Remote operation and maintenance system and method for power equipment based on digital twinborn technology

Abstract

The invention relates to the technical field of intelligent monitoring and operation and maintenance of power equipment, and discloses a remote operation and maintenance system and method for power equipment based on a digital twinborn technology, and the system comprises a data collection module, a digital twinborn modeling module, a state monitoring and diagnosis module, and an operation and maintenance decision module. The data acquisition module adopts a distributed grid architecture and optimizes sensor sampling by using a PPO algorithm; the digital twin modeling module fuses RLC voltage and a simplified heat dissipation equation, and model parameters are corrected through scaled Kalman filtering and aging factors; the state monitoring and diagnosis module realizes fault early warning based on multi-threshold detection and a Bayesian network; the operation and maintenance decision module adopts a genetic algorithm to generate an optimization operation and maintenance instruction; the method comprises the steps of adjusting parameters, updating a model, triggering calculation, sending a dimensional instruction and correcting model parameters. According to the invention, full-process intelligent management of power equipment is realized, the operation and maintenance cost is optimized, and stable operation of a power system is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring and operation and maintenance of power equipment, and in particular to a remote operation and maintenance system and method for power equipment based on digital twin technology. Background Art

[0002] In the early days, power equipment operation and maintenance primarily relied on manual inspections and scheduled maintenance, with fault diagnosis based on periodic sensor data collection and threshold comparison. Traditional systems employed fixed sampling frequencies, making them unable to adapt to load fluctuations. Coupled analysis of multi-source data remained limited to simple correlations, making it difficult to identify potential fault signatures under complex operating conditions. Initial applications of digital twin technology often employed simplified physical models, ignoring the impact of equipment aging on multi-physics coupling. Consequently, model prediction errors accumulated over time.

[0003] Over time, some systems have attempted to improve data validity by dynamically adjusting sampling strategies. However, existing methods rely heavily on empirical rules and lack a data-driven, closed-loop optimization mechanism. Current operations and maintenance systems suffer from numerous deficiencies in data collection, transmission, and processing. Data collection accuracy and efficiency are limited, and there is a lack of intelligent dynamic adjustment mechanisms. Sensor sampling frequencies and threshold ranges cannot be optimized in real time based on the specific needs of different devices and environments, and data security during transmission is difficult to ensure. Regarding data processing, existing systems lack a high level of intelligence, making it difficult to efficiently analyze and mine massive amounts of complex data, and unable to promptly detect hidden fault signs and potential risks within the data.

[0004] Chinese invention patent CN118862641A discloses a digital twin-based power equipment monitoring method and system. This system uses historical data to construct temperature difference and power influence functions, runs the digital twin model, and monitors faults. This invention only statically determines parameter influence functions through historical data regression analysis, with a fixed sampling strategy that cannot be adaptively adjusted in real time. It also relies solely on direct fault signals from the digital twin model to trigger shutdowns, lacking probabilistic modeling and multi-dimensional diagnosis of complex fault causes.

[0005] Chinese invention patent CN119761219A discloses a method for optimizing the operation of power equipment based on digital twins. This method uses a particle filter algorithm for state estimation and a particle swarm optimization algorithm for adjusting operating parameters. This invention's data acquisition strategy is fixed and cannot adapt to changes in equipment operating conditions, potentially leading to missed data or wasted computing power. The invention focuses solely on state estimation and parameter optimization for digital twin models, without involving interactive visualization or graphical presentation of risk levels.

[0006] Therefore, the present invention proposes a remote operation and maintenance system and method for power equipment based on digital twin technology. Summary of the Invention

[0007] The present invention provides a remote operation and maintenance system and method for power equipment based on digital twin technology. The digital twin technology is integrated with PPO, Bayesian network and genetic algorithm to construct a remote operation and maintenance system for power equipment. The algorithm dynamically optimizes data collection and diagnosis, combines the aging factor to correct the digital twin model, evaluates risks in multiple dimensions and generates the optimal operation and maintenance strategy, thereby realizing intelligent operation and maintenance.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0009] The present invention provides a remote operation and maintenance system and method for power equipment based on digital twin technology, comprising:

[0010] The data acquisition module adopts a distributed grid monitoring architecture, with edge computing nodes deployed in each grid. It uses multiple types of sensors to collect ambient temperature, humidity, wind speed, and equipment voltage, current, power, and load rate data. It uses the PPO algorithm to dynamically optimize the sensor sampling frequency and threshold range. The data acquisition module integrates AES encryption transmission function.

[0011] The digital twin modeling module constructs a digital twin model of power equipment that integrates the RLC voltage equation and the simplified heat dissipation equation; updates the model current and temperature by integrating real-time data through scalar Kalman filtering; introduces a dynamic correction parameter for the aging factor, and corrects the thermal time constant based on the insulation degradation coefficient α and the mechanical wear coefficient β;

[0012] The condition monitoring and diagnosis module adopts a dual diagnosis mechanism of multi-threshold dynamic detection and causal chain analysis; uses Bayesian networks to achieve probabilistic early warning of four typical faults of power equipment; and generates a risk map of power equipment based on the three-dimensional visualization interface of the digital twin model.

[0013] The operation and maintenance decision module generates an operation and maintenance instruction sequence based on a genetic algorithm that takes into account risk factors, operation and maintenance costs, and aging factors; transmits JSON format instructions via the HTTPS protocol and receives execution status feedback; uses a closed-loop learning unit to correct the Bayesian network weights based on the gradient of the difference between actual faults and predictions; and outputs a cross-grid collaborative operation and maintenance path.

[0014] The beneficial effects brought about by the technical solution provided by the present invention include at least:

[0015] The present invention dynamically adjusts the sensor sampling frequency and threshold range through the PPO algorithm to solve the problems of missed detection and redundancy in the traditional fixed sampling mode.

[0016] The present invention uses a distributed grid monitoring architecture and edge computing nodes to locally process data, reducing the load on the central platform, lowering data processing delays, and supporting real-time monitoring of large-scale equipment clusters.

[0017] The present invention accurately simulates the coupling relationship between the electrical performance and thermal characteristics of the device by integrating the RLC voltage equation and the simplified heat dissipation equation, thereby solving the model inaccuracy problem caused by ignoring the electrothermal interaction in the prior art.

[0018] The present invention introduces the insulation degradation coefficient α and the mechanical wear coefficient β to correct the thermal time constant k, so as to reflect the aging status of the equipment in real time.

[0019] The present invention utilizes multi-threshold dynamic detection and causal chain analysis technology, combined with Bayesian networks, to provide probabilistic early warning of equipment failures, discovering potential equipment failures in advance, enabling operation and maintenance personnel to arrange maintenance plans in advance, avoiding power outages caused by sudden equipment failures, and improving the reliability and power supply continuity of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 is a schematic diagram of the system architecture provided by an embodiment of the present invention;

[0022] Figure 2 Flowchart of the remote operation and maintenance method of power equipment based on digital twin technology provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0023] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0024] First embodiment

[0025] This embodiment provides a remote operation and maintenance system for power equipment based on digital twin technology, including:

[0026] Data acquisition module, digital twin modeling module, condition monitoring and diagnosis module, and operation and maintenance decision module;

[0027] The data acquisition module adopts a distributed grid monitoring architecture, with edge computing nodes deployed in each grid. It uses multiple types of sensors to collect ambient temperature, humidity, wind speed, and equipment voltage, current, power, and load rate data. It uses the PPO algorithm to dynamically optimize sensor sampling frequency and threshold range. The data acquisition module integrates AES encryption transmission function.

[0028] The digital twin modeling module builds a digital twin model of power equipment that integrates the RLC voltage equation and the simplified heat dissipation equation. It integrates real-time data through a scalar Kalman filter to update the model current and temperature. It also introduces a dynamic correction parameter for the aging factor to correct the thermal time constant based on the insulation degradation coefficient α and the mechanical wear coefficient β.

[0029] The condition monitoring and diagnosis module uses a dual diagnostic mechanism of multi-threshold dynamic detection and causal chain analysis; it uses a Bayesian network to achieve probabilistic early warning of four typical fault types of power equipment; and it generates a risk map of power equipment based on a three-dimensional visualization interface of the digital twin model.

[0030] The operation and maintenance decision-making module generates an operation and maintenance instruction sequence based on a genetic algorithm that takes into account risk factors, operation and maintenance costs, and aging factors. It transmits JSON-formatted instructions via the HTTPS protocol and receives execution status feedback. Through a closed-loop learning unit, it modifies the Bayesian network weights based on the gradient of the difference between actual faults and predictions. It then outputs a cross-grid collaborative operation and maintenance path.

[0031] Please refer to Figure 1 This is a schematic diagram of the architecture of the remote operation and maintenance system for power equipment based on digital twin technology provided by an embodiment of the present invention.

[0032] 1. Data Acquisition Module

[0033] Distributed grid monitoring unit, used to divide power equipment into multiple grids according to geographical areas, with each grid associated with an edge computing node;

[0034] Multiple types of sensors are used to collect environmental data and power equipment operation data; environmental data includes ambient temperature, humidity, and wind speed; operation data includes voltage, current, power, and equipment load rate of power equipment;

[0035] The encryption transmission unit uses the AES symmetric encryption algorithm to uniformly encrypt environmental data and operating data before transmitting them to the remote platform. The encryption transmission unit uses a dynamic key rotation mechanism, and the dynamic key is updated periodically.

[0036] Edge computing nodes use the PPO algorithm to perform real-time coupled analysis of environmental data and operational data. The PPO algorithm uses historical data and real-time data as input and dynamically generates the optimal sensor sampling frequency and threshold range through continuous learning and optimization.

[0037] The specific steps of the PPO algorithm include:

[0038] S1, data preprocessing, cleans the collected environmental data and operation data to remove outliers, missing values, and duplicate values; normalizes the cleaned data to unify data of different ranges into the [0,1] interval; performs wavelet denoising and Fourier transform on the normalized data to extract useful features;

[0039] S2, define reinforcement learning elements, compose the extracted useful features into a state vector; determine the adjustable values ​​of sensor sampling frequency and threshold range; design reward rules;

[0040] S3, initialize the model, select the PPO algorithm, and initialize its model parameters;

[0041] S4, interactive learning and reward calculation, the PPO algorithm model outputs the continuous action of the sensor sampling frequency and threshold adjustment value based on the current state vector; the sensor collects data according to the adjusted new parameters, and if a fault feature is detected, a positive reward is given; if a false alarm is detected, a negative reward is given, and the state, action, and reward are recorded in the experience buffer;

[0042] S5, strategy optimization, randomly selects historical data from the experience buffer and calculates the action advantage; through the PPO algorithm's unique clipping mechanism, it limits the strategy update range to avoid drastic adjustments that may lead to misjudgment;

[0043] S6, continuous iteration, repeating S4-S5, accumulating data and continuously optimizing the strategy until the fault capture rate and false alarm rate tend to be stable. The edge computing node calls the optimized strategy in real time and dynamically adjusts the sensor sampling frequency and threshold range.

[0044] It should be noted that the situations that trigger key updates include fixed cycles, single data transmission volume > 1GB, and edge nodes detecting abnormal access requests.

[0045] Fault characteristics include current surge, abnormal temperature gradient, insulation degradation signal, excessive mechanical vibration and arc discharge characteristics.

[0046] 2. Digital Twin Modeling Module

[0047] The physical structure mapping unit builds a digital twin model of the power equipment based on its geometric parameters and circuit topology, integrating the RLC voltage equation and simplified heat dissipation equation;

[0048] The calculation formula of the RLC voltage equation is:

[0049]

[0050] Where V(t) represents the total voltage at time t; R represents the resistance; I(t) represents the current at time t; L represents the inductance; C represents the capacitance;

[0051] The simplified heat dissipation equation is:

[0052] T(t)=T env +k·I 2 (t)·r

[0053] Where T(t) represents the temperature at time t; T env represents the ambient temperature; r represents the thermal resistance; k represents the thermal time constant;

[0054] The data fusion unit processes the encrypted transmitted current and temperature data through scalar Kalman filtering and updates I(t) and T(t) in the model;

[0055] Aging factor integration unit calculates the insulation degradation coefficient α based on the accumulated operating time, calculates the mechanical wear coefficient β based on the number of starts and stops, and corrects the resistance R and thermal time constant k;

[0056] The correction formula of the aging factor integrated unit for the thermal time constant k is:

[0057] k=k0·(1+γ·β)

[0058] Where k0 represents the initial thermal time constant; β is the mechanical wear coefficient; γ is the coupling coefficient, γ∈[0.01,0.99], which is adjusted by scalar Kalman filtering;

[0059] The calculation formula of mechanical wear coefficient is:

[0060] β=1-e -λN

[0061] Where e is the base of the natural logarithm; N is the number of times the equipment is started and stopped; and λ is the material attenuation coefficient.

[0062] It should be noted that different types of power equipment require different methods for obtaining geometric parameters and circuit topology. For example, for transformers, geometric parameters such as the number of winding turns and core size can be directly obtained from the device's design drawings. If these drawings are missing, laser scanning technology is used to perform a 3D scan of the transformer to obtain precise external dimensions. Nondestructive testing techniques are then used to determine the internal winding structure parameters.

[0063] Before entering the scalar Kalman filter, encrypted current and temperature data undergo data cleansing to remove outliers and duplicates. Wavelet denoising is then used to improve data quality. The denoised data is then normalized and mapped to the [0, 1] range for subsequent calculations.

[0064] The scalar Kalman filter dynamically adjusts the coupling coefficient γ by monitoring changes in the device's operating state. When the device is overloaded, the current and temperature change suddenly. At this time, the scalar Kalman filter adjusts γ in steps of 0.01 based on the rate of change of current and temperature.

[0065] 3. Condition Monitoring and Diagnosis Module

[0066] The multi-threshold dynamic detection unit compares the power equipment operation data, environmental data and the threshold range output by the digital twin model in real time based on the sampling frequency optimized by the edge computing node. The threshold range is dynamically adjusted through the PPO algorithm.

[0067] The causal chain analysis unit constructs a causal network of environmental data, equipment operation data, and fault data, and calculates the probability of typical faults using a Bayesian network. The Bayesian network contains no more than 20 key nodes and covers four typical fault modes: insulation breakdown, mechanical fatigue, loose connections, and arc discharge.

[0068] The hierarchical early warning unit is used to receive the real-time data output by the multi-threshold dynamic detection unit, compare it with the threshold range, and the fault probability data output by the causal chain analysis unit, perform early warning logic processing, and output the following content:

[0069] Transmitting a warning level identification signal to the risk map unit. The warning level identification includes three states: red warning, yellow warning, and green normal;

[0070] Push a structured alarm information package to the operation and maintenance decision module. The information package contains at least the equipment number, warning level, trigger time, and associated parameter abnormality fields;

[0071] Early warning logic processing includes:

[0072] When the real-time data exceeds the upper threshold and the failure probability P>0.7, a red warning is triggered;

[0073] When the real-time data exceeds the upper threshold or the failure probability P>0.4, a yellow warning is triggered;

[0074] For other conditions, maintain the green normal state;

[0075] The risk map unit, based on the received warning level identification, marks the risk level of power equipment in red, yellow and green in the three-dimensional visualization interface of the digital twin model; red indicates high risk, yellow indicates medium risk, and green indicates low risk.

[0076] It should be noted that the PPO algorithm dynamically adjusts the threshold range based on the device load factor. When the device load factor is below 30%, the algorithm prioritizes reducing the sensitivity of the current and voltage thresholds to avoid false alarms under low load conditions. When the load factor exceeds 70%, the temperature threshold range is narrowed to improve the accuracy of high-temperature fault detection. During the data comparison process, parallel computing technology is used to group environmental data and device operating data and compare them against the threshold simultaneously. If the voltage and current data of the same device are collected by different edge computing nodes, the nodes synchronize the data every 5 seconds via the MQTT protocol to ensure consistency in threshold judgments.

[0077] When constructing the causal network, the PC-stable algorithm was used for causal discovery. The original data was first discretized using equal-frequency binning, dividing the continuous variable into five intervals. The 18 key nodes with the highest correlation with four typical fault types were screened. The Bayesian network training data was derived from historical equipment operation records and fault reports. This training data was further processed to extract key features. The Bayesian network was trained using the expectation-maximization algorithm to determine the conditional probability distribution of the key nodes.

[0078] When real-time data is missing or erroneous, the hierarchical warning unit will complete and correct the data. For missing data, interpolation or prediction methods based on historical data are used to complete the data. For data errors, data validation rules are used to determine and correct the data.

[0079] 4. Operation and Maintenance Decision Module

[0080] The instruction optimization unit generates an operation and maintenance instruction sequence through a genetic algorithm based on the warning level output by the risk map unit. The fitness function of the genetic algorithm is:

[0081] F=ω1·A+ω2·B

[0082] Where ω1 and ω2 represent weight coefficients; A represents the risk coefficient; B represents the operation and maintenance cost, B∈[0,1]; the risk coefficient value includes: red warning corresponds to 1.0, yellow warning corresponds to 0.5, and green normal corresponds to 0.1; the operation and maintenance cost includes labor cost and equipment downtime loss cost;

[0083] The remote interface unit transmits operation and maintenance instructions to the device terminal via the HTTPS protocol and AES encryption, and receives feedback from the terminal on the operation and maintenance execution status. Operation and maintenance instructions are in JSON format and include the device number, instruction content, priority, and designated operation and maintenance personnel information. Execution status includes success, failure, and delay.

[0084] The closed-loop learning unit is used to modify the weight parameters of the Bayesian network using the gradient descent method based on the difference between the operation and maintenance execution results and the predicted faults. The modification formula is:

[0085] Δ∈=η·(y 实际 -y 预测 )

[0086] Among them, η is the learning rate; y 实际 Indicates actual fault, 1 means yes, 0 means no; y 预测 represents the predicted failure probability;

[0087] The collaborative optimization unit is used to calculate the optimal operation and maintenance path through the genetic algorithm based on the grid division results of the distributed grid monitoring unit. The fitness function of the genetic algorithm is:

[0088]

[0089] Among them, δ represents the equipment aging coefficient, which is a weighted fusion of the insulation degradation coefficient α and the mechanical wear coefficient β.

[0090] It should be noted that labor costs are calculated based on the level and working hours of the operation and maintenance personnel.

[0091] Equipment downtime loss costs are estimated based on the type of equipment and the duration of downtime.

[0092] Example of specific implementation steps of genetic algorithm:

[0093] Using binary encoding, each operation and maintenance task is represented by a binary bit to indicate whether it is executed. For example, if there are 5 operation and maintenance tasks, the code of an individual task may be [1,0,1,1,0], indicating that tasks 1, 3, and 4 are executed, and tasks 2 and 5 are not executed;

[0094] Randomly generate 50 individuals as the initial population;

[0095] Using the roulette wheel selection method, the probability of each individual being selected is calculated based on its fitness value. The higher the fitness value, the greater the probability of being selected;

[0096] The crossover probability is set to 0.8. Two individuals are randomly selected and a new individual is generated by single-point crossover. For example, if individual A = [1, 0, 1, 1, 0] and individual B = [0, 1, 0, 1, 1], and the crossover point is the third position, then the new individuals generated after crossover are A' = [1, 0, 0, 1, 1] and B' = [0, 1, 1, 1, 0];

[0097] The mutation probability is set to 0.01. Randomly select a binary bit of an individual and perform the inversion operation. For example, the individual [1,0,1,1,0] may become [1,0,0,1,0] after mutation;

[0098] When the optimal fitness value of the population no longer improves for 20 consecutive generations, the algorithm is terminated.

[0099] Example of genetic algorithm steps to find the optimal path:

[0100] Using path coding, each individual represents an operation and maintenance path. For example, if there are 4 devices, the code of an individual may be [1,3,4,2], which means that the operation and maintenance personnel visit devices 1, 3, 4, and 2 in sequence.

[0101] Randomly generate 30 individuals as the initial population;

[0102] Based on the operation and maintenance path represented by each individual, the sum of the aging coefficients of all devices on the path is calculated as the fitness value. The higher the fitness value, the better the path.

[0103] The selection operation uses the tournament selection method, randomly selecting three individuals from the population each time, and the individual with the highest fitness value is selected to advance to the next generation. The crossover operation uses the sequential crossover method, and the mutation operation uses the exchange mutation method, which randomly exchanges the order of two devices in the path;

[0104] When the maximum number of iterations reaches 100, the algorithm is terminated and the individual with the highest fitness value is output as the optimal operation and maintenance path.

[0105] Second embodiment

[0106] This embodiment provides a remote operation and maintenance method for power equipment based on digital twin technology. The method is applied in a power system covering multiple transformers, transmission lines and ancillary equipment.

[0107] Please refer to Figure 2 Flowchart of the remote operation and maintenance method of power equipment based on digital twin technology provided in an embodiment of the present invention.

[0108] The specific steps include:

[0109] S1. Run the PPO algorithm through the edge computing node to dynamically adjust the sensor sampling frequency and alarm threshold;

[0110] The PPO algorithm was deployed on each grid edge computing node, and parameters were optimized using grid search. The learning rate range was set to 0.0001-0.001, and the discount factor range was set to 0.95-0.99. Data was collected every 50ms, and outliers were removed using the 3σ principle.

[0111] For example, a transformer operates at a sampling frequency of 10 seconds during normal operation. When the PPO algorithm determines it is nearing full load based on data such as the load factor, it increases the sampling frequency to 2 seconds and adjusts the current alarm threshold from 1.2 times the rated current to 1.25 times. This adjustment improves the transformer's fault detection rate by 30% and reduces the false alarm rate by 25%.

[0112] S2. Input the AES-encrypted current and temperature data into the scalar Kalman filter to update the RLC equation parameters and thermodynamic state variables in the digital twin model;

[0113] Using the AES-256 encryption algorithm, keys are generated based on device identifiers and timestamps. Dynamic key update trigger conditions include:

[0114] Periodic update, default is 24 hours;

[0115] The number of single-node data packet retransmissions is greater than 3 times;

[0116] The key has been used > 100,000 times.

[0117] The edge computing node decrypts the encrypted data after receiving it and inputs it into the scalar Kalman filter.

[0118] For a transformer with an initial resistance of 10Ω, an inductance of 0.5H, and a capacitance of 0.001F, filtered data was substituted into the RLC and heat dissipation equations to update the parameters. After the update, the error between the model output temperature and the actual measured temperature was controlled within ±2°C.

[0119] S3. When the failure probability P output by the Bayesian network is greater than 0.7, the genetic algorithm is triggered to calculate the operation and maintenance instructions;

[0120] The Bayesian network was trained using over 100,000 pieces of historical operational data, covering equipment status at the time of the failure, environmental conditions, and maintenance records. After data preprocessing, the causal relationships and conditional probability distributions of 15 key nodes were determined.

[0121] When the probability of a transformer failure, P, exceeds 0.7, the genetic algorithm is triggered, and the population size is set to the square root of the number of currently online devices. Assuming a population size of 50, a crossover probability of 0.8, and a mutation probability of 0.01, the algorithm generates an O&M instruction based on the risk factor (red alert A = 1.0), O&M cost (B = 0.6), and aging factor: arrange for a professional to conduct an on-site inspection within 2 hours and prepare a backup transformer. This instruction is given the highest priority.

[0122] S4. Send a JSON data packet containing the device number, command content, priority, and designated operation and maintenance personnel information to the target device via the HTTPS protocol;

[0123] The JSON data packet format is as follows:

[0124] json

[0125] {

[0126] "device_id":"001",

[0127] "instruction":"Immediately carry out emergency inspection and prepare spare transformers",

[0128] "priority":"high",

[0129] "maintainer": "Zhang San"}

[0130] The operation and maintenance decision module sends a data packet and automatically retransmits it three times if a network problem occurs.

[0131] S5, model modification, modifying the Bayesian network weight parameters according to the difference between the actual fault and the predicted result;

[0132] The mean square error (MSE) is used to calculate the difference between the actual fault and the predicted result. For example, for an insulation breakdown fault, the actual label is 1 and the predicted probability is 0.8, so the MSE is 0.04.

[0133] The Bayesian network weights were modified using gradient descent at a learning rate of 0.01, with regular validation to prevent overfitting. After multiple revisions, the accuracy of 100 fault predictions increased from 80% to 85%.

[0134] The practical application of the remote operation and maintenance system and method for power equipment based on digital twin technology in this embodiment can improve the operation and maintenance efficiency and reliability of power equipment and ensure the stable operation of the power system.

[0135] Furthermore, it should be noted that the present invention may be provided as a method, apparatus, or computer program product. Thus, embodiments of the present invention may take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention may take the form of a computer program product embodied on one or more computer-usable storage media containing computer-usable program code.

[0136] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the process in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0137] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for implementing the process in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0138] It should also be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. The terms "comprises," "includes," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or terminal device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or terminal device. In the absence of further restrictions, an element defined by the phrase "comprises a ..." does not exclude the presence of additional identical elements in the process, method, article, or terminal device that includes the element.

[0139] Finally, it should be noted that the above is a preferred embodiment of the present invention. It should be noted that although the preferred embodiment of the present invention has been described, it is clear that those skilled in the art, once they understand the basic inventive concept of the present invention, can make various improvements and modifications without departing from the principles of the present invention. Such improvements and modifications should also be considered as within the scope of protection of the present invention. Therefore, the appended claims are intended to be interpreted as including the preferred embodiment and all changes and modifications that fall within the scope of the embodiments of the present invention.

Claims

1. A remote operation and maintenance system for power equipment based on digital twin technology, characterized in that: include: Data acquisition module, digital twin modeling module, condition monitoring and diagnosis module, and operation and maintenance decision module; The data acquisition module adopts a distributed grid monitoring architecture, with edge computing nodes deployed in each grid. It uses multiple types of sensors to collect ambient temperature, humidity, wind speed, and equipment voltage, current, power, and load rate data. It uses the PPO algorithm to dynamically optimize the sensor sampling frequency and threshold range. The data acquisition module integrates AES encryption transmission function. The digital twin modeling module constructs a digital twin model of power equipment that integrates the RLC voltage equation and the simplified heat dissipation equation; updates the model current and temperature by integrating real-time data through scalar Kalman filtering; introduces a dynamic correction parameter for the aging factor, and corrects the thermal time constant based on the insulation degradation coefficient α and the mechanical wear coefficient β; The state monitoring and diagnosis module adopts a dual diagnosis mechanism of multi-threshold dynamic detection and causal chain analysis; it uses Bayesian networks to achieve probabilistic early warning of four typical faults of power equipment; Generate risk maps of power equipment based on the three-dimensional visualization interface of the digital twin model; The operation and maintenance decision module generates an operation and maintenance instruction sequence based on a genetic algorithm that takes into account risk factors, operation and maintenance costs, and aging factors; transmits JSON format instructions via the HTTPS protocol and receives execution status feedback; and uses a closed-loop learning unit to modify the Bayesian network weights based on the gradient difference between actual faults and predictions. Output cross-grid collaborative operation and maintenance path.

2. The remote operation and maintenance system for electric power equipment based on digital twin technology according to claim 1 is characterized in that: The data acquisition module includes: Distributed grid monitoring unit, used to divide power equipment into multiple grids according to geographical areas, with each grid associated with an edge computing node; Multiple types of sensors are used to collect environmental data and power equipment operating data; the environmental data includes the ambient temperature, humidity and wind speed; the operating data includes the voltage, current, power and equipment load rate of the power equipment; The encryption transmission unit uses the AES symmetric encryption algorithm to uniformly encrypt the environmental data and operating data and transmit them to the remote platform; the encryption transmission unit uses a dynamic key rotation mechanism, and the dynamic key is periodically updated; The edge computing node uses the PPO algorithm to perform real-time coupled analysis of environmental data and operational data. The PPO algorithm takes historical data and real-time data as input, and dynamically generates the optimal sensor sampling frequency and threshold range through continuous learning and optimization.

3. The remote operation and maintenance system for electric power equipment based on digital twin technology according to claim 2 is characterized in that: The specific steps of the PPO algorithm include: S1, data preprocessing, cleans the collected environmental data and operation data to remove outliers, missing values, and duplicate values; normalizes the cleaned data to unify data of different ranges into the [0,1] interval; performs wavelet denoising and Fourier transform on the normalized data to extract useful features; S2, define reinforcement learning elements, compose the extracted useful features into a state vector; determine the adjustable values ​​of sensor sampling frequency and threshold range; design reward rules; S3, initialize the model, select the PPO algorithm, and initialize its model parameters; S4, interactive learning and reward calculation, the PPO algorithm model outputs the continuous action of the sensor sampling frequency and threshold adjustment value based on the current state vector; the sensor collects data according to the adjusted new parameters, and if a fault feature is detected, a positive reward is given; if a false alarm is detected, a negative reward is given, and the state, action, and reward are recorded in the experience buffer; S5, strategy optimization, randomly selects historical data from the experience buffer and calculates the action advantage; through the PPO algorithm's unique clipping mechanism, it limits the strategy update range to avoid drastic adjustments that may lead to misjudgment; S6, continuous iteration, repeating S4-S5, accumulating data and continuously optimizing the strategy until the fault capture rate and false alarm rate tend to be stable. The edge computing node calls the optimized strategy in real time and dynamically adjusts the sensor sampling frequency and threshold range.

4. The remote operation and maintenance system for electric power equipment based on digital twin technology according to claim 1 is characterized in that: The digital twin modeling module includes: The physical structure mapping unit builds a digital twin model of the power equipment based on its geometric parameters and circuit topology, integrating the RLC voltage equation and simplified heat dissipation equation; The calculation formula of the RLC voltage equation is: Where V(t) represents the total voltage at time t; R represents the resistance; I(t) represents the current at time t; L represents the inductance; C represents the capacitance; The calculation formula of the simplified heat dissipation equation is: T(t)=T env +k·I 2 (t)·r Where T(t) represents the temperature at time t; T env represents the ambient temperature; r represents the thermal resistance; k represents the thermal time constant; The data fusion unit processes the encrypted transmitted current and temperature data through scalar Kalman filtering and updates I(t) and T(t) in the model; The aging factor integration unit calculates the insulation degradation coefficient α based on the accumulated operating time, calculates the mechanical wear coefficient β based on the number of starts and stops, and corrects the resistance R and thermal time constant k.

5. The remote operation and maintenance system for electric power equipment based on digital twin technology according to claim 4 is characterized in that: The correction formula of the aging factor integrated unit for the thermal time constant k is: k=k0·(1+γ·β) Where k0 represents the initial thermal time constant; β is the mechanical wear coefficient; γ is the coupling coefficient, γ∈[0.01,0.99], which is adjusted by scalar Kalman filtering; The calculation formula of the mechanical wear coefficient is: β=1-e-λN Where e is the base of the natural logarithm; N is the number of times the equipment is started and stopped; and λ is the material attenuation coefficient.

6. The remote operation and maintenance system for electric power equipment based on digital twin technology according to claim 1 is characterized in that: The condition monitoring and diagnosis module includes: The multi-threshold dynamic detection unit compares the power equipment operation data, environmental data and the threshold range output by the digital twin model in real time based on the sampling frequency optimized by the edge computing node. The threshold range is dynamically adjusted through the PPO algorithm. A causal chain analysis unit constructs a causal network of environmental data, equipment operation data, and fault data, and calculates the probability of typical faults using a Bayesian network containing no more than 20 key nodes, covering four typical fault modes: insulation breakdown, mechanical fatigue, loose connectors, and arc discharge. The hierarchical early warning unit is used to receive the real-time data output by the multi-threshold dynamic detection unit, compare it with the threshold range, and the fault probability data output by the causal chain analysis unit, perform early warning logic processing, and output the following content: Transmitting a warning level identification signal to the risk map unit, wherein the warning level identification includes three states: red warning, yellow warning, and green normal; Pushing a structured alarm information package to the operation and maintenance decision module, the information package at least includes the device number, warning level, trigger time and associated parameter abnormality item fields; The early warning logic processing includes: When the real-time data exceeds the upper threshold and the failure probability P>0.7, a red warning is triggered; When the real-time data exceeds the upper threshold or the failure probability P>0.4, a yellow warning is triggered; For other conditions, maintain the green normal state; The risk map unit, based on the received warning level identification, marks the risk level of power equipment in red, yellow and green in the three-dimensional visualization interface of the digital twin model; red indicates high risk, yellow indicates medium risk, and green indicates low risk.

7. The remote operation and maintenance system for electric power equipment based on digital twin technology according to claim 1 is characterized in that: The operation and maintenance decision module includes: The instruction optimization unit generates an operation and maintenance instruction sequence through a genetic algorithm based on the warning level output by the risk map unit. The fitness function of the genetic algorithm is: F=ω1·A+ω2·B Wherein, ω1 and ω2 represent weight coefficients; A represents the risk coefficient; B represents the operation and maintenance cost, B∈[0,1]; the value of the risk coefficient includes: red warning corresponds to 1.0, yellow warning corresponds to 0.5, and green normal corresponds to 0.1; the operation and maintenance cost includes labor cost and equipment downtime loss cost; The remote interface unit transmits operation and maintenance instructions to the device terminal via the HTTPS protocol and AES encryption, and receives the operation and maintenance execution status feedback from the terminal; the operation and maintenance instructions are in JSON format and include the device number, instruction content, priority, and designated operation and maintenance personnel information; the execution status includes success, failure, and delay; The closed-loop learning unit is used to modify the weight parameters of the Bayesian network using the gradient descent method based on the difference between the operation and maintenance execution results and the predicted faults. The modification formula is: Δ∈=η·(and 实际 -and 预测 ) Among them, η is the learning rate; y 实际 Indicates actual fault, 1 means yes, 0 means no; y 预测 represents the predicted failure probability; The collaborative optimization unit is used to calculate the optimal operation and maintenance path through a genetic algorithm based on the grid division results of the distributed grid monitoring unit. The fitness function of the genetic algorithm is: Among them, δ represents the equipment aging coefficient, which is a weighted fusion of the insulation degradation coefficient α and the mechanical wear coefficient β.

8. A remote operation and maintenance method for power equipment based on digital twin technology, characterized in that: Including steps: S1, runs the PPO algorithm through the edge computing node to dynamically adjust the sensor sampling frequency and alarm threshold; S2, inputs the AES-encrypted current and temperature data into the scalar Kalman filter to update the RLC equation parameters and thermodynamic state variables in the digital twin model; S3, when the failure probability P output by the Bayesian network is greater than 0.7, the genetic algorithm is triggered to calculate the operation and maintenance instructions; S4, sending a JSON data packet containing the device number, command content, priority, and designated operation and maintenance personnel information to the target device via the HTTPS protocol; S5, model modification, modifies the Bayesian network weight parameters according to the difference between the actual fault and the predicted results.

Images