Adaptive maintaining method and system for time-varying consistency of digital twin components of power transformation equipment

Abstract

The invention provides a power transformation equipment digital twinborn part time-varying consistency self-adaptive maintaining method and system, and relates to the technical field of power grids, and the method comprises the steps: collecting the real-time operation data of power transformation equipment, constructing a physical model, generating a digital twinborn model, and obtaining the initial state parameters of parts through deep learning training. And establishing a time-varying feature vector matrix based on the initial state parameters, performing real-time updating by using an adaptive Kalman filtering algorithm to obtain a dynamic feature model, and calculating a state deviation value of the digital twin component and the physical equipment. And when the deviation value exceeds a preset threshold value, triggering a self-adaptive consistency keeping mechanism, generating a state compensation parameter through deep reinforcement learning, applying the state compensation parameter to a digital twin model, and realizing dynamic synchronization of the digital twin component and the physical equipment. According to the invention, the state consistency of the digital twinning component and the physical equipment can be adaptively maintained, and the precision and reliability of digital twinning are improved.

Description

Technical Field

[0001] The present invention relates to power grid technology, and in particular to a method and system for adaptively maintaining time-varying consistency of digital twin components of power transformation equipment. Background Art

[0002] As power systems become increasingly complex and intelligent, the reliable operation of substation equipment is crucial to ensuring power supply. Traditional substation equipment operation and maintenance methods mainly rely on manual inspections and regular maintenance, which are inefficient and difficult to detect potential faults in a timely manner. Digital twin technology provides a new solution for the intelligent operation and maintenance of substation equipment. By building a digital twin model corresponding to the physical equipment, real-time monitoring of equipment status, fault diagnosis and predictive maintenance can be achieved.

[0003] However, the existing digital twin technology still has some defects and shortcomings when applied to substation equipment: 1. It is difficult to maintain the state consistency between the digital twin model and the physical device for a long time. Due to the complex operating environment of substation equipment and the influence of various factors, its state parameters will change over time. Traditional digital twin model construction methods usually only consider the initial state of the equipment and are difficult to adapt to the time-varying characteristics of the equipment, resulting in a gradual increase in the state deviation between the digital twin model and the physical device, affecting the accuracy and reliability of the model.

[0004] 2. It is difficult to effectively process the multi-source heterogeneous data of substation equipment. There are many types of operating data of substation equipment, including voltage, current, temperature, vibration, etc., and the data sources and formats are different. Traditional digital twin model construction methods are usually difficult to effectively integrate and process these multi-source heterogeneous data, making it difficult for the model to fully reflect the operating status of the equipment.

[0005] 3. Lack of adaptive model update mechanism. The operating status of substation equipment will change over time, and the digital twin model also needs to be constantly updated to maintain consistency with the physical equipment. Traditional digital twin model update methods usually rely on manual intervention or fixed time periods, which are difficult to adapt to the dynamic changes in equipment status, resulting in untimely or too frequent model updates, affecting the efficiency and accuracy of the model. Summary of the invention

[0006] The embodiments of the present invention provide a method and system for adaptively maintaining the time-varying consistency of digital twin components of substation equipment, which can solve the problems in the prior art.

[0007] According to a first aspect of the embodiments of the present invention, A method for adaptively maintaining the time-varying consistency of digital twin components of substation equipment is provided, including: Collecting real-time operation data of the substation equipment, the real-time operation data including voltage data, current data, temperature data, vibration data and load data of the substation equipment; constructing a physical model of the substation equipment based on the real-time operation data, and generating a digital twin model of the substation equipment according to the physical model; inputting the real-time operation data into the digital twin model, training the digital twin model through a deep learning algorithm, and obtaining initial state parameters of the digital twin components of the substation equipment; Based on the initial state parameters, a time-varying eigenvector matrix of the digital twin component of the substation equipment is established, wherein the time-varying eigenvector matrix includes equipment operation state characteristics, equipment performance attenuation characteristics, and equipment failure characteristics; the time-varying eigenvector matrix is ​​updated in real time using an adaptive Kalman filter algorithm to obtain a dynamic feature model of the digital twin component of the substation equipment; and the state deviation value between the digital twin component of the substation equipment and the physical device is calculated according to the dynamic feature model; When the state deviation value exceeds the preset deviation threshold, the adaptive consistency maintenance mechanism is triggered; the adaptive consistency maintenance mechanism generates state compensation parameters through a deep reinforcement learning algorithm, and applies the state compensation parameters to the digital twin model to achieve dynamic synchronization between the digital twin components of the substation equipment and the physical equipment.

[0008] Based on the real-time operation data, a physical model of the substation equipment is constructed, and a digital twin model of the substation equipment is generated according to the physical model, including: The real-time operation data is constructed as a state vector, wherein the state vector includes the voltage data, the current data, the temperature data, the vibration data and the load data; based on the state vector, a physical model of the substation equipment is constructed in combination with a device physical parameter matrix and an environmental constraint matrix, wherein the device physical parameter matrix includes device geometric dimension parameters, material physical property parameters, electrical property parameters and thermodynamic property parameters, and the environmental constraint matrix includes environmental temperature parameters, humidity parameters and electromagnetic field strength parameters; The physical model is converted into a digital twin model by a digital mapping conversion function, wherein the digital mapping conversion function discretizes the physical quantity, digitally converts the geometric model, maps the dynamic characteristic parameters, and digitally expresses the constraint conditions of the physical model; A model error between the digital twin model and the physical model is calculated, an error compensation function is generated based on the model error, and the error compensation function is applied to the digital twin model to obtain an optimized digital twin model.

[0009] Inputting the real-time operation data into the digital twin model, training the digital twin model through a deep learning algorithm, and obtaining the initial state parameters of the digital twin components of the substation equipment include: Constructing a hybrid architecture of a multi-layer perceptron and a recurrent neural network, wherein the hybrid architecture of the multi-layer perceptron and the recurrent neural network is used to extract the time series features of the real-time operation data of the substation equipment, wherein the time series features are modeled by a long short-term memory network, and the long short-term memory network realizes effective modeling of the time series data through the synergy of the input gate, the forget gate and the memory unit; A multi-head self-attention mechanism is introduced to enhance the model's ability to focus on key features, and to achieve adaptive feature extraction of input data through weighted calculation of query matrix, key matrix and value matrix; The hybrid architecture is trained using an adaptive learning rate optimization algorithm, which improves the convergence speed and stability of the model by dynamically adjusting the learning rate. After the training is completed, the initial state parameters of the digital twin components of the substation equipment are extracted based on the test data set, and the initial state parameters are used to characterize the current operating state of the substation equipment; The reliability of the initial state parameters is verified by a cross-validation method, and the cross-validation method evaluates the generalization ability of the model through multiple training and validation processes to ensure the accuracy and reliability of the initial state parameters.

[0010] The time-varying eigenvector matrix is ​​updated in real time by using an adaptive Kalman filter algorithm to obtain a dynamic feature model of the digital twin component of the substation equipment; and the state deviation value between the digital twin component of the substation equipment and the physical device is calculated according to the dynamic feature model, including: Based on the time-varying eigenvector matrix, state prediction is performed through a state transfer matrix and a control input matrix to obtain a priori state estimation value, and a priori error covariance matrix is ​​calculated at the same time. The priori error covariance matrix is ​​combined with a process noise covariance matrix to evaluate prediction accuracy; Calculate the Kalman gain according to the prior state estimate and the prior error covariance matrix, in combination with the observation matrix and the observation noise covariance matrix, wherein the Kalman gain is used to balance the weight of the predicted value and the observed value; The priori state estimate is corrected by using the Kalman gain to obtain a posterior state estimate, and an error covariance matrix is ​​updated, wherein the posterior state estimate constitutes a dynamic characteristic model of the digital twin component of the substation equipment; The dynamic feature model is compared with the actual state vector of the physical device to calculate the state deviation value, which is used to evaluate the consistency between the digital twin model and the physical device and serve as the basis for model updating.

[0011] When the state deviation value exceeds the preset deviation threshold, the adaptive consistency maintenance mechanism is triggered; the adaptive consistency maintenance mechanism generates state compensation parameters through a deep reinforcement learning algorithm, and applies the state compensation parameters to the digital twin model to achieve dynamic synchronization between the digital twin components of the substation equipment and the physical equipment, including: Monitor the state deviation value between the digital twin model of the substation equipment and the physical equipment, where the state deviation value is used to quantify the difference between the digital twin model and the physical equipment. When the state deviation value exceeds a preset deviation threshold, an adaptive consistency maintenance mechanism is triggered; In the adaptive consistency maintenance mechanism, a deep reinforcement learning algorithm is used to generate state compensation parameters. The deep reinforcement learning algorithm determines the optimal state compensation parameters to minimize the state deviation value through continuous learning of the environmental state and strategy optimization. The state compensation parameters are used to adjust the dynamic characteristics of the digital twin model; The state compensation parameters are applied to the digital twin model of the substation equipment, and the dynamic characteristics of the digital twin model are adjusted to keep it consistent with the actual state of the physical equipment, thereby achieving dynamic synchronization between the digital twin components of the substation equipment and the physical equipment.

[0012] The state compensation parameters are generated by using a deep reinforcement learning algorithm. The deep reinforcement learning algorithm determines the optimal state compensation parameters to minimize the state deviation value through continuous learning of the environment state and strategy optimization, including: Construct a state space, an action space and a reward function, wherein the state space is composed of the physical device state, the digital twin state and the state deviation value, the action space is defined as a set of state compensation parameters, and the reward function is used to evaluate the effect of each action and maximize the cumulative reward; A deep reinforcement learning model is constructed using a dual neural network structure, which is used to estimate the state-action value function and optimize the compensation strategy through a policy gradient method, which updates the policy parameters by calculating the gradient of the policy objective function; An experience replay mechanism is used to store transfer samples. The experience replay mechanism stores four-tuple samples of state, action, reward and next state by constructing an experience pool, and updates the target network parameters through a soft update rule. The soft update rule is used to smoothly update the target network parameters. Generate state compensation parameters based on the optimized strategy, wherein the state compensation parameters are generated by selecting the optimal action that can maximize the state-action value function, and converting the optimal action into a state compensation amount through a compensation mapping function, wherein the state compensation amount is used to adjust the dynamic characteristics of the digital twin model; The learning rate is dynamically adjusted through an adaptive learning mechanism. The adaptive learning mechanism dynamically adjusts the learning rate through an exponential decay function to adapt to different learning stages, and defines a convergence condition to ensure a gradual decrease in the state deviation value. The convergence condition is used to judge the convergence and stability of the algorithm.

[0013] A second aspect of an embodiment of the present invention provides a system for adaptively maintaining time-varying consistency of digital twin components of power transformation equipment, including: The first unit is used to collect real-time operation data of the substation equipment, wherein the real-time operation data includes voltage data, current data, temperature data, vibration data and load data of the substation equipment; based on the real-time operation data, a physical model of the substation equipment is constructed, and a digital twin model of the substation equipment is generated according to the physical model; the real-time operation data is input into the digital twin model, and the digital twin model is trained by a deep learning algorithm to obtain the initial state parameters of the digital twin components of the substation equipment; The second unit is used to establish a time-varying eigenvector matrix of the digital twin component of the substation equipment based on the initial state parameters, wherein the time-varying eigenvector matrix includes equipment operation state characteristics, equipment performance attenuation characteristics, and equipment failure characteristics; use an adaptive Kalman filter algorithm to update the time-varying eigenvector matrix in real time to obtain a dynamic characteristic model of the digital twin component of the substation equipment; and calculate the state deviation value between the digital twin component of the substation equipment and the physical device according to the dynamic characteristic model; The third unit is used to trigger the adaptive consistency maintenance mechanism when the state deviation value exceeds a preset deviation threshold; the adaptive consistency maintenance mechanism generates state compensation parameters through a deep reinforcement learning algorithm, and applies the state compensation parameters to the digital twin model to achieve dynamic synchronization between the digital twin components of the substation equipment and the physical equipment.

[0014] A third aspect of the embodiments of the present invention An electronic device is provided, comprising: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0015] A fourth aspect of the embodiments of the present invention is: A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0016] The beneficial effects of this application are as follows: Improve the accuracy of the digital twin model of substation equipment. Through real-time operation data and deep learning algorithms, the operating status of substation equipment can be accurately captured and an accurate digital twin model can be established, thereby improving the accuracy and reliability of the model.

[0017] Realize dynamic synchronization between the digital twin components of substation equipment and physical equipment. Using the adaptive Kalman filter algorithm and deep reinforcement learning algorithm, the digital twin model can be updated and compensated in real time to ensure that the status of the digital twin components is consistent with that of the physical equipment, achieving dynamic synchronization.

[0018] Enhance the operation and maintenance capabilities of substation equipment. Through digital twin technology, the operating status of substation equipment can be monitored in real time, potential faults can be discovered in a timely manner, and predictive maintenance can be performed, thereby improving the reliability and safety of substation equipment and reducing maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of a method for adaptively maintaining time-varying consistency of a digital twin component of a substation equipment according to an embodiment of the present invention; Figure 2 It is a structural schematic diagram of a system for adaptively maintaining time-varying consistency of digital twin components of substation equipment according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0022] Figure 1 Schematic diagram of the process of the method for adaptively maintaining the time-varying consistency of the digital twin components of the substation equipment according to the embodiment of the present invention. Figure 1 As shown, the method includes: S101. Collect real-time operation data of the substation equipment, wherein the real-time operation data includes voltage data, current data, temperature data, vibration data and load data of the substation equipment; construct a physical model of the substation equipment based on the real-time operation data, and generate a digital twin model of the substation equipment according to the physical model; input the real-time operation data into the digital twin model, train the digital twin model through a deep learning algorithm, and obtain the initial state parameters of the digital twin components of the substation equipment; S102. Based on the initial state parameters, establish a time-varying feature vector matrix of the digital twin component of the substation equipment, wherein the time-varying feature vector matrix includes equipment operation state characteristics, equipment performance attenuation characteristics, and equipment failure characteristics; use an adaptive Kalman filter algorithm to update the time-varying feature vector matrix in real time to obtain a dynamic feature model of the digital twin component of the substation equipment; calculate the state deviation value between the digital twin component of the substation equipment and the physical device according to the dynamic feature model; S103. When the state deviation value exceeds the preset deviation threshold, the adaptive consistency maintenance mechanism is triggered; the adaptive consistency maintenance mechanism generates state compensation parameters through a deep reinforcement learning algorithm, and applies the state compensation parameters to the digital twin model to achieve dynamic synchronization between the digital twin components of the substation equipment and the physical equipment.

[0023] In an optional implementation, constructing a physical model of the substation equipment based on the real-time operation data, and generating a digital twin model of the substation equipment according to the physical model includes: The real-time operation data is constructed as a state vector, wherein the state vector includes the voltage data, the current data, the temperature data, the vibration data and the load data; based on the state vector, a physical model of the substation equipment is constructed in combination with a device physical parameter matrix and an environmental constraint matrix, wherein the device physical parameter matrix includes device geometric dimension parameters, material physical property parameters, electrical property parameters and thermodynamic property parameters, and the environmental constraint matrix includes environmental temperature parameters, humidity parameters and electromagnetic field strength parameters; The physical model is converted into a digital twin model by a digital mapping conversion function, wherein the digital mapping conversion function discretizes the physical quantity, digitally converts the geometric model, maps the dynamic characteristic parameters, and digitally expresses the constraint conditions of the physical model; A model error between the digital twin model and the physical model is calculated, an error compensation function is generated based on the model error, and the error compensation function is applied to the digital twin model to obtain an optimized digital twin model.

[0024] The method for constructing a digital twin model of substation equipment aims to build a high-precision digital twin model based on real-time operation data.

[0025] First, collect the real-time operation data of the substation equipment, including voltage data, current data, temperature data, vibration data and load data. For example, collect the real-time operation data of a transformer, the voltage is 110kV, the current is 1000A, the temperature is 40℃, the vibration frequency is 50Hz, and the load is 80%. Construct this data into a state vector to describe the real-time operation status of the substation equipment.

[0026] Next, the physical model of the substation equipment is constructed by combining the equipment physical parameter matrix and the environmental constraint matrix. The equipment physical parameter matrix includes equipment geometric size parameters, such as the height, width and length of the transformer; material physical property parameters, such as the magnetic permeability of the transformer core and the resistivity of the winding; electrical property parameters, such as the rated voltage and rated current of the transformer; and thermodynamic property parameters, such as the heat dissipation coefficient of the transformer. The environmental constraint matrix includes ambient temperature parameters, such as the ambient temperature of the substation; humidity parameters, such as the ambient humidity of the substation; and electromagnetic field strength parameters, such as the electromagnetic field strength around the substation. For example, assuming that the ambient temperature is 25°C, the humidity is 60%, and the electromagnetic field strength is 100V / m. Based on the state vector, the equipment physical parameter matrix and the environmental constraint matrix, the physical model of the substation equipment is constructed using the principles of physics and engineering experience. The model can describe the physical characteristics and operating laws of the substation equipment.

[0027] Then, the physical model is converted into a digital twin model through a digital mapping conversion function. The digital mapping conversion function includes the following steps: physical quantity discretization, such as discretizing continuous voltage values ​​into a finite number of voltage levels; geometric model digital conversion, such as converting the three-dimensional geometric model of the transformer into a digital model; dynamic characteristic parameter mapping, such as mapping the frequency response characteristics of the transformer into a digital model; and constraint condition digital expression, such as digital expression of constraint conditions such as ambient temperature, humidity, and electromagnetic field strength. Through these steps, various aspects of the physical model are converted into digital form to construct a digital twin model of the substation equipment.

[0028] Finally, the model error between the digital twin model and the physical model is calculated. For example, the difference between the voltage value predicted by the digital twin model and the actual measured voltage value is compared. An error compensation function is generated based on the model error, and the error compensation function is applied to the digital twin model to obtain an optimized digital twin model, thereby improving the accuracy of the digital twin model. For example, if it is found that the voltage value predicted by the digital twin model is consistently 2% lower than the actual measured voltage value, an error compensation function can be constructed to adjust the voltage value predicted by the digital twin model upward by 2%.

[0029] Beneficial effects: 1. Improve model accuracy: Optimize the digital twin model through the error compensation function, reduce the model error, improve the model accuracy, and more accurately reflect the actual operating status of the substation equipment.

[0030] 2. Enhance model reliability: Build a digital twin model based on real-time operation data and physical models to ensure the reliability and credibility of the model and provide a reliable reference for the operation and maintenance of substation equipment.

[0031] 3. Improve simulation efficiency: Converting the physical model into a digital twin model allows simulation analysis to be performed in a digital environment, avoiding the high cost and long cycle of physical experiments and improving simulation efficiency.

[0032] In an optional implementation, the real-time operation data is input into the digital twin model, and the digital twin model is trained by a deep learning algorithm to obtain the initial state parameters of the digital twin components of the substation equipment, including: Constructing a hybrid architecture of a multi-layer perceptron and a recurrent neural network, wherein the hybrid architecture of the multi-layer perceptron and the recurrent neural network is used to extract the time series features of the real-time operation data of the substation equipment, wherein the time series features are modeled by a long short-term memory network, and the long short-term memory network realizes effective modeling of the time series data through the synergy of the input gate, the forget gate and the memory unit; A multi-head self-attention mechanism is introduced to enhance the model's ability to focus on key features, and to achieve adaptive feature extraction of input data through weighted calculation of query matrix, key matrix and value matrix; The hybrid architecture is trained using an adaptive learning rate optimization algorithm, which improves the convergence speed and stability of the model by dynamically adjusting the learning rate. After the training is completed, the initial state parameters of the digital twin components of the substation equipment are extracted based on the test data set, and the initial state parameters are used to characterize the current operating state of the substation equipment; The reliability of the initial state parameters is verified by a cross-validation method, and the cross-validation method evaluates the generalization ability of the model through multiple training and validation processes to ensure the accuracy and reliability of the initial state parameters.

[0033] The method for extracting initial state parameters of digital twin components of substation equipment analyzes the real-time operation data of substation equipment through a deep learning algorithm, builds a digital twin model, and extracts the initial state parameters that characterize the operating status of the equipment.

[0034] First, collect the real-time operation data of the substation equipment. This data can include multiple sensor data such as voltage, current, temperature, humidity, etc. For example, collect the real-time operation data of the No. 1 main transformer of a substation, including the voltage, current and temperature values ​​every second.

[0035] Then, the collected real-time operation data is input into the constructed digital twin model. The digital twin model is a virtual model based on the physical structure and operation mechanism of the substation equipment. For example, the 3D model of the No. 1 main transformer can be constructed using CAD software, and sensor nodes can be added to the model to simulate the structure and operation status of the real transformer.

[0036] Next, the digital twin model is trained using a deep learning algorithm. Specifically, a hybrid architecture of a multi-layer perceptron and a recurrent neural network is used to extract the time series features of the real-time operation data of the substation equipment. For example, the collected voltage, current, and temperature data sequence of the No. 1 main transformer is input into the hybrid architecture. The multi-layer perceptron is used to extract the features of each time step, and the recurrent neural network, especially the long short-term memory network (LSTM), is used to capture the time dependency between data sequences. The LSTM network effectively models time series data and learns the long-term dependency of data through the synergy of input gates, forget gates, and memory units.

[0037] In order to enhance the model's ability to focus on key features, a multi-head self-attention mechanism is introduced. The multi-head self-attention mechanism achieves adaptive feature extraction of input data by calculating the weighted sum between the query matrix, key matrix, and value matrix. For example, when processing the temperature data of the No. 1 main transformer, the self-attention mechanism can pay more attention to the moments when the temperature changes dramatically, thereby better capturing the trend of temperature changes.

[0038] During the training process, an adaptive learning rate optimization algorithm is used to train the hybrid architecture. The adaptive learning rate optimization algorithm can dynamically adjust the learning rate according to the loss function value during the training process to improve the convergence speed and stability of the model. For example, the Adam optimizer is used to dynamically adjust the learning rate according to the historical information of the gradient to speed up the training of the model.

[0039] After training is completed, the initial state parameters of the digital twin components of the substation equipment are extracted based on the test data set. These initial state parameters are used to characterize the current operating state of the substation equipment. For example, the initial state parameters of the No. 1 main transformer are extracted, including winding temperature, oil temperature, oil level and other parameters.

[0040] Finally, the reliability of the extracted initial state parameters is verified by the cross-validation method. The cross-validation method divides the data set into multiple subsets, and uses different subsets for training and verification, evaluates the generalization ability of the model, and ensures the accuracy and reliability of the initial state parameters. For example, the historical data of the No. 1 main transformer is divided into 5 parts, 4 of which are used to train the model each time, and 1 is used to verify the model. This is repeated 5 times, and finally the average performance index of the model is obtained to evaluate the generalization ability of the model.

[0041] The beneficial effects of this method are reflected in the following three aspects: 1. Improve accuracy: Through deep learning algorithms and digital twin technology, the initial state parameters of substation equipment can be extracted more accurately, thereby more accurately reflecting the actual operating status of the equipment.

[0042] 2. Enhance reliability: Using cross-validation method to evaluate the model can effectively improve the reliability and stability of initial state parameters and reduce errors.

[0043] 3. Improve efficiency: Through the adaptive learning rate optimization algorithm, the model training speed can be accelerated, the efficiency of parameter extraction can be improved, and the analysis time can be shortened.

[0044] In an optional implementation, the time-varying eigenvector matrix is ​​updated in real time using an adaptive Kalman filter algorithm to obtain a dynamic feature model of the digital twin component of the substation equipment; and the state deviation value between the digital twin component of the substation equipment and the physical device is calculated according to the dynamic feature model, including: Based on the time-varying eigenvector matrix, state prediction is performed through a state transfer matrix and a control input matrix to obtain a priori state estimation value, and a priori error covariance matrix is ​​calculated at the same time. The priori error covariance matrix is ​​combined with a process noise covariance matrix to evaluate prediction accuracy; Calculate the Kalman gain according to the prior state estimate and the prior error covariance matrix, in combination with the observation matrix and the observation noise covariance matrix, wherein the Kalman gain is used to balance the weight of the predicted value and the observed value; The priori state estimate is corrected by using the Kalman gain to obtain a posterior state estimate, and an error covariance matrix is ​​updated, wherein the posterior state estimate constitutes a dynamic characteristic model of the digital twin component of the substation equipment; The dynamic feature model is compared with the actual state vector of the physical device to calculate the state deviation value, which is used to evaluate the consistency between the digital twin model and the physical device and serve as the basis for model updating.

[0045] Obtain real-time operating data of substation equipment, including but not limited to current, voltage, temperature, humidity and other parameters. These data can be collected through sensor networks, SCADA systems and other channels. For example, collect real-time data of the No. 1 main transformer of a substation, including oil temperature of 60°C, winding temperature of 70°C, and ambient humidity of 50%. These data will be used as input to the Kalman filter algorithm to build and update the digital twin model.

[0046] The collected real-time data is preprocessed, including data cleaning, denoising, normalization and other operations to improve data quality and algorithm stability. For example, the sliding average filter method is used to remove noise in the oil temperature data, and all data are normalized by min-max to scale the data range to between 0 and 1. The preprocessed data will be used to construct the feature vector.

[0047] Construct a time-varying eigenvector matrix for the substation equipment. This matrix contains key features that characterize the equipment status and will be dynamically updated over time. For example, the pre-processed oil temperature, winding temperature, and ambient humidity form a three-dimensional eigenvector [0.8, 0.9, 0.5] and use it as the initial eigenvector matrix. This matrix will be updated in real time over time according to the Kalman filter algorithm.

[0048] Based on the time-varying eigenvector matrix at the current moment, the state is predicted through the state transfer matrix and the control input matrix to obtain the prior state estimate. The state transfer matrix describes the evolution of the system state over time, and the control input matrix describes the impact of the control input on the system state. For example, assuming that the state transfer matrix is ​​a unit matrix and the control input matrix is ​​a zero matrix, the prior state estimate is the same as the eigenvector matrix at the current moment, that is, [0.8, 0.9, 0.5]. At the same time, the prior error covariance matrix is ​​calculated to evaluate the prediction accuracy. Assume that the initial prior error covariance matrix is ​​a unit matrix with a diagonal element of 0.1.

[0049] Combine the process noise covariance matrix and the prior error covariance matrix to evaluate the prediction accuracy. The process noise covariance matrix describes the degree to which the system state is affected by random noise. Assume that the process noise covariance matrix is ​​a unit matrix with diagonal elements of 0.01. Add the prior error covariance matrix to the process noise covariance matrix to obtain the prediction error covariance matrix.

[0050] The Kalman gain is calculated based on the prior state estimate, the prior error covariance matrix, the observation matrix, and the observation noise covariance matrix. The observation matrix describes the relationship between the observation value and the system state, and the observation noise covariance matrix describes the degree to which the observation value is affected by random noise. Assume that the observation matrix is ​​a unit matrix and the observation noise covariance matrix is ​​a unit matrix with a diagonal element of 0.05. The Kalman gain is used to balance the weight of the predicted value and the observed value, and its value will be dynamically adjusted according to the prediction accuracy and observation accuracy.

[0051] The Kalman gain is used to correct the prior state estimate to obtain the posterior state estimate and update the error covariance matrix. The posterior state estimate constitutes the dynamic characteristic model of the digital twin component of the substation equipment. For example, assuming that the calculated Kalman gain is a unit matrix with a diagonal element of 0.5, the posterior state estimate is [0.825, 0.925, 0.525]. At the same time, the error covariance matrix is ​​updated for state prediction at the next moment.

[0052] Compare the dynamic feature model with the actual state vector of the physical device and calculate the state deviation value. The state deviation value is used to evaluate the consistency between the digital twin model and the physical device and serves as the basis for model update. For example, assuming that the actual state vector of the physical device is [0.85, 0.9, 0.5], the state deviation value is [-0.025, 0.025, 0.025].

[0053] The parameters of the Kalman filter algorithm, such as the state transfer matrix, control input matrix, process noise covariance matrix, and observation noise covariance matrix, are adjusted according to the state deviation value to improve the accuracy and stability of the digital twin model.

[0054] Beneficial effects: 1. Improved the accuracy and real-time performance of the digital twin model: Through the adaptive Kalman filter algorithm, the digital twin model can be updated in real time, enabling it to more accurately reflect the operating status of the physical equipment.

[0055] 2. Enhanced monitoring and prediction capabilities of substation equipment status: Based on the dynamic feature model, the future status of substation equipment can be predicted, thereby discovering potential failure risks in advance.

[0056] 3. Improved the operation and maintenance efficiency of substation equipment: Through the analysis of state deviation values, equipment abnormalities can be discovered in a timely manner, and maintenance personnel can be guided to perform targeted maintenance, thereby reducing downtime and maintenance costs.

[0057] In an optional implementation, when the state deviation value exceeds a preset deviation threshold, an adaptive consistency maintenance mechanism is triggered; the adaptive consistency maintenance mechanism generates a state compensation parameter through a deep reinforcement learning algorithm, and applies the state compensation parameter to the digital twin model to achieve dynamic synchronization between the digital twin component of the substation equipment and the physical device, including: Monitor the state deviation value between the digital twin model of the substation equipment and the physical equipment, where the state deviation value is used to quantify the difference between the digital twin model and the physical equipment. When the state deviation value exceeds a preset deviation threshold, an adaptive consistency maintenance mechanism is triggered; In the adaptive consistency maintenance mechanism, a deep reinforcement learning algorithm is used to generate state compensation parameters. The deep reinforcement learning algorithm determines the optimal state compensation parameters to minimize the state deviation value through continuous learning of the environmental state and strategy optimization. The state compensation parameters are used to adjust the dynamic characteristics of the digital twin model; The state compensation parameters are applied to the digital twin model of the substation equipment, and the dynamic characteristics of the digital twin model are adjusted to keep it consistent with the actual state of the physical equipment, thereby achieving dynamic synchronization between the digital twin components of the substation equipment and the physical equipment.

[0058] An adaptive consistency maintenance method for the digital twin model of substation equipment can monitor and automatically adjust the digital twin model in real time to keep it dynamically synchronized with the physical equipment, thereby improving the accuracy and reliability of the model.

[0059] First, obtain the real-time status data of the substation equipment entity. This data can be collected through sensors, intelligent electronic devices, etc., such as current, voltage, temperature, humidity, etc. Assume that the real-time current of a transformer entity is 150A, the voltage is 10kV, and the temperature is 40°C. At the same time, obtain the simulation status data of the corresponding digital twin model. Assume that the simulation current of the transformer digital twin model is 145A, the voltage is 9.9kV, and the temperature is 38°C.

[0060] Next, calculate the state deviation value between the digital twin model and the physical device. This can be achieved by comparing the real-time state data of the physical device with the simulated state data of the digital twin model. For example, the current deviation is 150A -145A = 5A, the voltage deviation is 10kV - 9.9kV = 0.1kV, and the temperature deviation is 40℃ - 38℃ = 2℃.

[0061] Then, it is determined whether the state deviation value exceeds the preset deviation threshold. The preset deviation threshold can be set according to the specific application scenario and device characteristics. Assume that the current deviation threshold is 10A, the voltage deviation threshold is 0.5kV, and the temperature deviation threshold is 5°C. In this case, the current, voltage, and temperature deviation values ​​do not exceed the preset thresholds.

[0062] If the state deviation value exceeds the preset deviation threshold, the adaptive consistency maintenance mechanism is triggered. In this case, assume that at a certain moment, the transformer real current suddenly increases to 165A, causing the current deviation to reach 165A - 145A = 20A, exceeding the preset threshold of 10A, and the adaptive consistency maintenance mechanism is triggered.

[0063] The core of the adaptive consistency maintenance mechanism is to use the deep reinforcement learning algorithm to generate state compensation parameters. The deep reinforcement learning algorithm continuously optimizes the strategy through interactive learning with the environment, and finally finds the optimal state compensation parameters that can minimize the state deviation value. The interaction process between the digital twin model and the physical device can be regarded as a deep reinforcement learning environment, and the state deviation value is used as a reward signal. For example, when the state deviation value decreases, a positive reward is given; when the state deviation value increases, a negative reward is given. In this way, the deep reinforcement learning algorithm can learn how to adjust the state compensation parameters to make the state of the digital twin model closer to the state of the physical device. Assume that after training with the deep reinforcement learning algorithm, the state compensation parameters obtained are: the current compensation parameter is +15A, the voltage compensation parameter is +0.08kV, and the temperature compensation parameter is +1℃.

[0064] Finally, the generated state compensation parameters are applied to the digital twin model of the substation equipment. The state compensation parameters are added to the original state data of the digital twin model to obtain the corrected state data. For example, if the current compensation parameter +15A is applied to the digital twin model, the corrected current value is 145A + 15A = 160A. In this way, the dynamic characteristics of the digital twin model can be adjusted to keep it consistent with the actual state of the physical device, and the dynamic synchronization of the digital twin components of the substation equipment and the physical device can be achieved.

[0065] Beneficial effects: 1. Improve model accuracy: Through the adaptive consistency maintenance mechanism, the digital twin model can track the state changes of physical equipment in real time, effectively reduce the deviation between the model and the entity, and significantly improve the accuracy and reliability of the model.

[0066] 2. Enhanced predictive capabilities: More precise digital twin models can more accurately predict the future state of substation equipment, providing a more reliable basis for equipment maintenance, fault diagnosis, and operation optimization.

[0067] 3. Improve operation and maintenance efficiency: Through real-time monitoring and prediction of digital twin models, potential problems can be discovered in advance, equipment operation strategies can be optimized, operation and maintenance efficiency can be improved, and operating costs can be reduced.

[0068] In an optional embodiment, a state compensation parameter is generated using a deep reinforcement learning algorithm, wherein the deep reinforcement learning algorithm determines the optimal state compensation parameter to minimize the state deviation value through continuous learning of the environment state and strategy optimization, including: Construct a state space, an action space and a reward function, wherein the state space is composed of the physical device state, the digital twin state and the state deviation value, the action space is defined as a set of state compensation parameters, and the reward function is used to evaluate the effect of each action and maximize the cumulative reward; A deep reinforcement learning model is constructed using a dual neural network structure, which is used to estimate the state-action value function and optimize the compensation strategy through a policy gradient method, which updates the policy parameters by calculating the gradient of the policy objective function; An experience replay mechanism is used to store transfer samples. The experience replay mechanism stores four-tuple samples of state, action, reward and next state by constructing an experience pool, and updates the target network parameters through a soft update rule. The soft update rule is used to smoothly update the target network parameters. Generate state compensation parameters based on the optimized strategy, wherein the state compensation parameters are generated by selecting the optimal action that can maximize the state-action value function, and converting the optimal action into a state compensation amount through a compensation mapping function, wherein the state compensation amount is used to adjust the dynamic characteristics of the digital twin model; The learning rate is dynamically adjusted through an adaptive learning mechanism. The adaptive learning mechanism dynamically adjusts the learning rate through an exponential decay function to adapt to different learning stages, and defines a convergence condition to ensure a gradual decrease in the state deviation value. The convergence condition is used to judge the convergence and stability of the algorithm.

[0069] The deep reinforcement learning algorithm is used to generate state compensation parameters to minimize the state deviation between the physical device and the digital twin model, thereby improving the accuracy and reliability of the digital twin. The specific implementation of this method is described in detail below: First, construct the various elements required for deep reinforcement learning. The state space consists of three parts: the real-time state of the physical device, the simulated state of the digital twin model, and the deviation value between the two. For example, if the temperature of the physical device is 25°C and the simulated temperature of the digital twin is 24°C, the state deviation value is 1°C. The action space is defined as a set of state compensation parameters that will be used to adjust the dynamic characteristics of the digital twin model. For example, the heat transfer coefficient can be used as a state compensation parameter. The design goal of the reward function is to maximize the cumulative reward and guide the agent to learn the optimal compensation strategy. A simple reward function can be defined as the negative of the state deviation value, that is, the smaller the deviation, the greater the reward. For example, when the state deviation value is 1°C, the reward is -1; when the state deviation value is 0.5°C, the reward is -0.5.

[0070] Next, a deep reinforcement learning model based on a dual neural network structure is constructed. The model contains two neural networks with the same structure but different parameters: one is used to estimate the state-action value function under the current policy, and the other is used as a target network to estimate the state-action value function under the target policy. The outputs of both networks are estimates of the expected cumulative rewards for taking different actions in a given state. The policy gradient method is used to optimize the compensation strategy. The core idea is to update the policy parameters by calculating the gradient of the policy objective function relative to the policy parameters, so that the strategy can be improved in the direction of obtaining greater cumulative rewards.

[0071] In order to improve the stability and efficiency of learning, the experience replay mechanism is introduced. The experience replay mechanism builds an experience pool to store the historical data of the interaction between the agent and the environment. These data are stored in the form of four-tuples, including the current state, the action taken, the reward obtained, and the next state. During the training process, a batch of samples are randomly selected from the experience pool to update the parameters of the neural network, avoiding the strong correlation between the data and improving the learning efficiency. The parameters of the target network are updated through the soft update rule, that is, the parameters of the current network are copied to the target network at a certain ratio, so that the parameter update of the target network is smoother, which is conducive to the stability of the algorithm.

[0072] Generate state compensation parameters based on the optimized strategy. Specifically, for the current physical device state and digital twin state, select the action that maximizes the state-action value function as the optimal action. Then, convert the optimal action into a state compensation through a compensation mapping function. For example, if the optimal action is to increase the heat transfer coefficient by 0.1, the state compensation is 0.1. Finally, apply the state compensation to the digital twin model to adjust its dynamic characteristics, such as increasing the heat transfer coefficient of the digital twin model by 0.1.

[0073] In order to adapt to different learning stages, an adaptive learning mechanism is used to dynamically adjust the learning rate. For example, an exponential decay function can be used to gradually reduce the learning rate. At the same time, a convergence condition is defined to ensure the gradual reduction of the state deviation value. For example, a threshold can be set, and when the state deviation value is lower than the threshold for multiple consecutive time steps, the algorithm is considered to have converged.

[0074] For example, suppose the temperature of the physical device varies between 20°C and 30°C, and the simulated temperature of the digital twin is also within this range. The initial state deviation value is 5°C. After training with the deep reinforcement learning algorithm, the final state deviation value is reduced to within 0.5°C.

[0075] The beneficial effects of this method are reflected in the following three aspects: First, the accuracy of digital twins is improved. By continuously learning and optimizing compensation strategies, the state deviation between the physical device and the digital twin model can be effectively reduced, allowing the digital twin to more accurately reflect the operating status of the physical device.

[0076] Second, the reliability of digital twins is enhanced. Through the setting of adaptive learning mechanism and convergence conditions, the stability and convergence of the algorithm are guaranteed, so that digital twins can run stably and long-term.

[0077] Third, the adaptability of digital twins is improved. Through the flexible definition of state space, action space and reward function, it can adapt to different types of physical devices and application scenarios, and has strong versatility and scalability.

[0078] Figure 2 Schematic diagram of the structure of the time-varying consistency adaptive maintenance system of the digital twin component of the substation equipment according to the embodiment of the present invention. Figure 2 As shown, the system comprises: The first unit is used to collect real-time operation data of the substation equipment, wherein the real-time operation data includes voltage data, current data, temperature data, vibration data and load data of the substation equipment; based on the real-time operation data, a physical model of the substation equipment is constructed, and a digital twin model of the substation equipment is generated according to the physical model; the real-time operation data is input into the digital twin model, and the digital twin model is trained by a deep learning algorithm to obtain the initial state parameters of the digital twin components of the substation equipment; The second unit is used to establish a time-varying eigenvector matrix of the digital twin component of the substation equipment based on the initial state parameters, wherein the time-varying eigenvector matrix includes equipment operation state characteristics, equipment performance attenuation characteristics, and equipment failure characteristics; use an adaptive Kalman filter algorithm to update the time-varying eigenvector matrix in real time to obtain a dynamic characteristic model of the digital twin component of the substation equipment; and calculate the state deviation value between the digital twin component of the substation equipment and the physical device according to the dynamic characteristic model; The third unit is used to trigger the adaptive consistency maintenance mechanism when the state deviation value exceeds a preset deviation threshold; the adaptive consistency maintenance mechanism generates state compensation parameters through a deep reinforcement learning algorithm, and applies the state compensation parameters to the digital twin model to achieve dynamic synchronization between the digital twin components of the substation equipment and the physical equipment.

[0079] According to a third aspect of the embodiments of the present invention, An electronic device is provided, comprising: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0080] A fourth aspect of the embodiments of the present invention is: A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0081] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for adaptively maintaining time-varying consistency of digital twin components of substation equipment, characterized in that: include: Collecting real-time operation data of the substation equipment, wherein the real-time operation data includes voltage data, current data, temperature data, vibration data and load data of the substation equipment; Based on the real-time operation data, a physical model of the substation equipment is constructed, and a digital twin model of the substation equipment is generated according to the physical model; the real-time operation data is input into the digital twin model, and the digital twin model is trained by a deep learning algorithm to obtain the initial state parameters of the digital twin components of the substation equipment; Based on the initial state parameters, a time-varying eigenvector matrix of the digital twin component of the substation equipment is established, wherein the time-varying eigenvector matrix includes equipment operation state characteristics, equipment performance attenuation characteristics, and equipment failure characteristics; the time-varying eigenvector matrix is ​​updated in real time using an adaptive Kalman filter algorithm to obtain a dynamic feature model of the digital twin component of the substation equipment; and the state deviation value between the digital twin component of the substation equipment and the physical device is calculated according to the dynamic feature model; When the state deviation value exceeds the preset deviation threshold, the adaptive consistency maintenance mechanism is triggered; the adaptive consistency maintenance mechanism generates state compensation parameters through a deep reinforcement learning algorithm, and applies the state compensation parameters to the digital twin model to achieve dynamic synchronization between the digital twin components of the substation equipment and the physical equipment.

2. The method according to claim 1, characterized in that: Based on the real-time operation data, a physical model of the substation equipment is constructed, and a digital twin model of the substation equipment is generated according to the physical model, including: The real-time operation data is constructed as a state vector, wherein the state vector includes the voltage data, the current data, the temperature data, the vibration data and the load data; based on the state vector, a physical model of the substation equipment is constructed in combination with a device physical parameter matrix and an environmental constraint matrix, wherein the device physical parameter matrix includes device geometric dimension parameters, material physical property parameters, electrical property parameters and thermodynamic property parameters, and the environmental constraint matrix includes environmental temperature parameters, humidity parameters and electromagnetic field strength parameters; The physical model is converted into a digital twin model by a digital mapping conversion function, wherein the digital mapping conversion function discretizes the physical quantity, digitally converts the geometric model, maps the dynamic characteristic parameters, and digitally expresses the constraint conditions of the physical model; A model error between the digital twin model and the physical model is calculated, an error compensation function is generated based on the model error, and the error compensation function is applied to the digital twin model to obtain an optimized digital twin model.

3. The method according to claim 1, characterized in that Inputting the real-time operation data into the digital twin model, training the digital twin model through a deep learning algorithm, and obtaining the initial state parameters of the digital twin components of the substation equipment include: Constructing a hybrid architecture of a multi-layer perceptron and a recurrent neural network, wherein the hybrid architecture of the multi-layer perceptron and the recurrent neural network is used to extract the time series features of the real-time operation data of the substation equipment, wherein the time series features are modeled by a long short-term memory network, and the long short-term memory network realizes effective modeling of the time series data through the synergy of the input gate, the forget gate and the memory unit; A multi-head self-attention mechanism is introduced to enhance the model's ability to focus on key features, and to achieve adaptive feature extraction of input data through weighted calculation of query matrix, key matrix and value matrix; The hybrid architecture is trained using an adaptive learning rate optimization algorithm, which improves the convergence speed and stability of the model by dynamically adjusting the learning rate. After the training is completed, the initial state parameters of the digital twin components of the substation equipment are extracted based on the test data set, and the initial state parameters are used to characterize the current operating state of the substation equipment; The reliability of the initial state parameters is verified by a cross-validation method, and the cross-validation method evaluates the generalization ability of the model through multiple training and validation processes to ensure the accuracy and reliability of the initial state parameters.

4. The method according to claim 1, characterized in that: The time-varying eigenvector matrix is ​​updated in real time by using an adaptive Kalman filter algorithm to obtain a dynamic feature model of the digital twin component of the substation equipment; and the state deviation value between the digital twin component of the substation equipment and the physical device is calculated according to the dynamic feature model, including: Based on the time-varying eigenvector matrix, state prediction is performed through a state transfer matrix and a control input matrix to obtain a priori state estimation value, and a priori error covariance matrix is ​​calculated at the same time. The priori error covariance matrix is ​​combined with a process noise covariance matrix to evaluate prediction accuracy; Calculate the Kalman gain according to the prior state estimate and the prior error covariance matrix, in combination with the observation matrix and the observation noise covariance matrix, wherein the Kalman gain is used to balance the weight of the predicted value and the observed value; The priori state estimate is corrected by using the Kalman gain to obtain a posterior state estimate, and an error covariance matrix is ​​updated, wherein the posterior state estimate constitutes a dynamic characteristic model of the digital twin component of the substation equipment; The dynamic feature model is compared with the actual state vector of the physical device to calculate the state deviation value, which is used to evaluate the consistency between the digital twin model and the physical device and serve as the basis for model updating.

5. The method according to claim 1, characterized in that: When the state deviation value exceeds the preset deviation threshold, the adaptive consistency maintenance mechanism is triggered; the adaptive consistency maintenance mechanism generates state compensation parameters through a deep reinforcement learning algorithm, and applies the state compensation parameters to the digital twin model to achieve dynamic synchronization between the digital twin components of the substation equipment and the physical equipment, including: Monitor the state deviation value between the digital twin model of the substation equipment and the physical equipment, where the state deviation value is used to quantify the difference between the digital twin model and the physical equipment. When the state deviation value exceeds a preset deviation threshold, an adaptive consistency maintenance mechanism is triggered; In the adaptive consistency maintenance mechanism, a deep reinforcement learning algorithm is used to generate state compensation parameters. The deep reinforcement learning algorithm determines the optimal state compensation parameters to minimize the state deviation value through continuous learning of the environmental state and strategy optimization. The state compensation parameters are used to adjust the dynamic characteristics of the digital twin model; The state compensation parameters are applied to the digital twin model of the substation equipment, and the dynamic characteristics of the digital twin model are adjusted to keep it consistent with the actual state of the physical equipment, thereby achieving dynamic synchronization between the digital twin components of the substation equipment and the physical equipment.

6. The method according to claim 1, characterized in that The state compensation parameters are generated by using a deep reinforcement learning algorithm. The deep reinforcement learning algorithm determines the optimal state compensation parameters to minimize the state deviation value through continuous learning of the environment state and strategy optimization, including: Construct a state space, an action space and a reward function, wherein the state space is composed of the physical device state, the digital twin state and the state deviation value, the action space is defined as a set of state compensation parameters, and the reward function is used to evaluate the effect of each action and maximize the cumulative reward; A deep reinforcement learning model is constructed using a dual neural network structure, which is used to estimate the state-action value function and optimize the compensation strategy through a policy gradient method, which updates the policy parameters by calculating the gradient of the policy objective function; An experience replay mechanism is used to store transfer samples. The experience replay mechanism stores four-tuple samples of state, action, reward and next state by constructing an experience pool, and updates the target network parameters through a soft update rule. The soft update rule is used to smoothly update the target network parameters. Generate state compensation parameters based on the optimized strategy, wherein the state compensation parameters are generated by selecting the optimal action that can maximize the state-action value function, and converting the optimal action into a state compensation amount through a compensation mapping function, wherein the state compensation amount is used to adjust the dynamic characteristics of the digital twin model; The learning rate is dynamically adjusted through an adaptive learning mechanism. The adaptive learning mechanism dynamically adjusts the learning rate through an exponential decay function to adapt to different learning stages, and defines a convergence condition to ensure a gradual decrease in the state deviation value. The convergence condition is used to judge the convergence and stability of the algorithm.

7. A time-varying consistency adaptive maintenance system for digital twin components of substation equipment, used to implement the method described in any one of claims 1 to 6, characterized in that: include: The first unit is used to collect real-time operation data of the substation equipment, wherein the real-time operation data includes voltage data, current data, temperature data, vibration data and load data of the substation equipment; based on the real-time operation data, a physical model of the substation equipment is constructed, and a digital twin model of the substation equipment is generated according to the physical model; the real-time operation data is input into the digital twin model, and the digital twin model is trained by a deep learning algorithm to obtain the initial state parameters of the digital twin components of the substation equipment; The second unit is used to establish a time-varying eigenvector matrix of the digital twin component of the substation equipment based on the initial state parameters, wherein the time-varying eigenvector matrix includes equipment operation state characteristics, equipment performance attenuation characteristics, and equipment failure characteristics; use an adaptive Kalman filter algorithm to update the time-varying eigenvector matrix in real time to obtain a dynamic characteristic model of the digital twin component of the substation equipment; and calculate the state deviation value between the digital twin component of the substation equipment and the physical device according to the dynamic characteristic model; The third unit is used to trigger the adaptive consistency maintenance mechanism when the state deviation value exceeds a preset deviation threshold; the adaptive consistency maintenance mechanism generates state compensation parameters through a deep reinforcement learning algorithm, and applies the state compensation parameters to the digital twin model to achieve dynamic synchronization between the digital twin components of the substation equipment and the physical equipment.

8. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images