Wind turbine variable pitch system reliability analysis method and system based on 5g communication

Abstract

The embodiment of the application provides a wind turbine variable pitch system reliability analysis method and system based on 5G communication, which uses a 5G communication network to obtain multi-source heterogeneous data streams of a wind turbine variable pitch system in real time, and constructs a dynamic feature vector capable of comprehensively representing the current operation state of the system. Subsequently, through a series of pre-trained lightweight machine learning agent models, the dynamic feature vector is interpreted into the instantaneous occurrence probability of each bottom basic event in the fault tree model in real time. Finally, these dynamically calculated probabilities are injected into the static fault tree logical structure together with the failure probability of the communication link in real time, thereby converting the traditional static reliability analysis model into an online diagnosis system capable of dynamically evolving with the working condition. The method solves the technical problem that the traditional reliability analysis method cannot reflect the actual risk of the equipment under real and variable working conditions due to the use of fixed failure probability, and realizes accurate and dynamic evaluation and prediction of system reliability.

Description

Technical Field

[0001] This invention relates to the field of wind power generation technology, specifically to a reliability analysis method and system for a wind turbine pitch system based on 5G communication. Background Technology

[0002] The pitch control system of a wind turbine is the core functional and safety protection unit of the wind turbine generator. Its main responsibility is to adjust the blade pitch angle according to real-time wind conditions to optimize energy capture efficiency and ensure the safety of the unit under extreme weather conditions. Given that wind turbines are typically deployed in remote areas with abundant wind resources but complex and variable operating conditions, the pitch control system is subjected to severe environmental challenges such as intense gusts and alternating high and low temperatures. These dynamically changing operating conditions and environmental factors cause the operating load and performance status of key components within the pitch control system to constantly change, resulting in a failure risk that is not a fixed static value but exhibits significant dynamic characteristics.

[0003] To meet these needs, the industry typically employs methods such as fault tree analysis to model and assess system reliability. However, a technical problem with traditional fault tree analysis in practical applications is that the probability of occurrence of each underlying basic failure event in the model is usually a fixed, static average value based on historical experience and long-term statistical data. This assumption of static probability is profoundly contradictory to the dynamic operating conditions, environmental influences, and gradual degradation processes of components faced by pitch systems in the real physical world. For example, when encountering strong gusts, the frequent large-angle movements of the pitch mechanism can cause a surge in the instantaneous load on the drive motor and bearings, at which point the probability of failure is far higher than the average level, and the static model cannot capture this second-level risk jump. Similarly, in low-temperature winter environments, factors such as decreased lubricant performance and performance degradation of electronic components can also lead to a dynamic increase in the failure rate of related components, which is also not reflected by the static model. Therefore, due to the static nature of their models, traditional reliability analysis methods cannot effectively utilize real-time operating data and are unable to accurately assess the true risk level of the system at a specific moment, resulting in untimely and inaccurate fault warnings, failing to meet the urgent needs of intelligent operation and maintenance in modern wind farms.

[0004] Therefore, an optimized reliability analysis scheme for wind turbine pitch systems is desired. Summary of the Invention

[0005] The present invention aims to at least solve one of the technical problems existing in the prior art, and provides a reliability analysis method and system for wind turbine pitch system based on 5G communication.

[0006] In a first aspect, embodiments of the present invention provide a reliability analysis method for a wind turbine pitch system based on 5G communication, comprising: Real-time acquisition of multi-source heterogeneous data from the wind turbine pitch system, including high-frequency sensor data, low-frequency operating condition data, and 5G communication network service quality data transmitted via 5G communication network; Time-domain and frequency-domain features are extracted from high-frequency sensor data and combined with low-frequency operating condition data and 5G communication network service quality data at the current moment to obtain the dynamic feature vector at the current moment. Based on the dynamic feature vector at the current moment, perform basic event probability proxy prediction to obtain the dynamic probability set at the current moment; To obtain the probability of communication failure at the current moment, communication link reliability modeling is performed on 5G communication network service quality data. Input the current dynamic probability set and the current communication failure probability into the static fault tree structure model to obtain the current top event probability and the current minimum cut set; Reliability assessment results are generated by performing a reliability assessment based on the current top event probability and the current minimum cut set.

[0007] Secondly, embodiments of the present invention provide a reliability analysis system for a wind turbine pitch system based on 5G communication, comprising: The real-time acquisition module is used to acquire multi-source heterogeneous data from the wind turbine pitch system in real time. The multi-source heterogeneous data includes high-frequency sensor data, low-frequency operating condition data, and 5G communication network service quality data transmitted through the 5G communication network. The current dynamic feature vector acquisition module is used to extract time-domain and frequency-domain features from high-frequency sensor data and combine them with low-frequency operating condition data and 5G communication network service quality data at the current moment to obtain the current dynamic feature vector. The basic event probability proxy prediction module is used to perform basic event probability proxy prediction based on the dynamic feature vector at the current moment to obtain the dynamic probability set at the current moment. The communication failure probability acquisition module is used to perform communication link reliability modeling on 5G communication network service quality data to obtain the communication failure probability at the current moment. The static fault tree structure model input module is used to input the current dynamic probability set and the current communication failure probability into the static fault tree structure model to obtain the current top event probability and the current minimum cut set. The reliability assessment result generation module is used to perform reliability assessment based on the current top event probability and the current minimum cut set to generate reliability assessment results.

[0008] Compared with existing technologies, this invention proposes a reliability analysis method for wind turbine pitch systems based on 5G communication. First, it utilizes the 5G communication network to acquire multi-source heterogeneous data streams from the wind turbine pitch system in real time and constructs a dynamic feature vector that comprehensively characterizes the current operating state of the system. Then, through a series of pre-trained lightweight machine learning proxy models, this dynamic feature vector is interpreted in real time into the instantaneous occurrence probabilities of each underlying basic event in the fault tree model. Finally, these dynamically calculated probabilities, along with the failure probabilities of the communication link, are injected in real time into the static fault tree logic structure, thereby transforming the traditional static reliability analysis model into an online diagnostic system that can dynamically evolve with operating conditions. This method solves the technical problem that traditional reliability analysis methods, due to their use of fixed failure probabilities, cannot reflect the actual risks of equipment under real and variable operating conditions, achieving accurate and dynamic assessment and prediction of system reliability. Attached Figure Description

[0009] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0010] Figure 1 A flowchart of a reliability analysis method for a wind turbine pitch system based on 5G communication according to an embodiment of the present invention; Figure 2 This is a data flow diagram illustrating the reliability analysis method for a wind turbine pitch system based on 5G communication according to an embodiment of the present invention. Figure 3 This is a flowchart illustrating the process of using a 5G-based wind turbine pitch system reliability analysis method according to an embodiment of the present invention to perform basic event probability proxy prediction based on the current dynamic feature vector to obtain a sequence of current dynamic probability sets. Figure 4 This is a flowchart illustrating the reliability analysis method for a wind turbine pitch system based on 5G communication according to an embodiment of the present invention, which inputs the current dynamic probability set and the current communication failure probability into a static fault tree structure model to obtain the current top event probability and the current minimum cut set. Figure 5 This is a block diagram of a reliability analysis system for a wind turbine pitch system based on 5G communication according to an embodiment of the present invention. Detailed Implementation

[0011] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0012] Unless otherwise specifically stated, the technical or scientific terms used in the embodiments of this invention should be understood in their ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains. The terms "comprising" or "including," as used in the embodiments of this invention, do not limit the shapes, numbers, steps, actions, operations, components, elements, and / or groups thereof mentioned, nor do they exclude the appearance or addition of one or more other different shapes, numbers, steps, actions, operations, components, elements, and / or groups thereof, or the inclusion of these.

[0013] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale, and techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail; however, where appropriate, the illustrated techniques, methods, and apparatus should be considered part of the specification. In all the examples shown and discussed herein, any other specific example may have different values. It should be noted that similar symbols and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0014] In the description of the embodiments of the present invention, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In the embodiments of the present invention, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in the embodiments of the present invention, as well as the features of different embodiments or examples.

[0015] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0016] Currently, reliability analysis methods for wind turbine pitch systems generally suffer from the drawback of static models. These methods often employ fixed failure probabilities based on historical statistics, failing to reflect the true risk level of the system under drastically changing operating conditions and harsh environments in real time, leading to delayed or failed early warnings. Therefore, this invention proposes a reliability analysis method for wind turbine pitch systems based on 5G communication. Specifically, this method first collects multi-source heterogeneous data from the wind turbine pitch system in real time, including high-frequency sensor data, low-frequency operating condition data, and 5G network service quality data. Next, it extracts time-domain and frequency-domain features characterizing the equipment's health status from the high-frequency data and fuses them with other real-time data to construct a dynamic feature vector that comprehensively describes the system's current operating status. Subsequently, it uses a series of pre-trained lightweight machine learning agent models to perform real-time inference on this dynamic feature vector to obtain the instantaneous failure probability of each underlying hardware component, forming a dynamic probability set. Simultaneously, this method also independently models the service quality of the 5G communication link, calculating the communication failure probability of control command transmission failure based on real-time packet loss rate and other data. Finally, the dynamically acquired hardware failure probabilities and communication failure probabilities are injected into a preset static fault tree logical structure model. The model is then rapidly solved from the bottom up using algorithms such as the minimum cut set method to obtain the top event failure probability and key risk combination of the entire system at the current moment. Based on this, accurate and real-time reliability assessment results and early warning signals are generated.

[0017] Figure 1 This is a flowchart of a reliability analysis method for a wind turbine pitch system based on 5G communication according to an embodiment of the present invention. Figure 2 This is a data flow diagram illustrating the reliability analysis method for a wind turbine pitch system based on 5G communication according to an embodiment of the present invention. Figure 1 and Figure 2As shown, the reliability analysis method and system for a wind turbine pitch system based on 5G communication according to an embodiment of the present invention includes the following steps: S100, real-time acquisition of multi-source heterogeneous data of the wind turbine pitch system, including high-frequency sensor data, low-frequency operating condition data, and 5G communication network service quality data transmitted through the 5G communication network; S200, extraction of time-domain and frequency-domain features from the high-frequency sensor data, and combination with the current low-frequency operating condition data and 5G communication network service quality data to obtain a current dynamic feature vector; S300, basic event probability proxy prediction based on the current dynamic feature vector to obtain a current dynamic probability set; S400, communication link reliability modeling based on the 5G communication network service quality data to obtain the current communication failure probability; S500, inputting the current dynamic probability set and the current communication failure probability into a static fault tree structure model to obtain the current top event probability and the current minimum cut set; S600, reliability assessment based on the current top event probability and the current minimum cut set to generate a reliability assessment result.

[0018] Specifically, in step S100, multi-source heterogeneous data from the wind turbine pitch system is collected in real time. This multi-source heterogeneous data includes high-frequency sensor data, low-frequency operating condition data, and 5G communication network quality of service data transmitted via a 5G communication network. It should be understood that, as the wind turbine pitch system is a complex system integrating multiple fields such as mechanics, electrical systems, and communications, its operational reliability is affected by the coupling of multiple factors, including the state of internal components, the external operating environment, and the quality of control signal transmission. Data from a single source cannot comprehensively characterize its dynamic risks. Therefore, in the technical solution of this invention, multi-source heterogeneous data from the wind turbine pitch system is collected in real time. This multi-source heterogeneous data includes high-frequency sensor data, low-frequency operating condition data, and 5G communication network quality of service data transmitted via a 5G communication network. This allows for the comprehensive and synchronous capture of all key information affecting system reliability from multiple dimensions, such as mechanical wear, electrical load, external stress, and communication link stability. This provides a complete and high-fidelity data foundation for subsequent construction of accurate dynamic feature vectors and reliability assessment, ensuring the comprehensiveness and accuracy of the analysis results.

[0019] More specifically, in a specific example of the present invention, the acquisition process of this multi-source heterogeneous data is implemented through the following steps. Specifically, high-frequency sensor data includes vibration signals, temperature signals, and current signals; low-frequency operating condition data includes wind speed, pitch angle, and power values; and 5G communication network service quality data includes signal strength, signal quality, network latency, and packet loss rate. First, high-frequency vibration and temperature sensors are deployed at key locations such as bearing housings and motor housings on the pitch drive chain of wind turbine blades in a high-altitude, cold mountain wind farm. The analog signals generated by these sensors are then converted to high-frequency digital signals using a high-speed analog-to-digital converter via an industrial data acquisition module. Second, the acquisition module is connected to a 5G communication terminal installed in the nacelle. This terminal utilizes 5G network slicing technology to establish a highly reliable, low-latency dedicated data channel, transmitting the acquired high-frequency sensor data in real-time and without loss to an edge computing server deployed at the bottom of the wind turbine tower or in the wind farm's main control room. Secondly, the edge computing server proactively synchronizes low-frequency operating condition data, such as wind speed, pitch angle command value, and generator active power value, from the wind turbine main control SCADA system via the industrial bus protocol. Simultaneously, it obtains 5G communication network service quality data for this data transmission link from the 5G communication terminal or network management platform, including signal strength, signal quality, network latency, and real-time packet loss rate. Finally, the edge computing server uses a unified time synchronization protocol to timestamp-align all received high-frequency, low-frequency, and network quality data, fusing data from different sources and frequencies into structured data frames with a unified time base, forming a continuous, multi-source heterogeneous data stream available for subsequent analysis.

[0020] Specifically, in step S200, time-domain and frequency-domain features are extracted from the high-frequency sensor data and combined with the low-frequency operating condition data and 5G communication network service quality data at the current moment to obtain the dynamic feature vector at the current moment. It should be understood that the raw multi-source heterogeneous data streams collected in the previous step, especially the high-frequency sensor signals, are massive in volume, extremely high in dimensionality, and contain a large amount of redundant information and noise, making them unsuitable as direct input to the reliability assessment model. Furthermore, data from different sources differ in format and physical meaning, making direct fusion analysis difficult. Therefore, in the technical solution of this invention, time-domain and frequency-domain features are further extracted from the high-frequency sensor data and combined with the low-frequency operating condition data and 5G communication network service quality data at the current moment to obtain the dynamic feature vector at the current moment. This transforms the unstructured raw signal into a set of structured numerical indicators that can accurately quantify the key states of the system, and organically combines the microscopic features reflecting the internal health status of components with the macroscopic data characterizing the external operating load and communication environment. In this way, a standardized, information-dense digital snapshot that can be directly processed by machine learning models can be generated. This snapshot comprehensively depicts the overall operating state of the pitch system at a specific moment, providing high-quality and strongly correlated input for subsequent accurate prediction of the instantaneous probability of each basic event.

[0021] More specifically, in a specific example of the present invention, extracting time-domain features and frequency-domain features from high-frequency sensor data includes: calculating the root mean square and kurtosis of the high-frequency sensor data to obtain time-domain features; and performing a fast Fourier transform on the high-frequency sensor data to extract the spectral energy at the fault characteristic frequencies of the bearings and gears as frequency-domain features.

[0022] Accordingly, the root mean square (RMS) and kurtosis of the high-frequency sensor data are calculated to obtain time-domain features. It should be understood that since the raw high-frequency vibration signal acquired from the sensor exists as a time-series waveform, it contains a large number of data points, but the information is relatively obscure, making it difficult to directly determine the specific health status of the equipment. Furthermore, different mechanical fault modes exhibit different patterns in the statistical characteristics of the signal. Therefore, in the technical solution of this invention, the RMS and kurtosis of the high-frequency sensor data are further calculated to obtain time-domain features, thereby quantifying the raw signal from the two key dimensions of vibration energy and impact characteristics. This transforms the complex vibration waveform into a numerical index highly sensitive to fault type, effectively distinguishing whether the equipment is in a state of uniform wear or has experienced an early localized impact fault, thus providing more diagnostically valuable feature input for subsequent fault probability prediction.

[0023] More specifically, in a concrete example of the present invention, the extraction process of this time-domain feature is as follows. First, the high-frequency digital signal stream continuously acquired from the pitch bearing vibration sensor is segmented at the edge computing node, i.e., divided into a series of data windows of fixed length, for example, each analysis window consists of 4096 sampling points. Next, for each data window, the root mean square value is calculated. Specifically, the amplitude of all sampling points within the window is squared, the arithmetic mean of all squared values ​​is calculated, and finally, the square root of this mean is taken. The resulting value is the energy intensity of the vibration signal within that time period. This value is highly sensitive to slowly changing faults such as accelerated overall bearing wear caused by strong wind loads. Simultaneously, for the same data window, the kurtosis value is calculated. Specifically, the fourth central moment of the amplitude of all sampling points within the window relative to the mean is calculated, and this is compared with the fourth power of the standard deviation. The resulting dimensionless parameter is the kurtosis. This value is extremely effective in measuring the strength of the impact component in the signal. When the bearing raceway experiences local spalling due to low-temperature brittleness, a strong impact pulse is generated, causing the kurtosis value to increase significantly. Finally, the calculated root mean square value and kurtosis value are used as two key temporal features at that moment and combined into the dynamic feature vector at the current moment.

[0024] Accordingly, a Fast Fourier Transform (FFT) is performed on the high-frequency sensor data to extract the spectral energy at the fault characteristic frequencies of bearings and gears as frequency domain features. It should be understood that relying solely on time domain features is sometimes insufficient to accurately distinguish the specific type and location of a fault, especially in rotating components such as bearings and gears, where different damage points generate impact signals with unique periodicity. These periodic features are often obscured by noise in complex time-domain waveforms. Therefore, in the technical solution of this invention, a Fast Fourier Transform is further performed on the high-frequency sensor data to extract the spectral energy at the fault characteristic frequencies of bearings and gears as frequency domain features. This converts the signal from the time domain to the frequency domain, revealing its inherent frequency structure, thereby enabling clear identification and quantification of the periodic vibration energy associated with specific physical faults. This allows for precise location and identification of the fault source, such as distinguishing between a bearing inner ring fault and a gear meshing abnormality, providing feature inputs with clear physical meaning and high diagnostic accuracy for reliability assessment models.

[0025] More specifically, in a concrete example of the present invention, the extraction process of the frequency domain features is as follows. First, based on the design drawings of the bearings and gearbox inside the pitch system, their precise geometric parameters are obtained. Combined with real-time speed information obtained from low-frequency operating data, the theoretical values ​​of the fault characteristic frequencies of each key component are pre-calculated, such as the bearing outer ring fault frequency, inner ring fault frequency, and gear meshing frequency. Next, a window function is applied to the same data window used for time-domain analysis to suppress spectral leakage, followed by a Fast Fourier Transform (FFT) algorithm. This transform operation converts the time-domain vibration signal into a frequency-domain spectrum, which shows the distribution of vibration energy at different frequencies. Finally, in the generated spectrum, the frequency points corresponding to the pre-calculated theoretical values ​​of each fault characteristic frequency are precisely located, and the amplitude or energy values ​​at these frequency points are extracted. These extracted energy values ​​at specific frequencies are considered as the frequency domain features at the current moment and are combined into a dynamic feature vector for subsequent analysis.

[0026] Specifically, in step S300, basic event probability proxy prediction is performed based on the current dynamic feature vector to obtain the current dynamic probability set. It should be understood that because there is a highly complex and nonlinear mapping relationship between the dynamic feature vector representing the system state and the final failure probability of each underlying component, it is difficult to establish a direct connection through a definite physical formula. However, the fault tree model, as the logical framework for reliability analysis, requires explicit basic event probabilities as its computational input. Therefore, in the technical solution of this invention, basic event probability proxy prediction is further performed based on the current dynamic feature vector to obtain the current dynamic probability set. This utilizes the powerful fitting and reasoning capabilities of the pre-trained machine learning proxy model to construct an efficient mapping channel from the multi-dimensional feature space to the failure probability space. This solution equips each basic event in the fault tree, such as accelerated bearing wear or motor winding overheating, with a dedicated lightweight proxy prediction model. Based on the input real-time feature vector, it quickly infers the instantaneous probability of the event occurring at the current moment. In this way, the characteristic data describing the physical state can be transformed into probability values ​​with clear statistical significance, generating a set of dynamic probability inputs that can be directly injected into the fault tree logic structure, providing a key quantitative basis for realizing the real-time dynamic solution of the entire reliability model.

[0027] Figure 3 This is a flowchart illustrating the method for reliability analysis of a wind turbine pitch system based on 5G communication according to an embodiment of the present invention, which involves performing basic event probability proxy prediction based on the current dynamic feature vector to obtain the current dynamic probability set. For example... Figure 3As shown, step S300 includes: S310, reasoning through each proxy model in the proxy model set to obtain the probability of occurrence of multiple basic failure events from the current dynamic feature vector; S320, combining the probability of occurrence of multiple basic failure events into the current dynamic probability set.

[0028] In step S310, the current dynamic feature vector is inferred through various surrogate models in the surrogate model set to obtain the probability of occurrence of multiple basic failure events. It should be understood that, due to the distinct underlying physical failure mechanisms and evolution patterns of different basic failure events in a pitch system, such as wear of mechanical components and overheating of electrical components, a single, general prediction model cannot accurately capture these diverse failure modes simultaneously. Therefore, in the technical solution of this invention, the current dynamic feature vector is further inferred through various surrogate models in the surrogate model set to obtain the probability of occurrence of multiple basic failure events, thereby deploying a specially trained and optimized prediction expert for each specific failure mode. Each surrogate model, such as a lightweight neural network model, is designed to be most sensitive to the specific physical quantity corresponding to it in the input feature vector, thereby enabling it to deeply learn and reproduce the complex nonlinear relationship from the initiation to the development of a specific fault. In this way, an independent and highly accurate probability assessment can be performed for each potential failure path, which not only improves the accuracy and robustness of the overall prediction, but also enhances the maintainability and scalability of the model, ultimately forming a comprehensive dynamic probability set output by multiple expert models.

[0029] More specifically, in a concrete example of the present invention, the proxy prediction process is implemented as follows. First, a proxy model set containing multiple pre-trained lightweight neural network models is loaded at the edge computing node. This set includes a bearing wear model and a motor overheating model. After obtaining a dynamic feature vector containing information such as vibration kurtosis, bearing fault frequency energy, motor temperature, motor current, and wind speed, this vector is simultaneously used as input to both models. Inference is performed in the bearing wear model. The network weights of this model are specially optimized to have higher sensitivity to vibration-related feature inputs, especially kurtosis and specific fault frequency energy values. After forward propagation calculation, its output layer generates a probability value, such as 0.02, which is the probability of the bearing wear aggravation event occurring at the current moment. At the same time, inference is performed in the motor overheating model. This model is more sensitive to the motor temperature and motor current values ​​in the feature vector. After calculating based on these input values, it outputs another probability value, such as 0.008, which represents the probability of the motor winding overheating event occurring at the current moment. Finally, the instantaneous occurrence probability values ​​output by these two models and all other models in the set are aggregated to form the dynamic probability set at the current moment, which can be used for subsequent fault tree analysis.

[0030] In step S320, the probabilities of multiple basic failure events are combined into a dynamic probability set for the current moment. It should be understood that since the inference process in the previous step outputs a series of discrete probability values ​​in parallel from multiple independent surrogate models, these values ​​are scattered in format and lack a direct correspondence with the basic events in the fault tree logical structure, making them unusable as a whole for subsequent reliability calculation modules. Therefore, in the technical solution of this invention, the probabilities of multiple basic failure events are further combined into a dynamic probability set for the current moment. This allows for the structured integration and encapsulation of the various independent prediction results, creating a unified and standardized data interface. This set explicitly pairs each basic event identifier defined in the fault tree with its corresponding real-time predicted probability value. This forms a snapshot that fully describes the probability state of all underlying failure modes at the current moment, providing a regular, complete, and unambiguous input data source for subsequent dynamic fault tree calculations, ensuring the consistency and integrity of data transmission between different analysis stages.

[0031] More specifically, in a concrete example of the present invention, the combination process is implemented as follows. First, a data object, such as a key-value pair mapping table or dictionary structure, is initialized in memory. The keys of this structure are predefined as unique identifiers for all basic events in the fault tree model, such as BE_BearingWear and BE_MotorOverheat. Next, when the instantaneous occurrence probability of 0.02 is obtained from the bearing wear proxy model, this value is associated with the key BE_BearingWear and stored in the mapping table. Simultaneously, when the probability value of 0.008 is obtained from the motor overheat proxy model, this value is associated with the key BE_MotorOverheat. This process continues until all proxy models in the set have completed inference and filled their respective output probability values ​​into the corresponding identifier positions in the mapping table. Finally, this complete mapping table containing all basic events and their corresponding instantaneous probability values ​​is identified as the dynamic probability set at the current moment and is passed to the next processing step as an independent, structured data unit.

[0032] Specifically, in step S400, communication link reliability modeling is performed on the 5G communication network service quality data to obtain the current communication failure probability. It should be understood that the overall reliability of a 5G-based pitch system depends not only on the health of its physical components but also heavily on the stability of the wireless communication link responsible for transmitting control commands and sensor data. This link itself is a potential source of failure; performance fluctuations or interruptions can directly lead to system instability, constituting an independent failure mode. Therefore, in the technical solution of this invention, communication link reliability modeling is further performed on the 5G communication network service quality data to obtain the current communication failure probability. This allows the real-time performance of the communication link to be quantified and integrated into the overall reliability analysis framework as a fundamental event. By establishing a mathematical relationship between network service quality indicators such as real-time packet loss rate and the probability of failure in critical command transmission, the reliability of the data channel is dynamically evaluated. This ensures that the final reliability assessment is comprehensive, reflecting both the physical degradation risk of hardware and the communication interruption risk of information interaction, thereby avoiding misjudgments of the overall system reliability due to neglecting communication failures and obtaining a more accurate risk assessment result covering the entire link from data acquisition to command execution.

[0033] More specifically, in a specific example of the present invention, performing communication link reliability modeling on 5G communication network service quality data to obtain the communication failure probability at the current moment includes: obtaining the real-time packet loss rate from the 5G communication network service quality data; and calculating the probability of instruction transmission failure based on the real-time packet loss rate to obtain the communication failure probability at the current moment.

[0034] Accordingly, the real-time packet loss rate is obtained from 5G communication network service quality data. It should be understood that among the many indicators for measuring communication network service quality, the packet loss rate is a core parameter directly representing the integrity and reliability of data transmission. Its value directly determines whether control commands or data information can be successfully delivered. Compared to other indicators such as network latency or signal strength, it has a more direct mathematical relationship with the probability of command transmission failure. Therefore, in the technical solution of this invention, the real-time packet loss rate is further obtained from 5G communication network service quality data to extract the most critical and representative quantitative input for communication link reliability modeling. This ensures that subsequent communication failure probability calculations are based on the most direct measurement of data transmission success or failure, thereby enabling the model to accurately reflect the real communication risks caused by instantaneous network fluctuations, especially in harsh electromagnetic environments.

[0035] More specifically, in a concrete example of the present invention, the process of obtaining the real-time packet loss rate is implemented as follows. First, an edge computing node deployed at the bottom of the wind turbine tower initiates a periodic query request to the management interface of the 5G communication terminal located in the nacelle via a standard network management protocol or an application programming interface provided by the equipment vendor. The purpose of this request is to obtain transmission statistics related to a specific network slice carrying the pitch system data stream. Next, key performance counter readings are parsed from the feedback of the 5G communication terminal, specifically the total number of data packets successfully sent by the transmitter and the total number of data packets successfully received by the receiver within a preset statistical time window, such as the past second. Finally, the number of lost data packets is obtained by subtracting the total number of sent packets from the total number of received packets, and then this number of lost packets is divided by the total number of sent packets to calculate the real-time packet loss rate within the time window. This calculated ratio, for example, 0.0001, is identified as the real-time packet loss rate at the current moment and used for the next calculation.

[0036] Accordingly, based on the real-time packet loss rate, the probability of instruction transmission failure is calculated to obtain the communication failure probability at the current moment. The probability of instruction transmission failure is calculated using the following formula, wherein the formula is:

[0037] in, For real-time packet loss rate, The number of consecutive data packets required to complete a critical control command. This represents the probability of a communication failure at the current moment.

[0038] It is understandable that the obtained real-time packet loss rate only describes the probability of a single data packet being lost during transmission. However, critical control commands of a pitch system, such as emergency stop or pitch test, often require multiple consecutive data packets. The command is only effective when all data packets are successfully received. Therefore, a clear mathematical model is needed to transform the single packet loss rate into the transmission failure rate of the entire command sequence. Thus, in the technical solution of this invention, the probability of command transmission failure is further calculated based on the real-time packet loss rate to obtain the current communication failure probability. This allows the basic network performance indicator to be upgraded into a reliability indicator that directly reflects the success or failure of critical control function execution through specific probability algorithms. This calculation clarifies that the condition for successful command transmission is that all N data packets are successfully transmitted, and defines all situations outside this scenario as transmission failure, thereby quantifying the failure risk of executing a complete command operation under the current network packet loss level. This generates a standardized probability value with clear physical meaning, which can be directly used as input for the basic event of communication link failure in fault tree analysis, ensuring the quantitative accuracy of communication risk in the overall system reliability assessment.

[0039] In step S500, the current dynamic probability set and the current communication failure probability are input into the static fault tree structure model to obtain the current top event probability and the current minimum cut set. It should be understood that since the occurrence probabilities of each basic event generated in the preceding steps are discrete risk indicators describing the underlying, independent components, and the top-level functional failure of the entire pitch system is the result of the complex coupling of these underlying events according to specific logical relationships, a structured logical model is needed to integrate and deduce these discrete risks. Therefore, in the technical solution of this invention, the current dynamic probability set and the current communication failure probability are further input into the static fault tree structure model to obtain the current top event probability and the current minimum cut set, thereby injecting the real-time quantified underlying event risks into a pre-constructed static fault tree topology structure that uses logical symbols such as AND gates and OR gates to describe the system fault propagation relationship. By employing fault tree quantitative analysis algorithms such as the minimum cut set method, the probability can be propagated and calculated from bottom to top between logic gates, ultimately solving for the comprehensive occurrence probability of the undesirable event at the top level of the system, such as pitch function failure, at the current moment. In this way, dispersed, micro-level component-level risks can be integrated into a single, macro-level system-level reliability metric, namely the top event probability. At the same time, by identifying the minimum cut set at the current moment, it is also possible to accurately locate the weakest link and the most critical combination of failure modes that lead to system failure, providing direct and clear quantitative guidance for risk assessment and maintenance decisions.

[0040] Figure 4This is a flowchart illustrating the dynamic fusion of operating condition features and vibration features based on an attention mechanism to obtain a fused feature vector of operating condition-vibration information, according to an embodiment of the reliability analysis method for a wind turbine pitch system based on 5G communication according to an embodiment of the present invention. Figure 4 As shown, in step S500, the dynamic probabilities of each current moment in the current moment dynamic probability set are injected into the static fault tree structure model, and the communication failure probability of the current moment is injected into the static fault tree structure model to obtain the basic event probability update fault tree structure model; in step S520, based on the logic gate propagation probability of the basic event probability update fault tree structure model, the minimum cut set method is used to perform bottom-up calculation to solve the current moment top event probability and the current moment minimum cut set.

[0041] In step S510, the current dynamic probabilities from the current dynamic probability set are injected into the static fault tree structure model, and the current communication failure probability is also injected into the static fault tree structure model to obtain a basic event probability updated fault tree structure model. It should be understood that since the pre-constructed static fault tree structure model only defines the logical dependencies between failure events of various system components, its underlying basic event nodes either do not contain probability values ​​in their initial state, or only contain a static probability value based on long-term statistics that cannot reflect real-time performance. The dynamic probability set and communication failure probability generated in the preceding steps are data independent of this logical structure. Therefore, in the technical solution of this invention, the current dynamic probabilities from the current dynamic probability set are further injected into the static fault tree structure model, and the current communication failure probability is also injected into the static fault tree structure model to obtain a basic event probability updated fault tree structure model. This dynamically binds real-time, quantified risk data with an abstract, static logical framework. This process assigns the most realistic occurrence probability at the current moment to each underlying basic event node in the fault tree. In this way, a general, static fault logic template can be transformed in real time into a fully quantitative, dynamic analysis model that accurately reflects the complete risk profile of the system under the current specific operating conditions, thus preparing the necessary data for the subsequent quantitative calculation of the probability of top events.

[0042] More specifically, in a concrete example of the present invention, the probability injection process is implemented as follows. First, a predefined data structure representing the fault logic of the pitch system is loaded into the calculation program. This structure contains unique identifiers for all basic event nodes, such as BE_BearingWear, BE_MotorOverheat, and BE_CommFailure, and the initial probability values ​​of these nodes are empty or set to zero. Next, each key-value pair in the current dynamic probability set is traversed. When the entry with the key BE_BearingWear is processed, its corresponding probability value of 0.02 is obtained. Then, the basic event node with the identifier BE_BearingWear is searched in the fault tree data structure, and its probability attribute is updated to 0.02. This operation is performed sequentially for all entries in the dynamic probability set, for example, updating the probability of the BE_MotorOverheat node to 0.008. Subsequently, the current communication fault probability value of 0.0005 is obtained independently, and the node with the identifier BE_CommFailure is found in the fault tree structure, and its probability attribute is updated to 0.0005. Finally, once all dynamic probability values ​​have been assigned to the corresponding basic event nodes in the fault tree, a fault tree structure model with fully updated basic event probabilities is obtained. This model is ready for the next step of quantitative calculation.

[0043] In step S520, the logic gate propagation probabilities of the fault tree structure model are updated based on the basic event probabilities. A bottom-up calculation using the minimum cut set method is then performed to calculate the current top event probability and the current minimum cut set. It should be understood that although the fault tree model after probability injection contains real-time quantified values ​​of all underlying risks, it does not yet reveal how these risks converge into the top-level, overall system failure risk through the system's logical structure, nor does it clearly identify which specific fault combinations are the main paths leading to system failure under the current operating conditions. Therefore, in the technical solution of this invention, the logic gate propagation probabilities of the fault tree structure model are further updated based on the basic event probabilities, and a bottom-up calculation using the minimum cut set method is performed to calculate the current top event probability and the current minimum cut set, thereby performing a quantitative solution for the entire dynamic risk model. This process identifies all minimum fault mode combinations that could lead to the top event (i.e., minimum cut sets) by analyzing the logical structure of the fault tree, and calculates the final occurrence probability of the top event based on these combinations and the injected dynamic probabilities, according to probability theory rules. In this way, we can complete the final deduction from dispersed component-level risks to integrated system-level risks, not only to obtain an overall, quantitative conclusion on system reliability, namely the top event probability, but also to provide specific, importance-ranked diagnostic information, namely the minimum cut set, thereby providing a basis for risk assessment decisions.

[0044] More specifically, in a concrete example of the present invention, the bottom-up calculation process is implemented as follows. First, the fault tree model that has undergone probability injection is structurally analyzed. An algorithm is used to traverse downwards from the top event node, identifying all basic event combinations that can lead to the top event based on the Boolean logic rules of AND and OR gates, and simplifying them into minimal cut sets. For example, the analysis shows that the minimal cut sets at the current moment are three independent events: BE_BearingWear, BE_MotorOverheat, and BE_CommFailure. Next, using known probability theory formulas, the probability of the top event is calculated based on these minimal cut sets. For a minimal cut set connected by multiple independent OR gate relationships, its top event probability can be approximated as the sum of the probabilities of each minimal cut set. Based on the previously injected probability values, the probability of the top event at the current moment is calculated as 0.02 plus 0.008 plus 0.0005, finally obtaining 0.0285. Finally, the calculation results are output, namely, the probability of the top event at the current moment is 0.0285, and the list of minimum cut sets at the current moment is also output in descending order of probability. The first item in the list is BE_BearingWear with a probability of 0.02, and the second item is BE_MotorOverheat with a probability of 0.008. This indicates that under the current severe working conditions caused by strong winds and low temperatures, bearing wear is the most significant source of risk to the system.

[0045] Specifically, in step S600, a reliability assessment is performed based on the current top event probability and the current minimum cut set to generate a reliability assessment result. It should be understood that since the top event probability and minimum cut set calculated in the previous step are quantitative, uninterpreted raw analytical data, they do not directly form a final judgment on the current reliability status of the system, nor are they transformed into specific suggestions that can directly guide operation and maintenance work. Therefore, in the technical solution of this invention, a further reliability assessment is performed based on the current top event probability and the current minimum cut set to generate a reliability assessment result, thereby comparing and interpreting the raw calculation results with preset operation and maintenance rules and security thresholds. This process aims to make a qualitative or quantitative judgment on the overall risk level of the current system and provide diagnostically valuable insights based on the key risk paths revealed by the minimum cut set. In this way, complex analytical data can be transformed into concise, clear, and highly operable final conclusions, such as a specific early warning signal or a diagnostic report containing the location of key risk sources, thereby achieving a complete closed loop from data collection to intelligent decision support, providing timely and accurate guidance for on-site operation and maintenance personnel.

[0046] More specifically, in a specific example of the present invention, reliability assessment is performed based on the current top event probability and the current minimum cut set to generate a reliability assessment result, including: generating a reliability warning signal and generating a reliability assessment result in response to the current top event probability being less than a preset threshold; generating a system real-time reliability time series curve based on the current top event probabilities at consecutive time points within a preset time window; and displaying the system real-time reliability time series curve on a reliability dashboard.

[0047] Accordingly, in response to the current top event probability being less than a preset threshold, a reliability warning signal is generated, along with a reliability assessment result. It should be understood that since the continuously calculated top event probability is a constantly changing numerical stream, without a clear judgment criterion, this value itself cannot automatically trigger an operational response. An automated mechanism is needed to distinguish between the normal operating state of the system and the abnormal state requiring manual intervention. Therefore, in the technical solution of this invention, in response to the current top event probability meeting the preset threshold condition, a reliability warning signal is generated, and a reliability assessment result is generated, thereby transforming continuous, quantitative risk assessment into discrete, actionable alarm decisions. By comparing the real-time calculated system failure risk with a pre-set safe operating baseline, automatic determination of the risk level can be achieved. This ensures that once the system reliability drops to an unacceptable level, an alarm can be issued immediately and automatically, thereby promptly notifying maintenance personnel of potential faults, providing decision support for proactive maintenance, avoiding unplanned downtime and catastrophic accidents, and achieving true predictive maintenance.

[0048] More specifically, in a concrete example of the present invention, the assessment and early warning process is implemented as follows. First, multiple risk thresholds are preset in the monitoring backend. For example, a top event probability exceeding 0.02 is set as a Level 1 early warning threshold, and exceeding 0.05 is set as a Level 2 early warning threshold. Next, the current top event probability value calculated in real time in the previous step, for example, 0.0285, is compared with these preset thresholds. In this scenario, the calculated probability value of 0.0285 has exceeded the Level 1 early warning threshold of 0.02. Subsequently, a Level 1 reliability early warning signal is generated. This signal can be a specific status code or message and is sent to the wind farm's central monitoring system. Finally, a detailed reliability assessment result containing this early warning signal is generated. This result not only indicates that the current risk level is a Level 1 early warning but also clearly states that the current top event probability is 0.0285, and includes the main reason for this early warning: a significantly increased bearing wear risk, as determined by minimum cut set analysis, contributing 0.02 to the failure probability. This complete result provides operations and maintenance personnel with a complete information chain from alarms to fault location.

[0049] Accordingly, based on the probability of the top event at the current moment within a preset time window, a real-time reliability time series curve of the system is generated; this curve is then displayed on a reliability dashboard. It should be understood that since single or discrete reliability assessments only reflect a snapshot of the system's health at a specific instant, they cannot reveal the degradation trend of equipment performance over time or the dynamic process of responding to changes in external operating conditions. Therefore, for long-term health management and predictive maintenance, the information dimensions are insufficient. Thus, in the technical solution of this invention, a real-time reliability time series curve of the system is further generated based on the probability of the top event at the current moment within a preset time window; this curve is then displayed on a reliability dashboard, thereby connecting a series of instantaneous, discrete assessment points into a continuous, visualized health status trajectory. By plotting the change in reliability (i.e., 1 minus the probability of the top event) over time as a curve, the system's performance degradation rate, response to specific events, and overall health trend can be intuitively displayed. This provides maintenance personnel with a powerful macro-level decision-making view that goes beyond single-point alarms, enabling them to conduct deeper analysis and longer-term predictions based on historical trends. This allows them to identify potential problems earlier, optimize maintenance plans, and achieve a shift from reactive response to proactive management.

[0050] More specifically, in a concrete example of the present invention, the generation and display process of the reliability curve is implemented as follows. First, after calculating the probability of the top event at the current moment, the system's real-time reliability value at the current moment is obtained by subtracting the probability value from 1. This reliability value and its corresponding timestamp are stored as a data point in the time series database for persistent storage. Next, on the reliability dashboard user interface provided to wind farm operation and maintenance personnel, when it is necessary to view the reliability trend of a specific pitch system, a query request is initiated to the time series database to obtain all reliability data points within a specified time window, such as the past 24 hours. Then, after receiving the returned data sequence, the dashboard's front-end visualization component renders it in a two-dimensional coordinate system, where the horizontal axis represents time and the vertical axis represents the reliability value. All data points are connected sequentially to form a smooth system real-time reliability time series curve. Finally, the dashboard not only displays this dynamically changing curve, but also plots a horizontal warning threshold line at the corresponding position on the vertical axis, such as 0.98. When the real-time curve crosses below this threshold line, maintenance personnel can intuitively see that the system has entered a warning state, and can make a more accurate judgment on the urgency and severity of the fault by combining the rate and shape of the curve's decline.

[0051] In summary, the reliability analysis method for wind turbine pitch system based on 5G communication according to embodiments of the present invention is explained. First, it utilizes the 5G communication network to acquire multi-source heterogeneous data streams of the wind turbine pitch system in real time and constructs a dynamic feature vector that comprehensively characterizes the current operating state of the system. Then, through a series of pre-trained lightweight machine learning proxy models, this dynamic feature vector is interpreted in real time into the instantaneous occurrence probabilities of each underlying basic event in the fault tree model. Finally, these dynamically calculated probabilities, along with the failure probabilities of the communication link, are injected in real time into the static fault tree logic structure, thereby transforming the traditional static reliability analysis model into an online diagnostic system that can dynamically evolve with operating conditions. This method solves the technical problem that traditional reliability analysis methods, due to their use of fixed failure probabilities, cannot reflect the actual risks of equipment under real and variable operating conditions, achieving accurate and dynamic assessment and prediction of system reliability.

[0052] Furthermore, a reliability analysis system for wind turbine pitch system based on 5G communication is also provided.

[0053] Figure 5 This is a block diagram of a 5G-based wind turbine pitch system reliability analysis system according to an embodiment of the present invention. Figure 5 As shown, the reliability analysis system 100 for a wind turbine pitch system based on 5G communication according to an embodiment of the present invention includes: a real-time acquisition module 110, used to acquire multi-source heterogeneous data of the wind turbine pitch system in real time, the multi-source heterogeneous data including high-frequency sensor data, low-frequency operating condition data and 5G communication network service quality data transmitted through a 5G communication network; a current-moment dynamic feature vector acquisition module 120, used to extract time-domain features and frequency-domain features from the high-frequency sensor data, and combine them with the current-moment low-frequency operating condition data and 5G communication network service quality data to obtain the current-moment dynamic feature vector; and a basic event probability proxy prediction module 130. The module 140 is used to perform basic event probability proxy prediction based on the current dynamic feature vector to obtain the current dynamic probability set; the communication failure probability acquisition module 140 is used to perform communication link reliability modeling on 5G communication network service quality data to obtain the current communication failure probability; the static fault tree structure model input module 150 is used to input the current dynamic probability set and the current communication failure probability into the static fault tree structure model to obtain the current top event probability and the current minimum cut set; the reliability assessment result generation module 160 is used to perform reliability assessment based on the current top event probability and the current minimum cut set to generate reliability assessment results.

[0054] As described above, the 5G communication-based wind turbine pitch system reliability analysis system 100 according to embodiments of the present invention can be deployed in the edge computing unit at the wind turbine site, such as a dedicated industrial control computer deployed inside the wind turbine nacelle or tower base, and can interact with high-frequency sensors installed on the pitch system, the wind turbine's SCADA monitoring system, and 5G communication terminals in real time. In one possible implementation, the 5G communication-based wind turbine pitch system reliability analysis system 100 according to embodiments of the present invention can be integrated into the wind turbine's condition monitoring and control system as an independent software module or hardware module. For example, the core model used for basic event probability prediction in this system, namely the surrogate model set, includes multiple lightweight neural network models for specific fault modes such as bearing wear and motor overheating. These models can be trained, validated, and optimized offline using massive amounts of historical data on the backend server of the wind farm control center, and the optimized model weight package can be distributed to the front-end edge computing unit. Of course, the complete process used to perform real-time online analysis in this system, including multi-source heterogeneous data acquisition, dynamic feature vector construction, basic event probability surrogate prediction, communication link reliability modeling, dynamic fault tree injection and resolution, and the final reliability assessment and early warning generation, can also be embedded in dedicated edge computing hardware, such as embedded processors or FPGA / GPU modules with AI acceleration functions inside the edge computing unit, to accelerate the processing of high-frequency data streams and the real-time inference process of the surrogate models, and ensure low-latency generation of the final reliability assessment results.

[0055] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A reliability analysis method for a wind turbine pitch system based on 5G communication, characterized in that, include: Real-time acquisition of multi-source heterogeneous data from the wind turbine pitch system, including high-frequency sensor data, low-frequency operating condition data, and 5G communication network service quality data transmitted via 5G communication network; Time-domain and frequency-domain features are extracted from high-frequency sensor data and combined with low-frequency operating condition data and 5G communication network service quality data at the current moment to obtain the dynamic feature vector at the current moment. Based on the dynamic feature vector at the current moment, perform basic event probability proxy prediction to obtain the dynamic probability set at the current moment; To obtain the probability of communication failure at the current moment, communication link reliability modeling is performed on 5G communication network service quality data. Input the current dynamic probability set and the current communication failure probability into the static fault tree structure model to obtain the current top event probability and the current minimum cut set; Reliability assessment results are generated by performing a reliability assessment based on the current top event probability and the current minimum cut set.

2. The reliability analysis method for wind turbine pitch system based on 5G communication according to claim 1, characterized in that, High-frequency sensor data includes vibration signals, temperature signals, and current signals; low-frequency operating condition data includes wind speed, pitch angle, and power values; and 5G communication network service quality data includes signal strength, signal quality, network latency, and packet loss rate.

3. The reliability analysis method for wind turbine pitch system based on 5G communication according to claim 1, characterized in that, Extracting time-domain and frequency-domain features from high-frequency sensor data, including: Calculate the root mean square and kurtosis of high-frequency sensor data to obtain time-domain features; Fast Fourier Transform is performed on high-frequency sensor data to extract the spectral energy at the fault characteristic frequencies of bearings and gears as frequency domain features.

4. The reliability analysis method for wind turbine pitch system based on 5G communication according to any one of claims 1 to 3, characterized in that, Based on the current dynamic feature vector, perform basic event probability proxy prediction to obtain the current dynamic probability set, including: The current dynamic feature vector is inferred through various proxy models in the proxy model set to obtain the probability of occurrence of multiple basic failure events; The probabilities of multiple basic failure events are combined into a dynamic probability set at the current moment.

5. The reliability analysis method for a wind turbine pitch system based on 5G communication according to any one of claims 1 to 3, characterized in that, To obtain the probability of communication failure at the current moment, communication link reliability modeling is performed on 5G communication network service quality data, including: Obtain real-time packet loss rate from 5G communication network service quality data; Based on the real-time packet loss rate, the probability of instruction transmission failure is calculated to obtain the communication failure probability at the current moment.

6. The reliability analysis method for wind turbine pitch system based on 5G communication according to claim 5, characterized in that, Based on the real-time packet loss rate, the probability of instruction transmission failure is calculated to obtain the communication failure probability at the current moment, including: The probability of instruction transmission failure is calculated using the following formula, where the formula is: in, For real-time packet loss rate, The number of consecutive data packets required to complete a critical control command. This represents the probability of a communication failure at the current moment.

7. The reliability analysis method for wind turbine pitch system based on 5G communication according to any one of claims 1 to 3, characterized in that, The current dynamic probability set and the current communication failure probability are input into the static fault tree structure model to obtain the current top event probability and the current minimum cut set, including: The dynamic probabilities of each current moment in the current moment dynamic probability set are injected into the static fault tree structure model, and the communication failure probability of the current moment is injected into the static fault tree structure model to obtain the basic event probability update fault tree structure model. The logic gate propagation probability of the fault tree structure model is updated based on the basic event probability. The minimum cut set method is used for bottom-up calculation to solve for the current top event probability and the current minimum cut set.

8. The reliability analysis method for wind turbine pitch system based on 5G communication according to any one of claims 1 to 3, characterized in that, Reliability assessment is performed based on the current top event probability and the current minimum cut set to generate reliability assessment results, including: In response to the current event probability being less than a preset threshold, a reliability warning signal is generated, and a reliability assessment result is also generated.

9. The reliability analysis method for a wind turbine pitch system based on 5G communication according to any one of claims 1 to 3, characterized in that, Also includes: Based on the probability of the top event at the current moment of consecutive time points within a preset time window, generate a real-time reliability time series curve of the system. The real-time reliability time series curve of the system is displayed on the reliability dashboard.

10. A reliability analysis system for a wind turbine pitch system based on 5G communication, characterized in that, include: The real-time acquisition module is used to acquire multi-source heterogeneous data from the wind turbine pitch system in real time. The multi-source heterogeneous data includes high-frequency sensor data, low-frequency operating condition data, and 5G communication network service quality data transmitted through the 5G communication network. The current dynamic feature vector acquisition module is used to extract time-domain and frequency-domain features from high-frequency sensor data and combine them with low-frequency operating condition data and 5G communication network service quality data at the current moment to obtain the current dynamic feature vector. The basic event probability proxy prediction module is used to perform basic event probability proxy prediction based on the dynamic feature vector at the current moment to obtain the dynamic probability set at the current moment. The communication failure probability acquisition module is used to perform communication link reliability modeling on 5G communication network service quality data to obtain the communication failure probability at the current moment. The static fault tree structure model input module is used to input the current dynamic probability set and the current communication failure probability into the static fault tree structure model to obtain the current top event probability and the current minimum cut set. The reliability assessment result generation module is used to perform reliability assessment based on the current top event probability and the current minimum cut set to generate reliability assessment results.

Images