A method for identifying a transformer area and binding a user based on a special transformer acquisition terminal

Abstract

The application relates to the technical field of power system transformer area management, and specifically discloses a transformer area identification and user binding method based on a special transformer acquisition terminal, which comprises the following steps: separating the harmonic interference generated by the start and stop of electric welders through a kurtosis index and time-frequency energy entropy fusion detection method, and extracting a fundamental wave voltage and a pure power signal; constructing a dynamic time warping model containing phase consistency and gradient feature constraints based on the fundamental wave voltage, calculating voltage trajectory similarity to determine electrical coupling; for users with electrical coupling, fusing dynamic weight power cross-correlation analysis to generate a comprehensive judgment value, combining historical topology change events, and using a space-time graph neural network to update the transformer area topology structure online, so that the dynamic determination of the transformer area to which the user belongs is realized; through multi-dimensional feature fusion and anti-interference algorithm design, the phase shift and waveform distortion problems caused by the electric welder harmonics are reduced, and the transformer area identification accuracy and real-time performance are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system transformer area management, and particularly relates to a transformer area identification and user binding method based on a special transformer terminal. BACKGROUND

[0002] Under the development background of smart grid and power internet of things, transformer area identification and user binding are the core basis for realizing accurate line loss calculation, load management and power quality monitoring. However, with the wide application of industrial field welding machines and other nonlinear loads, the high-order harmonic pollution of power grid voltage waveform generated in the start-stop process of the welding machines leads to the following technical bottlenecks of the traditional method:

[0003] The harmonic component in the start-stop process of the welding machine distorts the fundamental voltage and power signal, and the traditional single kurtosis detection or time-frequency analysis method cannot effectively distinguish the harmonic interference and the real topology change, resulting in an increase in the extraction error of the fundamental component and introducing distortion of the original data for subsequent analysis.

[0004] The prior art depends on full-map retraining or single switch signal triggering for transformer area topology updating, cannot respond to high-frequency disturbance events such as welding machine start-stop in real time, and does not combine the load state to check the authenticity of the topology change, which has the technical defect that the transient disturbance is misjudged as a topology change, and cannot meet the demand of dynamic management of industrial transformer areas.

[0005] Therefore, the present application provides a transformer area identification and user binding method based on a special transformer terminal. SUMMARY

[0006] The present application aims to provide a transformer area identification and user binding method based on a special transformer terminal to solve the above background problems.

[0007] The purpose of the present application can be achieved by the following technical solutions:

[0008] A transformer area identification and user binding method based on a special transformer terminal, comprising the following steps:

[0009] Harmonic separation processing is performed on the high-frequency sampling data of the special transformer terminal, a kurtosis index and time-frequency energy entropy fusion detection method is used to judge whether the start-stop of the welding machine causes harmonic interference, and if so, the fundamental voltage component and the pure power signal are extracted;

[0010] Based on the fundamental voltage component, the local distance of the user and the reference node is calculated, a voltage trajectory similarity model resistant to harmonic interference is constructed, the voltage trajectory similarity is obtained, and it is determined whether the user and the transformer area are electrically coupled;

[0011] If there is electrical coupling, the fusion dynamic weight power cross-correlation analysis is carried out in combination with the pure power signal to obtain a comprehensive judgment value, and whether the user and the transformer area need to be checked for topology update is determined based on the comprehensive judgment value;

[0012] If the topology update check is needed, a historical topology change event is obtained, the transformer area topology structure is updated online through a space-time graph neural network based on the comprehensive judgment value and the topology change event, and the transformer area to which the user belongs is determined.

[0013] As a further scheme of the application, the way of judging whether the start-stop of the electric welding machine causes harmonic interference is:

[0014] The characteristic harmonic frequency band when the electric welding machine starts and stops is extracted through a band-pass filter set to obtain a harmonic component waveform;

[0015] The kurtosis and time-frequency energy entropy of the harmonic component are obtained, and boundary interference analysis is carried out to obtain a boundary interference distance;

[0016] Based on the boundary interference distance, it is judged whether the start-stop of the electric welding machine causes harmonic interference, if so, the fundamental voltage component and the pure power signal are extracted, and if not, no processing is performed.

[0017] As a further scheme of the application, the way of performing the boundary interference analysis is:

[0018] Based on the kurtosis and time-frequency energy entropy of the harmonic component, a kurtosis-energy entropy feature vector is constructed;

[0019] The historical kurtosis-energy entropy data of the start-stop of the electric welding machine are obtained, a Gaussian mixture model is established, the harmonic interference boundary of the start-stop of the electric welding machine is set, and a harmonic interference boundary vector is constructed;

[0020] The kurtosis-energy entropy feature vector and the harmonic interference boundary vector are subjected to Euclidean distance calculation to obtain the boundary interference distance.

[0021] As a further scheme of the application, the way of judging whether the user and the transformer area exist electrical coupling is:

[0022] The fundamental voltage of the user and the reference node is collected, the fundamental voltage sequence of the user and the reference node is constructed respectively, the fundamental voltage sequence is subjected to time sequence alignment processing and amplitude standardization processing, and the fundamental local distance is calculated;

[0023] Based on the fundamental local distance, a DTW cumulative distance matrix is constructed, the phase consistency coefficient and the gradient feature are obtained, and a micro-section screening condition is constructed, the DTW distance meeting the micro-section screening condition is selected, and the voltage trajectory similarity is calculated;

[0024] Based on the voltage trajectory similarity, the strength of the electrical coupling of the transformer area is determined to obtain the determination result of whether the user and the transformer area exist electrical coupling.

[0025] As a further scheme of the present application: the way to construct the micro-segment screening condition is:

[0026] After the fundamental voltage sequence is processed by time sequence alignment and amplitude standardization, it is divided into equal-length micro-segments according to fixed time length;

[0027] The phase consistency coefficient and gradient feature of each equal-length micro-segment are calculated;

[0028] The micro-segment screening condition includes:

[0029] When DTW path searching, the micro-segment of is preferentially selected;

[0030] Wherein, is the phase consistency value of the mth micro-segment of the user and the nth micro-segment of the reference node respectively, is the phase threshold value, is the fundamental phase structure difference;

[0031] If the gradient feature difference between micro-segments is within the gradient threshold range, gradient constraint pruning is established and the DTW path is pruned.

[0032] As a further scheme of the present application: the way to determine whether the user and the transformer area need to be checked for topology update is:

[0033] The cross-correlation coefficient and fundamental voltage phase difference of the user power component and the reference node power component are obtained, and the sum is calculated to obtain the transformer area user similarity;

[0034] The voltage trajectory similarity and the transformer area user similarity are summed to obtain the comprehensive determination value;

[0035] Based on the comprehensive determination value, the correlation strength of the user and the transformer area is determined and analyzed to obtain the determination result of whether the user and the transformer area need to be checked for topology update.

[0036] As a further scheme of the present application: the way to obtain the cross-correlation coefficient of the user power component and the reference node power component is:

[0037] Based on the pure power signal, the fundamental active power, 3rd harmonic reactive power and 5th harmonic reactive power are obtained by separation processing, and the user power component containing the fundamental active power, 3rd harmonic reactive power and 5th harmonic reactive power is constructed;

[0038] Different hysteresis steps are set, and the cross-correlation coefficient of the user power component and the reference node power component under each hysteresis step is calculated.

[0039] As a further scheme of the present application: the way to determine the transformer area to which the user belongs is:

[0040] Obtain a historical topology change event, construct a time graph neural network based on an industrial transformer area characteristic, and determine a user's belonging transformer area.

[0041] As a further scheme of the application: the way of constructing the time graph neural network is:

[0042] Constructing an edge node and feature engineering;

[0043] Real-time detection and coding of topology change events;

[0044] Establishing an online inference framework of a spatio-temporal graph neural network;

[0045] Dynamically determining a user's belonging transformer area based on a correlation probability and a decision rule.

[0046] As a further scheme of the application: the way of dynamically determining a user's belonging transformer area based on the correlation probability and the decision rule is:

[0047] For each user node, output a probability distribution belonging to each transformer node of the transformer area through the graph neural network;

[0048] The decision rule includes:

[0049] Decision rule one, strong correlation direct binding: determining the user and the transformer area attribution based on the probability distribution and the comprehensive judgment value;

[0050] Decision rule two, conflict scene arbitration: when the difference of the probability distribution of multiple transformers is in a preset range, constructing a harmonic direction constraint to determine the attribution of the transformer area and the user.

[0051] The application has the following beneficial effects:

[0052] (1) The kurtosis index and time-frequency energy entropy fusion detection method is used to identify the harmonic interference caused by the start and stop of the electric welder, realize the extraction of the fundamental voltage component and the pure power signal, provide an anti-interference pure data basis for subsequent analysis, reduce the signal distortion problem caused by harmonic pollution in the nonlinear load scene, and improve the reliability of data preprocessing;

[0053] (2) A dynamic time warping model containing phase consistency and gradient feature constraint is constructed based on the fundamental voltage, the voltage trajectory similarity is quantified and the electrical coupling is judged through micro-segment feature matching and anti-harmonic rule design, which is conducive to suppressing the interference of phase shift and waveform distortion caused by harmonics on the matching result, improving the accuracy of voltage trajectory synchronization analysis, and providing a basis for transformer area electrical association judgment;

[0054] (3) The dynamic weight power cross-correlation analysis is fused, the power component weight is adaptively adjusted according to the electric welding machine load state, a comprehensive judgment value is generated in combination with the voltage trajectory similarity, the multi-dimension quantification of the user and the area association strength is realized, the misjudgment problem of a single index under a complex load scene is reduced, and the robustness and adaptability of the association strength judgment are enhanced; based on the comprehensive judgment value and the topology change event, the spatial-temporal graph neural network is used to update the area topology structure online, the local graph update strategy and the anti-interference checking mechanism driven by the event are used to dynamically determine the area to which the user belongs, the real-time and intelligentization of the topology management are realized, the area identification problem under the dynamic interference such as the electric welding machine start-stop is alleviated, and the accuracy and efficiency of the topology update are improved. BRIEF DESCRIPTION OF DRAWINGS

[0055] The application will be further described below in combination with the drawings.

[0056] Figure 1 is a flowchart of a user binding and area identification method based on a special terminal according to the present application;

[0057] Figure 2 is a judgment flowchart for judging whether a user and an area are in strong association according to the present application;

[0058] Figure 3 is a structural schematic diagram of a user binding and area identification system based on a special terminal according to the present application. DETAILED DESCRIPTION

[0059] The technical solutions in the embodiments of the present application will be clearly and completely described below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0060] Embodiment one: please refer to Figure 1 The present application is a user binding and area identification method based on a special terminal, which comprises the following steps:

[0061] Step one: the harmonic separation processing is performed on the high-frequency sampling data of the special terminal, the kurtosis index and time-frequency energy entropy fusion detection method is adopted, the harmonic interference caused by the electric welding machine start-stop is judged, and the fundamental voltage component and the pure power signal are extracted if the harmonic interference is caused;

[0062] The way of judging whether the electric welding machine start-stop causes harmonic interference by adopting the kurtosis index and time-frequency energy entropy fusion detection method is as follows:

[0063] Preferably, the voltage and current waveforms of the special terminal 10 kHz and above sampling rate are taken as high-frequency sampling data, and the high-frequency sampling data is decomposed in frequency domain by using iterative wavelet packet transform to separate the fundamental component (50 Hz) and each harmonic component (such as 3 times, 5 times, 7 times, etc.);

[0064] The characteristic harmonic frequency band when the electric welding machine starts and stops is extracted by the band-pass filter set to obtain the harmonic component waveform;

[0065] Preferably, the kurtosis K of the harmonic component is obtained by the formula:

[0066] Wherein, is the instantaneous sampling value of the harmonic component, is the mean value of the harmonic component, is the number of sampling points, is the standard deviation of the harmonic component, is the number of sampling points of the harmonic component.

[0067] The short-time Fourier transform is performed on the harmonic component to generate a time-frequency matrix .

[0068] The time-frequency energy distribution is obtained by the formula: .

[0069] Wherein, The harmonic component is the result of short-time Fourier transform (STFT), is the time-frequency matrix, the row represents the time t, the column represents the frequency f, and the element is a complex number representing the signal component at the corresponding time and frequency, is the global time-frequency energy sum.

[0070] Based on the time-frequency energy distribution , the time-frequency energy entropy is obtained by the equation:

[0071] Based on the kurtosis and the time-frequency energy entropy of the harmonic component, a kurtosis-energy entropy feature vector is constructed .

[0072] The historical kurtosis-energy entropy data of the electric welding machine start-stop is obtained, a Gaussian mixture model is established to set the harmonic interference boundary of the electric welding machine start-stop, and a harmonic interference boundary vector is constructed.

[0073] The Euclidean distance between the kurtosis-energy entropy feature vector and the harmonic interference boundary vector is calculated to obtain the boundary interference distance.

[0074] ​​​It can be understood that the way of setting the harmonic interference boundary of the electric welding machine start-stop by establishing a Gaussian mixture model is: first, collect the historical kurtosis-energy entropy data of the electric welding machine start-stop, train the Gaussian mixture model (GMM, determine the optimal component number by BIC) using the EM algorithm, and fit the data probability distribution; then, according to the 95% confidence threshold, the electric welding machine harmonic interference range is determined in the feature space, and the harmonic interference boundary vector is constructed according to the harmonic interference range; the boundary interference distance is a matching quantitative index of the current harmonic feature and the electric welding machine start-stop historical harmonic feature boundary, which reflects whether the current harmonic belongs to the interference caused by the electric welding machine start-stop;

[0075] Based on the boundary interference distance, it is judged whether the electric welding machine start-stop causes harmonic interference, if so, the fundamental voltage component and the pure power signal are extracted;

[0076] If no harmonic interference is caused, no processing is performed.

[0077] Preferably, the boundary interference distance and the preset boundary interference distance threshold (the Euclidean distance mean of the interference samples and the non-interference samples is calculated by training the GMM with 500 groups of historical kurtosis-energy entropy data of the electric welding machine start-stop, and the 1.2 times of the mean is taken as the boundary interference distance threshold) are compared, and if the boundary interference distance is higher than or equal to the preset boundary interference distance threshold, it is considered that the electric welding machine start-stop causes harmonic interference.

[0078] Those skilled in the art can understand that the way of extracting the fundamental voltage component and the pure power signal is: when it is determined that the current harmonic is the interference caused by the electric welding machine start-stop through the boundary interference distance, the frequency range of the electric welding machine harmonic is located based on the previous time-frequency energy distribution (such as the 3rd and 5th frequency bands where the pulse harmonics are concentrated), and the time-frequency masking technology is used: the harmonic components in the frequency range are applied with attenuation weight (dynamically adjusted according to the energy proportion) in the time-frequency matrix, and the fundamental voltage component is restored by inverse transformation; the same harmonic positioning and suppression operation is performed on the current signal, the fundamental current is extracted, the harmonic power of the electric welding machine is separated from the total power through the instantaneous product of the fundamental voltage and the fundamental current, and finally the pure power signal free from harmonic pollution is obtained.

[0079] Step two: based on the fundamental voltage component, the fundamental local distance between the user and the reference node is calculated, the voltage trajectory similarity model against harmonic interference is constructed, the voltage trajectory similarity is obtained, and it is determined that there is an electrical coupling between the user and the transformer area;

[0080] The acquisition method of the user and the reference node is:

[0081] Preferably, the terminal monitoring point of the electric welding machine or the special transformer of the electric equipment is taken as the user node;

[0082] The stable node not affected by the harmonic interference of the electric welding machine is taken as the reference node;

[0083] The skilled in the art can understand that the user node collects the voltage and current high-frequency sampling data of the node through the special transformer terminal, which is used to analyze the harmonic interference of the electric welding machine start-stop. The voltage trajectory of the node will be directly affected by the start-stop of the electric welding machine, containing obvious harmonic components;

[0084] The reference node refers to the reference monitoring point in the power system for comparison, that is, the stable node not affected by the harmonic interference of the electric welding machine, such as the monitoring point near the bus of the substation, the point of common coupling, or other load nodes in the same power grid not affected by the electric welding machine;

[0085] The voltage waveform of the reference node is used as a reference signal for comparison with the voltage trajectory of the user node. The similarity of the two is calculated by the fundamental local distance to quantify the degree of harmonic interference on the voltage of the user node;

[0086] Wherein, the way to calculate the fundamental local distance between the user and the reference node is:

[0087] The fundamental voltage of the user and the reference node is collected, and the fundamental voltage sequence of the user and the reference node is constructed, and the time sequence alignment processing and amplitude normalization processing are performed on the fundamental voltage sequence;

[0088] Preferably, the distance formula is: The fundamental local distance is obtained ;

[0089] Wherein, represents the fundamental voltage of the user at time i and the reference node at time j after time sequence alignment processing and amplitude normalization processing;

[0090] Wherein, are the voltage change rates of the user and the reference node, respectively, is the amplitude difference weight, ;

[0091] Wherein, the way to construct the voltage trajectory similarity model resistant to harmonic interference is:

[0092] Based on the fundamental local distance The DTW cumulative distance matrix is constructed, the phase consistency coefficient and the gradient feature are obtained, and the micro-segment screening condition is constructed, and the DTW distance meeting the micro-segment screening condition is selected;

[0093] Wherein, the way to obtain the phase consistency coefficient and the gradient feature is:

[0094] After the time sequence alignment processing and amplitude normalization processing of the fundamental voltage sequence, the equal-length micro-segments are divided according to the fixed time length;

[0095] The phase consistency coefficient and the gradient feature of each equal-length micro-segment are calculated;

[0096] It needs to be explained that after the voltage signal is divided into micro-segments by equal-length time window, the phase information of each frequency component is obtained by Fourier transform decomposition for each micro-segment, which is substituted into the Kovesi model, and the alignment degree of each frequency phase relative to the fundamental phase is quantified to calculate the phase consistency coefficient of the micro-segment, which represents the disorder degree of the phase structure under harmonic interference;

[0097] The voltage time sequence of the micro-segment is synchronized, the amplitude change rate of adjacent sampling points is calculated by difference method, and the mean and variance statistics of the gradient are extracted as gradient features to reflect the severity of waveform mutation;

[0098] Among them, the micro-segment screening conditions include:

[0099] Micro-segment screening condition one: when searching for the DTW path, preferentially select the micro-segment ;

[0100] Among them, and are the phase consistency values of the mth micro-segment of the user and the nth micro-segment of the reference node respectively, is a preset phase threshold, is the fundamental phase structure difference, which is used to suppress the interference of the phase offset caused by the harmonic on the overall matching; Take 0.8-0.9 (the preferred value is 0.85), according to the statistical phase consistency coefficient of the fundamental voltage micro-segment of 1000 groups of electric welding machine start-stop, take 95% quantile (i.e. 95% of the phase consistency coefficient of normal micro-segment ≥0.8); if the power of the electric welding machine in the transformer area is >50kW, ;

[0101] Micro-segment screening condition two, if the gradient feature difference between micro-segments is , establish gradient constraint pruning and prune the DTW path;

[0102] Among them, is the gradient threshold preset by the person skilled in the art according to the actual measurement data of the user side voltage fluctuation (such as the maximum gradient of 100 groups of voltage mutation data is 0.08kV / ms), is the gradient feature value of the mth micro-segment of the user and the nth micro-segment of the reference;

[0103] It should be noted that the way to establish gradient constraint pruning and prune the DTW path is: after completing the time alignment processing and amplitude standardization processing of the fundamental voltage sequence, and dividing into equal-length micro-sections according to a fixed time length, first calculate the gradient features of each micro-section, and then realize pruning by comparing the size relationship between the gradient feature difference of any two to-be-matched micro-sections (such as the mth micro-section of the user and the nth micro-section of the reference node) and the preset gradient threshold; if the gradient feature difference exceeds the preset gradient threshold, it indicates that this group of micro-sections has obvious waveform distortion due to welding machine harmonic interference, which does not meet the requirement of data stability for voltage trajectory similarity calculation, therefore, the gradient constraint pruning mechanism is established, and this group of micro-sections is excluded from the candidate range of subsequent DTW (dynamic time warping) path search, only the micro-sections with gradient feature difference within the preset gradient threshold range and relatively stable waveform are retained to participate in the DTW path search, so as to reduce the interference of waveform distortion caused by harmonics on the calculation result of voltage trajectory similarity;

[0104] Among them, the two micro-section screening conditions need to be executed cooperatively, not one of them;

[0105] Based on the DTW distance meeting the micro-section screening condition, the voltage trajectory similarity is calculated;

[0106] Preferably, the way to calculate the voltage trajectory similarity based on the selected DTW distance meeting the micro-section screening condition is: through the formula: The voltage trajectory similarity S is obtained;

[0107] Among them, is the global DTW distance of the micro-section screening condition, is the reference DTW distance corresponding to the historical data without harmonic interference (such as the welding machine is not started and stopped);

[0108] Based on the voltage trajectory similarity S, the strength of the electrical coupling of the transformer area is determined, if the voltage trajectory similarity S is higher than the preset voltage trajectory similarity threshold, it is determined that the user and the transformer area exist electrical coupling;

[0109] If the voltage trajectory similarity S is lower than the preset voltage trajectory similarity threshold, it is determined that the user and the transformer area do not exist electrical coupling;

[0110] It can be understood that the function of determining that the user and the transformer area exist electrical coupling is:

[0111] Function one, build the physical judgment reference of transformer area topology association, the electrical coupling determination directly verifies whether the user and the transformer area exist electrical connection relationship from the electrical physical law level through the phase synchronization and power mutual correlation characteristics of the voltage trajectory, and establishes the determination standard of whether the transformer area belongs to the same electrical network for the transformer area identification.

[0112] Effect two, quantifying the dynamic strength level of the association between the user and the transformer area, and the strength of the electrical coupling quantifies the closeness of the user to the transformer area: users close to the transformer are more strongly coupled, and non-linear loads can temporarily change the coupling characteristics due to harmonic interference, which helps to reduce false positives in complex load scenarios.

[0113] Effect three, supporting dynamic perception and updating of transformer area topology, in grid operation, events such as welding machine start-stop and switch position change will change the electrical coupling relationship in real time; real-time determination of electrical coupling can capture dynamic changes in topology: transient coupling fluctuations trigger interference verification (such as welding machine transient filtering), permanent coupling changes drive the local update of the topology of the space-time graph neural network, adapting to the scene needs of industrial transformer area high-frequency load disturbance and topology adjustment.

[0114] Embodiment two: as shown in the figure, a transformer area identification and user binding method based on a special transformer acquisition terminal, further comprising the following steps: Figure 1

[0115] Step three: if there is electrical coupling, conduct cross-correlation analysis combined with the pure power signal to obtain a comprehensive determination value, determine the association strength between the user and the transformer area, if the user and the transformer area are in strong association, if not, perform topology update verification;

[0116] The cross-correlation analysis method is as follows:

[0117] Based on the pure power signal, separate processing is performed to obtain the fundamental active power, 3rd harmonic reactive power, and 5th harmonic reactive power, and a user power component containing the fundamental active power, 3rd harmonic reactive power, and 5th harmonic reactive power is constructed;

[0118] Set different lag steps, calculate the cross-correlation coefficient of the user power component and the reference node power component at each lag step ;

[0119] Preferably, the cross-correlation coefficient is obtained by the formula: ;

[0120] It can be understood that the cross-correlation coefficient reflects the linear correlation between the kth power component on the user side and the kth power component on the reference node side at a time lag; the value range of the lag step is 0~20 steps, and the corresponding time interval of each step is 0.01 seconds, matching the sampling rate of 10 kHz of the special transformer terminal (0.01 seconds corresponding to 100 sampling points); step 0 represents no time lag, and step 20 represents a lag of 0.2 seconds, covering the typical time delay range of welding machine power fluctuations;

[0121] ​​Wherein, k=1, 3, 5, respectively represent the number of times of the active power, the 3rd harmonic reactive power, the 5th harmonic reactive power;

[0122] X k is the user power component, Y k is the reference node power component, respectively, the standard deviation of the kth power on the user side, the standard deviation of the kth power on the reference node side, the time offset;

[0123] is the covariance, quantifying the consistency of the linear trend of

[0124] It needs to be explained that the reference node has no electric welding machine load by default, and the harmonic power approaches 0;

[0125] The correlation coefficient of the user power component and the reference node power component and the phase difference of the fundamental wave voltage are summed to obtain the substation user similarity;

[0126] Preferably, the substation user similarity is obtained by the formula: ;

[0127] Wherein, respectively: the preset weight coefficient of the kth power component, the weight coefficient of the phase difference, the phase difference of the fundamental wave voltage between the user and the reference node;

[0128] The voltage trajectory similarity S and the substation user similarity are summed to obtain a comprehensive judgment value;

[0129] Preferably, the summing processing mode is: the comprehensive judgment value Pd is obtained by the equation:

[0130] Wherein, Both are preset dynamic weights, and satisfy: When the electric welding machine power is ≤50kW, α=0.5, β=0.5; when the electric welding machine power is >50kW, α=0.6, β=0.4;

[0131] It can be understood that the physical meaning of the comprehensive judgment value is: in the scene of the existence of electric welding machine start-stop harmonic interference in the industrial substation, by fusing the voltage trajectory similarity and the substation user similarity, the electrical physical coupling tightness and the load operation homology between the user and the substation are quantified;

[0132] ​​The voltage trajectory similarity is based on the phase consistency and gradient characteristics of the fundamental voltage, and reflects whether they are in the same physical connection nature of the electrical network; the transformer area user similarity is based on the cross-correlation coefficient of the pure power signal and the fundamental voltage phase difference, and embodies the synergy of the electric welding machine characteristic load in energy transmission and operation state. The comprehensive judgment value analyzes the real association of the user and the transformer area from the two dimensions of physical connection and energy synergy, provides a physical basis for distinguishing transient harmonic interference and real topology change, supports the judgment and dynamic update of transformer area attribution;

[0133] As shown in Figure 2 , based on the comprehensive judgment value, the association strength of the user and the transformer area is determined, and whether the user and the transformer area are in strong association is determined. If not, the topology update verification is performed;

[0134] If the user and the transformer area are in strong association, the user and the transformer area are associated and bound;

[0135] For example, the strong association threshold is set to 0.8 (based on 200 groups of user data confirmed transformer area attribution statistics), when the comprehensive judgment value of the user and the transformer area is 0.87, it is determined that they are in strong association; if it is only 0.61, the topology update verification process is triggered.

[0136] It can be understood that the purpose of associating and binding the user and the transformer area is:

[0137] Purpose one: supporting transformer area line loss accounting and governance, by clearly defining the electrical association boundary of the user and the transformer area, dividing the calculation range of transformer area power supply-user power consumption, providing real electrical connection basis for time period loss decomposition (such as loss anomaly of electric welding machine start-stop period) and loss reduction measures (such as harmonic governance, reactive power compensation), reducing the line loss problem caused by traditional file error.

[0138] Purpose two: beneficial to load characteristic analysis and dynamic regulation of distribution network, based on the binding relationship, the harmonic load and power fluctuation law of the user (such as industrial user with electric welding machine) can be traced directionally, supporting transformer area load prediction, overload warning and power quality governance, providing association benchmark for topology update verification, making the distribution network topology model and the actual electrical connection dynamic alignment, and ensuring the accuracy of fault location, power failure information reporting and other businesses.

[0139] Step four: if topology update verification is needed, the historical topology change event is obtained, the comprehensive judgment value and the topology change event are used to update the transformer area topology structure online through the space-time graph neural network, and the user's transformer area is determined;

[0140] The way to obtain the topology change event is:

[0141] When the switch position change event is detected, the synchronous check of the corresponding period of the electric welder start-stop state is performed, and if the electric welder is in the start-stop state when the event occurs, the time difference between the event and the load disturbance is calculated;

[0142] It needs to be explained that the switch state signal is collected through the SCADA system and the intelligent switch device, the hardware dry contact trigger is used, the sliding window difference and Kalman filtering algorithm are used to eliminate jitter interference, the current mutation synchronicity verification is combined, and the switch position change event is determined;

[0143] Based on the time difference between the event and the load disturbance, it is determined whether it is a topology change;

[0144] For example, the time difference is less than 50ms: it is determined that the transient disturbance is caused by the start-stop of the electric welder, and it is not a topology change; otherwise, the electric welder interference is excluded, and it is determined to be a topology change;

[0145] Among them, the way of updating the substation topology structure online through the space-time graph neural network is:

[0146] Preferably, the time graph neural network is constructed based on the characteristics of the industrial substation to determine the substation to which the user belongs;

[0147] Among them, the way of constructing the time graph neural network is:

[0148] S401, construct edge nodes and feature engineering;

[0149] Preferably, the nodes include: transformer nodes V T , namely the monitoring point of the low-voltage side of the substation transformer, and the features include the rated capacity and the short-circuit impedance;

[0150] User nodes V U ; the electric welder terminal of the special transformer, and the features include the comprehensive determination value and the load type (electric welder or ordinary load);

[0151] Topology event nodes V E ;

[0152] The edges include: electrical connection edges E el , event influence edges E e ; switch state change, DG access and other events, and the features include the event type and the occurrence time;

[0153] The feature engineering includes node features and edge features, and the node features are composed of the fundamental wave voltage amplitude, the line impedance, the comprehensive determination value and the historical topology label;

[0154] It needs to be explained that the historical topology label is calculated from the substation attribution probability of the past monitoring period;

[0155] The edge features are composed of the cross-correlation coefficient, the voltage trajectory similarity and the event type code;

[0156] S402, real-time detection and coding of topology change events;

[0157] Preferably, when an event is detected, a topology event node V E is generated, and the graph structure is updated: switch off: delete the corresponding electrical connection edge E el , add event impact edge E e and mark the affected users;

[0158] Welder start-stop: update the load characteristics (such as harmonic content change) of the corresponding user node V U , recalculate the electrical connection edge weight;

[0159] DG access: add DG node, calculate the electrical connection edge weight between DG node and transformer node;

[0160] It needs to be explained that the DG node refers to the electrical connection point of the distributed power supply access to the distribution network. When the DG node is added, the base wave power cross-correlation coefficient, the base wave voltage phase difference, and the power transmission time delay and other characteristics of the DG node and the transformer node are calculated, and the edge weight representing the electrical connection tightness is generated by fusion;

[0161] S403, establish an online inference framework of the space-time graph neural network;

[0162] Preferably, a relational graph convolution network is used to process different types of nodes and edges, aggregate neighbor node features through multi-layer convolution, calculate the time sequence dependence of topology changes through a gated recurrent unit, and realize the online inference framework through an incremental learning mechanism;

[0163] S404, dynamically determine the user's belonging substation based on the correlation probability and decision rules;

[0164] For each user node V U , the probability distribution of belonging to each substation transformer node V T is output through the graph neural network;

[0165] Wherein, T represents the substation, and U represents the user;

[0166] Preferably, the probability distribution of each substation transformer node V T is obtained through the equation: ;

[0167] It can be understood that the probability distribution of each substation transformer node V T ​​Quantitative description of transformer node electrical parameters, distribution interval reflects parameter fluctuation range, and system average result can be used for evaluating typical operation state of transformer node, providing risk evaluation basis based on probability statistics for distribution line loss prediction, voltage quality management and distribution network planning;

[0168] wherein, is an embedding vector of a user node, and Z is the number of transformer nodes in a transformer area; is an embedding vector of a transformer area Ta, and is an embedding vector of a transformer area Tb;

[0169] It should be noted that the user embedding vector includes the user voltage trajectory similarity S and the transformer area user similarity; the embedding vector of the transformer area includes the transformer area voltage trajectory reference similarity and the transformer area power reference similarity;

[0170] The embedding vector of the transformer area (including Ta and Tb) includes the transformer area voltage trajectory reference similarity and the transformer area power reference similarity;

[0171] wherein, the standard electrical characteristic template is extracted from the historical normal operation data of the transformer node by the same method as calculating the user side similarity, and the transformer area voltage trajectory reference similarity and the transformer area power reference similarity are extracted from the standard electrical characteristic template;

[0172] a and b are both numbers of transformer nodes in the transformer area;

[0173] wherein, the decision rule includes:

[0174] Decision rule one, strong association direct binding: if > 0.8, and the comprehensive judgment value > 0.7, it is judged that the user belongs to the transformer area ;

[0175] Decision rule two, conflict scene arbitration: when the difference between the probability distributions of multiple transformers is less than 0.1, a harmonic direction constraint is constructed to determine the attribution of the transformer area and the user;

[0176] For example, the way to construct the harmonic direction constraint is to calculate the third harmonic phase difference between the user and each transformer;

[0177] Select the transformer with the third harmonic phase difference less than 10 degrees and the probability distribution value maximum as the attribution.

[0178] Embodiment three: as shown in the figure, a transformer area identification and user binding system based on a special transformer acquisition terminal, further includes the following modules: Figure 3

[0179] ​​An interference analysis module is used for harmonic separation processing of high-frequency sampling data of the special transformer terminal, adopts a kurtosis index and time-frequency energy entropy fusion detection method to judge whether the welding machine start-stop causes harmonic interference, and if so, extracts the fundamental voltage component and pure power signal;

[0180] A coupling analysis module is used for calculating the fundamental local distance of the user and the reference node based on the fundamental voltage component, constructing a voltage trajectory similarity model against harmonic interference, obtaining the voltage trajectory similarity, and determining whether there is electrical coupling between the user and the transformer area;

[0181] An association judgment module is used for fusion dynamic weight power cross-correlation analysis combined with the pure power signal to obtain a comprehensive determination value, judge the association strength between the user and the transformer area, and if the user and the transformer area are in strong association, if not, perform topology update verification;

[0182] An update verification module is used for obtaining historical topology change events, updating the transformer topology structure online through a space-time graph neural network based on the comprehensive determination value and the topology change events, and determining the transformer area to which the user belongs if the topology update verification is needed.

[0183] The above describes one embodiment of the present application in detail, but the content described is only the preferred embodiment of the present application and cannot be considered as limiting the scope of the present application. Any equivalent changes and improvements made according to the scope of the present application should still belong to the scope of the present application.

Claims

1. A method for identifying a transformer area and binding a user based on a special variable acquisition terminal, characterized in that: The method comprises the following steps: Harmonic separation processing is performed on high-frequency sampling data of the special transformer terminal, a kurtosis index and a time-frequency energy entropy fusion detection method are used to determine whether the welding machine start-stop causes harmonic interference, if so, fundamental voltage components and pure power signals are extracted; Based on the fundamental voltage components, the fundamental local distance between the user and the reference node is calculated, a voltage trajectory similarity model resistant to harmonic interference is constructed, the voltage trajectory similarity is obtained, and it is determined whether the user and the transformer station are electrically coupled; If there is electrical coupling, cross-correlation analysis is performed in combination with the pure power signals to obtain a comprehensive determination value, and it is determined whether the user and the transformer station need to be checked for topology update based on the comprehensive determination value; If the topology update check is needed, historical topology change events are obtained, the comprehensive determination value and the topology change events are used to update the transformer station topology online through a space-time graph neural network to determine the transformer station to which the user belongs; The way to determine that the user and the transformer station are electrically coupled is: The fundamental voltage of the user and the reference node is collected, the fundamental voltage sequence of the user and the reference node is constructed respectively, the fundamental voltage sequence is processed for time sequence alignment and amplitude standardization, and the fundamental local distance is calculated; Based on the fundamental local distance, a DTW cumulative distance matrix is constructed, a phase consistency coefficient and a gradient feature are obtained and a micro-segment screening condition is constructed, the DTW distance meeting the micro-segment screening condition is selected, and the voltage trajectory similarity is calculated; Based on the voltage trajectory similarity, the strength of the electrical coupling of the transformer station is determined to obtain the determination result of whether the user and the transformer station are electrically coupled; The way to construct the micro-segment screening condition is: After the fundamental voltage sequence is processed for time sequence alignment and amplitude standardization, equal-length micro-segments are divided according to a fixed time length; The phase consistency coefficient and the gradient feature of each equal-length micro-segment are calculated; The micro-segment screening condition includes: When searching for a DTW path, microsegments of are preferred. wherein, respectively, the phase consistency value of the mth micro-segment of the user and the nth micro-segment of the reference node, is a preset phase threshold value, is a fundamental phase structure difference; If the gradient feature difference between the micro-segments is within a preset gradient threshold range, a gradient constraint pruning is established, and the DTW path is pruned.

2. The method for identifying a transformer area and binding a user based on a special collection terminal according to claim 1, characterized in that: The way to determine whether the welding machine start-stop causes harmonic interference is: The characteristic harmonic frequency band when the welding machine starts and stops is extracted through a band-pass filter set to obtain a harmonic component waveform; The kurtosis and time-frequency energy entropy of the harmonic component are obtained, and boundary interference analysis is performed to obtain a boundary interference distance; Based on the boundary interference distance, it is determined whether the welding machine start-stop causes harmonic interference, if so, the fundamental voltage components and the pure power signals are extracted, and if not, no processing is performed.

3. The method according to claim 2, wherein the method is characterized in that: The way to perform the boundary interference analysis is: Based on the kurtosis and time-frequency energy entropy of the harmonic component, a kurtosis-energy entropy feature vector is constructed; The historical kurtosis-energy entropy data of the welding machine start-stop is obtained, a Gaussian mixture model is established to set the harmonic interference boundary of the welding machine start-stop, and a harmonic interference boundary vector is constructed; The kurtosis-energy entropy feature vector and the harmonic interference boundary vector are calculated for Euclidean distance to obtain the boundary interference distance.

4. The method according to claim 1, characterized in that: The way to determine whether the user and the transformer station need to be checked for topology update is: The cross-correlation coefficient of the user power component and the reference node power component and the fundamental voltage phase difference are obtained, and the sum is calculated to obtain the transformer station user similarity; The voltage trajectory similarity and the transformer station user similarity are summed to obtain a comprehensive determination value. Based on the comprehensive judgment value, the association strength of the user and the transformer area is determined and analyzed to obtain a judgment result of whether the user and the transformer area need to be checked for topology update.

5. The method according to claim 4, wherein the method is characterized in that: The way of obtaining the cross-correlation coefficient of the user power component and the reference node power component is: Based on the pure power signal, the fundamental active power, 3rd harmonic reactive power and 5th harmonic reactive power are obtained through separation processing, and a user power component containing the fundamental active power, 3rd harmonic reactive power and 5th harmonic reactive power is constructed; Different hysteresis steps are set, and the cross-correlation coefficient of the user power component and the reference node power component under each hysteresis step is calculated.

6. The method according to claim 1, characterized in that: The way of determining the transformer area to which the user belongs is: Obtain historical topology change events, construct a time graph neural network based on the characteristics of industrial transformer areas, and determine the transformer area to which the user belongs.

7. The method according to claim 6, characterized in that: The way of constructing the time graph neural network is: Constructing edge nodes and feature engineering; Real-time detection and coding of topology change events are performed; An online inference framework of the space-time graph neural network is established; The transformer area to which the user belongs is dynamically determined based on the association probability and the decision rule.

8. The method according to claim 7, characterized in that: The way of dynamically determining the transformer area to which the user belongs based on the association probability and the decision rule is: For each user node, the probability distribution of belonging to each transformer node is output through the graph neural network; The decision rule includes: Strong association direct binding: determining the user and the transformer area based on the probability distribution and the comprehensive judgment value; Conflict scene arbitration: when the difference of the probability distribution of multiple transformers is in a preset range, a harmonic direction constraint is constructed to determine the ownership of the transformer area and the user.

Images