Operation and maintenance robot for monitoring quality of wind power gear oil and purifying and filling based on Internet of Things

Abstract

The invention relates to the technical field of operation and maintenance of wind generating sets, and discloses a wind power gear oil quality monitoring and purification filling operation and maintenance robot based on the Internet of Things, which comprises an operation and maintenance robot and a digital center layer, the operation and maintenance robot integrates an oil path and valve body module, a filter module, a filler module, a high-precision sensor module and a local control center module, and the digital center layer comprises an AI decision engine; the method comprises the following steps that the operation and maintenance robot periodically collects oil intrinsic data and reports the oil intrinsic data; the digital center layer fuses the multi-modal data, drives an AI decision engine to analyze and generate an operation and maintenance decision instruction; and the operation and maintenance robot receives the instruction and autonomously executes operations such as standby, purification or oil change. In the purification operation, the system calculates the filtering gradient in real time and reports the filtering gradient to optimize the model. Predictive maintenance of the wind power gear oil is realized, the operation and maintenance cost is reduced, and the operation reliability of a wind turbine generator is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wind turbine generator set operation and maintenance, and in particular to an Internet of Things-based wind power gear oil quality monitoring and purification and filling operation and maintenance robot. Background Art

[0002] Wind turbine gearboxes are their core transmission components, and the quality of their lubricant is directly related to the unit's operating efficiency, reliability, and service life. However, due to the complex field conditions of wind turbines and the internal working conditions of gearboxes, traditional manual sampling and inspection or fixed-cycle oil change strategies have many drawbacks.

[0003] On the one hand, manual inspections and sampling are costly and inefficient, and cannot reflect the oil status in real time, resulting in delayed maintenance or excessive maintenance. On the other hand, simply changing the oil based on the number of operating hours often fails to fully utilize the actual performance life of the oil, resulting in a waste of resources. At the same time, once the oil deteriorates unexpectedly during the cycle, it may cause premature wear or even failure of the gearbox, leading to unplanned downtime, seriously affecting the wind farm's power generation revenue.

[0004] Existing technologies still face challenges in achieving real-time, accurate, and intelligent monitoring, evaluation, and on-demand maintenance of gear oil. There is a lack of an integrated solution that can deeply integrate data collection, status assessment, and autonomous operation and maintenance.

[0005] Therefore, the present invention proposes a wind power gear oil quality monitoring and purification filling operation and maintenance robot based on the Internet of Things to solve the shortcomings of the existing technology. Summary of the Invention

[0006] In response to the shortcomings of the existing technology, the present invention provides a wind turbine gear oil quality monitoring and purification and filling operation and maintenance robot based on the Internet of Things, which solves the problems of poor real-time performance, inaccurate maintenance, waste of resources, and lack of integrated intelligent decision-making and autonomous execution capabilities in the operation and maintenance of wind turbine gearbox lubricating oil.

[0007] The first aspect of the present invention provides a wind power gear oil quality monitoring and purification and filling operation and maintenance robot based on the Internet of Things.

[0008] The robot includes: oil circuit and valve body module, filter module, filler module, high-precision sensor module, local control center module and digital hub layer.

[0009] The oil circuit and valve body module has a built-in motor pump and multiple sets of electromagnets, which are used to build oil circulation, purification, filling and waste discharge paths according to instructions.

[0010] The filter module has pipelines connected to the oil circuit and the valve body module, and is internally provided with a coarse filter and a fine filter for filtering the oil in the circulation path.

[0011] The filler module, whose pipeline is also connected to the oil circuit and the valve body module, is used to fill new oil into the gearbox.

[0012] The high-precision sensor module is used to collect key physical and chemical properties of the oil. It includes an oil quality sensor C3 for monitoring the oil's water activity (AW) and cleanliness level (NC), and a pressure sensor located at the inlet and outlet of the filter module to monitor real-time differential pressure.

[0013] The local control center module is electrically connected to the oil circuit and valve module and the high-precision sensor module. It is configured to: collect intrinsic oil data output by the high-precision sensor module; transmit this data to the digital hub layer via the IoT communication module; receive operation and maintenance decision instructions issued by the digital hub layer; and, based on these operation and maintenance decision instructions, control the oil circuit and valve module, the filter module, and the filler module to perform corresponding operation and maintenance operations.

[0014] The digital hub layer, as a remote data processing and decision-making unit, is used to perform the following functions: obtaining mechanical dynamic data and operating condition data from the main control SCADA system of the wind turbine generator set; fusing the received oil intrinsic data, the mechanical dynamic data and the operating condition data to form multimodal data; the digital hub layer has a built-in AI decision engine, which drives the pre-trained comprehensive health index CHI model to process the multimodal data to generate the operation and maintenance decision instructions.

[0015] In an optional technical solution, to accurately predict oil life, the AI ​​decision engine also runs a pre-trained Remaining Useful Life (RUL) model. This RUL model uses the time series health index output by the Comprehensive Health Index (CHI) model as input data and outputs the oil's Remaining Useful Life (RUL) through time series analysis and prediction. This RUL prediction serves as a key basis for generating the final operation and maintenance decision instructions.

[0016] In an optional technical solution, in order to strike a balance between multiple operation and maintenance objectives, the AI ​​decision engine generates the operation and maintenance decision instructions by solving a multi-objective optimization problem. The multi-objective optimization problem aims to find an optimal operation and maintenance decision. , so that the overall operation and maintenance cost function Minimize. This problem can be specifically expressed as follows: ; Where, is the comprehensive health index, for the stated remaining useful life, which are inputs to decision making; Represents the energy consumption of operation and maintenance operations; Represents resource consumption of consumables such as filter elements; Represents synergy with the overall external maintenance plan of the wind farm; , , , is the weight coefficient of each optimization objective; , , , is the corresponding cost function. By solving this optimization problem, we can output operation and maintenance decision instructions that take into account equipment health, operation and maintenance energy consumption, resource consumption, and maintenance coordination.

[0017] In an optional technical solution, in order to dynamically monitor the filter performance degradation process, when the local control center module controls the execution of the purification operation, the module is also used to monitor the real-time pressure difference data monitored by the pressure sensor at high frequency. Calculate filter gradient The filter gradient The calculation formula is: ; The filter gradient Characterizes the rate of change of the filter's clogging degree over time during the purification process. After calculation, the local control center module will filter the gradient The digital core layer receives the data and filters the gradient data. The multimodal data is incorporated into the multimodal data for processing by the comprehensive health index (CHI) model, thereby enabling the model to more accurately assess the health status of the system.

[0018] In an optional technical solution, in order to improve the generalization ability and accuracy of the AI ​​decision-making model, the digital hub layer is also used to evaluate the model uncertainty of the AI ​​decision-making engine online. When the monitored model uncertainty value exceeds a preset threshold, it indicates that the model has low confidence in its current judgment. At this time, the digital hub layer will generate and issue active diagnostic instructions. After receiving the instruction, the local control center module controls the robot to perform specific exploratory operations to obtain new data that can specifically eliminate the uncertainty of the current model, and upload the new data for iterative optimization of the model.

[0019] In an optional technical solution, the operation and maintenance operations controlled and executed by the local control center module can be specifically defined as the following three states: When the water activity AW and the cleanliness level NC measured by the oil quality sensor C3 meet the first preset condition, it is determined that the oil quality is in good condition, and all functional modules of the robot are controlled to enter a standby state.

[0020] When the water activity AW and the cleanliness level NC do not meet the first preset condition but meet the second preset condition, it is determined that the oil quality is deteriorated but can be purified, and the oil circuit, valve body module and filter module are controlled to perform a cyclic purification operation.

[0021] When the water activity AW or the cleanliness level NC meets a third preset condition, it is determined that the oil quality has been severely deteriorated or contaminated, and the oil circuit, valve body module, and filler module are controlled to perform an oil change operation of draining the waste oil and filling with new oil.

[0022] A second aspect of the present invention provides a method for monitoring the quality of wind power gear oil and for purifying, filling, and maintaining it based on the Internet of Things. The method is applied to the robot described in any of the aforementioned technical solutions and comprises the following steps: The local control center module collects oil intrinsic data through the high-precision sensor module connected to it, and sends the oil intrinsic data to the digital central layer through the Internet of Things.

[0023] The digital central layer obtains mechanical dynamic data and operating condition data from the fan master control SCADA system, and fuses the oil intrinsic data, mechanical dynamic data and operating condition data into multimodal data.

[0024] The digital hub layer drives its built-in AI decision engine to run the pre-trained comprehensive health index CHI model to process the multimodal data and generate operation and maintenance decision instructions.

[0025] The local control center module receives the operation and maintenance decision instruction and controls each functional module of the robot to execute the operation and maintenance operation corresponding to the instruction.

[0026] Prior to generating the operation and maintenance decision instruction, the method further includes running a pre-trained RUL model. This model uses the time series health index output by the CHI model as input to predict the RUL of the oil. The generation of the operation and maintenance decision instruction is further based on this predicted RUL.

[0027] The step of generating the operation and maintenance decision instructions is specifically accomplished by solving a multi-objective optimization problem. This multi-objective optimization problem uses the comprehensive health index (CHI) and remaining useful life (RUL) as decision inputs, and comprehensively evaluates equipment health, operation and maintenance energy consumption, consumable resource consumption, and coordination with external wind farm maintenance plans to output the final operation and maintenance decision instructions.

[0028] The method also includes the step of calculating a filtration gradient based on the pressure difference data monitored by the pressure sensor at high frequency during the purification operation; and incorporating the calculated filtration gradient into the multimodal data as dynamic process data for use by the comprehensive health index CHI model during processing.

[0029] The present invention provides an IoT-based wind power gear oil quality monitoring and purification and filling operation and maintenance robot. It has the following beneficial effects: 1. This invention constructs a comprehensive health index (CHI) model and a remaining useful life (RUL) model, and integrates oil intrinsic data, mechanical dynamic data, and operating condition data to form multimodal data for comprehensive analysis. This allows for accurate assessment of the health status of wind turbine gear oil and related components and prediction of future decline trends. This predictive maintenance strategy changes the previous operation and maintenance model that relied on fixed cycles or post-event responses. It can identify potential risks and provide early warnings before failures occur, improving the foresight and planning of operation and maintenance, thereby ensuring the long-term stable operation of wind turbine generator sets.

[0030] 2. This invention introduces a decision-making mechanism for solving multi-objective optimization problems, taking into account multiple dimensions such as equipment health, O&M energy consumption, consumable resource consumption, and coordination with external wind farm maintenance plans to generate optimal O&M decision instructions. This approach avoids the problems of excessive O&M costs or insufficient maintenance that can result from traditional single-threshold judgments. It maximizes the overall economic benefits of O&M activities while ensuring equipment safety and reliability, transforming O&M decisions from being solely driven by technology to being driven by both technology and economics.

[0031] 3. This invention proposes a dynamic monitoring method for calculating filtration gradients. This method feeds dynamic process data reflecting filter performance degradation during purification operations into a comprehensive health index model, significantly improving the accuracy and real-time performance of system status assessments. Furthermore, by assessing model uncertainty and triggering proactive diagnostic instructions, the robot can autonomously acquire new data to address model blind spots. This creates a data-driven self-learning and iterative optimization loop, significantly enhancing the generalization and reliability of the AI ​​decision-making engine.

[0032] 4. This invention integrates high-precision sensors, execution modules, and a local control center module into an integrated operation and maintenance robot, and collaborates with a remote digital hub layer through the Internet of Things. This creates a complete automated, intelligent, closed-loop operation and maintenance system. This system not only significantly reduces the frequency of manual inspections and on-site maintenance, alleviating the need for personnel to operate in high-risk environments, but also avoids human interference through standardized, autonomous operating procedures, ensuring efficient and consistent operation and maintenance of wind turbine gear oil quality monitoring, purification, and refueling. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 Schematic diagram of the overall system architecture of the present invention; Figure 2 This is a schematic diagram of the internal structure and functional modules of the operation and maintenance robot of the present invention; Figure 3 This is a hydraulic principle diagram of the oil circuit system of the operation and maintenance robot of the present invention; Figure 4 This is a flow chart of the operation and maintenance method of the present invention.

[0034] Among them, 100, operation and maintenance robot; 110, oil circuit and valve body module; 120, filter module; 130, filler module; 140, high-precision sensor module; 150 local control center module; 160, digital central layer; 161, AI decision engine. DETAILED DESCRIPTION

[0035] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0036] Reference Figure 1 An embodiment of the present invention provides an IoT-based wind turbine gear oil quality monitoring, purification, and refueling operation and maintenance system. This system physically comprises an integrated operation and maintenance robot 100 deployed at the wind turbine generator site, and a digital backbone layer 160 deployed on a remote server. The operation and maintenance robot 100 and the digital backbone layer 160 establish an interactive link for data and instructions via the IoT communication network, forming a closed-loop control system that integrates data collection, intelligent analysis, decision-making, and autonomous execution.

[0037] Reference Figure 2The hardware system of the maintenance robot 100 is integrated into a unified housing. It includes an oil circuit and valve module 110, a filter module 120, a filler module 130, a high-precision sensor module 140, and a local control center module 150. These modules work together to perform specific physical maintenance operations.

[0038] The digital backbone layer 160 is the remote data processing and decision-making unit of the system, and its core is an AI decision engine 161. This engine is responsible for receiving and processing the data uploaded by the operation and maintenance robot 100, and generating operation and maintenance decision instructions and sending them to the operation and maintenance robot 100. The generation of operation and maintenance decisions is achieved by solving a multi-objective optimization problem, which aims to find an optimal operation and maintenance decision. , so that a comprehensive cost function minimize.

[0039] In order to achieve a more accurate dynamic perception of the system status, the present invention also introduces the concept of filtration gradient. During the purification operation, the system uses the real-time pressure difference data across the filter module 120 monitored by the pressure sensor in the high-precision sensor module 140. , calculate the filter gradient .

[0040] Reference Figure 4 The embodiment of the present invention provides a wind power gear oil quality monitoring and purification and filling operation and maintenance method based on the Internet of Things, which includes the following steps: S1. The local control center module 150 controls the high-precision sensor module 140 to collect oil intrinsic data, and sends the data to the digital hub layer 160 through the Internet of Things.

[0041] S2. The digital central layer 160 obtains mechanical dynamic data and operating condition data from the wind turbine master control SCADA system, and fuses these two data with the received oil intrinsic data to form multimodal data.

[0042] S3. The digital hub layer 160 drives its AI decision engine 161 to run the pre-trained comprehensive health index CHI model and remaining service life RUL model, processes the multimodal data, and finally generates operation and maintenance decision instructions by solving the aforementioned multi-objective optimization problem.

[0043] S4. The local control center module 150 receives the operation and maintenance decision instructions issued by the digital hub layer 160, and controls the oil circuit and valve body module 110, the filter module 120 and the filler module 130 and other hardware according to the instructions to perform corresponding operation and maintenance operations.

[0044] Reference Figure 2 and Figure 3, the operation and maintenance robot 100 and its internal functional modules in the embodiment of the present invention are described in detail. The operation and maintenance robot 100 integrates all functional modules into a unified housing to form an integrated device, which is easy to install and deploy in the narrow space of a wind turbine generator set.

[0045] Reference Figure 3 The oil circuit and valve body module 110 is the core execution unit for realizing various flow path switching of oil inside the robot and between the robot and the external gearbox oil tank. The module includes a motor pump and a group of solenoid valves, specifically including the first solenoid valve Z1, the second solenoid valve Z2, the third solenoid valve Z3, the fourth solenoid valve Z4 and the fifth solenoid valve Z6. The motor pump provides power for fluid circulation in the entire oil circuit system. Each solenoid valve acts as an electrically controlled switch, and its power-on and power-off states are controlled by the local control center module 150. Through different combined actions, a fluid circuit that meets specific operation and maintenance needs is constructed. For example, a detection circuit for oil state sampling, a circulation filtration circuit for oil purification, a discharge circuit for waste oil discharge, and a filling circuit for new oil filling or sensor calibration can be constructed.

[0046] The inlet and outlet of the filter module 120 are connected to corresponding pipeline ports on the oil circuit and valve body module 110. Inside the module, a coarse filter and a fine filter are arranged in series along the oil flow direction. The coarse filter removes larger particles from the oil to protect the downstream fine filter and sensors. The fine filter performs high-precision filtration, removing tiny suspended particles from the oil. It is a core component that performs purification operations and improves the cleanliness level of the oil.

[0047] The filler module 130 houses an independent oil reservoir for storing fresh gear oil. Its outlet is connected to the oil circuit and valve module 110, and its connection to the main oil circuit is controlled by the fifth solenoid valve Z6. This module's function is to refill the wind turbine gearbox with fresh oil during oil changes, or during system initialization, injecting a small amount of fresh oil upon command to flush and calibrate the oil quality sensor in the high-precision sensor module 140.

[0048] The high-precision sensor module 140 is used to obtain key physical and chemical parameters that characterize the oil and system status in real time. This module includes: The oil quality sensor C3 is installed on the main detection pipeline and is used to monitor the water activity AW and cleanliness level NC of the oil flowing through it in real time.

[0049] The differential pressure sensor has two pressure measuring points respectively arranged at the inlet and outlet pipes of the filter module 120, and is used to monitor the real-time pressure difference at both ends of the filter module 120 at high frequency. The pressure difference data is used to calculate the filtration gradient. basis.

[0050] The temperature sensor, whose probe extends into the gearbox oil tank, is used to monitor the real-time temperature of the oil and provide a basis for judging the temperature conditions for subsequent operations.

[0051] The liquid level sensor is also installed in the gearbox oil tank to monitor the real-time liquid level of the oil tank to prevent it from falling below the lower limit during oil discharge operation and to prevent it from rising above the upper limit during oil filling operation.

[0052] The local control center module 150 is the local control core of the operation and maintenance robot 100 and can be implemented as a programmable logic controller (PLC) or embedded system. This module establishes electrical connections through electrical signal lines with the motor pump and all solenoid valves in the oil circuit and valve module 110, the injector module 130, and all sensors in the high-precision sensor module 140. Its functions include receiving and processing all sensor signals from the high-precision sensor module 140; executing operation and maintenance decision instructions issued by the remote digital hub layer and converting them into low-level timing control signals for the motor pump and solenoid valves; and packaging the collected raw sensor data and transmitting it to the remote digital hub layer via its built-in IoT communication unit.

[0053] Reference Figure 1 , the digital backbone layer 160 and its core functions in the embodiment of the present invention are described in detail. The digital backbone layer 160 is deployed in the form of a cloud server cluster or edge computing node, serving as the remote data processing and decision-making unit of the system, and an AI decision engine 161 runs inside it.

[0054] The primary function of the digital backbone layer 160 is to perform multimodal data fusion. The specific process is as follows: The digital hub layer 160 stably receives the oil intrinsic data uploaded by the local control center module 150 of the operation and maintenance robot 100 through the Internet of Things interface. This data includes water activity AW, cleanliness level NC, oil temperature and real-time pressure difference. At the same time, the digital hub layer 160 establishes communication with the main control SCADA system of the wind turbine generator set through the data interface, and obtains mechanical dynamic data related to the gearbox operating status (such as gearbox vibration, torque, speed) and operating condition data (such as wind speed, generator active power). Ultimately, the digital hub layer 160 time-aligns, unifies the format and structures the oil intrinsic data, mechanical dynamic data and operating condition data of different sources, formats and frequencies, and integrates them into a unified multimodal data set as a standardized input for subsequent AI model analysis and processing.

[0055] The AI ​​decision engine 161 internally runs a pre-trained Comprehensive Health Index (CHI) model. This model (e.g., a deep neural network model) receives the multimodal dataset as input and, by learning and analyzing the nonlinear relationships within the multi-dimensional, heterogeneous data, outputs a quantized, normalized Comprehensive Health Index (CHI). This CHI value comprehensively reflects the overall health of the wind turbine gearbox oil system at the current moment. The model continuously outputs CHI values ​​at a preset frequency, forming a time series health index that characterizes the evolution of the system's health status over time.

[0056] The AI ​​decision engine 161 also runs a pre-trained Remaining Useful Life (RUL) model. This model, based on a time series prediction algorithm (e.g., a long short-term memory (LSTM) network or a gated recurrent unit (GRU), receives as input the time series health index output by the comprehensive health index (CHI) model. By analyzing the historical decline trajectory and decay pattern of the health index, the RUL model predicts the remaining effective service life required for the oil to return from its current state to a performance level below a preset failure threshold, i.e., the remaining useful life (RUL) of the oil. This RUL prediction provides a critical, forward-looking basis for subsequent operation and maintenance decisions.

[0057] The AI ​​decision engine 161 generates the final operation and maintenance decision instructions by solving a multi-objective optimization problem. This process aims to balance multiple operation and maintenance goals such as equipment health, economic costs, and planning coordination. The multi-objective optimization problem is specifically expressed as follows: ; Where, is the operation and maintenance decision variable to be solved, and its value range covers all preset operation and maintenance operations (for example: standby, perform purification, perform oil change); Yes and decision The goal of optimization is to find the decision that minimizes the value of the associated comprehensive cost function. ; is the current health index output by the comprehensive health index model; In the form of , it means that the worse the health status, the higher the cost; It is an executive decision The corresponding expected energy consumption; It is an executive decision The resulting consumption of consumable resources, such as the reduction in the life of the filter element; It is a decision An indicator of the compatibility of the O&M activity represented with the existing external maintenance plan for the wind farm area. If the activity can be incorporated into the existing plan, the cost is low; if unplanned maintenance is required, the cost is high. , , , It is the preset weight coefficient of each optimization goal. Its value is configured according to the overall operation and maintenance strategy of the wind farm and is used to adjust the priority between different optimization goals. , , , are respective cost functions used to convert inputs of different physical meanings (such as health index, energy consumption joules, number of consumables, and synergy score) into a unified, additively comparable cost metric.

[0058] By solving the optimization problem, the AI ​​decision engine 161 can output an optimal operation and maintenance decision instruction that takes into account multiple factors under current conditions.

[0059] Reference Figure 4 , and combined with Figures 1 to 3 , the overall system workflow of one embodiment of the present invention is described in detail. This process closely combines the local autonomous operation of the operation and maintenance robot 100 with the remote intelligent decision-making of the digital backbone layer 160.

[0060] During the initialization and self-calibration phase, the system performs initialization and self-calibration when the maintenance robot 100 is first connected to the wind turbine gearbox oil system, or after the fine filter element in the filter module 120 has been replaced. The local control center module 150 controls the filler module 130 and energizes the fifth solenoid valve Z6. A small amount of new oil is injected from the filler module 130 into the oil circuit to flush the sensing surface of the oil quality sensor C3 in the high-precision sensor module 140, eliminating interference from residual contaminants or old oil. After flushing, this new calibration oil flows back to the gearbox oil tank through a dedicated pipeline.

[0061] During the periodic monitoring and data reporting phase, after initialization is complete, the system enters periodic monitoring mode. The local control center module 150 automatically wakes up according to a preset time period (e.g., 24 hours). The system first checks the preconditions for operation, for example, the oil temperature measured by the temperature sensor is not lower than a preset starting temperature threshold (e.g., 20°C), and the liquid level measured by the liquid level sensor is within the normal range. Once the conditions are met, the local control center module 150 starts the motor pump in the oil circuit and valve body module 110, and controls the first solenoid valve Z1 and the third solenoid valve Z3 to be energized, establishing an oil detection path. The oil in the gearbox is sucked in and flows through the oil quality sensor C3. The local control center module 150 collects oil intrinsic data (water activity AW, cleanliness level NC) and other sensor data, and reports this data to the digital hub layer 160 via the Internet of Things.

[0062] During the cloud-based analysis and decision-making phase, after receiving the data reported by the operation and maintenance robot 100, the digital hub layer 160 drives its AI decision engine 161 to perform analysis and decision-making. First, the comprehensive health index CHI model and the remaining service life RUL model are run to process the multimodal data that integrates the oil intrinsic data and SCADA data, and output the current CHI value and the predicted RUL value. Subsequently, the AI ​​decision engine 160 uses CHI and RUL as key inputs and substitutes them into the aforementioned multi-objective optimization problem model for solution, ultimately generating a clear operation and maintenance decision instruction, which is one of standby, purification, or oil change.

[0063] During the local autonomous operation and maintenance execution phase, the local control center module 150 receives and parses the operation and maintenance decision instructions issued by the digital hub layer 160, and controls each hardware module to perform corresponding physical operations according to the instructions.

[0064] Case 1 (Standby): When the received command indicates standby, the corresponding first preset condition is met. That is, the digital hub layer 160 determines that the oil is in good health (for example, the CHI value is above a high threshold and the RUL value meets long-term operating requirements). The local control center module 150 then powers off the motor pump and all solenoid valves, and the maintenance robot 100 enters a low-power standby state until the next monitoring cycle.

[0065] Case 2 (Purification Operation): When the received instruction is to perform purification, the corresponding second pre-set condition is met. The digital central layer 160 determines that the oil has deteriorated but its performance can be restored through filtration. The local control center module 150 energizes the first solenoid valve Z1, the second solenoid valve Z2, and the fourth solenoid valve Z4, establishing a circulation purification loop. Oil is drawn from the gearbox oil tank, flows through the coarse filter and the fine filter, and then passes through the oil quality sensor C3 for real-time monitoring before returning to the oil tank.

[0066] During the purification process, the local control center module 150 collects data from the differential pressure sensor at high frequency and calculates the filter gradient in real time according to the following formula: : ; Calculated filter gradient This dynamic process data is continuously reported to the digital hub layer 160 for real-time correction of the CHI model. If the reading of the oil quality sensor C3 still does not meet the preset cleanliness target after a preset purification time (e.g., one hour), the system will record the maintenance need to replace the fine filter and will re-initialize the process to continue the purification operation until the oil quality meets the target and enters the standby state.

[0067] Case 3 (Oil Change): When the received instruction is for an oil change, the corresponding third preset condition is met, meaning the digital central layer 160 determines that the oil is severely degraded or contaminated (for example, the CHI value falls below a low threshold, the RUL value is extremely short, the water activity AW exceeds the safety limit, or the cleanliness level NC is so poor that it cannot be restored through purification). The local control center module 150 first controls the first solenoid valve Z1 and the second solenoid valve Z2 to switch to the waste oil drain position, draining the waste oil in the gearbox into a designated waste oil tank. During the draining process, the level sensor continuously monitors the tank level. Once the level reaches the preset lower limit, the draining operation is immediately stopped to prevent the gearbox from running dry. After the waste oil is drained, the system performs flushing, calibration, and fresh oil filling. During the fresh oil filling process, the level sensor monitors again and stops filling when the level approaches the preset upper limit to prevent overflow. After the oil change is complete, the system enters the purification process to circulate and filter the new oil to ensure optimal operation.

[0068] Active diagnosis and model iteration. At any stage of system operation, if the AI ​​decision engine 161 of the digital central layer 160 evaluates that the prediction results of its internal model (CHI or RUL model) for the current working conditions have high uncertainty (that is, the model confidence is lower than the preset threshold) during analysis, it indicates that the current data is at the cognitive boundary of the model. At this time, the digital central layer 160 will not issue conventional operation and maintenance instructions, but will generate and issue an active diagnosis instruction. After receiving the instruction, the local control center module 150 will control the robot to perform specific exploratory operations, such as adjusting the speed of the motor pump to change the oil flow rate, or performing more intensive short-cycle sampling. These operations are designed to obtain new data that can effectively reduce model uncertainty. After the new data is uploaded, it will be used for online updates or offline retraining of the AI ​​model, thereby realizing the self-evolution and continuous optimization of the decision-making capabilities of the entire system.

[0069] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. The wind power gear oil quality monitoring and purification filling operation and maintenance robot based on the Internet of Things is characterized by: include: Oil circuit and valve body module, including motor pump and multiple sets of electromagnets; a filter module, including a coarse filter and a fine filter, connected to the oil circuit and valve body module, for filtering the oil during the purification operation; A filler module, connected to the oil circuit and valve body module, for adding new oil during the oil change operation; A high-precision sensor module, including an oil quality sensor C3 for monitoring water activity AW and cleanliness level NC, and a pressure sensor for monitoring the pressure difference across the filter module; A local control center module is electrically connected to the oil circuit and valve body module and the high-precision sensor module. The local control center module is used to collect oil intrinsic data from the high-precision sensor module and send it to the digital central layer through the Internet of Things, and receive operation and maintenance decision instructions sent by the digital central layer; According to the operation and maintenance decision instructions, control the oil circuit and valve body module, filter module and filler module to perform corresponding operation and maintenance operations; The digital central layer has a built-in AI decision engine, which is used to obtain mechanical dynamic data and operating condition data from the wind turbine master control SCADA system, fuse the oil intrinsic data, mechanical dynamic data and operating condition data into multimodal data, and drive the AI ​​decision engine to run the pre-trained comprehensive health index CHI model to process the multimodal data and generate operation and maintenance decision instructions.

2. The wind power gear oil quality monitoring and purification filling operation and maintenance robot based on the Internet of Things according to claim 1 is characterized in that: The AI ​​decision engine also runs a pre-trained remaining service life RUL model; the remaining service life RUL model uses the time series health index output by the comprehensive health index CHI model as input to predict the remaining service life RUL of the oil, and uses the remaining service life RUL as the basis for generating the operation and maintenance decision instructions.

3. The wind power gear oil quality monitoring and purification filling operation and maintenance robot based on the Internet of Things according to claim 2 is characterized in that: The AI ​​decision engine generates operation and maintenance decision instructions by solving a multi-objective optimization problem; the multi-objective optimization problem uses the comprehensive health index CHI and the remaining service life RUL as decision inputs, and comprehensively evaluates equipment health, operation and maintenance energy consumption, consumable resource consumption, and coordination with the wind farm's external maintenance plan to output the operation and maintenance decision instructions.

4. The wind power gear oil quality monitoring and purification filling operation and maintenance robot based on the Internet of Things according to claim 1 is characterized in that: When the local control center module controls the execution of the purification operation, the local control center module is also used to calculate the filtration gradient based on the pressure difference data monitored by the pressure sensor at high frequency, and send the filtration gradient as dynamic process data to the digital central layer; the digital central layer incorporates the filtration gradient into the multimodal data for processing by the comprehensive health index CHI model.

5. The wind power gear oil quality monitoring and purification filling operation and maintenance robot based on the Internet of Things according to claim 1 is characterized in that: The digital hub layer is also used to evaluate the model uncertainty of the AI ​​decision engine; and when the model uncertainty exceeds a preset threshold, it generates and issues active diagnostic instructions, and the local control center module executes exploratory operations corresponding to the active diagnostic instructions to obtain new data for eliminating model uncertainty.

6. The wind power gear oil quality monitoring and purification filling operation and maintenance robot based on the Internet of Things according to claim 1 is characterized in that: The local control center module controls and executes corresponding operation and maintenance operations including: When the water activity AW and cleanliness level NC measured by the oil quality sensor C3 meet the first preset condition, all modules are controlled to enter the standby state; When the water activity AW and the cleanliness level NC meet a second preset condition, controlling the oil circuit, valve body module, and filter module to perform a purification operation; When the water activity AW or the cleanliness level NC meets a third preset condition, the oil circuit and valve body module and the filler module are controlled to perform an oil change operation of draining waste oil and filling oil.

7. A wind power gear oil quality monitoring, purification, filling and operation and maintenance method based on the Internet of Things, applied to the robot according to any one of claims 1 to 6, characterized in that: The method comprises: S1. The local control center module collects oil intrinsic data through a high-precision sensor module and sends the oil intrinsic data to the digital hub layer through the Internet of Things; S2. The digital hub layer obtains mechanical dynamic data and operating condition data from the wind turbine master control SCADA system, and fuses the oil intrinsic data, mechanical dynamic data, and operating condition data into multimodal data; S3. The digital hub layer drives the AI ​​decision engine to run the pre-trained comprehensive health index (CHI) model to process the multimodal data and generate operation and maintenance decision instructions; S4. The local control center module receives the operation and maintenance decision instructions and controls each module of the robot to perform corresponding operation and maintenance operations.

8. The method for monitoring the quality of wind power gear oil and purifying and filling it according to claim 7, characterized in that: In step S3, the digital hub layer drives the AI ​​decision engine to run the pre-trained comprehensive health index CHI model to process the multimodal data and generate operation and maintenance decision instructions, and the following steps are also included before the step: Run the pre-trained RUL model; The remaining service life RUL model uses the time series health index output by the comprehensive health index CHI model as input to predict the remaining service life RUL of the oil, and uses the remaining service life RUL as the judgment basis for generating the operation and maintenance decision instructions.

9. The method for monitoring the quality of wind power gear oil and purifying and filling it according to claim 8, characterized in that: In step S3, the digital hub layer drives the AI ​​decision engine to run the pre-trained comprehensive health index CHI model to process the multimodal data and generate operation and maintenance decision instructions, including the following steps: Generate operation and maintenance decision instructions by solving multi-objective optimization problems; The multi-objective optimization problem uses the comprehensive health index (CHI) and the remaining service life (RUL) as decision inputs, and comprehensively evaluates equipment health, operation and maintenance energy consumption, consumable resource consumption, and coordination with external maintenance plans of the wind farm.

10. The method for monitoring the quality of wind power gear oil and purifying and filling and operating and maintaining it based on the Internet of Things according to claim 7, characterized in that: The method further comprises: During the purification operation, the filtration gradient is calculated based on the pressure difference data monitored by the pressure sensor at high frequency; The filtered gradient is incorporated into the multimodal data as dynamic process data for processing by the comprehensive health index (CHI) model.

Images