Fusion modeling method and device suitable for feeder terminal and electronic equipment

Abstract

The invention provides a fusion modeling method and device suitable for a feeder terminal and electronic equipment, and belongs to the technical field of power distribution networks. The method comprises: acquiring an electrical characteristic value of a power distribution network; inputting the electrical characteristic value into a fault classification model to obtain a fault classification result output by the fault classification model; wherein the fault classification model is obtained based on training of at least one of historical time domain features, historical frequency domain features and historical time-frequency domain features; any one of the historical time domain feature, the historical frequency domain feature and the historical time-frequency domain feature is obtained by performing fault judgment on historical data of the power distribution network in a set time period based on a mechanism model and performing feature extraction based on a data model. According to the method, the generalization and interpretability of the mechanism model and the nonlinear fitting and feature extraction capability of the data model can be fully utilized through fusion modeling of the mechanism model and the data model, and the accuracy and credibility of a fault classification result are improved while the model calculation efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network, and in particular to a fusion modeling method applicable to a feeder terminal, a fusion modeling device applicable to a feeder terminal, an electronic device, a machine-readable storage medium and a computer program product. Background Art

[0002] Line fault diagnosis consists of two steps: fault classification and fault location. However, when the control mechanism handles the fault, it is difficult to control the emergency response to the fault, and it is also difficult to accurately and efficiently start the live line inspection, resulting in a long time-consuming fault troubleshooting, low fault handling efficiency, and high manpower and material costs, which affects the safe and stable operation of the power grid and cannot provide users with continuous and reliable power supply.

[0003] Existing methods use pure mechanism models for fault diagnosis but suffer from insufficient accuracy; and use pure data models for fault diagnosis but suffer from weak generalization and poor interpretability. Summary of the invention

[0004] The purpose of the embodiments of the present invention is to provide a fusion modeling method suitable for feeder terminals, a fusion modeling device suitable for feeder terminals, an electronic device, a machine-readable storage medium and a computer program product, so as to solve the problem of insufficient accuracy in fault diagnosis using pure mechanism models; the defects of using pure data models for fault diagnosis such as weak generalization and poor interpretability.

[0005] In order to achieve the above object, an embodiment of the present invention provides a fusion modeling method applicable to a feeder terminal, including:

[0006] Obtain electrical characteristic values ​​of the distribution network;

[0007] Inputting the electrical characteristic value into a fault classification model to obtain a fault classification result output by the fault classification model;

[0008] Among them, the fault classification model is obtained by training based on at least one of historical time domain features, historical frequency domain features and historical time and frequency domain features; the historical time domain features, the historical frequency domain features and any one of the historical time and frequency domain features are obtained by performing fault judgment based on the mechanism model and feature extraction based on the data model based on the historical data of the distribution network within a set time period.

[0009] Optionally, the fault classification model is trained by the following steps:

[0010] Obtain historical data of the distribution network within a set time period;

[0011] Input the historical data into the mechanism model to obtain the fault diagnosis result output by the mechanism model; the mechanism model is used to determine whether a fault occurs in the historical data;

[0012] When the fault diagnosis result indicates the existence of a fault, extract the fault information eigenvalue from the historical data;

[0013] Input the fault information eigenvalue into the data model to obtain at least one of the historical time domain feature, historical frequency domain feature, and historical time-frequency domain feature output by the data model;

[0014] Based on at least one of the historical time domain feature, historical frequency domain feature, and historical time-frequency domain feature, perform model training to obtain a fault classification model.

[0015] Optionally, the historical time domain feature includes the ratio of the maximum zero-sequence current calculated based on the fault information eigenvalue to the maximum current change, or the standard deviation of the waveform distortion degree calculated based on the fault information eigenvalue.

[0016] Optionally, the historical frequency domain feature is the degree of zero-sequence current distortion obtained by performing Fourier transform on the zero-sequence current of the fault information eigenvalue.

[0017] Optionally, the historical time-frequency domain feature is the time-frequency domain feature obtained by performing wavelet transform on the zero-sequence current of the fault information eigenvalue.

[0018] Optionally, the historical time-frequency domain feature is represented by the following formula:

[0019]

[0020] Among them, C5 to C11 represent 7 historical time-frequency domain features, M1, M total 、M mid 、M k respectively represent intermediate calculation parameters, f(t) represents the zero-sequence current waveform of the fault information eigenvalue, t0 represents the initial time of f(t), ttotal represents the termination time of the selected waveform, ts (t0 < ts < ttotal) represents the fault occurrence time, tw represents the end time of the fault transient process t mid ,represents the intermediate time of f(t), t k represents t mid and t total between the selected time, used to distinguish the time-frequency characteristics of the zero-sequence current before and after the transient process, tm represents the peak time of f(t), th represents the first half peak of the zero-sequence current after tm, n represents the number of wavelet transform layers, E i represents in [t s ,t w], E1 represents the energy of the i-th wavelet transform signal of the zero-sequence current in the period of [t s , t w ], ej represents the energy of the jth wavelet transform signal of the zero-sequence current in the period [t0, tk], e1 represents the energy of the first wavelet transform signal of the zero-sequence current in the period [t0, tk], U0 represents the line voltage before the fault, I m Indicates the current amplitude, M s Represents the maximum slope of f(t).

[0021] On the other hand, an embodiment of the present invention further provides a fusion modeling device applicable to a feeder terminal, comprising:

[0022] An acquisition module, used for acquiring electrical characteristic values ​​of the distribution network;

[0023] A classification module, used for inputting the electrical characteristic value into a fault classification model to obtain a fault classification result output by the fault classification model;

[0024] Among them, the fault classification model is obtained by training based on at least one of historical time domain features, historical frequency domain features and historical time and frequency domain features; the historical time domain features, the historical frequency domain features and any one of the historical time and frequency domain features are obtained by performing fault judgment based on the mechanism model and feature extraction based on the data model based on the historical data of the distribution network within a set time period.

[0025] Optionally, the fault classification model is trained by the following steps:

[0026] Obtain historical data of the distribution network within a set time period;

[0027] The historical data is input into a mechanism model to obtain a fault diagnosis result output by the mechanism model; the mechanism model is used to determine whether a fault occurs in the historical data;

[0028] When the fault diagnosis result indicates that a fault exists, extracting a fault information characteristic value from the historical data;

[0029] Inputting the fault information characteristic value into a data model to obtain at least one of a historical time domain characteristic, a historical frequency domain characteristic, and a historical time-frequency domain characteristic output by the data model;

[0030] Model training is performed based on the historical time domain features, the historical frequency domain features, and at least one of the historical time and frequency domain features to obtain a fault classification model.

[0031] Optionally, the historical time-domain feature includes the ratio of the maximum zero-sequence current calculated based on the fault information eigenvalue to the maximum current change, or the standard deviation of the waveform distortion degree calculated based on the fault information eigenvalue.

[0032] Optionally, the historical frequency-domain feature is the degree of zero-sequence current distortion obtained by performing a Fourier transform on the zero-sequence current of the fault information eigenvalue.

[0033] Optionally, the historical time-frequency domain feature is the time-frequency domain feature obtained by performing a wavelet transform on the zero-sequence current of the fault information eigenvalue.

[0034] Optionally, the historical time-frequency domain feature is represented by the following formula:

[0035]

[0036] where C5 to C11 represent 7 historical time-frequency domain features, M1, M total , M mid , M k respectively represent intermediate calculation parameters, f(t) represents the zero-sequence current waveform of the fault information eigenvalue, t0 represents the initial time of f(t), ttotal represents the end time of the selected waveform, ts (t0 < ts < ttotal) represents the fault occurrence time, tw represents the end time of the fault transient process t mid , represents the intermediate time of f(t), t k represents t mid and t total the selected time between, used to distinguish the time-frequency characteristics of the zero-sequence current before and after the transient process, tm represents the peak time of f(t), th represents the first half of the peak of the zero-sequence current after tm, n represents the number of wavelet transform layers, E i represents the energy of the i-th layer wavelet transform signal of the zero-sequence current in the period of [t s , t w , E1 represents the energy of the first layer wavelet transform signal of the zero-sequence current in the period of [t s , t w , ej represents the energy of the j-th layer wavelet transform signal of the zero-sequence current in the period of [t0, tk], e1 represents the energy of the first layer wavelet transform signal of the zero-sequence current in the period of [t0, tk], U0 represents the line voltage before the fault, I m represents the current amplitude, M s represents the maximum slope of f(t).

[0037] On the other hand, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the above-mentioned fusion modeling method applicable to the feeder terminal is implemented.

[0038] On the other hand, the present invention also provides a machine-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the above-mentioned fusion modeling method applicable to feeder terminals is implemented.

[0039] On the other hand, the present invention also provides a computer program product, including a computer program, which implements the above-mentioned fusion modeling method applicable to feeder terminals when executed by a processor.

[0040] Through the above technical solution, the embodiment of the present invention obtains a fault classification model based on historical time domain features, historical frequency domain features, and at least one of the historical time and frequency domain features; wherein the historical time domain features, the historical frequency domain features, and any one of the historical time and frequency domain features are obtained by performing fault judgment based on the mechanism model and feature extraction based on the data model based on the historical data of the distribution network within a set time period. Therefore, the embodiment of the present invention can make full use of the generalization and interpretability of the mechanism model and the nonlinear fitting and feature extraction capabilities of the data model in line fault diagnosis through the fusion modeling of the mechanism model and the data model, thereby improving the accuracy and credibility of the fault classification results while improving the model calculation efficiency.

[0041] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0043] Figure 1 It is one of the flow charts of the fusion modeling method applicable to the feeder terminal provided by the present invention;

[0044] Figure 2 This is the second flow chart of the fusion modeling method applicable to the feeder terminal provided by the present invention;

[0045] Figure 3 It is a structural schematic diagram of a fusion modeling device applicable to a feeder terminal provided by the present invention;

[0046] Figure 4 It is one of the structural schematic diagrams of the electronic device provided by the present invention;

[0047] Figure 5 This is the second structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0048] The specific implementation of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.

[0049] Method Embodiment

[0050] Please refer to Figure 1 , an embodiment of the present invention provides a fusion modeling method applicable to a feeder terminal, comprising:

[0051] Step 100: Obtain electrical characteristic values ​​of the distribution network.

[0052] The fusion modeling method applicable to feeder terminals of the embodiment of the present invention is applied to electronic equipment, specifically to feeder terminals (Feeder Terminal Unit, FTU) of distribution networks. The feeder terminals of the embodiment of the present invention collect line three-phase current, voltage, zero-sequence current, frequency, temperature and other data in real time, and after filtering out bad data through the 3σ criterion, extract the time domain characteristics, frequency domain characteristics and time-frequency domain characteristics of the three-phase current, voltage, zero-sequence current, frequency, temperature and other data as electrical characteristic values ​​of the distribution network.

[0053] Among them, the time domain feature characterizes the characteristics of the physical quantities such as voltage and current in the distribution network changing over time. For example, when the distribution network fails, the time domain waveform of the fault current will show specific changes, such as sudden increase, waveform distortion, etc. For example, the time domain feature can be the ratio of the maximum zero-sequence current to the maximum current change over a period of time. By analyzing these time domain features, the rapid identification and positioning of distribution network faults can be achieved. The frequency domain feature characterization is to convert the physical quantities in the distribution network from the time domain to the frequency domain, and then analyze its characteristics at different frequencies. For example, the frequency domain feature can be the degree of zero-sequence current distortion calculated by Fourier transform of the zero-sequence current. The time-frequency domain feature is to combine the time domain feature with the frequency domain feature, taking into account both the change of the physical quantity over time and its characteristics at different frequencies. For example, the frequency domain feature can be a time-frequency feature based on wavelet transform.

[0054] Step 200: Input the electrical characteristic value into a fault classification model to obtain a fault classification result output by the fault classification model.

[0055] The electrical characteristic value of the feeder terminal is input into the fault classification model to obtain the fault classification result output by the fault classification model. The fault classification model is obtained by training based on at least one of the historical time domain features, the historical frequency domain features and the historical time and frequency domain features; the historical time domain features, the historical frequency domain features and any one of the historical time and frequency domain features are obtained by performing fault judgment based on the mechanism model and feature extraction based on the data model based on the historical data of the distribution network within a set time period.

[0056] The historical data of the distribution network within the set time period can be the three-phase current, voltage, zero-sequence current, frequency, temperature and other data of the distribution network in the past period of time. Figure 2 , the embodiment of the present invention uses the existing mechanism model to perform fault diagnosis, inputs historical data into the model, uses the mechanism model to quickly determine whether a fault exists, and reproduces the historical data to extract fault information characteristic values. For example, the operating status of the distribution network can be modeled and analyzed based on the mass balance equation, the energy balance equation, the momentum balance equation, etc. The collected historical data is compared with the parameters in the mechanism model to analyze whether the historical data meets the expectations of normal operation. If the historical data deviates from the expected value in the mechanism model, this may indicate the presence of a fault. The normalized fault information characteristic value is then used as the input of the data model, which can provide a high-quality data sample for the data model. The data model outputs at least one of historical time domain features, historical frequency domain features, and historical time and frequency domain features.

[0057] The fault classification model of the embodiment of the present invention can be constructed based on various existing machine learning models. For example, in one embodiment, the fault classification model can be constructed based on an extreme learning machine (ELM). For example, the steps for constructing a fault classification model based on ELM are as follows: The embodiment of the present invention takes the waveform characteristics of the historical data of the distribution network within a set time period as input, so the number of neurons in the input layer is set to the number of features. Combined with the empirical formula, the number of hidden layer neurons is set to 13. Different fault types are used as outputs, and the number of neurons in the output layer is set to 6. The activation function of the hidden layer neurons is selected as the Sigmoid function, and the output vector is processed by the Softmax function to convert the value of each element in the output vector to between 0 and 1.

[0058] Traditional line fault diagnosis mostly uses methods based on frequency domain theory to perform fault diagnosis, and obtains the characteristics of various fault conditions through heuristic rules, so that the logical judgment criteria and setting values ​​can be preset in advance and compared with the input differential current to realize the diagnosis of the fault type. The embodiment of the present invention obtains a fault classification model based on historical time domain features, historical frequency domain features, and at least one feature of historical time and frequency domain features; wherein the historical time domain features, the historical frequency domain features, and any one of the historical time and frequency domain features are obtained by performing fault judgment based on the mechanism model and feature extraction based on the data model based on the historical data of the distribution network within a set time period. Therefore, the embodiment of the present invention can make full use of the generalization and interpretability of the mechanism model and the nonlinear fitting and feature extraction capabilities of the data model through the fusion modeling of the mechanism model and the data model in line fault diagnosis, thereby improving the accuracy and credibility of the fault classification results while improving the model calculation efficiency.

[0059] In other aspects of the embodiments of the present invention, the fault classification model is trained by the following steps:

[0060] Step 10: Obtain historical data of the distribution network within a set time period.

[0061] For example, the feeder terminal obtains the historical data of the three-phase current, voltage, zero-sequence current, frequency, and temperature of the distribution network in the past month.

[0062] Step 20: input the historical data into a mechanism model to obtain a fault diagnosis result output by the mechanism model; the mechanism model is used to determine whether a fault occurs in the historical data.

[0063] Step 30: When the fault diagnosis result indicates that a fault exists, extract a fault information characteristic value from the historical data.

[0064] After the bad data in the historical data is filtered out by the 3σ criterion, the historical data is input into the mechanism model by using the mechanism modeling method to obtain the fault diagnosis result output by the mechanism model; thus, the fault is quickly identified, and when the fault diagnosis result indicates the existence of a fault, the fault information characteristic value is extracted from the historical data. The fault information characteristic value is the historical data with the fault.

[0065] Step 40: Input the fault information characteristic value into a data model to obtain at least one of the historical time domain characteristics, historical frequency domain characteristics, and historical time-frequency domain characteristics output by the data model.

[0066] Step 50: Perform model training based on the historical time domain features, the historical frequency domain features, and at least one of the historical time and frequency domain features to obtain a fault classification model.

[0067] The feeder terminal inputs the fault information feature value into the data model to obtain at least one of the historical time domain features, historical frequency domain features, and historical time and frequency domain features output by the data model. Model training is performed based on the historical time domain features, the historical frequency domain features, and at least one of the historical time and frequency domain features to obtain a fault classification model. The model training process of the fault classification model can use an existing model training method, which is not described in detail here.

[0068] Through the fault classification model composed of the mechanism model and data model of the embodiment of the present invention, the results analysis such as fault analysis, fault location and self-healing control are realized according to the real-time electrical characteristic values. The present invention can cooperate with the fault indicator acquisition unit, the pole circuit breaker, and the primary PT to realize the functions of fault indicator management, line electrical monitoring, fault analysis, fault location, fault isolation, etc. in the form of APP. It can quickly respond to the faults occurring in the overhead lines and isolate the fault source in time, restore power supply to the non-fault area, and improve the power supply reliability. At the same time, the present invention cooperates with the multifunctional sensor kits such as the inclination angle of the overhead line tower, ice coating, cable temperature, video monitoring, etc. to realize the function of the line environment monitoring APP, and conduct real-time monitoring and hazard warning of the operating environment of the overhead distribution lines.

[0069] In other aspects of the embodiments of the present invention, due to the complex relationship between fault symptoms and fault causes, it is often difficult to meet the requirements of fault diagnosis by only taking the change of a single information component as the basis for diagnosis. Therefore, multiple information on the tower line is considered to perform fault diagnosis. Thus, the embodiments of the present invention perform model training based on at least one of the historical time domain features, the historical frequency domain features, and the historical time and frequency domain features to obtain a fault classification model.

[0070] The historical time domain characteristics include a ratio of a maximum zero-sequence current to a maximum current change calculated based on the fault information characteristic value, or a standard deviation of a waveform distortion calculated based on the fault information characteristic value.

[0071] The fault information characteristic value may be the historical current data of the fault occurring within a period of time (e.g., within a day), and the maximum zero-sequence current calculated based on the fault information characteristic value may reflect the overall change state of the fault relative to the fault medium. The maximum change of the current during the initial transient stage indicates the dynamic state of the fault. Therefore, the ratio of the maximum zero-sequence current to the maximum current change calculated based on the fault information characteristic value may reflect the characteristic information of the fault, and the fault classification model may be trained by selecting the historical time domain features of the ratio of the maximum zero-sequence current to the maximum current change calculated based on the fault information characteristic value, which is beneficial to improving the accuracy and reliability of the fault classification model in line fault diagnosis.

[0072] Even in the steady state of the fault, the steady-state characteristics of faults caused by different reasons are different. The embodiment of the present invention uses the randomness of waveform distortion to describe the steady-state content, and defines it as the standard deviation of the waveform distortion calculated based on the fault information characteristic value. The fault information characteristic value can be the historical current data of the fault within a period of time (for example, within a day). The standard deviation of the waveform distortion is an indicator to measure the degree of deviation of the waveform from the standard sine wave. The formula is the standard deviation calculation formula, and the numerical value is the value of the discrete point of the waveform (voltage amplitude). Therefore, the standard deviation of the waveform distortion calculated based on the fault information characteristic value can reflect the steady-state characteristics of the fault. By selecting the historical time domain characteristics of the standard deviation of the waveform distortion calculated based on the fault information characteristic value to train the fault classification model, it is beneficial to improve the accuracy and reliability of the fault classification model in line fault diagnosis.

[0073] In other aspects of the embodiments of the present invention, the historical frequency domain feature is a zero-sequence current distortion degree obtained by calculating the zero-sequence current of the fault information characteristic value through Fourier transform.

[0074] The degree of zero-sequence current distortion is extracted by fast Fourier transform analysis of the fault zero-sequence current of different types of fault information characteristic values. The fault information characteristic value can be the historical current data of the fault within a period of time (for example, within a day). That is, the zero-sequence current equivalent to the fault information characteristic value is a waveform, and the waveform is Fourier transformed to form the expression of the zero-sequence current in the frequency domain, that is, the degree of zero-sequence current distortion. The historical frequency domain characteristics of the zero-sequence current distortion degree can reveal the frequency distribution and frequency components of physical quantities in the distribution network, which is of great significance for analyzing problems such as harmonics and interharmonics in the distribution network. By selecting the zero-sequence current distortion degree obtained by Fourier transforming the zero-sequence current of the fault information characteristic value to train the fault classification model, it is beneficial to improve the accuracy and reliability of the fault classification model in line fault diagnosis.

[0075] In other aspects of the embodiments of the present invention, the historical time-frequency domain features are time-frequency domain features obtained by calculating the zero-sequence current of the fault information characteristic value through wavelet transformation.

[0076] The waveforms of different types of faults differ in amplitude, energy at different times, and energy in different frequency bands. Wavelet transform has good adaptability to the time-frequency window and can reflect the characteristic information of the waveform at the initial stage of the fault. According to the wavelet transform, the zero-sequence current of the fault information characteristic value is calculated by wavelet transform and seven features (C5–C11) are extracted from the time-frequency domain features as shown in the following formula:

[0077] The historical time-frequency domain characteristics are expressed by the following formula:

[0078]

[0079]

[0080] Among them, C5 to C11 represent 7 historical time-frequency domain features, M1, M total , M mid , M k respectively represent intermediate calculation parameters, f(t) represents the zero-sequence current waveform of the fault information eigenvalue, t0 represents the initial time of f(t), ttotal represents the termination time of the selected waveform, ts (t0 < ts < ttotal) represents the fault occurrence time, tw represents the end time of the fault transient process t mid , represents the intermediate time of f(t), t k represents t mid and t total The selected time between is used to distinguish the time-frequency characteristics of the zero-sequence current before and after the transient process. tm represents the peak time of f(t), th represents the first half peak of the zero-sequence current after tm, n represents the number of wavelet transform layers, E i represents the energy of the i-th layer wavelet transform signal of the zero-sequence current in the period of [t s , t w , E1 represents the energy of the first layer wavelet transform signal of the zero-sequence current in the period of [t s , t w , ej represents the energy of the j-th layer wavelet transform signal of the zero-sequence current in the period of [t0, tk], e1 represents the energy of the first layer wavelet transform signal of the zero-sequence current in the period of [t0, tk], U0 represents the line voltage before the fault, I m represents the current amplitude, M s represents the maximum slope of f(t). The fault information eigenvalue can be historical current data with faults occurring within a period of time (such as within a day).

[0081] Since the time-frequency domain features obtained by wavelet transform calculation of the zero-sequence current of the fault information eigenvalue can reflect the characteristic information of the waveform in the initial stage of the fault. By selecting the historical time-frequency domain features obtained by wavelet transform calculation of the zero-sequence current of the fault information eigenvalue to train the fault classification model, it is beneficial to improve the accuracy and reliability of the fault classification model for line fault diagnosis.

[0082] In other embodiments of the present invention, two or more features including the ratio of the maximum zero-sequence current calculated based on the fault information characteristic value to the maximum current change, the standard deviation of the waveform distortion calculated based on the fault information characteristic value, the degree of zero-sequence current distortion calculated based on the fault information characteristic value through Fourier transform, and the time-frequency domain features calculated based on the zero-sequence current through wavelet transform can be used to train the fault classification model. By simultaneously training the model based on two or more features, namely, performing fault diagnosis by considering multiple information components, it is beneficial to further improve the accuracy and reliability of line fault diagnosis by the fault classification model.

[0083] Device Embodiment

[0084] Please refer to Figure 3 On the other hand, an embodiment of the present invention further provides a fusion modeling device applicable to a feeder terminal, comprising:

[0085] An acquisition module 301 is used to acquire electrical characteristic values ​​of a distribution network;

[0086] A classification module 302, used for inputting the electrical characteristic value into a fault classification model to obtain a fault classification result output by the fault classification model;

[0087] Among them, the fault classification model is obtained by training based on at least one of historical time domain features, historical frequency domain features and historical time and frequency domain features; the historical time domain features, the historical frequency domain features and any one of the historical time and frequency domain features are obtained by performing fault judgment based on the mechanism model and feature extraction based on the data model based on the historical data of the distribution network within a set time period.

[0088] Optionally, the fault classification model is trained by the following steps:

[0089] Obtain historical data of the distribution network within a set time period;

[0090] The historical data is input into a mechanism model to obtain a fault diagnosis result output by the mechanism model; the mechanism model is used to determine whether a fault occurs in the historical data;

[0091] When the fault diagnosis result indicates that a fault exists, extracting a fault information characteristic value from the historical data;

[0092] Inputting the fault information characteristic value into a data model to obtain at least one of a historical time domain characteristic, a historical frequency domain characteristic, and a historical time-frequency domain characteristic output by the data model;

[0093] Model training is performed based on at least one of the historical time-domain features, the historical frequency-domain features, and the historical time-frequency domain features to obtain a fault classification model.

[0094] Optionally, the historical time-domain features include the ratio of the maximum zero-sequence current to the maximum current change calculated based on the fault information eigenvalue, or the standard deviation of the waveform distortion degree calculated based on the fault information eigenvalue.

[0095] Optionally, the historical frequency-domain feature is the degree of zero-sequence current distortion obtained by performing a Fourier transform on the zero-sequence current of the fault information eigenvalue.

[0096] Optionally, the historical time-frequency domain feature is the time-frequency domain feature obtained by performing a wavelet transform on the zero-sequence current of the fault information eigenvalue.

[0097] Optionally, the historical time-frequency domain feature is represented by the following formula:

[0098]

[0099] where C5 to C11 represent 7 historical time-frequency domain features, M1, M total , M mid , M k respectively represent intermediate calculation parameters, f(t) represents the zero-sequence current waveform of the fault information eigenvalue, t0 represents the initial time of f(t), ttotal represents the termination time of the selected waveform, ts (t0 < ts < ttotal) represents the fault occurrence time, tw represents the end time of the fault transient process t mid , represents the intermediate time of f(t), t k represents t mid and t total the selected time between, used to distinguish the time-frequency characteristics of the zero-sequence current before and after the transient process, tm represents the peak time of f(t), th represents the first half peak of the zero-sequence current after tm, n represents the number of wavelet transform layers, E i represents the energy of the i-th layer wavelet transform signal of the zero-sequence current in the period of [t s , t w , E1 represents the energy of the first layer wavelet transform signal of the zero-sequence current in the period of [t s , t w , ej represents the energy of the j-th layer wavelet transform signal of the zero-sequence current in the period of [t0, tk], e1 represents the energy of the first layer wavelet transform signal of the zero-sequence current in the period of [t0, tk], U0 represents the line voltage before the fault, I m represents the current amplitude, M s represents the maximum slope of f(t).

[0100] The fusion modeling device suitable for feeder terminals includes a processor and a memory. The above-mentioned acquisition module and classification module, etc. are stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize corresponding functions.

[0101] The processor includes a kernel, which calls the corresponding program unit from the memory. There can be one or more kernels.

[0102] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0103] Figure 4 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4 As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430 and a communication bus 440, wherein the processor 410, the communication interface 420 and the memory 430 communicate with each other through the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute a fusion modeling method applicable to a feeder terminal, the method comprising: obtaining an electrical characteristic value of a distribution network; inputting the electrical characteristic value into a fault classification model to obtain a fault classification result output by the fault classification model; wherein the fault classification model is obtained by training based on at least one of historical time domain features, historical frequency domain features and historical time and frequency domain features; the historical time domain features, the historical frequency domain features and any one of the historical time and frequency domain features are obtained by performing fault judgment based on a mechanism model and extracting features based on a data model based on historical data of the distribution network within a set time period.

[0104] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0105] In one embodiment, please refer to Figure 5 The electronic device can be a feeder terminal, which includes a main control core board 10, an AC acquisition module 20, a switch input / output module 30, a communication module 40, a data storage module 50, a power module 60, and a line loss module (not shown), etc., supports telemetry, telesignaling pulse and remote control functions, supports 4G interface, Ethernet interface, and has RS-232 and RS-485 wired communication modes. In addition, the feeder terminal also includes a temperature measurement module, a clock module (model RX8025T) and a display module.

[0106] The main function of the AC acquisition module 20 is to send the voltage and current at the node of the distribution network to be tested to the A / D converter (AD73360 A / D converter) for subsequent processing through the current transformer and the voltage transformer. It can synchronously collect the three-phase voltage, current and neutral current, with a sampling accuracy of 0.5 and a sampling frequency of 4.8KHz, and also has a fault filtering function.

[0107] The acquisition of switch quantity is the basis of remote signaling function. Remote signaling is used to transmit the status of circuit breakers, switch positions and other switch and system status. Switch quantity output is the basis of remote control function. Only feeder terminals with output function can realize remote control of switch devices such as relays and photoelectric couplers. The switch quantity input / output module 30 of the feeder terminal of the embodiment of the present invention is designed to realize 4-way digital quantity input and 4-way digital quantity output to meet the needs of remote signaling and remote control. The output circuit requires the control relay to respond quickly and prevent false operation during power-on and power-off processes and complex electromagnetic environments. The input circuit requires a reasonable action voltage and a matching interference suppression and absorption circuit to prevent remote signaling from changing position or sending errors under certain external interference. The switch circuit is divided into two parts, input and output, and the two signals need to be designed separately. Among them, the input signal is used to obtain the contact position signal of the connected control relay, photoelectric coupler (photoelectric coupler with model TPL521-4), etc., for equipment status monitoring, and the output signal is used to control the field equipment.

[0108] In order to expand the applicable occasions of the feeder terminal, the communication module 40 of the feeder terminal in the embodiment of the present invention supports three transmission modes: 4G module (4G module of model SCMG401), Ethernet module (4G module of model LAN8720) and RS-232 / RS-485 wired communication mode. The Ethernet interface supports the configuration of the transmission mode between the destination host through auto-negotiation, and supports the Auto-MDIX automatic flip function to adapt to two types of network cables: direct connection and crossover.

[0109] The built-in flash memory of the main control core board 10 is 1MB. The data storage module 50 of the embodiment of the present invention can be expanded with a 4MB SPI interface FLASH chip AT25DF321A to record the current, voltage and switch opening and closing status when a distribution line fails, as well as the terminal configuration information and recording files.

[0110] The power module 60 should be able to output voltage according to actual needs to ensure effective power supply to the feeder terminal, communication equipment and switch operation. In order to ensure the reliability of power supply and the continuity of feeder terminal operation, there are at least two output circuits. The device is powered by an AC 220V power supply, and two 12V batteries are connected in series as a backup power supply. The battery management uses an integrated charge and discharge management module, which can automatically charge, discharge, activate and protect.

[0111] The following introduces the software design of the feeder terminal

[0112] The software part of the feeder terminal should be able to collect, process and control the data on the line in real time and report the result information to the main distribution station. According to the hardware circuit design, the software of the feeder terminal logically divides various information collection, human-machine interface functions, I / O control functions and communication functions into multiple tasks that run concurrently, and defines the interaction process between multiple tasks. The operating system manages the tasks and switches the status. The software platform is a non-real-time and real-time "dual operating system" platform. The non-real-time operating system serves as a platform for edge computing development such as data management, statistics, and analysis, and the real-time operating system serves as a platform for real-time business development such as data collection, protection, and control.

[0113] The operation management system is designed as follows:

[0114] Based on two sets of operating systems, non-real-time and real-time, the real-time operating system realizes the tailorable characteristics of the operating system, providing a full set of components from multi-task real-time kernel to middleware and peripheral drivers. Each task is mainly scheduled by event-driven, assisted by priority scheduling to achieve data interaction between different tasks. The main functions are: initialization, time stamp service, task monitoring and device management, event recording, waveform recording and storage, etc. The non-real-time operating system focuses on communication, monitoring, control, management, etc., and develops APPs with corresponding functions respectively.

[0115] The data acquisition and processing module program design is as follows:

[0116] In order to improve the accuracy of data collection at the feeder terminal, the number of sampling points is set during each sampling cycle. After a cycle of sampling is completed, the data is immediately processed in the FFT unit, the required electrical measurements are quickly calculated and analyzed, and any faults are reported immediately.

[0117] The switch input and output program design is as follows:

[0118] There are a large number of signals in the form of 0 and 1 in the power distribution line. The input switch quantity signals are mainly the status of the equipment on the line and the remote control instructions in the remote power distribution. The output switch quantity signals mainly include the opening and closing signals of the switch, the remote control instructions in the power distribution and the equipment alarm signals.

[0119] The communication program is designed as follows:

[0120] Includes uplink communication program and 101 / 104 protocol.

[0121] Uplink communication protocol: adapt to various types of uplink communication module 40 solutions.

[0122] 104 Protocol: Data service transmission.

[0123] 101 Protocol: Data service transmission.

[0124] The extensions are as follows:

[0125] Based on the basic software program, the feeder terminal develops advanced applications with extended functions according to the characteristic requirements, including the following functions: overcurrent protection, overload protection, control circuit disconnection fault, quick-break protection, reclosing, zero-sequence current protection, feeder automation, etc., as well as functions including the above-mentioned fusion modeling method suitable for feeder terminals.

[0126] The embodiment of the present invention proposes a fusion modeling method suitable for feeder terminals that combines a mechanism model and a data model, which improves the model calculation efficiency while ensuring the accuracy and credibility of the modeling results. The embodiment of the present invention also proposes a feeder terminal, the hardware circuit of the feeder terminal is designed in a modular manner, the software part of the feeder terminal is a non-real-time operating system that implements edge computing services such as data management, statistics, and analysis, and the real-time operating system implements real-time services such as data collection, protection, and control.

[0127] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a machine-readable storage medium. When the computer program is executed by a processor, the computer can execute a fusion modeling method applicable to a feeder terminal, the method including: obtaining electrical characteristic values ​​of a distribution network; inputting the electrical characteristic values ​​into a fault classification model to obtain a fault classification result output by the fault classification model; wherein the fault classification model is trained based on at least one of historical time domain features, historical frequency domain features, and historical time and frequency domain features; the historical time domain features, the historical frequency domain features, and any one of the historical time and frequency domain features are obtained by performing fault judgment based on a mechanism model and extracting features based on the data model based on the historical data of the distribution network within a set time period.

[0128] On the other hand, the present invention also provides a machine-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute a fusion modeling method applicable to a feeder terminal, the method comprising: obtaining electrical characteristic values ​​of a distribution network; inputting the electrical characteristic values ​​into a fault classification model to obtain a fault classification result output by the fault classification model; wherein the fault classification model is trained based on at least one of historical time domain features, historical frequency domain features, and historical time and frequency domain features; the historical time domain features, the historical frequency domain features, and any one of the historical time and frequency domain features are obtained by performing fault judgment based on a mechanism model and extracting features based on the data model based on the historical data of the distribution network within a set time period.

[0129] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0130] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0131] 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 make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fusion modeling method applicable to a feeder terminal, characterized in that: include: Obtain electrical characteristic values ​​of the distribution network; Inputting the electrical characteristic value into a fault classification model to obtain a fault classification result output by the fault classification model; Among them, the fault classification model is obtained by training based on at least one of historical time domain features, historical frequency domain features and historical time and frequency domain features; the historical time domain features, the historical frequency domain features and any one of the historical time and frequency domain features are obtained by performing fault judgment based on the mechanism model and feature extraction based on the data model based on the historical data of the distribution network within a set time period.

2. The fusion modeling method applicable to feeder terminals according to claim 1, characterized in that: The fault classification model is trained by the following steps: Obtain historical data of the distribution network within a set time period; The historical data is input into a mechanism model to obtain a fault diagnosis result output by the mechanism model; the mechanism model is used to determine whether a fault occurs in the historical data; When the fault diagnosis result indicates that a fault exists, extracting a fault information characteristic value from the historical data; Inputting the fault information characteristic value into a data model to obtain at least one of a historical time domain characteristic, a historical frequency domain characteristic, and a historical time-frequency domain characteristic output by the data model; Model training is performed based on the historical time domain features, the historical frequency domain features, and at least one of the historical time and frequency domain features to obtain a fault classification model.

3. The fusion modeling method applicable to feeder terminals according to claim 1 or 2, characterized in that: The historical time domain characteristics include a ratio of a maximum zero-sequence current to a maximum current change calculated based on the fault information characteristic value, or a standard deviation of a waveform distortion calculated based on the fault information characteristic value.

4. The fusion modeling method applicable to feeder terminals according to claim 1 or 2, characterized in that: The historical frequency domain feature is the degree of zero-sequence current distortion obtained by calculating the zero-sequence current of the fault information characteristic value through Fourier transform.

5. The fusion modeling method applicable to feeder terminals according to claim 1 or 2, characterized in that: The historical time-frequency domain features are time-frequency domain features obtained by calculating the zero-sequence current of the fault information characteristic value through wavelet transformation.

6. The fusion modeling method applicable to feeder terminals according to claim 5, characterized in that: The historical time-frequency domain characteristics are expressed by the following formula: Among them, C5 to C11 represent 7 historical time-frequency domain features, M1, M total , M mid , M k respectively represent intermediate calculation parameters, f(t) represents the zero-sequence current waveform of the fault information feature value, t0 represents the initial time of f(t), ttotal represents the termination time of the selected waveform, ts (t0 < ts < ttotal) represents the fault occurrence time, tw represents the end time t mid of the fault transient process, represents the intermediate time of f(t), t k represents t mid and t total the selected time between them, used to distinguish the time-frequency characteristics of the zero-sequence current before and after the transient process, tm represents the peak time of f(t), th represents the first half peak of the zero-sequence current after tm, n represents the number of wavelet transform layers, E i represents the energy of the i-th layer wavelet transform signal of the zero-sequence current in the period of [t s , t w , E1 represents the energy of the first layer wavelet transform signal of the zero-sequence current in the period of [t s , t w , e j represents the energy of the j-th layer wavelet transform signal of the zero-sequence current in the period of [t0, tk], e1 represents the energy of the first layer wavelet transform signal of the zero-sequence current in the period of [t0, tk], U0 represents the line voltage before the fault, I m represents the current amplitude, M s represents the maximum slope of f(t).

7. A fusion modeling device suitable for a feeder terminal, characterized in that: include: An acquisition module, used for acquiring electrical characteristic values ​​of the distribution network; A classification module, used for inputting the electrical characteristic value into a fault classification model to obtain a fault classification result output by the fault classification model; Among them, the fault classification model is obtained by training based on at least one of historical time domain features, historical frequency domain features and historical time and frequency domain features; the historical time domain features, the historical frequency domain features and any one of the historical time and frequency domain features are obtained by performing fault judgment based on the mechanism model and feature extraction based on the data model based on the historical data of the distribution network within a set time period.

8. The fusion modeling device applicable to a feeder terminal according to claim 7, characterized in that: The fault classification model is trained by the following steps: Obtain historical data of the distribution network within a set time period; The historical data is input into a mechanism model to obtain a fault diagnosis result output by the mechanism model; the mechanism model is used to determine whether a fault occurs in the historical data; When the fault diagnosis result indicates that a fault exists, extracting a fault information characteristic value from the historical data; Inputting the fault information characteristic value into a data model to obtain at least one of a historical time domain characteristic, a historical frequency domain characteristic, and a historical time-frequency domain characteristic output by the data model; Model training is performed based on the historical time domain features, the historical frequency domain features, and at least one of the historical time and frequency domain features to obtain a fault classification model.

9. The fusion modeling device applicable to a feeder terminal according to claim 7 or 8, characterized in that: The historical time domain characteristics include a ratio of a maximum zero-sequence current to a maximum current change calculated based on the fault information characteristic value, or a standard deviation of a waveform distortion calculated based on the fault information characteristic value.

10. The fusion modeling device applicable to a feeder terminal according to claim 7 or 8, characterized in that: The historical frequency domain feature is the degree of zero-sequence current distortion obtained by calculating the zero-sequence current of the fault information characteristic value through Fourier transform.

11. The fusion modeling device applicable to a feeder terminal according to claim 7 or 8, characterized in that: The historical time-frequency domain features are time-frequency domain features obtained by calculating the zero-sequence current of the fault information characteristic value through wavelet transformation.

12. The fusion modeling device applicable to a feeder terminal according to claim 11, characterized in that: The historical time-frequency domain characteristics are expressed by the following formula: Among them, C5 to C11 represent 7 historical time-frequency domain features, M1, M total , M mid , M k respectively represent intermediate calculation parameters, f(t) represents the zero-sequence current waveform of the fault information feature value, t0 represents the initial time of f(t), ttotal represents the end time of the selected waveform, ts (t0 < ts < ttotal) represents the fault occurrence time, tw represents the end time t mid of the fault transient process, represents the intermediate time of f(t), t k represents t mid and t total The selected time between them is used to distinguish the time-frequency characteristics of the zero-sequence current before and after the transient process. tm represents the peak time of f(t), th represents the first half peak of the zero-sequence current after tm, n represents the number of wavelet transform layers, E i represents the energy of the i-th layer wavelet transform signal of the zero-sequence current in the period of [t s , t w , E1 represents the energy of the first layer wavelet transform signal of the zero-sequence current in the period of [t s , t w , ej represents the energy of the j-th layer wavelet transform signal of the zero-sequence current in the period of [t0, tk], e1 represents the energy of the first layer wavelet transform signal of the zero-sequence current in the period of [t0, tk], U0 represents the line voltage before the fault, I m represents the current amplitude, M s represents the maximum slope of f(t).

13. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the fusion modeling method applicable to the feeder terminal according to any one of claims 1 to 6 is implemented.

14. A machine-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the fusion modeling method applicable to a feeder terminal according to any one of claims 1 to 6 is implemented.

15. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the fusion modeling method applicable to a feeder terminal according to any one of claims 1 to 6 is implemented.

Images