintenance task is calculated, and opportunistic maintenance decisions are executed.

[0012] A hierarchical collaborative optimization scheduling model is constructed to generate an initial operation and maintenance scheduling scheme. The hierarchical collaborative optimization scheduling model aims to minimize the total operation and maintenance cost within the simulation period. The first layer adopts a heuristic optimization algorithm to prioritize the construction of the skeleton route of the water surface operation and maintenance vehicle for units with emergency maintenance tasks. The second layer is based on the marginal cost-benefit evaluation rule to dynamically optimize and insert non-emergency opportunity maintenance tasks into the skeleton route.

[0013] According to the initial operation and maintenance scheduling plan, based on the dynamic priority scoring strategy, the aerial inspection vehicle is scheduled to inspect the units whose dynamic priority scores exceed the threshold, and obtain real-time observation data reflecting the actual degree of degradation; the estimated degradation status of the target unit components is corrected, and a local dynamic replanning mechanism is initiated based on the cognitive bias before and after the correction; the surface operation and maintenance vehicle and the aerial inspection vehicle are controlled to execute the operation and maintenance operations of the offshore wind farm according to the replanned scheduling instructions.

[0014] Furthermore, the physical evolution model of the unit components is a component-specific competing failure model, which includes mutually coupled soft failure models and hard failure models, wherein:

[0015] The soft failure model uses a non-homogeneous gamma process with an introduced state acceleration term to describe the degradation increment of unit components. The degradation increment of a unit component per unit time is determined by the product of the basic gamma increment and the state acceleration term, which is a nonlinear characteristic reflecting the accelerated accumulation of damage when the unit component approaches the failure threshold. The cumulative degradation amount of the unit component is obtained by continuously accumulating the degradation increment. When the cumulative degradation amount reaches the set wear limit, a soft failure is determined to have occurred.

[0016] The hard failure model uses a dynamic stress-strength interference model to describe the residual shock resistance of unit components. The residual shock resistance of unit components decreases with the increase of cumulative degradation, and the external environmental impact load borne by unit components changes dynamically with real-time meteorological conditions. When the external environmental impact load borne by the unit components at any time exceeds the current residual shock resistance, stress-strength interference occurs, and hard failure is determined to have occurred.

[0017] Soft failures and hard failures are competing failures. When either one occurs first, the corresponding unit component is determined to have suffered a functional failure.

[0018] Furthermore, the magnitude of the basic gamma increment is determined by the equivalent physical stress borne by the unit components; the equivalent physical stress is calculated based on meteorological forecast data through a multiphysics coupled stress mapping function, specifically:

[0019] For the blade subsystem, its equivalent physical stress is modeled as a square function of the ambient wind speed; for the gearbox system, its equivalent physical stress is modulated by the wind-wave coupling effect and is determined by the rated operating condition reference value, the wind speed influence term, and the coupling weight coefficient of the wave load; for the generator subsystem, its equivalent physical stress is modeled as a square function of the ambient wind speed.

[0020] Furthermore, the rolling time-domain simulation predicts the maintenance needs of each unit's components within a set future time window, determines the maximum operational capacity of surface maintenance vehicles by combining meteorological forecast data, and performs preemptive scheduling based on supply and demand matching analysis. The specific preemptive repair tasks are as follows:

[0021] Within each rolling time-domain simulation cycle, continuous effective operation time windows that meet the vehicle's safe operation threshold are selected based on meteorological forecast data. Combining the number of available surface maintenance vehicles and the average operation efficiency of surface maintenance vehicles for maintenance operations, the maximum physical task capacity for each future evolution day is evaluated. If maintenance demand exceeds the maximum physical task capacity for a future evolution day due to meteorological blocking or resource congestion, it is considered that the future evolution day has a capacity overflow, triggering preemptive scheduling. Tasks are sorted in descending order of delay penalty cost, and the maintenance task with the highest delay penalty cost in the future evolution day is preemptively removed from the continuous effective operation time window of the current evolution day as an emergency repair task.

[0022] Furthermore, the game rule based on the waste value of remaining lifetime and the expected failure risk cost calculates the delay penalty cost of each potential maintenance task and executes the opportunistic maintenance decision in the following specific manner:

[0023] After confirming that there are maintenance tasks to be performed on the target unit, iterate through other components in the target unit that are in a non-urgent state; quantify the remaining life wasted value caused by performing maintenance in advance, and the expected failure risk cost of postponing maintenance to be performed separately in the future; if and only if the expected failure risk cost is greater than the remaining life wasted value, the corresponding non-urgent component is added to the maintenance task list of the current evolution day.

[0024] Furthermore, the first layer employs a heuristic optimization algorithm to prioritize the construction of the skeleton routes for surface maintenance vehicles for units with emergency maintenance tasks. Specifically:

[0025] For emergency maintenance tasks triggered by physical conditions and preemptive scheduling of emergency repair tasks, candidate skeleton routes for surface maintenance vehicles to perform maintenance operations are constructed. A sliding window optimization mechanism is introduced to perform a sliding search on each candidate skeleton route within the continuous effective operation time window of the current evolution day to lock the optimal route departure time that minimizes the total downtime power generation loss. The minimum total downtime power generation loss is used as the fitness of the candidate skeleton route, and a heuristic optimization algorithm is used to iteratively optimize the candidate skeleton route with the minimum fitness as the skeleton route for surface maintenance vehicles to perform maintenance operations.

[0026] The heuristic optimization algorithm is a genetic algorithm, which uses an integer encoding mechanism. Chromosomes are represented as task allocation vectors with a length equal to the available resources of the system, used to map maintenance tasks to available surface maintenance vehicles. During the iteration process of the genetic algorithm, a tournament selection strategy is used to select and retain elite individuals, a uniform mask crossover operator is used to generate offspring, and a random reset mutation operator is applied to dynamically adjust the task configuration or change it to a postponed state. The iteration is repeated until the preset maximum number of iterations is reached or the fitness converges.

[0027] Furthermore, the second layer, based on marginal cost-benefit evaluation rules, dynamically optimizes and inserts non-urgent opportunity maintenance tasks into the backbone route, specifically as follows:

[0028] After determining the backbone route, a dynamic optimization insertion strategy is used to traverse the opportunity maintenance tasks. The additional marginal cost generated by inserting the opportunity maintenance task into any node of the current backbone route is calculated. Only when the sum of the additional marginal cost and the remaining lifetime wastage value is lower than the estimated cost of the future independent maintenance of the opportunity maintenance task, the opportunity maintenance task is confirmed to be inserted and the backbone route sequence is locally smoothed.

[0029] Furthermore, the total operation and maintenance cost includes: the dispatch cost of surface maintenance vehicles and aerial inspection vehicles, navigation energy consumption cost, maintenance material replacement cost, inherent maintenance downtime power generation loss during maintenance, and downtime power generation loss due to functional loss during the period from unit damage to repair.

[0030] Furthermore, the dynamic priority scoring strategy is as follows:

[0031] The inspection priority score of the units is determined by combining physical failure risk and cognitive uncertainty. The inspection priority score is calculated by the time interval since the last inspection, the current estimated degradation status and the preset risk sensitivity factors. The offshore wind turbines with high inspection priority scores are prioritized for inspection.

[0032] Furthermore, the process of correcting the estimated degradation state of the target unit components, and using a posterior state update rule to correct it based on the cognitive bias before and after the correction, involves initiating a local dynamic replanning mechanism.

[0033] The state posterior update rule is a Bayesian inference mechanism; after obtaining real-time observation data of the target unit components, the estimated degradation state of each component of the target unit is reshaped through Bayesian update, and the unobserved time of the unit is reset to zero; if the cognitive deviation of any component in the target unit before and after the update exceeds a preset threshold, it is determined that an information surprise phenomenon has occurred, triggering a local dynamic replanning algorithm to reallocate on-site transport capacity, identifying the maintenance task corresponding to the component as a new emergency maintenance task, and partially inserting it into the execution route of the current surface maintenance vehicle.

[0034] Based on the above technical solution, the present invention has at least the following beneficial effects:

[0035] 1. This invention utilizes an opportunistic maintenance strategy to perform hitch-on maintenance on nearby units during a single sea voyage, significantly improving the economies of scale in operation and maintenance logistics.

[0036] 2. This invention introduces an active perception and dynamic replanning mechanism for UAVs, which can promptly detect and correct hidden random damage that the physical model cannot capture, effectively avoiding serious failures caused by cognitive biases and avoiding high downtime losses.

[0037] 3. Through capacity-aware simulation, this invention can identify future weather-related blockade risks in advance and move tasks to the current window for emergency repairs, thereby enhancing the robustness of the operation and maintenance system under severe sea conditions such as typhoons and cold waves. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the overall process of the method for opportunistic maintenance and path co-optimization of offshore wind farms in extreme weather proposed in this invention.

[0039] Figure 2 This is a multi-vehicle Gantt chart for opportunity maintenance collaborative scheduling in this embodiment of the invention;

[0040] Figure 3This is a graph showing the evolution of the cognitive state of a target unit component and the active perception correction process of an airborne inspection vehicle in an embodiment of the present invention.

[0041] Figure 4 This is a Gantt chart of a preemptive scheduling scheme for dealing with extreme weather blocking in this embodiment of the invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0043] Although the steps in this invention are arranged by reference numerals, this is not intended to limit the order of the steps. Unless the order of the steps is explicitly stated or the execution of a step requires other steps as a basis, the relative order of the steps can be adjusted. It is understood that the term "and / or" as used herein refers to and covers any and all possible combinations of one or more of the associated listed items.

[0044] like Figure 1 As shown, the proposed method for opportunistic maintenance and path optimization of offshore wind farms in extreme weather conditions includes the following steps:

[0045] S1. Environmental Perception and Physical State Extrapolation: Acquire real-time monitoring data and weather forecast data of offshore wind farms; in this embodiment, before the start of a decision cycle for opportunistic maintenance and path co-optimization, the latest weather forecast data is received to obtain real-time wind speed. and Wave High Then, based on the pre-set physical evolution model of the turbine components and combined with meteorological forecast data, the estimated degradation status of each turbine component in the offshore wind farm is updated. And determine whether any functional failure has occurred in any of the unit's components; among which For unit serial number, For component serial number, For the current time;

[0046] In a preferred embodiment, the preset physical evolution model of the turbine components is a component-specific competitive failure model, and separate competitive failure models with independent parameters are constructed for different components of the wind turbine. The competitive failure model includes a soft failure model and a hard failure model that are coupled together, wherein:

[0047] The soft failure model employs a non-homogeneous gamma process with an introduced state acceleration term to describe the degradation increment of unit components. The degradation increment of a unit component per unit time is determined by the product of the base gamma increment and the state acceleration term. The state acceleration term reflects the nonlinear characteristic of accelerated damage accumulation as the unit component approaches the failure threshold. The magnitude of the base gamma increment is determined by the equivalent physical stress borne by the unit component. The equivalent physical stress is calculated based on meteorological forecast data using a multiphysics coupled stress mapping function, specifically:

[0048] For the blade subsystem, its equivalent physical stress is modeled as a square function of the ambient wind speed; for the gearbox system, its equivalent physical stress is modulated by the wind-wave coupling effect and is determined by the rated operating condition reference value, the wind speed influence term, and the coupling weight coefficient of the wave load; for the generator subsystem, its equivalent physical stress is modeled as a square function of the ambient wind speed. In this embodiment, the equivalent physical stress mapping for different components is as follows:

[0049] The equivalent compressive stress of the blade is ;

[0050] Gearbox withstands wind and wave coupled stress ;

[0051] The aging stress of generator insulation is ;

[0052] in, The reference wind speed under rated operating conditions. This is the reference wave height under rated operating conditions. The coupling weighting coefficient for wave loads;

[0053] The formula for the degradation increment is then expressed as:

[0054] ;

[0055] Among them, the basic increment Follows a gamma distribution, state acceleration term This reflects the physical phenomenon of exponentially accelerated damage. Represents the acceleration sensitivity coefficient. For time intervals, For degradation increment;

[0056] The cumulative degradation amount of unit components is obtained by continuously accumulating the degradation increment. When the cumulative degradation amount reaches the set wear limit, a soft failure is determined to have occurred.

[0057] The hard failure model uses a dynamic stress-strength interference model to describe the residual shock resistance of unit components; the residual shock resistance of unit components decreases with increasing cumulative degradation, and the formula for residual shock resistance is expressed as:

[0058] ;

[0059] in This indicates the rated impact strength of the unit's components at the time of manufacture. Indicates the strength attenuation coefficient of the component;

[0060] External environmental impact loads on unit components The formula for the external environmental impact load borne by different components as real-time meteorological conditions dynamically change is expressed as follows:

[0061] External environmental impact loads on the blades ;

[0062] External environmental impact loads borne by the gearbox ;

[0063] External environmental impact loads on generators ;

[0064] in, Indicates in The maximum instantaneous gust speed within the time window at which the time is located. Indicates in Maximum wave height within the time window at that moment. The reference wind speed under rated operating conditions. This refers to the reference wave height under rated operating conditions. This is the wave impact amplification factor;

[0065] When the external environmental impact load on the unit components exceeds the current remaining impact resistance at any given time, i.e. Stress-strength interference occurs, indicating a hard failure.

[0066] At the same time, soft failure and hard failure are competing failures. When either one occurs first, the corresponding unit component is determined to have suffered a functional failure.

[0067] Unlike existing technologies that typically only consider the cumulative natural wear and tear of equipment or simple environmental stress, the method proposed in this invention deeply couples the "soft failure" caused by material fatigue with the instantaneous "hard failure" induced by extreme sea conditions (such as typhoons and cold waves), revealing the objective fact that aging components are more vulnerable to the same extreme weather impacts. Based on this competitive failure model, this invention overcomes the limitations of traditional fixed threshold alarms, accurately predicting the risk of instantaneous breakage of the unit in impending severe sea conditions. This allows the system to proactively perform emergency repairs on high-risk tasks before weather lockouts occur, fundamentally avoiding catastrophic failures and significantly reducing the operation and maintenance costs of offshore wind power throughout its entire lifecycle.

[0068] S2. Meteorological capacity perception and preemptive estimation: Based on rolling time-domain simulation, the maintenance needs of each unit component within a future set time window are predicted. Combined with meteorological forecast data, the maximum operable capacity of the water surface maintenance vehicle is determined. Preemptive scheduling is carried out based on supply and demand matching analysis to preemptively carry out emergency repair tasks.

[0069] As a preferred implementation, the preemptive scheduling in this embodiment is specifically as follows:

[0070] Within each rolling time-domain simulation cycle, for the future... On each evolution day, based on meteorological forecast data, a continuous and effective operating time window that meets the vehicle's safe operation thresholds (including the vehicle's safe operation wind speed threshold and wave height threshold) is selected. Combined with the number of available surface maintenance vehicles Average operational efficiency for maintenance work on surface transport vehicles Assess the maximum daily physical mission capacity for each future evolutionary day. The formula is expressed as:

[0071] ;

[0072] If the future Maintenance needs arise on each evolutionary day due to weather closures or resource congestion. When the maximum physical task capacity for the day is exceeded, i.e. If capacity overflow occurs on the future evolution day, preemptive scheduling is triggered. Tasks are sorted in descending order of delay penalty cost. The maintenance task with the highest delay penalty cost on the future evolution day is preemptively removed from the current evolution day's continuous effective operation time window and treated as a preemptive maintenance task.

[0073] It should be noted that "rolling time domain simulation" in this application refers to "simulation" within a "rolling time domain". "Rolling time domain" refers to a continuously advancing time window; "simulation" refers to simulating and extrapolating the state over a future period of time, and in this embodiment, it refers to rehearsing the maintenance requirements of each unit component.

[0074] Simultaneously, based on the game rules of the remaining life wastage value and the expected failure risk cost, the delay penalty cost of each potential maintenance task is calculated, and the opportunity maintenance decision based on the internal cost game is executed.

[0075] In a preferred embodiment, after confirming that there are maintenance tasks to be performed on the target unit, the other components within the target unit that are not in an emergency state are traversed; the remaining life wastage value caused by prematurely performing maintenance is quantified, as well as the expected failure risk cost incurred by postponing maintenance to be performed separately in the future; the remaining life wastage value and expected failure risk costs The formula is expressed as:

[0076] ;

[0077] ;

[0078] in, For components Preventive maintenance costs, This is an estimate of the logistics and basic downtime costs for a single independent voyage required for future repairs of this component at sea. To map the current degradation state to a probability function of hard failure, the corresponding non-urgent component is added to the maintenance task list for the current evolution day if and only if the expected failure risk cost exceeds the remaining lifetime wastage value.

[0079] Through capacity-aware pre-simulation and preemptive scheduling in step S2, the method proposed in this invention can identify future weather lockout risks in advance and move tasks to the current window for emergency repairs, thereby enhancing the robustness of the operation and maintenance system under severe sea conditions such as typhoons and cold waves.

[0080] S3. Layered scheduling planning for air-sea coordination: Constructing a layered collaborative optimization scheduling model to generate an initial operation and maintenance scheduling scheme; the layered collaborative optimization scheduling model aims to minimize the total operation and maintenance cost within the simulation period;

[0081] In a preferred embodiment, the total maintenance cost Z comprises the dispatch cost of surface maintenance vehicles and aerial inspection vehicles, navigation energy consumption cost, maintenance material replacement cost determined by the maintenance type triggered by the degradation state of unit components, inherent maintenance downtime power generation loss during maintenance, and failure downtime power generation loss due to functional loss during the period from unit damage to repair; the formula aimed at minimizing the total maintenance cost is expressed as follows:

[0082] ;

[0083] in, This represents the set of available surface maintenance vehicles; This represents the set of all nodes, including the onshore homeport and wind turbine nodes; This represents the set of wind turbine nodes for all candidate maintenance tasks on that day. This represents the fixed cost of dispatching a maintenance vehicle per trip. As a decision variable, if the vehicle If dispatched to sea, the value is 1; otherwise, it is 0. This represents the energy cost per unit distance traveled. Represents a node To the node The sailing distance; For path decision variables, if the vehicle directly from the node sail to the node If the value is 1, then the value is 1; otherwise, the value is 0. As a decision variable, if the unit If a task is strategically shelved (postponed) by the system, the value is 1; otherwise, the value is 0. Indicates at node The actual material costs incurred in performing maintenance; Indicates the unit The start time of maintenance work; The dynamic downtime loss function is a nonlinear integral of the actual start time of maintenance. This represents the maximum penalty value set for postponing high-risk / urgent tasks.

[0084] Since the opportunity maintenance and route coordination optimization problem of offshore wind farms has NP-hard characteristics, in this embodiment, the hierarchical coordination optimization scheduling model adopts a two-layer coordination optimization strategy. Its specific configuration is as follows: the first layer adopts a heuristic optimization algorithm to prioritize the construction of the skeleton route of the water surface maintenance vehicle for the units containing emergency maintenance tasks.

[0085] In a preferred implementation, in this embodiment, the first layer of optimization targets emergency maintenance tasks triggered by physical conditions and preemptive scheduling of emergency maintenance tasks, constructing candidate skeleton routes for surface maintenance vehicles to perform maintenance operations; a sliding window optimization mechanism is introduced, which performs a sliding search on each candidate skeleton route within the continuous effective operation time window of the current evolution day, locking the optimal route departure time that minimizes the total downtime power generation loss, so that the entire process meets meteorological constraints and minimizes downtime loss; the minimum total downtime power generation loss is used as the fitness of the candidate skeleton route, and a heuristic optimization algorithm is used to iteratively optimize the candidate skeleton route with the minimum fitness as the skeleton route for surface maintenance vehicles to perform maintenance operations;

[0086] The heuristic optimization algorithm is a genetic algorithm, which uses an integer encoding mechanism. Chromosomes are represented as task allocation vectors with a length equal to the available resources of the system, used to map maintenance tasks to available surface maintenance vehicles. During the iteration process of the genetic algorithm, a tournament selection strategy is used to select and retain elite individuals, a uniform mask crossover operator is used to generate offspring, and a random reset mutation operator is applied to dynamically adjust the task configuration or change it to a postponed state. The iteration is repeated until the preset maximum number of iterations is reached or the fitness converges.

[0087] The second layer, based on marginal cost-benefit evaluation rules, dynamically optimizes and inserts non-urgent opportunity maintenance tasks into the backbone route; as a preferred implementation, in this embodiment, the second layer optimization specifically involves:

[0088] After determining the skeleton route through the first layer of optimization, a dynamic optimization insertion strategy is adopted to traverse the opportunity maintenance tasks. The additional marginal cost generated by inserting the opportunity maintenance task into any node of the current skeleton route is calculated. Only when the sum of the additional marginal cost and the remaining lifetime wastage value is lower than the estimated cost of the future independent maintenance of the opportunity maintenance task, the insertion of the opportunity maintenance task is confirmed and the skeleton route sequence is locally smoothed.

[0089] When obtaining the initial scheduling plan, this method uses the aforementioned hierarchical collaborative optimization scheduling model to decouple emergency tasks from non-emergency tasks, greatly compressing the search space of heuristic algorithms and solving the industry pain points of large-scale multi-vehicle path planning easily falling into the "curse of dimensionality" and extremely slow solution under extreme weather conditions. At the same time, in terms of minimizing the total operation and maintenance cost, the system uses the marginal cost-benefit evaluation rule to accurately nest opportunistic maintenance tasks into the determined skeleton route and downtime window. Under the premise of not violating meteorological safety and not increasing additional ship dispatch costs, it maximizes the economies of scale of multi-task "carpooling to sea" and achieves the ultimate reduction of global logistics dispatch costs and downtime losses.

[0090] S4. According to the initial operation and maintenance scheduling plan, based on the dynamic priority scoring strategy, the airborne inspection vehicle is scheduled to inspect the units whose dynamic priority scores exceed the threshold, and real-time observation data reflecting their true degradation level is obtained; as a preferred implementation, in this embodiment, the dynamic priority scoring strategy is:

[0091] The inspection priority score for the unit is calculated by integrating physical failure risk and cognitive uncertainty. The inspection priority score is determined by the time interval since the last inspection, the current estimated degradation status, and preset risk sensitivity factors. No. Typhoon turbine inspection priority score The formula is:

[0092] ;

[0093] in, For the first The time interval since the last inspection of the typhoon generator. Based on preset risk sensitivity factors, priority is given to inspecting offshore wind turbines with high inspection priority scores (i.e., high-risk and high-uncertainty targets).

[0094] Real-time observation data is acquired, and the estimated degradation state of the target unit components is corrected using a state posterior update rule. A local dynamic replanning mechanism is then initiated based on the cognitive bias before and after the correction. Specifically, in this embodiment:

[0095] The state posterior update rule is a Bayesian inference mechanism; real-time observation data of the target unit components are obtained. Then, the cognitive state distribution is reshaped through Bayesian updates, as expressed by the formula:

[0096] ;

[0097] in, This represents the prior estimate of the degraded state value obtained by extrapolation from the physical model before acquiring observational data. The normalized dynamic fusion weights are jointly determined by the static confidence level of the sensors in real-time observation data and the dynamic confidence level of the physical model. This represents the posterior estimated degradation state value of the target unit component calculated after fusion.

[0098] The unobserved time of the unit is reset to zero. If the cognitive deviation of any component in the target unit before and after the update exceeds a preset threshold, an information surprise phenomenon is determined, triggering a local dynamic replanning algorithm to reallocate on-site resources. The maintenance task corresponding to that component is identified as a new emergency maintenance task and partially inserted into the current execution route of the surface maintenance vehicle. Finally, the surface maintenance vehicle and the aerial inspection vehicle are controlled to execute the operation and maintenance work of the offshore wind farm according to the replanned scheduling instructions.

[0099] By introducing the UAV active perception and dynamic replanning mechanism in step S4, the method proposed in this invention can promptly detect and correct hidden random damage that the physical model cannot capture, effectively avoiding serious failures caused by cognitive biases and avoiding high downtime losses.

[0100] The above is the complete process of the method proposed in this invention. To verify the reliability and superiority of the method proposed in this invention, this embodiment constructs a simulation environment containing 78 wind turbine units (represented by T0-T77 in the attached figures), configures 3 conventional manned maintenance vessels (CTVs, i.e., 3 surface maintenance vehicles) and 1 unmanned aerial vehicle (UAV, i.e., 1 space inspection vehicle), and conducts a 90-day full-cycle rolling time-domain simulation. The meteorological generation model superimposes random disturbances and extreme weather, and sets up four sets of comparative experiments: passive correction (benchmark 1), fixed threshold (benchmark 2), static planning only (benchmark 3), and no-opportunity maintenance (benchmark 4). The simulation results are shown in Table 1:

[0101] Table 1 Simulation Results

[0102]

[0103] As shown in the table, the total cost of the offshore wind farm maintenance solution obtained by using the method proposed in this invention is reduced to US$1,570,679, which is about 96.4% lower than baseline 1. It completely avoids the downtime loss of about US$2.04 million and the replacement cost of US$41 million, achieving zero-failure operation, which is 46.4% lower than baseline 2, and effectively curbs the waste of resources caused by frequent ship dispatch.

[0104] By implementing 36 opportunity maintenances, with an average of 2.79 parts repaired per voyage, the free-riding effect was effectively leveraged, reducing the logistics and time costs of a single voyage by approximately 45%, and further reducing the total cost by 19.4% compared to benchmark 4.

[0105] Combined with relevant multi-ship scheduling Gantt charts (such as...) Figure 2 As shown in the Gantt chart for day 41, this invention, based on dynamic marginal cost assessment, greedily inserts non-urgent tasks into the backbone route. Multiple tasks share the high fixed dispatch and sailing costs, resulting in logistics savings far exceeding the material waste caused by premature replacements. This effectively increases the density of single-trip operations and avoids capacity fragmentation. Simultaneously, internal opportunity maintenance is performed on the suboptimal components of the target aircraft. Physically, tower preparation time is shared, and scheduling perfectly reuses the unavoidable downtime window of the main task, minimizing additional downtime losses.

[0106] In the 90-day full-cycle verification, benchmark 4, excluding the opportunistic repair mechanism, required 45 sea voyages to complete 69 repairs. However, relying on internal and external collaboration, this invention completed 78 repairs in only 28 sea voyages. This is the core driving force that drives a 19.4% reduction in total cost without significantly increasing material costs.

[0107] Combination Figure 3 The cognitive evolution trajectory of a high-risk crew, as shown, often lags behind actual wear and tear (dashed line) in natural evolution, where prior model estimates (dashed line) tend to be outpaced. This system utilizes an airborne inspection vehicle to accurately intercept abrupt changes caused by hidden random shocks and decisively triggers six local dynamic replanning operations to correct the flight path. Compared to baseline 3, this mechanism successfully intercepts sudden failure risks, avoids costly hard failures, and validates the high reliability premium.

[0108] Combination Figure 4 The diagram shows the preemptive scheduling Gantt chart for day 59. In the lead-up to a prolonged weather lockout (such as a severe cold wave), this invention identifies the risk of mismatch between supply and demand of transport capacity in advance through capacity-aware pre-simulation. The system breaks through the traditional fixed threshold limitation and directs surface vehicles to operate at full capacity within the available time window, preemptively repairing components that are about to be damaged. This scheduling completely avoids catastrophic failures and high downtime losses caused by weather lockouts, demonstrating excellent robustness in extreme weather.

[0109] In summary, this invention proposes a collaborative optimization method for opportunistic maintenance and path selection in offshore wind farms under extreme weather conditions. This method addresses existing problems in offshore wind farm operation and maintenance, such as divergent perceptions of component degradation states, complex failure modes influenced by extreme weather, and a lack of coordination in the scheduling of sea, land, and air resources under dynamic weather windows. The proposed method collaboratively optimizes multiple air and sea resources and maintenance timing, significantly improving operational robustness under extreme weather conditions and reducing total system cost.

[0110] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0111] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can fetch and execute instructions from, an instruction execution system, apparatus or device).

[0112] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for coordinated optimization of opportunistic maintenance and path selection in offshore wind farms under extreme weather conditions, characterized in that, Specifically, the following steps are included: Based on the physical evolution model of the turbine components, combined with real-time monitoring data and meteorological forecast data obtained from offshore wind farms, the estimated degradation status of each turbine component in the offshore wind farm is updated, and it is determined whether each turbine component has experienced functional failure. Based on rolling time-domain simulation, the maintenance needs of each unit component within a future set time window are predicted. Combined with meteorological forecast data, the maximum operational capacity of the water surface maintenance vehicle is determined. Based on supply and demand matching analysis, preemptive scheduling is carried out to pre-position emergency repair tasks. Based on the game rule of the remaining life wastage value and the expected failure risk cost, the delay penalty cost of each potential maintenance task is calculated, and opportunistic maintenance decisions are executed. A hierarchical collaborative optimization scheduling model is constructed to generate an initial operation and maintenance scheduling scheme. The hierarchical collaborative optimization scheduling model aims to minimize the total operation and maintenance cost within the simulation period. The first layer adopts a heuristic optimization algorithm to prioritize the construction of the skeleton route of the water surface operation and maintenance vehicle for units with emergency maintenance tasks. The second layer is based on the marginal cost-benefit evaluation rule to dynamically optimize and insert non-emergency opportunity maintenance tasks into the skeleton route. The first layer employs a heuristic optimization algorithm, prioritizing the construction of the skeleton route for surface maintenance vehicles for units with emergency maintenance tasks. Specifically: For emergency maintenance tasks triggered by physical conditions and preemptive scheduling of emergency repair tasks, candidate skeleton routes for surface maintenance vehicles to perform maintenance operations are constructed. A sliding window optimization mechanism is introduced to perform a sliding search on each candidate skeleton route within the continuous effective operation time window of the current evolution day to lock the optimal route departure time that minimizes the total downtime power generation loss. The minimum total downtime power generation loss is used as the fitness of the candidate skeleton route, and a heuristic optimization algorithm is used to iteratively optimize the candidate skeleton route with the minimum fitness as the skeleton route for surface maintenance vehicles to perform maintenance operations. The heuristic optimization algorithm is a genetic algorithm, which uses an integer encoding mechanism. Chromosomes are represented as task allocation vectors with a length equal to the available resources in the system, used to map maintenance tasks to available surface maintenance vehicles. During the iteration process of the genetic algorithm, a tournament selection strategy is used to select and retain elite individuals, a uniform mask crossover operator is used to generate offspring, and a random reset mutation operator is applied to dynamically adjust the task configuration or change it to a postponed state. The iteration is repeated until the preset maximum number of iterations is reached or the fitness converges. According to the initial operation and maintenance scheduling plan, based on the dynamic priority scoring strategy, the aerial inspection vehicle is scheduled to inspect the units whose dynamic priority scores exceed the threshold, and obtain real-time observation data reflecting the actual degree of degradation; the estimated degradation status of the target unit components is corrected, and a local dynamic replanning mechanism is initiated based on the cognitive bias before and after the correction; the surface operation and maintenance vehicle and the aerial inspection vehicle are controlled to execute the operation and maintenance operations of the offshore wind farm according to the replanned scheduling instructions.

2. The method for opportunistic maintenance and path coordination optimization of offshore wind farms in extreme weather conditions, as described in claim 1, is characterized in that... The physical evolution model of the unit components is a component-specific competing failure model, which includes mutually coupled soft failure models and hard failure models, wherein: The soft failure model uses a non-homogeneous gamma process with an introduced state acceleration term to describe the degradation increment of unit components. The degradation increment of a unit component per unit time is determined by the product of the basic gamma increment and the state acceleration term, which is a nonlinear characteristic reflecting the accelerated accumulation of damage when the unit component approaches the failure threshold. The cumulative degradation amount of the unit component is obtained by continuously accumulating the degradation increment. When the cumulative degradation amount reaches the set wear limit, a soft failure is determined to have occurred. The hard failure model uses a dynamic stress-strength interference model to describe the residual shock resistance of unit components. The residual shock resistance of unit components decreases with the increase of cumulative degradation, and the external environmental impact load borne by unit components changes dynamically with real-time meteorological conditions. When the external environmental impact load borne by the unit components at any time exceeds the current residual shock resistance, stress-strength interference occurs, and hard failure is determined to have occurred. Soft failures and hard failures are competing failures. When either one occurs first, the corresponding unit component is determined to have suffered a functional failure.

3. The method for coordinated optimization of opportunity maintenance and path for offshore wind farms in extreme weather conditions, as described in claim 2, is characterized in that... The magnitude of the basic gamma increment is determined by the equivalent physical stress borne by the unit components; the equivalent physical stress is calculated based on meteorological forecast data through a multiphysics coupled stress mapping function, specifically: For the blade subsystem, its equivalent physical stress is modeled as a square function of the ambient wind speed; for the gearbox system, its equivalent physical stress is modulated by the wind-wave coupling effect and is determined by the rated operating condition reference value, the wind speed influence term, and the coupling weight coefficient of the wave load; for the generator subsystem, its equivalent physical stress is modeled as a square function of the ambient wind speed.

4. The method for opportunistic maintenance and path coordination optimization of offshore wind farms in extreme weather conditions, as described in claim 1, is characterized in that... Based on rolling time-domain simulation, the maintenance needs of each unit component within a future set time window are predicted. Combined with meteorological forecast data, the maximum operational capacity of surface maintenance vehicles is determined. Based on supply-demand matching analysis, preemptive scheduling is implemented. The specific preemptive repair tasks are as follows: Within each rolling time-domain simulation cycle, based on meteorological forecast data, continuous and effective operation time windows that meet the vehicle's safe operation threshold are selected. Combining the number of available surface maintenance vehicles and the average operation efficiency of surface maintenance vehicles in maintenance operations, the maximum physical task capacity of each future evolution day is evaluated. If, on a future evolution day, the maintenance demand exceeds the maximum physical task capacity of that day due to weather blocking or resource congestion, it is considered that there is a capacity overflow on that future evolution day, triggering preemptive scheduling. Tasks are sorted in descending order of delay penalty cost, and the maintenance task with the highest delay penalty cost on that future evolution day is preemptively removed from the current evolution day's continuous effective operation time window and treated as a preemptive maintenance task.

5. The method for opportunistic maintenance and path coordination optimization of offshore wind farms in extreme weather conditions, as described in claim 1, is characterized in that... The game rule based on the waste value of remaining lifetime and the cost of expected failure risk calculates the penalty cost of postponing each potential maintenance task and executes the opportunistic maintenance decision as follows: After confirming that there are maintenance tasks to be performed on the target unit, iterate through other components within the target unit that are not in an emergency state; Quantify separately the wasted value of remaining life caused by premature maintenance, and the expected failure risk cost of postponing maintenance to be performed separately in the future; Non-urgent components shall be added to the maintenance task list for the current evolution day only if the expected failure risk cost is greater than the remaining lifetime wastage value.

6. The method for opportunistic maintenance and path coordination optimization of offshore wind farms in extreme weather conditions, as described in claim 1, is characterized in that... The second layer, based on marginal cost-benefit evaluation rules, dynamically optimizes and inserts non-urgent opportunity maintenance tasks into the backbone route, specifically as follows: After determining the backbone route, a dynamic optimization insertion strategy is used to traverse the opportunity maintenance tasks. The additional marginal cost generated by inserting the opportunity maintenance task into any node of the current backbone route is calculated. Only when the sum of the additional marginal cost and the remaining lifetime wastage value is lower than the estimated cost of the future independent maintenance of the opportunity maintenance task, the opportunity maintenance task is confirmed to be inserted and the backbone route sequence is locally smoothed.

7. The method for coordinated optimization of opportunity maintenance and path for offshore wind farms in extreme weather conditions, as described in claim 1, is characterized in that... The total operation and maintenance cost includes: the dispatch cost of surface maintenance vehicles and aerial inspection vehicles, navigation energy consumption cost, maintenance material replacement cost, inherent maintenance downtime power generation loss during maintenance, and downtime power generation loss due to functional loss during the period from unit damage to repair.

8. A method for opportunistic maintenance and path coordination optimization of offshore wind farms in extreme weather conditions, as described in claim 1, is characterized in that... The dynamic priority scoring strategy is as follows: The inspection priority score of the units is determined by combining physical failure risk and cognitive uncertainty. The inspection priority score is calculated by the time interval since the last inspection, the current estimated degradation status and the preset risk sensitivity factors. The offshore wind turbines with high inspection priority scores are prioritized for inspection.

9. A method for coordinated optimization of opportunistic maintenance and path for offshore wind farms in extreme weather conditions, as described in claim 1, is characterized in that... The estimated degradation state of the target unit components is corrected by initiating a local dynamic replanning mechanism based on the cognitive bias before and after correction, and then using a posterior state update rule for correction. Specifically: The state posterior update rule is a Bayesian inference mechanism; after obtaining real-time observation data of the target unit components, the estimated degradation state of each component of the target unit is reshaped through Bayesian update, and the unobserved time of the unit is reset to zero; if the cognitive deviation of any component in the target unit before and after the update exceeds a preset threshold, it is determined that an information surprise phenomenon has occurred, triggering a local dynamic replanning algorithm to reallocate on-site transport capacity, identifying the maintenance task corresponding to the component as a new emergency maintenance task, and partially inserting it into the execution route of the current surface maintenance vehicle.

Images