A method for constructing digital twin of circuit breaker

Abstract

The present disclosure discloses a method for constructing a digital twin of a circuit breaker, comprising the following steps: S100: constructing a multi-domain physical simulation model of a circuit breaker; S200: constructing a data-driven model of a circuit breaker; S300: dynamically interacting and optimizing a circuit breaker entity with a multi-domain coupled physical simulation model and a data-driven model in real time to obtain a digital twin of the circuit breaker.

Description

Technical Field

[0001] The present disclosure belongs to the intersection of power equipment fault diagnosis, Internet of Things technology, and artificial intelligence algorithms, and specifically relates to a method for constructing a digital twin of a circuit breaker. Background Art

[0002] Big data, sensors and artificial intelligence technologies are rapidly developing intelligent status monitoring technology, a means of detecting the operating status of power equipment and assisting maintenance. As a switching device that can carry and break normal circuit currents and break fault currents within a specified time in the power system, medium and high voltage circuit breakers provide an important guarantee for the stability of the power transmission and distribution links of the power system. Therefore, they have received extensive attention in the fields of power equipment fault diagnosis, intelligent online monitoring and predictive maintenance. However, due to the serious lack of online perception data accumulation of circuit breakers in the current power industry and the difficulty of data processing, and the fact that most of the circuit breaker fault diagnosis models proposed by the academic community are based on offline simulation experimental data, such a trade-off has led to poor results in the actual application of the obtained models. However, if the load test of "circuit breaker breaking megawatt load" is to be carried out, the conditions of high voltage and high current must be met at the same time, which requires expensive supporting equipment; and improving the offline fault diagnosis model of circuit breakers requires a large amount of data as support, and hundreds or even thousands of circuit breaker load breaking tests will seriously affect the stability of the regional power grid, and are not economical and operable.

[0003] Digital twin refers to the construction of a virtual entity in the digital world that is identical to an object in the physical world through digital means, so as to achieve design optimization, analysis and prediction, and response control of the physical entity. With the continuous development of Internet of Things technology, computer-aided design and simulation technology, artificial intelligence and virtual reality technology, more industrial products and equipment have been endowed with intelligent features, and the concept of digital twin has gradually been applied to the complete product life cycle stages such as equipment design, equipment manufacturing, equipment use and maintenance, and equipment recycling. Summary of the invention

[0004] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a method for constructing a digital twin of a circuit breaker. The twin constructed based on this method can simulate the actual operating conditions of the circuit breaker for online testing, and can obtain a large amount of close-to-real data to improve the algorithm quality of the data-driven model.

[0005] To achieve the above objectives, the present disclosure provides the following technical solutions:

[0006] A method for constructing a digital twin of a circuit breaker comprises the following steps:

[0007] S100: Construct a multi-domain physical simulation model of circuit breaker;

[0008] S200: Build a circuit breaker data-driven model;

[0009] S300: Dynamically interact and optimize the circuit breaker entity with the multi-domain coupled physical simulation model and the data-driven model in real time to obtain the digital twin of the circuit breaker.

[0010] Preferably, in step S100, the multi-domain physical simulation model of the circuit breaker includes: a three-dimensional structural simulation model of the circuit breaker, a simulation model of the circuit breaker electromechanical system, a simulation model of the circuit breaker control loop and a simulation model of the circuit breaker insulation structure.

[0011] Preferably, in step S200, the circuit breaker data-driven model includes an electromechanical system fault diagnosis model and a circuit breaker remaining service life prediction model.

[0012] Preferably, the construction of the electromechanical system fault diagnosis model includes the following steps:

[0013] S201: Building a test platform based on the circuit breaker body to simulate different operating states of the circuit breaker;

[0014] S202: Based on the test platform, the circuit breaker is opened and closed under different operating conditions, and the corresponding operating coil current signal and moving iron core linear travel displacement signal are collected at the same time;

[0015] S203: performing data preprocessing on the operating coil current signal and the moving iron core linear travel displacement signal under different operating states, respectively, extracting time domain features from the preprocessed signals, respectively, and combining the time domain features to obtain feature vectors under different operating states;

[0016] S204: taking the feature vector and its corresponding running state as a feature label set;

[0017] S205: Divide the feature label set into a training set and a test set, use the training set to train the electromechanical system fault diagnosis model, and use the test set to test the model. If the test accuracy exceeds 90%, the electromechanical system fault diagnosis model is constructed; otherwise, adjust the parameters to retrain the model.

[0018] Preferably, the construction of the circuit breaker remaining service life prediction model includes the following steps:

[0019] S2001: Apply artificial dirt with a certain value of salt density, insoluble dirt ash density and humidity to the main circuit bushing of the circuit breaker;

[0020] S2002: applying a constant rated voltage across the main circuit of the circuit breaker, and collecting leakage current signals and high-frequency current sensor signals;

[0021] S2003: changing the test variables, and repeatedly executing steps S2001 and S2002 to obtain a data set containing leakage current signals and high-frequency current sensor signals corresponding to different test variables, until the main circuit bushing has a flashover;

[0022] S2004: extracting characteristic quantities of leakage current signals and high-frequency current sensor signals in the data set;

[0023] S2005: Screen the characteristic quantities to obtain variables of the circuit breaker remaining service life prediction model, and fit the variables to complete the model construction.

[0024] The present disclosure also provides a circuit breaker digital twin construction device, comprising:

[0025] The first building module is used to build a multi-domain physical simulation model of a circuit breaker;

[0026] The second building module is used to build a circuit breaker data-driven model;

[0027] The interactive optimization module is used to dynamically interact and optimize the circuit breaker entity with the multi-domain coupled physical simulation model and the data-driven model in real time to obtain the digital twin of the circuit breaker.

[0028] Preferably, the second building block comprises:

[0029] The first construction submodule is used to construct a fault diagnosis model for an electromechanical system;

[0030] The second construction submodule is used to construct a circuit breaker remaining service life prediction model.

[0031] The present disclosure also provides a storage medium, in which a plurality of instructions are stored, and the instructions are suitable for being loaded and executed by a processor:

[0032] Construct a multi-domain physical simulation model of circuit breaker;

[0033] Build a data-driven model for circuit breakers;

[0034] The circuit breaker entity is dynamically interacted and optimized in real time with the multi-domain coupled physical simulation model and the data-driven model to obtain the digital twin of the circuit breaker.

[0035] Compared with the prior art, the beneficial effects brought about by the present disclosure are: based on this digital twin, many researches that are difficult to carry out in reality can be carried out, such as using probabilistic simulation models to simulate various complex fault conditions that cannot be carried out on site, obtaining signal characteristics, and storing them in historical databases for data retrieval when the equipment is actually put into operation; based on limited, easy-to-arrange, direct data such as sensor signal data and historical databases, through the coupling model of physical simulation and machine learning, indirect monitoring of information that is difficult to perceive can be achieved to achieve the effect of "virtual sensor"; online testing of the actual working conditions of the circuit breaker is simulated, which greatly reduces the test cost, obtains a large amount of close-to-real data to improve the algorithm quality of the data-driven model, and is conducive to the practical application of the artificial intelligence model. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a flow chart of a method for constructing a digital twin of a circuit breaker provided by an embodiment of the present disclosure;

[0037] Figure 2 is a structural schematic diagram of a circuit breaker physical simulation model provided by another embodiment of the present disclosure;

[0038] Figure 3 is a schematic diagram of constructing a circuit breaker data-driven model provided by another embodiment of the present disclosure;

[0039] Figure 4 It is a schematic diagram of the interaction between a circuit breaker entity and its digital twin provided by another embodiment of the present disclosure. DETAILED DESCRIPTION

[0040] The following will refer to the attached Figures 1 to 4 Specific embodiments of the present disclosure are described in detail. Although specific embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0041] It should be noted that certain words are used in the specification and claims to refer to specific components. Those skilled in the art should understand that technicians may use different nouns to refer to the same component. This specification and claims do not use the difference in nouns as a way to distinguish components, but use the difference in the functions of the components as the criterion for distinction. As mentioned throughout the specification and claims, "including" or "comprising" is an open term, so it should be interpreted as "including but not limited to". The subsequent description of the specification is a preferred embodiment of the present disclosure, but the description is based on the general principles of the specification and is not used to limit the scope of the present disclosure. The scope of protection of the present disclosure shall be determined by the attached claims.

[0042] To facilitate understanding of the embodiments of the present disclosure, further explanation will be given below using specific embodiments as examples in conjunction with the accompanying drawings, and each of the accompanying drawings does not constitute a limitation on the embodiments of the present disclosure.

[0043] In one embodiment, if Figure 1 As shown, the present disclosure proposes a method for constructing a digital twin of a circuit breaker, comprising the following steps:

[0044] S100: Construct a multi-domain physical simulation model of circuit breaker;

[0045] In this step, if Figure 2 As shown, firstly, according to the actual structural parameters of the circuit breaker, a three-dimensional structural model of the circuit breaker is constructed based on Solidworks software and saved as a model file, and the model file is imported into MATLAB, and then the three-dimensional structural model of the circuit breaker is constructed with the help of Simulink Simscape Multibody module library; according to the structural parameters of the permanent magnetic actuator of the circuit breaker, the electromechanical system of the circuit breaker is simulated by using ANSYS Maxwell software, and the simulation model of the electromechanical system of the circuit breaker can be obtained by combining with MATLAB Simulink; according to the control loop design of the circuit breaker, the simulation model of the circuit breaker control loop is constructed based on the Stateflow module library in MATLAB Simulink; according to the insulation structure and insulation material parameters of the circuit breaker, the material state evolution model of the insulating sleeve and insulating rod of the circuit breaker and the stress finite element model of the mechanical structure can be realized by COMSOL software.

[0046] The three-dimensional structure model of the circuit breaker, the electromechanical system simulation model, the control loop simulation model and the insulation structure simulation model constructed above constitute a multi-domain physical simulation model of the circuit breaker.

[0047] S200: Build a circuit breaker data-driven model;

[0048] In this step, the constructed circuit breaker data-driven model mainly includes: electromechanical system fault diagnosis model and circuit breaker remaining service life prediction model. The electromechanical system fault diagnosis model is mainly aimed at the circuit structure and mechanical structure of the circuit breaker, and is mainly established based on a large amount of data collected from offline fault simulation tests; the remaining service life prediction model is mainly aimed at the insulation structure of the circuit breaker, and requires the external characteristic signal and characteristic data of the insulation structure under long-term power-on test.

[0049] S300: Dynamically interact and optimize the circuit breaker entity with the multi-domain coupled physical simulation model and the data-driven model in real time to obtain the digital twin of the circuit breaker.

[0050] The above embodiments constitute a complete technical solution of the present disclosure. The system-level physical simulation model constructed in this embodiment takes into account the coupling between simulation models in multiple physical fields, comprehensively analyzes multi-field and deep-level physical mechanisms from the establishment of the model, and uses artificial intelligence models to solve the technical difficulties of multiple degrees of freedom of system equations, differences in time scales and coupling ranges, and realizes the deep integration of multiple physical models.

[0051] Furthermore, by constructing a digital twin of a circuit breaker, it is possible to characterize and model the ultra-realistic state of the circuit breaker. Based on the digital twin of the circuit breaker, the complex fault operating state of the circuit breaker can be simulated, and destructive tests can be set up to simulate online breaking tests, so as to obtain the physical and operating characteristics of the circuit breaker under extreme operating conditions, form a circuit breaker fault feature database and a full life cycle database, and provide a big data foundation for the development of predictive maintenance algorithms for circuit breakers. For example, considering the economy and the stability of the regional power grid, large-scale load breaking tests of megawatt-level circuit breakers will not be carried out in practice, and offline fault simulation tests are often used instead, which are different from the actual operating state. The multi-domain physical simulation model in the digital twin of the circuit breaker can simulate the relationship between electricity and mechanics when the circuit breaker is broken, and can obtain signal data such as the operating coil current when the circuit breaker breaks different capacity loads. For another example, for safety reasons, in actual offline fault simulation tests, testers will not design combined faults, but will only simulate single fault states. The fault diagnosis model trained by the database composed of signal features of single fault states cannot effectively identify combined faults. However, the use of digital twins of circuit breakers can easily simulate combined faults, form a combined fault feature database, and train fault diagnosis models, thereby effectively identifying the true fault state of the circuit breaker.

[0052] In another embodiment, if Figure 3 As shown, the construction of the electromechanical system fault diagnosis model includes the following steps:

[0053] S201: Building a test platform based on the circuit breaker body to simulate different operating states of the circuit breaker, including normal operating state, electrical fault state and mechanical fault state;

[0054] In this step, the test platform includes a circuit breaker entity, a controller, a signal acquisition system and a host computer. According to the physical structure of the circuit breaker, different types and degrees of fault states are manually set, and the host computer sends action instructions to the controller to control the circuit breaker to open. The signal acquisition system collects the test data when the circuit breaker is in action, uploads it to the host computer, and then displays it in the LabVIEW software, while calling the MATLAB program to perform feature processing and subsequent work on the signal.

[0055] S202: Based on the test platform, the circuit breaker is opened and closed under different operating conditions, and the corresponding operating coil current signal and moving iron core linear travel displacement signal are collected at the same time;

[0056] S203: performing data preprocessing on the operating coil current signal and the moving iron core linear travel displacement signal under different operating states, respectively, extracting time domain features from the preprocessed signals, respectively, and combining the time domain features to obtain feature vectors under different operating states;

[0057] In this step, the current signal of the operating coil and the linear displacement signal of the moving iron core are preprocessed through the following steps: conversion from the signal data receiving side to the real physical quantity (the existing linear interpolation algorithm can be used), abnormal value processing (abnormal values ​​include +Inf value, -Inf value and NaN value, etc., which can be processed by the existing moving average algorithm) and data smoothing processing (the existing local linear regression algorithm can be used). Through preprocessing, the noise in the signal can be eliminated, which is convenient for the subsequent feature extraction process.

[0058] Furthermore, the time domain features extracted from the pre-processed operating coil current signal include: the first peak current value, the trough current value, the second peak current value, the core moving time, the core movement time, the contact movement time, the arc cut-off time, and the total time. The time domain features extracted from the pre-processed moving core linear travel displacement signal include: contact opening distance, overtravel, total travel, closing (opening) time, average closing (opening) speed, and average closing (opening) speed at full opening distance.

[0059] Furthermore, the above time domain features are concatenated into a feature vector in a specified order.

[0060] S204: taking the feature vector and its corresponding running state as a feature label set;

[0061] In this step, illustratively, the feature tag set is shown in Table 1:

[0062]

[0063] S205: Divide the feature label set into a training set and a test set, use the training set to train the electromechanical system fault diagnosis model, and use the test set to test the model effect. If the model test accuracy exceeds 90%, the electromechanical system fault diagnosis model is constructed; otherwise, adjust the parameters to retrain the model.

[0064] In this step, the electromechanical system fault diagnosis model adopts a fully connected neural network model, which has an input layer (14 neurons), two hidden layers (29 and 15 neurons), and an input layer (7 neurons). The activation function of the input layer and hidden layer neurons is the Sigmoid function, which is expressed as:

[0065]

[0066] The activation function of the output layer is the Softmax function, expressed as:

[0067]

[0068] Among them, z i represents the i-th fault classification, z j represents the output value of the jth neuron in the output layer of the neural network, Indicates the total number of fault categories.

[0069] In another embodiment, the construction of the circuit breaker remaining service life prediction model comprises the following steps:

[0070] S2001: According to the artificial pollution test platform of standard GB / T 4585-2004 "Artificial pollution test for high-voltage insulators for AC systems", artificial pollution with a certain value of salt density, insoluble dirt ash density and humidity is applied to the main circuit bushing of the circuit breaker;

[0071] S2002: applying a constant rated voltage across the main circuit of the circuit breaker, and collecting leakage current signals and high-frequency current sensor signals;

[0072] S2003: changing the test variables, repeating steps S2001 and S2002, obtaining a data set containing leakage current signals and high-frequency current sensor signals corresponding to different variable parameters, until the main circuit bushing has a flashover;

[0073] S2004: Each test data in the data set consists of test serial number, test time, test variables and sensor characteristic quantities, among which the test variables include ESDD value, NSDD value and humidity; the sensor characteristic quantities are extracted from the leakage current signal and the high-frequency current sensor signal respectively. The leakage current signal characteristic quantities include: time domain effective value, number of super-threshold pulses, super-threshold pulse duration, frequency domain fundamental wave proportion, third harmonic proportion, fifth harmonic proportion and waveform distortion rate; the high-frequency current sensor signal characteristic quantities include: maximum discharge amplitude, average discharge amplitude, maximum number of pulses, average number of pulses, pulse amplitude variance, pulse delay variance, skewness, steepness, number of local peaks, delay Weibull shape, delay Boolean scale and amplitude asymmetry.

[0074] S2005: Based on the test variables, the k-means clustering method is used for clustering, and each test data is classified and standardized according to the cluster center point. Then, the change trend of each feature quantity over time is analyzed, and the feature quantity with the most obvious change over time is selected as the variable of the circuit breaker remaining service life prediction model, and the functional relationship between the variable and time is fitted to complete the model construction.

[0075] In this step, the curve of the variable changing over time must satisfy the exponential accumulation model shown below:

[0076]

[0077] in, represents the intercept, which is a constant term; and is the random variable to be fitted, which determines the slope of the model, where It follows a log-normal distribution, Obey Gaussian distribution; is the noise term, subject to .

[0078] In addition, the circuit breaker entity is dynamically interacted with the multi-domain coupled physical simulation model and the data-driven model and optimized in real time. Figure 4 The invention is illustrated by the following specific examples.

[0079] For example, the operating circuit current of the circuit breaker theoretically obeys the following formula:

[0080]

[0081] Among them, U is the operating voltage; R is the operating circuit resistance; is the starting inductance, which is obtained from the finite element analysis model of the electromagnet and has a large difference from the actual value. It is difficult to measure in practice. Therefore, the resistance value R of the circuit breaker operating circuit can be controlled in the laboratory. When the circuit breaker is actuated, the operating coil current signal is collected and the least squares fitting method is used to Parameter estimation is performed to compare with existing simulation models The values ​​are compared to modify the optimization simulation model.

[0082] In another embodiment, the present disclosure further provides a circuit breaker digital twin construction device, comprising:

[0083] The first building module is used to build a multi-domain physical simulation model of a circuit breaker;

[0084] The second building module is used to build a circuit breaker data-driven model;

[0085] The interactive optimization module is used to dynamically interact and optimize the circuit breaker entity with the multi-domain coupled physical simulation model and the data-driven model in real time to obtain the digital twin of the circuit breaker.

[0086] In another embodiment, the second building block includes:

[0087] The first construction submodule is used to construct a fault diagnosis model for an electromechanical system;

[0088] The second construction submodule is used to construct a circuit breaker remaining service life prediction model.

[0089] In another embodiment, the present disclosure further provides a storage medium, in which a plurality of instructions are stored, and the instructions are suitable for being loaded and executed by a processor:

[0090] Construct a multi-domain physical simulation model of circuit breaker;

[0091] Build a data-driven model for circuit breakers;

[0092] The circuit breaker entity is dynamically interacted and optimized in real time with the multi-domain coupled physical simulation model and the data-driven model to obtain the digital twin of the circuit breaker.

[0093] The above embodiments are only for illustrating the technical concept and features of the present disclosure, and their purpose is to enable people familiar with the technology to understand the content of the present disclosure and implement it accordingly, and they cannot be used to limit the protection scope of the present disclosure. Any equivalent changes or modifications made according to the spirit of the present disclosure should be included in the protection scope of the present disclosure.

Claims

1. A method for constructing a digital twin of a circuit breaker, comprising the following steps: S100: Build a multi-domain physical simulation model of a circuit breaker, including: According to the actual structural parameters of the circuit breaker, a three-dimensional structural model of the circuit breaker is constructed based on Solidworks software and saved as a model file. The model file is then imported into MATLAB, and the three-dimensional structural model of the circuit breaker is constructed with the help of the Simulink Simscape Multibody module library. According to the structural parameters of the permanent magnetic actuator of the circuit breaker, the electromechanical system of the circuit breaker is simulated using ANSYS Maxwell software, and the simulation model of the electromechanical system of the circuit breaker can be obtained by combining it with MATLAB Simulink. According to the control loop design of the circuit breaker, a simulation model of the circuit breaker control loop is constructed based on the Stateflow module library in MATLAB Simulink. According to the insulation structure and insulation material parameters of the circuit breaker, the material state evolution model of the insulating sleeve and insulating pull rod of the circuit breaker and the stress finite element model of the mechanical structure can be realized through COMSOL software. S200: Build a circuit breaker data-driven model, including: Construct a fault diagnosis model for the electromechanical system. This model is based on an offline fault simulation test platform. It collects a large number of current signals of the operating coil and linear displacement signals of the moving iron core under different operating conditions. After data preprocessing and feature extraction, the signals are combined into feature vectors and trained using a neural network to obtain a high-precision fault diagnosis model. A circuit breaker remaining service life prediction model is constructed. The model extracts, screens and fits the data by applying artificial dirt, applying a constant rated voltage and collecting leakage current and high-frequency current sensor signals to predict the remaining service life of the circuit breaker. S300: Dynamically interact and optimize the circuit breaker entity with the multi-domain coupled physical simulation model and the data-driven model in real time to obtain the digital twin of the circuit breaker.

2. The method according to claim 1, wherein: In step S100, the multi-domain physical simulation model of the circuit breaker includes: a three-dimensional structure simulation model of the circuit breaker, a simulation model of the electromechanical system of the circuit breaker, a simulation model of the control loop of the circuit breaker, and a simulation model of the insulation structure of the circuit breaker.

3. The method according to claim 1, wherein: In step S200, the circuit breaker data driven model includes an electromechanical system fault diagnosis model and a circuit breaker remaining service life prediction model.

4. The method according to claim 3, wherein: The construction of the electromechanical system fault diagnosis model includes the following steps: S201: Building a test platform based on the circuit breaker body to simulate different operating states of the circuit breaker; S202: Based on the test platform, the circuit breaker is opened and closed under different operating conditions, and the corresponding operating coil current signal and moving iron core linear travel displacement signal are collected at the same time; S203: performing data preprocessing on the operating coil current signal and the moving iron core linear travel displacement signal under different operating states, respectively, extracting time domain features from the preprocessed signals, respectively, and combining the time domain features to obtain feature vectors under different operating states; S204: taking the feature vector and its corresponding running state as a feature label set; S205: Divide the feature label set into a training set and a test set, use the training set to train the electromechanical system fault diagnosis model, and use the test set to test the model. If the test accuracy exceeds 90%, the electromechanical system fault diagnosis model is constructed; otherwise, adjust the parameters to retrain the model.

5. The method according to claim 3, wherein: The construction of the circuit breaker remaining service life prediction model includes the following steps: S2001: Apply artificial dirt with a certain value of salt density, insoluble dirt ash density and humidity to the main circuit bushing of the circuit breaker; S2002: applying a constant rated voltage across the main circuit of the circuit breaker, and collecting leakage current signals and high-frequency current sensor signals; S2003: changing the test variables, and repeatedly executing steps S2001 and S2002 to obtain a data set containing leakage current signals and high-frequency current sensor signals corresponding to different test variables, until the main circuit bushing has a flashover; S2004: extracting characteristic quantities of leakage current signals and high-frequency current sensor signals in the data set; S2005: Screen the characteristic quantities to obtain variables of the circuit breaker remaining service life prediction model, and fit the variables to complete the model construction.

6. A circuit breaker digital twin construction device for implementing the method according to claim 1, comprising: The first building module is used to build a multi-domain physical simulation model of a circuit breaker; The second building module is used to build a circuit breaker data-driven model; The interactive optimization module is used to dynamically interact and optimize the circuit breaker entity with the multi-domain coupled physical simulation model and the data-driven model in real time to obtain the digital twin of the circuit breaker.

7. The device according to claim 6, wherein: The second building block comprises: The first construction submodule is used to construct a fault diagnosis model for an electromechanical system; The second construction submodule is used to construct a circuit breaker remaining service life prediction model.

8. A storage medium having a plurality of instructions stored therein, wherein the instructions are suitable for being loaded and executed by a processor: Construct a multi-domain physical simulation model of circuit breaker; Build a data-driven model for circuit breakers; The circuit breaker entity is dynamically interacted and optimized in real time with the multi-domain coupled physical simulation model and the data-driven model to obtain the digital twin of the circuit breaker.

Images