Method and system for optimizing maintenance decision of gear box of wind turbine in shaggy environment

Abstract

This invention discloses a method and system for optimizing maintenance decisions for wind turbine gearboxes in barren desert environments. The method involves: first, acquiring the online remaining life prediction value, inspection information, and environmental conditions of the gearbox; then, fusing the prediction and inspection information through a dual-path calibration network to perform posterior probability calibration on the point-value lifespan and introducing survival consistency constraints to obtain the remaining lifespan probability distribution; next, calculating the interval failure risk based on this distribution, solving for the maximum safe maintenance interval that meets a preset risk threshold, and forming a risk-feasible maintenance window; simultaneously, constructing an environmental accessibility function based on wind speed, temperature, dust, and road conditions, and establishing a time-varying operation and maintenance cost model that includes deployment costs, operating costs, downtime losses, and collaborative maintenance discounts; finally, within the risk-feasible window, solving for the optimal maintenance time and action with the minimum total cost through a rolling optimization strategy. This invention can significantly improve the reliability and economy of maintenance decisions in complex desert environments.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance and predictive maintenance technology for wind power equipment, specifically to a method and system for optimizing maintenance decisions for wind turbine gearboxes, and more specifically, to a method and system for optimizing maintenance decisions for wind turbine gearboxes in desert and barren environments. Background Technology

[0002] As onshore wind turbines continue to be deployed to large-capacity, remote areas and complex environments, the gearbox, as a critical component in the transmission chain with a high risk of failure and high maintenance costs, directly affects the unit's availability, downtime losses, and the efficiency of operation and maintenance resource allocation. Especially in wind farm environments such as deserts, Gobi, wastelands, high-altitude cold regions with large temperature differences and fluctuating road conditions, maintenance activities are affected not only by the degree of component degradation, but also by wind speed, ambient temperature, dust, road conditions, and the accessibility of operation and maintenance teams.

[0003] In existing technologies, some methods directly use point-value remaining life results to schedule maintenance. Although this can provide some trend reference, point-value prediction lacks an effective expression of prediction bias and uncertainty, making it difficult to directly support the selection of maintenance timing in high-risk scenarios. While some methods consider maintenance costs, they are usually approximated by fixed or average costs, making it difficult to characterize the significant time-varying characteristics of deployment costs, downtime losses, and collaborative maintenance benefits in desert environments.

[0004] Furthermore, existing technologies generally suffer from the problem of separating predictive models from maintenance decision-making models. That is, the output of the remaining lifetime model remains only at the analysis level, without being further transformed into risk-feasibility windows and actual work order suggestions. This makes it difficult for the algorithm to form a closed-loop application on wind farm operation and maintenance platforms. Therefore, there is an urgent need for a technical solution suitable for barren and desert environments that can integrate remaining lifetime a posteriori calibration, risk constraints, time-varying cost modeling, and rolling maintenance decision-making to enhance the engineering applicability and deployment value of maintenance decisions. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for optimizing maintenance decisions for wind turbine gearboxes in barren and desert environments. This method addresses the problems in existing wind turbine gearbox maintenance decision-making processes, such as the uncertainty of remaining life prediction results making them difficult to use directly for decision-making, unclear maintenance risk boundaries, difficulty in accurately characterizing the time-varying nature of operation and maintenance costs in barren and desert environments, and difficulty in dynamically optimizing maintenance timing. Using this invention can significantly improve the reliability, economy, and engineering applicability of wind turbine gearbox maintenance decisions in complex environments.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] In a first aspect, the present invention provides a method for optimizing maintenance decisions of wind turbine gearboxes in desert environments, the method comprising: S1. Obtain online remaining life prediction results, inspection information, environmental status information, and operation and maintenance cost information of wind turbine gearboxes; S2. Input the online remaining lifetime prediction results and inspection information into the dual-path calibration network, perform posterior calibration on the online remaining lifetime prediction results, and obtain the calibrated remaining lifetime probability distribution; wherein, the dual-path calibration network includes a mean calibration path and an evidence strength modeling path; S3. In the post-calibration process, a survival consistency constraint is introduced so that the remaining lifetime result after calibration satisfies the overall non-incremental evolution law with the equipment degradation process. S4. Based on the calibrated remaining lifetime probability distribution, calculate the failure probability of the wind turbine gearbox at each time point, and further construct the interval failure risk within the future candidate maintenance interval. S5. Based on the preset risk threshold, determine the maximum safe maintenance interval that meets the risk constraints at the current decision moment, and obtain the risk-feasible maintenance window; S6. Construct an environmental reachability function based on the environmental state information, and establish a time-varying operation and maintenance cost model in combination with the maintenance action type; the time-varying operation and maintenance cost model includes dispatch cost, on-site operation cost, and power outage loss; S7. Within the risk-feasible maintenance window, construct a rolling optimization objective function based on the time-varying operation and maintenance cost model and the interval failure risk, and solve for the optimal maintenance time. S8. Output the corresponding maintenance action according to the optimal maintenance time to guide the preventive maintenance, corrective maintenance or replacement maintenance of the wind turbine gearbox.

[0008] Furthermore, in step S1, the environmental status information includes at least one or more of wind speed, ambient temperature, dust intensity index, and road accessibility index; the operation and maintenance cost related information includes at least one or more of benchmark deployment cost, benchmark operation cost, maintenance period, electricity price information, wind curtailment information, and collaborative maintenance scale.

[0009] Furthermore, in step S2, the mean calibration path is used to correct the systematic bias of the online remaining lifetime prediction results, and the evidence strength modeling path is used to fuse inspection information and characterize the credibility change of the online remaining lifetime prediction results, thereby obtaining a remaining lifetime probability distribution that takes into account both the prediction mean and uncertainty.

[0010] Furthermore, in step S3, the survival consistency constraint is used to limit the remaining lifetime result after calibration from showing a reverse increase in the time dimension that is inconsistent with the equipment degradation law, thereby ensuring that the calibration result meets the physical consistency of the wind turbine gearbox remaining lifetime decreasing as a whole during operation.

[0011] Furthermore, in step S4, the interval failure risk is used to characterize the probability that at least one failure event will occur between the current decision time and the candidate maintenance time, so as to directly introduce the calibrated remaining lifetime uncertainty into the maintenance decision-making process.

[0012] Furthermore, in step S5, the maximum safe maintenance interval is the maximum time length for which maintenance can be delayed under a given risk threshold, and the risk-feasible maintenance window is determined jointly by the current decision time and the maximum safe maintenance interval.

[0013] Furthermore, in step S6, the environmental accessibility function is used to characterize the comprehensive impact of environmental conditions on the feasibility of maintenance deployment and the difficulty of construction organization; in the time-varying operation and maintenance cost model, the deployment cost and on-site operation cost are dynamically adjusted with changes in environmental accessibility, and the power generation loss due to downtime is determined based on the maintenance period, wind turbine power level, electricity price fluctuations, and wind curtailment situation.

[0014] Furthermore, step S6 also includes constructing a collaborative maintenance discount model to characterize the unit maintenance cost reduction effect brought about by the sharing of transportation resources, hoisting equipment, maintenance personnel and on-site support resources when multiple wind turbine units are jointly maintained.

[0015] Furthermore, the collaborative maintenance discount model is environment-dependent and is used to describe the weakening effect of the desert environment on the joint maintenance efficiency of multiple units.

[0016] Furthermore, in step S7, the rolling optimization objective function aims to minimize the total maintenance cost within the risk-feasible window, and is repeatedly solved as the online remaining life prediction results, inspection information, environmental status information, and cost parameters are updated, so as to achieve dynamic updates of maintenance timing.

[0017] Further, step S2 specifically includes: S21. The online remaining lifetime prediction result is used as the initial lifetime estimation input mean calibration path to correct the deviation of the original remaining lifetime prediction result and obtain the calibrated remaining lifetime mean; wherein, the mean calibration path is used to reduce the fluctuation and local shift of the prediction result in the later stage of degradation while maintaining the basic consistency of the original degradation trend. S22. Input the online remaining lifetime prediction results and inspection information into the evidence strength modeling path to obtain evidence strength parameters corresponding to the current degradation stage and inspection status; wherein, the evidence strength modeling path is used to characterize the change in the credibility of the remaining lifetime prediction results and reflect the moderating effect of inspection information on prediction uncertainty. S23. The calibrated mean remaining lifetime and the evidence strength parameter are jointly mapped to construct the posterior probability distribution of the remaining lifetime of the wind turbine gearbox at the current moment, which is used to simultaneously characterize the central trend and uncertainty range of the remaining lifetime prediction results.

[0018] Further, step S3 specifically includes: S31. Based on the remaining lifetime results after calibration at adjacent time points, determine whether the remaining lifetime at subsequent time points shows an abnormal increase that is inconsistent with the degradation pattern. S32. When an abnormal increase in the calibrated remaining lifetime result is detected, apply a survival consistency constraint penalty to the corresponding result to limit the non-physical fluctuation of the remaining lifetime result in the time dimension. S33. Introduce the survival consistency constraint into the posterior calibration process so that the remaining lifetime result after calibration satisfies the physical consistency of decreasing as the equipment degrades, thereby improving the interpretability and reliability of the calibration results in maintenance decisions.

[0019] Further, step S4 specifically includes: S41. Based on the calibrated remaining lifetime probability distribution, calculate the single-moment failure probability of the wind turbine gearbox reaching the failure threshold at each discrete moment. S42. Based on the single-time failure probability at each time point within the candidate maintenance interval, construct the interval failure risk from the current decision time to the candidate maintenance time where at least one failure event occurs. S43. The interval failure risk is used as a safety characterization index in maintenance decision-making to reflect the cumulative failure risk level faced by the wind turbine gearbox during delayed maintenance.

[0020] Further, step S5 specifically includes: S51. Compare the interval failure risk with a preset risk threshold to screen candidate maintenance periods that meet the risk constraints. S52. Under the condition of satisfying the preset risk threshold, determine the maximum length of time that the maintenance can be delayed at the current decision moment, as the maximum safe maintenance interval; S53. Construct a risk-feasible maintenance window based on the current decision time and the maximum safe maintenance interval to limit the search range for subsequent rolling optimization.

[0021] Further, step S6 specifically includes: S61. Based on wind speed, ambient temperature, dust intensity index and road accessibility index, construct an environmental accessibility function to characterize the comprehensive impact of current environmental conditions on the feasibility of maintenance deployment and the difficulty of on-site construction organization. S62. Based on the environmental accessibility function, establish a deployment cost model so that the deployment cost can change dynamically as environmental conditions deteriorate. S63. Based on the environmental accessibility function and maintenance action type, establish a field operation cost model so that the field operation cost can reflect the cost changes caused by the decrease in operation efficiency and the increase in human resource input in the desert environment. S64. Based on the maintenance period, wind turbine power level, electricity price fluctuations and wind curtailment, establish a power outage loss model to characterize the economic losses caused by unit shutdown at different maintenance times. S65. Construct a collaborative maintenance discount model based on the scale of joint maintenance to characterize the unit maintenance cost reduction effect caused by the sharing of transportation resources, hoisting equipment, maintenance personnel and on-site support resources when multiple wind turbines are jointly maintained. S66. Combining the aforementioned deployment cost model, on-site operation cost model, power outage loss model, and collaborative maintenance discount model, establish a time-varying operation and maintenance cost model for the wind turbine gearbox under different candidate maintenance times.

[0022] Further, step S7 specifically includes: S71. Within the risk-feasible maintenance window, with the goal of minimizing the total maintenance cost, construct a rolling optimization objective function that includes the risk-weighted cost of preventive maintenance and corrective maintenance after failure. S72. Jointly evaluate the time-varying maintenance costs and interval failure risks corresponding to each candidate maintenance time within the risk-feasible maintenance window, and solve for the optimal waiting time for maintenance at the current decision time. S73. Determine the optimal maintenance time for the current cycle based on the optimal waiting maintenance time, and output the corresponding maintenance action suggestion.

[0023] Further, step S8 specifically includes: S81. Send the optimal maintenance time and corresponding maintenance action to the wind farm operation and maintenance execution terminal to guide the scheduling of maintenance personnel, spare parts resources and hoisting resources; S82. In the next decision cycle, reacquire the updated online remaining lifetime prediction results, inspection information, environmental status information, and operation and maintenance cost information. S83. Based on the updated input information, repeat the posterior calibration, risk assessment, cost modeling and rolling optimization process to achieve dynamic updating and adaptive optimization of the timing of wind turbine gearbox maintenance.

[0024] Secondly, embodiments of the present invention also provide a wind turbine gearbox maintenance decision optimization system in a desert environment. This system applies the above-mentioned method to achieve dynamic updating and adaptive optimization of wind turbine gearbox maintenance timing. The system includes: Data acquisition module: used to acquire online remaining life prediction results, inspection information, environmental status information, and operation and maintenance cost information of wind turbine gearboxes; The post-calibration module is used to input online remaining lifetime prediction results and inspection information into the dual-path calibration network and output the calibrated remaining lifetime probability distribution. Risk assessment module: used to calculate failure probability, interval failure risk, and maximum safe maintenance interval based on the calibrated remaining lifetime probability distribution; Environment Modeling Module: Used to construct environmental reachability functions based on environmental state information; Cost calculation module: used to establish a time-varying operation and maintenance cost model based on environmental accessibility function combined with operation and maintenance cost related information and maintenance action type; Rolling optimization module: used to find the optimal maintenance time within the risk-feasible maintenance window; Decision output module: Used to output corresponding maintenance actions and send maintenance suggestions to the wind farm operation and maintenance execution terminal.

[0025] Thirdly, embodiments of the present invention also provide an electronic device, including a processor and a memory, wherein the memory stores machine-executable instructions that can be executed by the processor, and the processor executes the machine-executable instructions to implement the above-described method.

[0026] As can be seen from the above technical solutions, compared with the prior art, the method and system disclosed in this invention have the following beneficial effects: 1. This invention extends the traditional point-value remaining lifetime prediction results into a probability distribution form that can characterize uncertainty by integrating online remaining lifetime prediction results with inspection information through a post-calibration mechanism, thereby improving the stability and reliability of remaining lifetime results at the maintenance decision-making level. 2. By introducing survival consistency constraints, this invention ensures that the calibrated remaining lifetime results conform to the equipment degradation and evolution laws, avoids abnormal fluctuations in the prediction results over time, and improves the physical rationality of the lifetime prediction results. 3. This invention constructs the failure probability and interval failure risk based on the calibrated remaining lifetime probability distribution, and further solves the maximum safe maintenance interval, so that the maintenance timing selection is transformed from experience-triggered to feasible region optimization under risk constraints, thereby improving the safety of maintenance decisions. 4. This invention constructs an environmental accessibility function and a time-varying operation and maintenance cost model for desert and Gobi environments, which enables a comprehensive description of factors such as wind speed, temperature, dust, road conditions, electricity price fluctuations and wind curtailment, making the maintenance economic analysis more closely resemble real engineering scenarios. 5. By introducing a multi-unit collaborative maintenance discount mechanism and a rolling optimization strategy, this invention enables dynamic adjustment of maintenance intervals based on equipment status, environmental conditions, and cost levels. It can reduce long-term operation and maintenance costs while meeting risk constraints, and has strong engineering promotion value.

[0027] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0028] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0031] Figure 1 This is a schematic diagram of the overall process of the wind turbine gearbox maintenance decision optimization method provided by the present invention in a desert environment. Figure 2 A schematic diagram comparing the online remaining life prediction results and the post-calibration results of the wind turbine gearbox provided by the present invention; wherein (a) shows the online remaining life prediction results and (b) shows the post-calibration results. Figure 3 This is a schematic diagram illustrating the evolution relationship between inspection information and evidence strength parameters provided by the present invention. Figure 4A schematic diagram illustrating the single-moment failure probability evolution based on the calibrated remaining lifetime probability distribution provided by this invention; Figure 5 A schematic diagram of the environmental reachability function and its components provided by the present invention; Figure 6 A schematic diagram illustrating the evolution trajectory of the gearbox's post-inspection remaining life under different maintenance actions provided by the present invention; Figure 7 This is a schematic diagram of the rolling optimization solution process under risk constraints provided by the present invention; Figure 8 A schematic diagram illustrating the coupling relationship between expected maintenance cost and interval failure risk provided by this invention; Figure 9 The structural block diagram of the wind turbine gearbox maintenance decision optimization system provided by the present invention is shown below. Figure 10 This is a schematic diagram of the electronic device structure provided by the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0033] In the description of this invention, it should be noted that some processes described in this application specification and drawings include multiple operations that appear in a specific order. However, it should be clearly understood that these operations may be performed in any order or in parallel. Furthermore, various numbers are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0034] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0035] Example 1: Reference Figure 1 As shown in the figure, this invention discloses an optimization method for wind turbine gearbox maintenance decisions in a desert environment, including: S1. Obtain online remaining life prediction results, inspection information, environmental status information, and operation and maintenance cost information of wind turbine gearboxes; S2. Input the online remaining lifetime prediction results and inspection information into the dual-path calibration network, perform posterior calibration on the online remaining lifetime prediction results, and obtain the calibrated remaining lifetime probability distribution; wherein, the dual-path calibration network includes a mean calibration path and an evidence strength modeling path; S3. In the post-calibration process, a survival consistency constraint is introduced so that the remaining lifetime result after calibration satisfies the overall non-incremental evolution law with the equipment degradation process. S4. Based on the calibrated remaining lifetime probability distribution, calculate the failure probability of the wind turbine gearbox at each time point, and further construct the interval failure risk within the future candidate maintenance interval. S5. Based on the preset risk threshold, determine the maximum safe maintenance interval that meets the risk constraints at the current decision moment, and obtain the risk-feasible maintenance window; S6. Construct an environmental reachability function based on the environmental state information, and establish a time-varying operation and maintenance cost model by combining operation and maintenance cost related information and maintenance action type; S7. Within the risk-feasible maintenance window, construct a rolling optimization objective function based on the time-varying operation and maintenance cost model and the interval failure risk, and solve for the optimal maintenance time. S8. Output the corresponding maintenance action according to the optimal maintenance time to guide the preventive maintenance, corrective maintenance or replacement maintenance of the wind turbine gearbox.

[0036] This embodiment is applied to large-scale onshore wind farms, and is particularly suitable for intelligent operation and maintenance scenarios of wind turbines in complex operating environments such as deserts, Gobi, desert edges, high-altitude and large temperature difference areas. In these scenarios, wind turbine gearboxes are subjected to the combined effects of random wind loads, alternating torque, dust erosion, and significant temperature fluctuations over long periods, resulting in a degradation process with obvious uncertainty and time-varying characteristics. Simultaneously, maintenance activities are also affected by factors such as weather windows, road conditions, on-site deployment difficulties, and fluctuations in downtime losses, making traditional maintenance methods that rely on fixed maintenance intervals or point-value remaining life thresholds insufficient to meet risk control and economic operation requirements.

[0037] The operation and maintenance team deploys the method of this invention on the wind farm's remote intelligent operation and maintenance platform or centralized control center: The system periodically receives the online remaining life prediction results of each wind turbine gearbox, manual inspection results, meteorological monitoring data, road condition information, and electricity price information; then, it performs a post-hoc calibration on the original remaining life prediction results and constructs a remaining life probability distribution that can characterize uncertainty by combining inspection evidence; on this basis, it further assesses the failure risk level in the future period of time, and comprehensively considers environmental accessibility, deployment costs, construction costs, power generation losses due to downtime, and opportunities for joint maintenance of multiple units, and automatically generates the current optimal maintenance time and maintenance action suggestions.

[0038] For example, when a turbine's gearbox has entered the mid-to-late stage of degradation, although the online life prediction result has not yet fallen below a fixed threshold, inspection information shows minor anomalies. The system will reduce the confidence of the corresponding prediction result through a dual-path calibration network, widening the posterior remaining life distribution and thus increasing the cumulative failure risk within the future waiting period. If wind speeds are high, dust is increased, and road accessibility is reduced in the next two days, the system will not directly recommend immediate deployment. Instead, it will prioritize searching for a feasible window with more favorable environmental conditions and lower overall costs within the risk threshold. If other turbines have similar maintenance needs during the same period, the system will further trigger a collaborative maintenance discount mechanism, prioritizing the generation of joint maintenance recommendations. Based on this, maintenance personnel can prepare spare parts, hoisting resources, and maintenance teams in advance, completing gearbox maintenance within the planned maintenance window, thereby avoiding potential serious failures and unplanned downtime. The entire process requires no manual calculation, achieving closed-loop optimization from "life prediction" to "risk assessment" to "maintenance decision output," significantly improving the operational availability and maintenance economy of wind farms.

[0039] The implementation steps of this embodiment are described in detail below: According to S1, first obtain the online remaining life prediction results, inspection information, environmental status information, and operation and maintenance cost information of the wind turbine gearbox.

[0040] This embodiment addresses remote intelligent operation and maintenance scenarios for wind farms in complex environments such as deserts, high-altitude areas, large temperature differences, and strong sandstorms. The system is deployed at the wind farm's central control center or intelligent operation and maintenance platform. It periodically receives online remaining life prediction results, manual inspection results, environmental status parameters, and cost parameters for each wind turbine gearbox. This information is then time-aligned and used as input for subsequent post-calibration, risk assessment, and maintenance optimization. The online remaining life prediction results represent the initial estimate of the gearbox's remaining life at the current moment; inspection information provides external prior verification evidence; environmental status information represents the external constraints for maintenance deployment; and operation and maintenance cost-related information reflects the actual economic efficiency of maintenance.

[0041] In this embodiment, the environmental state information can be represented as an environmental state vector:

[0042] in, t Indicates the current moment; Indicates time t wind speed; Indicates time t Ambient temperature; Indicates time t Dust intensity index; Indicates time tThe road accessibility index. The parentheses "( t ")" indicates that the quantity is at time "(". t The function is E, which represents the set of multidimensional environmental variables. This environmental state vector will serve as the input to the subsequent environmental reachability function and time-varying cost model.

[0043] According to S2, the online remaining lifetime prediction results are calibrated a posteriori.

[0044] In this embodiment, the core of step S2 is to construct a dual-path calibration network (DPCS-Net) that integrates online remaining lifetime prediction results and inspection information with survival consistency constraints. This expands the traditional point-value remaining lifetime results into a probability distribution form, so as to simultaneously characterize the prediction center trend and prediction uncertainty. Figure 2 A comparison diagram between online remaining lifetime prediction results and posterior calibration results is provided: Figure 2 In part (a), the original predicted remaining lifetime curve shows significant fluctuations in the later stages of the lifetime. Figure 2 In part (b), after calibration by DPCS-Net, the calibration curve introduces an uncertainty interval while maintaining the overall degradation trend. This indicates that this embodiment does not output only a single point value, but rather outputs a posterior distribution that can be used for risk decision-making.

[0045] set up t The normalized remaining lifetime prediction value output by the always-on online prediction model is The actual remaining lifespan is Inspection information is Given the prediction results and inspection information, the true remaining lifespan follows a Beta posterior distribution, expressed as:

[0046] Its probability density expression is:

[0047] Among them, subscript t This indicates that the variable is taken from the first... t A discrete time point; and For the Beta distribution at time... t Two shape parameters; the vertical bar "|" indicates a conditional distribution, that is, given... and The true remaining lifetime distribution under the given conditions; the symbol "~" indicates compliance; It is the Gamma function; p (·) represents the probability density.

[0048] The reason for using the Beta distribution in this embodiment is that the normalized remaining lifetime is defined between 0 and 1, while the Beta distribution is naturally defined in a unit interval, which is suitable for characterizing the uncertainty of the remaining lifetime. Figure 2 The light blue uncertainty band in part (b) is derived from the variance expression of the posterior distribution.

[0049] To avoid direct confrontation and To address the numerical instability issues introduced by modeling, this embodiment employs a mean-evidence strength parameterization approach. Let the posterior mean be... The strength of evidence is Then we have:

[0050] Under this parameterization, the mean and variance of the Beta distribution are as follows:

[0051] in, This represents the mean; Indicates variance.

[0052] Furthermore, shape parameters can be and The only certainty is: .

[0053] Combination Figure 3 It can be seen that when the inspection information is at a low level for a long time in the early stage of the life cycle (i.e., the value of the inspection information is 0), the inspection information will be affected. A larger value indicates that the model has a high confidence level in the prediction results; as the degree of inspection anomaly increases, As the posterior distribution gradually decreases, it becomes more dispersed, and the uncertainty increases accordingly. Figure 3 The blue curve represents the strength of evidence. The purple curve represents inspection information. The evolution relationship between the two indicates that inspection information mainly affects posterior uncertainty by adjusting the strength of evidence, rather than directly causing abrupt changes in the mean.

[0054] The first path in DPCS-Net is the mean calibration path, which aims to correct system biases without disrupting the original predicted degradation trend. The "original predicted degradation trend" refers to the overall decline in online remaining lifetime predictions as equipment operating time increases; that is, the prediction results decrease as the equipment degradation process progresses. The posterior mean of the calibrated remaining lifetime can be written as:

[0055] in, This is a gate function used to limit the calibration amplitude; Represents the prediction bias drift function; symbol " The symbol "" indicates multiplication. This structure ensures that the calibration results are consistent with the predicted results in terms of the overall degradation trend.

[0056] from Figure 2 The comparison shows that the original prediction curve (i.e., the curve of "original predicted remaining lifetime" in the figure) fluctuates in the later stage. After calibration by the above formula, the posterior mean curve is smoother and closer to the actual degradation trajectory, indicating that the mean calibration path can effectively suppress local bias in the later stage of lifetime.

[0057] The second path in DPCS-Net is the evidence strength modeling path, used to characterize the evolution of the confidence level of the prediction results. Evidence strength can be modeled as a joint function of the prediction results and inspection information, which can be written as:

[0058] in, Indicates by t The strength of the basic evidence determined by the degradation stage at any moment, with the superscript "0" indicating the basic item or the benchmark item before the fusion inspection, rather than the power; For the inspection evidence weighting coefficient and , where the subscript " I This indicates that the parameter is related to inspection information; Indicates the inspection information The mapping function is used to characterize the degree of inspection anomalies.

[0059] It should be noted that inspection information is not directly corrected. Instead, it is through correction Change the concentration of the posterior distribution. Figure 3 The inspection information in the middle of the system showed a jump around approximately 350h and 450h, while The significant decline reflects this mechanism.

[0060] According to S3, a survival consistency constraint is introduced.

[0061] To ensure the physical plausibility of the calibrated remaining lifetime over time, this embodiment introduces a survival consistency constraint, ensuring that the average remaining lifetime after calibration exhibits a non-increasing trend, which can be expressed as:

[0062] Among them, subscript t +1 indicates the current time. t The next adjacent discrete time point; inequality sign " "This means that the average remaining lifetime at subsequent times is not greater than that at the current time"; "Indicates any" t The constraint reflects the physical law that the remaining life of the gearbox will not increase unnecessarily during normal degradation. During training, this constraint is implemented through a soft penalty term, which can be written as:

[0063] in, The subscript "mono" is an abbreviation for monotonic, which means "monotonic constraint loss". Indicates all moments t Summation; This indicates taking the larger value between 0 and the value within the parentheses. When the value is 0, a positive penalty is applied when there is an abnormal increase.

[0064] Combining the mean calibration path, the evidence strength modeling path, and the survival consistency constraint, the training objective can be expressed as:

[0065] in, Represents the total loss function; The subscript "nll" is an abbreviation for negative log-likelihood, representing the negative log-likelihood loss of the Beta distribution; The subscript "mono" also indicates a monotonicity constraint, corresponding to These are constraint weighting coefficients used to balance statistical fit and physical consistency.

[0066] According to S4, calculate the failure probability and construct the interval failure risk.

[0067] After obtaining the posterior distribution, step S4 further maps the probabilistic remaining lifetime result to the risk space. Figure 4 The evolution of failure probability based on the remaining lifetime probability distribution after calibration is presented. From... Figure 4 It is evident that in the early stages of life... and All are relatively high, with a posteriori failure probability close to zero; as decline, The probability of posterior failure decreases, but increases rapidly at the end of the lifespan and exceeds the risk reference threshold, indicating that the amplification of risk comes from the combined effect of "mean degradation + increased uncertainty".

[0068] Let the failure threshold be Define a single-moment failure event as "moment". t If the actual remaining lifetime is not higher than the failure threshold, then the failure probability at a single moment is Beta, distributed in the interval [0, 1]. The cumulative probability, i.e.:

[0069] in, Subscript This indicates failure, or malfunction. Represents a probability operator; Subscript This also indicates that the service is invalid; The parameter is and The cumulative distribution function of the Beta distribution, where the variable before the semicolon ";" is the independent variable and the distribution parameter after it is the distribution parameter.

[0070] At this moment of decision-making If the plan is to be in Maintenance is carried out in a timely manner, including Indicates the length of time waiting for repair. Then, an interval event that fails at least once within the waiting interval can be defined as follows:

[0071] The corresponding interval failure risk can be written as:

[0072] in, express arrive At least one failure event must occur within the interval; Union operation; Indicates the time interval of the event; The subscript "int" is an abbreviation for interval, indicating the risk of interval.

[0073] According to S5, solve for the maximum safe maintenance interval.

[0074] Given a risk threshold The maximum safe maintenance interval is defined as the maximum time length during which maintenance can be delayed while ensuring that the risk of failure within the interval does not exceed the threshold. It can be written as:

[0075] in, The subscript "max" indicates the maximum allowed value. This indicates that the maximum safe maintenance interval depends on the current decision-making time. ;symbol" "Indicates that the conditions are met" All The maximum value in.

[0076] Figure 7The solution results for the rolling optimization under risk constraints are shown, where the purple curve represents... The green curve represents the optimal waiting time. .from Figure 7 It is evident that as equipment degradation deepens, The continuous shrinking indicates that the maximum waiting time allowed by risk constraints is becoming shorter and shorter; while The fact that it remains within the safety boundary indicates that rolling optimization searches for the economically optimal repair time within the risk-feasible region, rather than minimizing costs independently outside the safety boundary.

[0077] Construct an environment reachability function and a time-varying operation and maintenance cost model according to S6.

[0078] In this embodiment, the environmental accessibility function is used to comprehensively characterize the impact of wind speed, temperature, dust, and road conditions on the difficulty of maintenance deployment and the efficiency of on-site construction organization. Multiple soft-gating functions are employed to construct... A ( t To characterize the inhibitory effect of extreme environmental factors on maintenance activities, its expression is:

[0079] in, A ( t () indicates time t Overall environmental accessibility; subscript v , D , T , q These correspond to four components: wind speed, dust, temperature, and road accessibility, respectively. For the Sigmoid function; , , These represent the wind speed sensitivity coefficient, dust sensitivity coefficient, and temperature deviation sensitivity coefficient, respectively. , , These represent the wind speed threshold, dust threshold, and allowable temperature difference threshold, respectively. Indicates the reference ambient temperature; Represents the absolute value function; This indicates the road status item. Figure 5 A schematic diagram of the environmental accessibility function and its components is given. When a certain component decreases significantly, the total accessibility also decreases synchronously, reflecting that the desert environment weakens the feasibility of maintenance.

[0080] In complex environments such as deserts and Gobi, maintenance deployment costs can be expressed as a nonlinear function of environmental accessibility:

[0081] in, Indicates time t The dispatch cost; the subscript "disp" is an abbreviation for dispatch, indicating dispatch; This represents the baseline deployment cost under ideal conditions, with the superscript "0" indicating the ideal conditions. To prevent the lower limit constant from being too small in the denominator; This represents the environmental amplification factor. The formula indicates that when... A ( t The cost of deployment will increase non-linearly as environmental conditions deteriorate.

[0082] On-site operation costs can be expressed as:

[0083] in, Indicates at time t Execute action u The cost of on-site operations; the subscript "work" indicates an operation; Indicates action u The baseline operating cost under ideal conditions, with the superscript "0" indicating that it is under ideal conditions; Indicates the environmental sensitivity coefficient; u This indicates the type of maintenance action, such as preventative maintenance, corrective maintenance, or replacement maintenance. This formula illustrates that the worse the environment, the lower the construction efficiency and the greater the input of labor and equipment.

[0084] In complex environments, the duration of the same maintenance action can be affected by environmental factors. Let the action... u At any moment t The downtime period is:

[0085] in, Indicates action u At any moment t The actual downtime; This indicates the baseline construction period; the superscript "0" indicates the ideal environment. This represents the sensitivity coefficient to the amplification of the construction period. Based on this time-varying downtime, the downtime power generation loss can be expressed as:

[0086] in, The subscript "down" indicates a system shutdown; the minimum integration time is the start time of maintenance. t The upper limit is the time when the shutdown ends. ; Indicates time The electricity price coefficient; This indicates the curtailment factor, with the subscript "curt" being an abbreviation for curtailment. Indicates wind speed The corresponding wind turbine power curve; This indicates that the time interval is integrated over the downtime.

[0087] For joint maintenance of multiple units, this embodiment introduces a collaborative maintenance discount factor:

[0088] in, The subscript "coop" is an abbreviation for cooperation, indicating collaborative maintenance. m This indicates the scale of joint maintenance, specifically the number of gearboxes being maintained simultaneously in the same batch. Indicates the scale effect index; This represents the environmental attenuation coefficient. The above formula shows that as... m Increased costs per unit can be reduced through economies of scale; however, when the environment deteriorates, A ( t When the efficiency of collaborative maintenance decreases, the efficiency of collaborative maintenance will be weakened.

[0089] In summary, a single unit at time t Take action u The total cost at that time can be written as:

[0090] Where the subscript "tot" represents the total cost; variables m Explicitly written into the function's independent variables, this indicates that the total cost is affected by the scale of joint maintenance. Figure 5 The fluctuations in accessibility in the medium environment are the direct source of fluctuations in time-varying costs.

[0091] According to S7, construct the rolling optimization objective function and solve for the optimal maintenance time.

[0092] Within the feasible risk window, this embodiment dynamically selects the repair time with the lowest cost through rolling optimization. Let's assume that at time... Decision-making, if waiting If preventative maintenance is performed after a certain number of time units, the expected cost function can be written as:

[0093] in, Indicates the current decision-making moment Waiting The expected cost; This indicates the cost of preventative maintenance. This indicates the probability of failure occurring during the waiting period. This represents the cost of corrective maintenance. The subscript "PM" stands for preventive maintenance, and the subscript "CM" stands for corrective maintenance. The product in the second item represents the risk-weighted term of "probability of failure during the waiting period" and "corrective maintenance cost after failure".

[0094] The optimal waiting time can be defined as:

[0095] in," "" represents the optimal solution; argmin represents the independent variable that minimizes the objective function; constraint interval This indicates that the search is only conducted within the risk-feasible window.

[0096] Figure 7 Showing and The contrast relationship. Figure 8 This further illustrates the coupled changes in expected cost and interval failure risk: in the early to mid-life, interval failure risk is low, and expected cost changes slowly; in the later life, interval failure risk rapidly approaches the upper limit of risk constraints, and if maintenance is further delayed, expected cost rises sharply. Therefore, this embodiment achieves a dynamic balance between risk constraints and economic efficiency by identifying regions where risk and cost deteriorate simultaneously through rolling optimization.

[0097] According to S8, output maintenance actions.

[0098] Get the optimal waiting time The optimal maintenance time can then be expressed as:

[0099] in, For optimal maintenance time, This is the current decision-making moment. The system will [represent the current moment]. The corresponding maintenance actions are sent to the operation and maintenance execution terminal for the purpose of arranging personnel, spare parts, and hoisting resources.

[0100] See Figure 6The diagram illustrates the evolution trajectory of the gearbox's posterior remaining life under different maintenance actions. The light red dashed line represents the warning threshold; the green dashed line represents the PM threshold, triggered when the gearbox's remaining life degradation curve reaches 0.2; and the blue dashed line represents the RP threshold, triggered when the gearbox's remaining life degradation curve reaches 0. When the posterior remaining life decreases to the preventative maintenance range, the system prioritizes outputting the PM action; if the degradation further deepens and approaches the replacement threshold, the system outputs the RP action. The PM and RP markers in the diagram indicate that this embodiment does not simply trigger mechanically at the threshold, but rather dynamically reshapes the posterior remaining life trajectory under the combined constraints of risk, environment, and cost, thereby delaying failure and controlling total cost.

[0101] In summary, this embodiment expands online remaining lifetime prediction results from single-point values ​​to probability distributions by constructing a DPCS-Net calibration mechanism based on a Beta posterior distribution; it achieves an explicit expression of the maintenance decision risk boundary through joint modeling of single-moment failure probability, interval failure risk, and maximum safe maintenance interval; and it enables dynamic adjustment of maintenance timing based on degradation status and operating environment through the coordinated design of environmental accessibility functions, time-varying cost models, and rolling optimization objective functions. Figures 2 to 8 As can be seen, this embodiment can not only improve the prediction of jitter in the later stage of life and introduce reasonable uncertainty expression, but also output more engineering-executable maintenance strategies in complex environments.

[0102] Example 2: Reference Figure 9 As shown in the figure, an embodiment of the present invention discloses a wind turbine gearbox maintenance decision optimization system in a desert environment, comprising: Data acquisition module: used to acquire online remaining life prediction results, inspection information, environmental status information, and operation and maintenance cost related information of wind turbine gearboxes, and to perform time alignment, format unification, and preprocessing on the multi-source heterogeneous data; the environmental status information includes wind speed, ambient temperature, dust intensity index, and road accessibility index, and the operation and maintenance cost related information includes benchmark deployment cost, benchmark operation cost, maintenance period, electricity price information, wind curtailment information, and collaborative maintenance scale; The posterior calibration module is used to input the online remaining lifetime prediction results and inspection information into a dual-path calibration network to perform posterior calibration on the online remaining lifetime prediction results and obtain the calibrated remaining lifetime probability distribution. The dual-path calibration network includes a mean calibration path and an evidence strength modeling path. The mean calibration path is used to correct the systematic bias of the original remaining lifetime prediction results, and the evidence strength modeling path is used to fuse inspection information and characterize the change in the credibility of the prediction results. Consistency constraint module: used to introduce survival consistency constraints in the post-calibration process, so that the remaining lifetime result after calibration meets the overall non-incremental evolution law of the equipment degradation process, so as to ensure that the remaining lifetime result conforms to the physical consistency of the wind turbine gearbox degradation process. Risk assessment module: used to calculate the failure probability of wind turbine gearbox at each time point based on the calibrated remaining life probability distribution, and further construct the interval failure risk within the future candidate maintenance interval; also used to solve the maximum safe maintenance interval according to the preset risk threshold and generate risk-feasible maintenance window; Environmental Modeling Module: Used to construct an environmental accessibility function based on wind speed, ambient temperature, dust intensity index, and road accessibility index, in order to characterize the comprehensive impact of current environmental conditions on the feasibility of maintenance deployment and the difficulty of on-site construction organization; Cost calculation module: used to establish a time-varying operation and maintenance cost model based on the environmental accessibility function and maintenance action type; the time-varying operation and maintenance cost model includes a dispatch cost model, an on-site operation cost model, and a power outage loss model; the cost calculation module is also used to construct a collaborative maintenance discount model based on the scale of joint maintenance, so as to characterize the unit maintenance cost reduction effect formed when multiple wind turbine units are jointly maintained; Rolling optimization module: Used to construct a rolling optimization objective function based on the time-varying operation and maintenance cost model and the interval failure risk within the risk-feasible maintenance window, and solve for the optimal waiting time and optimal maintenance time under the current decision cycle; Decision output module: It is used to output the corresponding maintenance action according to the optimal maintenance time, and send maintenance suggestions to the wind farm operation and maintenance execution terminal to guide the scheduling of maintenance personnel, spare parts resources and hoisting resources.

[0103] The system can be deployed on the wind farm control center server or cloud-based intelligent operation and maintenance platform. It can automatically trigger calculations according to preset time cycles or when inspection information or meteorological conditions change, and output maintenance suggestions, planning windows and resource scheduling information to the operation and maintenance execution terminal.

[0104] This embodiment applies to a predictive decision-making platform for intelligent operation and maintenance of wind turbine gearboxes in barren and desert environments. It includes a data acquisition module, a post-calibration module, a consistency constraint module, a risk assessment module, an environmental modeling module, a cost calculation module, a rolling optimization module, and a decision output module. This platform integrates online remaining life prediction, inspection information fusion, failure risk assessment, environmental accessibility modeling, and maintenance decision optimization into a unified processing flow. After all modules work together, the remaining life prediction results of the gearbox can be expanded from point values ​​to a probability distribution, and further generate risk-feasible maintenance windows, optimal maintenance times, maintenance action suggestions, and resource scheduling information. This information is used to assist in the formulation of maintenance plans, spare parts preparation, and shift scheduling for gearboxes in barren and desert wind farms.

[0105] In summary, this invention discloses a method and system for optimizing maintenance decisions of wind turbine gearboxes in desert environments. The method includes: acquiring online remaining lifetime predictions, inspection information, environmental status information, and operation and maintenance cost-related parameters of the wind turbine gearbox; constructing a dual-path calibration network integrating online prediction results and inspection information, performing posterior probability calibration on the point-value remaining lifetime to obtain a posterior distribution of remaining lifetime that satisfies survival consistency constraints; calculating the single-moment failure probability and interval failure risk based on the posterior distribution, and solving for the maximum safe maintenance interval under a given risk threshold; constructing an environmental accessibility function combining wind speed, temperature, dust, and road conditions, and further establishing a time-varying operation and maintenance cost model including deployment costs, operating costs, downtime power generation losses, and collaborative maintenance discounts; and finally, determining the optimal maintenance time and maintenance action within a risk-feasible window using a rolling optimization strategy. This invention expands the remaining life prediction results from point values ​​to a probability distribution, enabling simultaneous prediction deviation correction, uncertainty quantification, and risk-constrained decision-making. By introducing environmental accessibility and collaborative maintenance mechanisms, it achieves adaptive optimization of wind turbine gearbox maintenance intervals in desert and Gobi environments. It can be directly applied to the intelligent operation and maintenance platform for wind turbine gearboxes in desert and Gobi areas, generating risk-feasible maintenance windows, optimal maintenance times, and maintenance action suggestions, and assisting in maintenance plan formulation and resource scheduling.

[0106] Example 3: Reference Figure 10 As shown, this embodiment of the invention also provides an electronic device that can perform the above-described method. The electronic device may include a processor 10, a memory 11, a communication bus 12, and a communication interface 13, and may also include a computer program stored in the memory 11 and capable of running on the processor 10.

[0107] In some embodiments, the processor 10 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits packaged with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 10 is the control unit of the electronic device, connecting various components of the entire electronic device through various interfaces and lines. It executes programs or modules stored in the memory 11 and calls data stored in the memory 11 to perform various functions of the electronic device and process data.

[0108] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, electronic devices, or computer program products, etc. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0109] It should be noted that the word "comprising" does not exclude the presence of components or steps not listed in the claims. The words "a" or "an" preceding a component do not exclude the presence of a plurality of such components. This invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer.

[0110] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0111] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for optimizing maintenance decisions of wind turbine gearboxes in a desert environment, characterized in that, The method includes: Obtain online remaining life prediction results, inspection information, environmental status information, and operation and maintenance cost information for wind turbine gearboxes; The online remaining lifetime prediction results and inspection information are input into the dual-path calibration network to perform posterior calibration on the online remaining lifetime prediction results, and the calibrated remaining lifetime probability distribution is obtained. Introducing survival consistency constraints during the post-calibration process ensures that the remaining lifetime after calibration exhibits an overall non-increasing trend. Based on the calibrated remaining lifetime probability distribution, the failure probability of the wind turbine gearbox is calculated, and the interval failure risk is constructed. Based on the preset risk threshold, determine the maximum safe maintenance interval at the current decision moment and obtain the risk-feasible maintenance window; An environmental reachability function is constructed based on the environmental state information, and a time-varying operation and maintenance cost model is established by combining operation and maintenance cost-related information and maintenance action types. Within the risk-feasible maintenance window, a rolling optimization objective function is constructed based on the time-varying operation and maintenance cost model and the interval failure risk to solve for the optimal maintenance time; Output the corresponding maintenance action based on the optimal maintenance time.

2. The method according to claim 1, characterized in that, In this method, the environmental status information obtained includes at least one or more of wind speed, ambient temperature, dust intensity index, and road accessibility index; the operation and maintenance cost related information obtained includes at least one or more of deployment cost, operation cost, maintenance period, electricity price information, wind curtailment information, and collaborative maintenance scale.

3. The method according to claim 1, characterized in that, In this method, the dual-path calibration network includes: The mean calibration path is used to correct systematic biases in online remaining lifetime prediction results; The evidence strength modeling path is used to integrate inspection information and characterize the credibility changes of online remaining lifetime prediction results, so as to obtain a remaining lifetime probability distribution that takes into account both the prediction mean and uncertainty.

4. The method according to claim 1, characterized in that, The survival consistency constraint restricts the remaining lifetime result after calibration from increasing in the opposite direction to the equipment degradation pattern over time, so that the calibration result meets the physical consistency of the remaining lifetime of the wind turbine gearbox decreasing as a whole during operation.

5. The method according to claim 1, characterized in that, The interval failure risk is used to characterize the probability that at least one failure event will occur between the current decision time and the candidate maintenance time, and the calibrated remaining lifetime uncertainty is directly introduced into the maintenance decision-making process.

6. The method according to claim 1, characterized in that, The maximum safe maintenance interval is the maximum time length for which maintenance can be delayed under a given risk threshold; the risk-feasible maintenance window is determined based on the current decision time and the maximum safe maintenance interval.

7. The method according to claim 1, characterized in that, The environmental accessibility function characterizes the comprehensive impact of environmental conditions on the feasibility of maintenance deployment and the difficulty of construction organization; the time-varying operation and maintenance cost model includes deployment cost, on-site operation cost and power generation loss during downtime; wherein, the deployment cost and the on-site operation cost are dynamically adjusted with changes in environmental accessibility, and the power generation loss during downtime is determined based on the maintenance period, wind turbine power level, electricity price fluctuations and wind curtailment situation.

8. The method according to claim 7, characterized in that, In this method, a collaborative maintenance discount model is also constructed to characterize the unit maintenance cost reduction effect brought about by resource sharing when multiple wind turbine units are jointly maintained. The collaborative maintenance discount model is environmentally dependent and describes the weakening effect of desert and barren environments on the efficiency of joint maintenance of multiple units. In combination with the collaborative maintenance discount model, a time-varying operation and maintenance cost model is established, which includes dispatch cost, on-site operation cost, power outage loss and collaborative maintenance discount.

9. The method as described in claim 1, characterized in that, The rolling optimization objective function aims to minimize the total maintenance cost within the risk-feasible window. It is repeatedly solved as the online remaining life prediction results, inspection information, environmental status information, and cost parameters are updated, and the optimal maintenance time is dynamically updated.

10. A decision-making optimization system for wind turbine gearbox maintenance in a desert environment, characterized in that, The system applies the method as described in any one of claims 1–9, the system comprising: Data acquisition module: used to acquire online remaining life prediction results, inspection information, environmental status information, and operation and maintenance cost information of wind turbine gearboxes; Posterior calibration module: used to perform posterior calibration on the online remaining lifetime prediction results and output the calibrated remaining lifetime probability distribution; Risk assessment module: used to calculate failure probability, interval failure risk, and maximum safe maintenance interval based on the calibrated remaining lifetime probability distribution; Environment Modeling Module: Used to construct environmental reachability functions based on environmental state information; Cost calculation module: used to establish a time-varying operation and maintenance cost model based on environmental accessibility function combined with operation and maintenance cost related information and maintenance action type; Rolling optimization module: used to find the optimal maintenance time within the risk-feasible maintenance window; Decision output module: Used to output the corresponding maintenance action based on the optimal maintenance time, and send maintenance suggestions to the wind farm operation and maintenance execution terminal.

Images