An electricity accounting management system based on artificial intelligence

Abstract

The present invention relates to the technical field of power management, and particularly relates to a power consumption accounting management system based on artificial intelligence, including: a dynamic power consumption accounting module: constructing a dynamic power calculation model based on the historical operation data of user equipment and outputting a theoretical power value; a multi-dimensional accounting difference analysis module: performing three-dimensional difference analysis on the theoretical power value and the measured power value of a metering terminal to generate a difference coefficient matrix at the device level; an abnormal area correlation and positioning module: generating a heat map of abnormal probabilities of an abnormal current propagation path and positioning the physical area of abnormal power consumption; an adaptive verification strategy generation module: configuring verification parameters according to the heat map of abnormal probabilities. The present invention can effectively identify highly concealed abnormal sources such as multi-point collaborative power theft and inverter failures, and can also output a heat map of probabilities and the main propagation path, improving the spatial accuracy and structural coherence of abnormal traceability, and breaking through the bottleneck that the existing methods cannot distinguish noise disturbances from substantial abnormalities.

Description

Technical Field

[0001] The present invention relates to the technical field of power management, and particularly to a power consumption accounting management system based on artificial intelligence. Background Art

[0002] With the large-scale access of distributed energy sources, energy storage systems and power electronic devices, the electricity consumption behavior on the user side shows highly dynamic, non-linear and strongly disturbed characteristics. The traditional electricity consumption accounting method based on a fixed model has been difficult to adapt to the complex operating environment. In the existing electricity management systems, the following technical bottlenecks generally exist:

[0003] On the one hand, current electricity consumption calculations mostly rely on the accumulation of active power under steady state or a fixed value model based on simplified assumptions, and cannot accurately reflect the changes in the operating states of devices under multiple working conditions. Especially when facing the charge and discharge pulses of energy storage devices, the reverse current at night of photovoltaic inverters or the harmonic distortion generated by power electronic devices, there are obvious calculation deviations.

[0004] On the other hand, the means for identifying abnormal electricity consumption mainly rely on static threshold alarms or historical comparison methods, and it is difficult to capture hidden abnormal events with weak periodicity, high-frequency disturbances or cross-node collaborative behaviors. At the same time, the existing verification mechanisms usually mainly rely on manual verification, lacking intelligent scheduling and dynamic tracking capabilities for high-risk areas, resulting in low efficiency of abnormal event tracing and poor positioning accuracy.

[0005] In addition, there is currently no effective association between the mainstream systems in electricity consumption accounting and the power grid topology, making it difficult to integrate abnormal characteristics with the physical connection relationships of devices, resulting in the fragmentation of the data layer and the physical layer, and seriously affecting the reliability of the overall decision-making closed-loop. Summary of the Invention

[0006] The present invention provides a power consumption accounting management system based on artificial intelligence, which integrates an intelligent power consumption accounting and abnormal positioning method that models historical operation data, perceives real-time electrical characteristics, and analyzes spatial topology structures, so as to achieve fine modeling of complex electricity consumption behaviors, accurate identification of abnormal propagation paths, and adaptive strategy verification of key areas, thereby improving the accuracy, responsiveness and intelligent level of the power management system.

[0007] A power consumption accounting management system based on artificial intelligence includes:

[0008] A dynamic power consumption accounting module: constructs a dynamic power consumption calculation model based on the historical operation data of user devices, and in real time collects the actual voltage waveforms, current harmonic components and power pulse sequences of current electricity-consuming devices, and outputs a theoretical power consumption value through the dynamic power consumption calculation model;

[0009] Multi - dimensional accounting difference analysis module: Conduct three - dimensional difference analysis on the theoretical power value and the measured power value of the metering terminal. The three - dimensional difference analysis includes time - domain cumulative deviation, frequency - domain energy offset, and pulse sequence similarity, and generates a difference coefficient matrix at the device level;

[0010] Abnormal area correlation and location module: Input the difference coefficient matrix into the spatial topology analysis network, combine the grid node impedance parameters and the user equipment location information, generate a heat map of abnormal probabilities of the abnormal current propagation path, and locate the physical area of abnormal power consumption;

[0011] Adaptive verification strategy generation module: Configure verification parameters according to the heat map of abnormal probabilities, including:

[0012] Enable the high - frequency power quality monitoring mode for high - probability abnormal areas;

[0013] Apply impedance perturbation tests to adjacent node devices.

[0014] Optionally, the dynamic power accounting module specifically includes:

[0015] Historical data feature mining: Conduct multi - dimensional decomposition processing on the historical operation data of user equipment, including: extracting the energy ratio of the steady - state component and the transient component of the voltage waveform using adaptive wavelet packet decomposition, establishing a correlation matrix between the current harmonic components and the equipment load rate through harmonic coupling analysis, and aligning the power pulse sequence patterns under different working conditions using the dynamic time warping algorithm;

[0016] Dynamic power calculation model construction: Input the processed historical operation data into the hybrid prediction network. The hybrid prediction network includes:

[0017] An LSTM time - series prediction unit based on the equipment aging curve, used to generate a basic power prediction value, a fully - connected network embedded with a frequency - domain attention mechanism to correct the prediction deviation caused by harmonic distortion, and a pulse feature compensator to dynamically adjust the prediction weight according to the real - time collected power pulse sequence;

[0018] Synchronously capture the zero - crossing distortion rate and phase jitter parameters of the voltage waveform, the 3 - 39th harmonic content and phase - angle distribution of the current harmonics, and the amplitude mutation gradient and time - interval entropy value of the power pulse through the heterogeneous data acquisition unit deployed at the edge side;

[0019] Theoretical value calculation: Input the real - time collected data into the dynamic power calculation model to obtain the theoretical power value.

[0020] Optionally, in the calculation of the theoretical power value, it includes performing:

[0021] Adaptive filtering processing based on the equipment operation stage to eliminate the measurement noise caused by environmental temperature and humidity;

[0022] Integrate the correlation features of voltage, current and pulse through spatiotemporal feature fusion algorithm;

[0023] The output includes the theoretical power value within the normal operating fluctuation range, and the theoretical power value is dynamically updated as the equipment ages.

[0024] Optionally, the multi-dimensional accounting difference analysis module specifically includes:

[0025] Time domain cumulative deviation calculation: The theoretical power value and the measured power value are compared in a sliding manner in a preset time window, and the dynamic time warping algorithm is used to align the asynchronously sampled power sequence, calculate the cumulative deviation in each window, and generate a time domain deviation vector;

[0026] Frequency domain energy offset detection: Decompose the current harmonic components of the theoretical value and the measured value using the frequency domain decomposition algorithm, calculate the energy spectrum density ratio, extract the energy offset index of each harmonic, and construct the frequency domain offset vector;

[0027] Pulse sequence similarity evaluation: Based on the structure matching algorithm, the distance distribution between the theoretical value and the measured pulse sequence is matched, the phase synchronization error of the pulse amplitude mutation point is calculated, the KL divergence of the pulse interval distribution is quantified, and the pulse similarity vector is generated;

[0028] Generation of difference coefficient matrix: The time domain deviation vector, frequency domain offset vector, and pulse similarity vector are tensor-concatenated, and dimension differences are eliminated through normalization processing weighted by feature importance, and a third-order difference coefficient matrix with the output dimension of [device number × timestamp × difference type] is output.

[0029] Optionally, the frequency domain energy offset detection specifically includes:

[0030] In the frequency domain analysis stage, the corresponding current harmonic components in the theoretical and measured electrical quantity sequences are first extracted, and wavelet packet decomposition is used to perform multi-scale frequency band analysis on each group of harmonic signals to extract the energy distribution characteristics within the preset sensitive frequency band (3kHz-15kHz). The sensitive frequency band is selected to cover the operating frequency range of typical power electronic equipment. After the extraction is completed, the energy density of the theoretical and measured data within the sensitive frequency band is quantitatively calculated respectively, and the offset index of each order of harmonics is extracted based on the relative offset degree of the two. The energy offset results of all orders are summarized to construct a frequency domain offset vector.

[0031] Optionally, in the pulse sequence similarity evaluation, the structure matching algorithm adopts the Hausdorff distance algorithm to perform position matching on the mutation point sets in the two sequences and identify the phase synchronization error.

[0032] Optionally, the abnormal area association positioning module specifically includes:

[0033] Power grid topology modeling: Construct a topology graph of the power grid node connection relationship according to the user equipment location information, mark the line impedance parameters between each node, and superimpose the reverse current constraint conditions of the distributed power access points in the topology graph to generate an electrified topology model including an impedance matrix and a node admittance matrix;

[0034] Abnormal propagation simulation: Map the difference coefficient matrix to the corresponding nodes of the power grid topology model; perform abnormal propagation deduction based on the graph neural network (GNN). The calculation of the abnormal propagation deduction includes:

[0035] i. Calculate the attenuation factor of the abnormal current according to the node impedance parameters;

[0036] ii. Capture cross-regional abnormal correlation features through the multi-head attention mechanism;

[0037] iii. Use the Monte Carlo method to simulate the diffusion path of the abnormal current in the topology network;

[0038] Probability distribution generation: Count the occurrence frequency of the abnormal current in each line during the simulated propagation, calculate the abnormal current residence probability value in combination with the line impedance parameters, generate an abnormal probability heat distribution map covering the entire network, and mark the set of suspicious lines whose probability values exceed the preset abnormal residence probability threshold;

[0039] Physical area positioning: Perform spatial clustering analysis on the abnormal probability heat distribution map to identify abnormal probability aggregation areas; delimit the physical boundaries of abnormal electricity consumption according to the user equipment location and the power grid topology connection relationship.

[0040] Optionally, the abnormal area association positioning module further includes outputting including suspicious device identifiers and the main abnormal propagation path, where:

[0041] The suspicious device identifiers are based on the power grid nodes connected by the set of suspicious lines, bind the power grid nodes to the actual user equipment to form a set of suspicious device identifiers, indicating potential abnormal power sources or affected terminals;

[0042] The main abnormal propagation path records the node paths and their sequences experienced in each round of propagation during the abnormal diffusion process simulated by the Monte Carlo method. Among all the simulated paths, count the occurrence frequency of each path, and select the path sequence with the highest cumulative occurrence frequency as the main abnormal propagation path. The output main abnormal propagation path sequence is a structured and ordered list of nodes, reflecting the main propagation trajectory of abnormal information in the power grid.

[0043] Optionally, the adaptive verification strategy generation module specifically includes:

[0044] When the abnormal probability value in a certain area of the abnormal probability heat distribution map exceeds the preset abnormal residence probability threshold, a monitoring mode switching instruction is sent to the metering terminal to which the area belongs, and the following operations are performed:

[0045] i. Increase the voltage / current sampling frequency to 5-10 times the original frequency;

[0046] ii. Synchronously enable the harmonic component real-time tracking mode to capture the phase mutations of 3-39th sub-harmonics;

[0047] iii. Deploy a transient event detector on the edge side to record the voltage sag / swell waveform segments;

[0048] The impedance perturbation test is performed as follows: For the adjacent node devices with the largest change in the abnormal probability gradient in the abnormal probability heat distribution map, apply multi-band impedance perturbations.

[0049] Optionally, the multi-band impedance perturbation includes:

[0050] i. Inject a 0.1-2 kHz swept test signal through a controllable current source;

[0051] ii. Theoretical impedance response calculation: Based on the topological model and impedance matrix , calculate the theoretical response impedance spectrum;

[0052] Obtain the measured impedance response: Record the voltage of each node after injecting the test signal to obtain the measured response impedance spectrum;

[0053] iii. Abnormal offset determination: Calculate the Euclidean distance between the theoretical response impedance spectrum and the measured response impedance spectrum;

[0054] iv. Compare the Euclidean distances of the measured impedance spectrum and the theoretical impedance spectrum, calculate the abnormal node offset percentage. When the abnormal node offset percentage exceeds the second threshold (12%-18%), it is marked as an "impedance anomaly associated device". The impedance offset is the relative deviation of the measured impedance value Zmeas from the theoretically expected impedance Zref.

[0055] Advantages of the present invention:

[0056] In the present invention, by introducing a historical feature mining mechanism based on wavelet decomposition, harmonic coupling analysis and dynamic time warping, a hybrid power model integrating LSTM time series prediction, frequency domain attention correction and pulse compensation is constructed, which improves the dynamic response modeling ability for complex electricity consumption behaviors (such as energy storage device charging and discharging, variable load fluctuations). Compared with the traditional static power model, the proposed method can accurately reflect the power fluctuation laws under equipment aging, harmonic pollution and non-steady state operation conditions, realize high-precision theoretical power output with self-adaptive correction ability, and provide a traceable dynamic benchmark for subsequent difference analysis.

[0057] The present invention proposes a graph neural network anomaly propagation modeling method that integrates a difference coefficient matrix and a power grid topology structure. By introducing an impedance attenuation factor, a multi-head attention mechanism, and Monte Carlo diffusion simulation, an anomaly recognition system that synergistically combines electrical attributes and data-driven algorithms is constructed. This mechanism can not only effectively identify highly concealed anomaly sources such as multi-point collaborative power theft and inverter failures, but also output a probability heat map and the main propagation path, improving the spatial accuracy and structural coherence of anomaly tracing, and breaking through the bottleneck that existing methods cannot distinguish between noise disturbances and actual anomalies.

[0058] Based on the anomaly area positioning results, the present invention further constructs a gradient-based intelligent verification strategy system. Through the conditional activation of a high-frequency power quality monitoring mode and a multi-band impedance perturbation injection mechanism, refined multi-dimensional verification of key areas is achieved. It can dynamically adjust the sampling frequency, monitoring dimension, and detection intensity to ensure that verification resources are allocated according to risk priority, improving the overall verification efficiency and positioning accuracy of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] To more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only those of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0060] Figure 1 It is a schematic diagram of the management system module for the embodiments of the present invention;

[0061] Figure 2 It is a schematic diagram of the adaptive verification strategy generation module for the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0062] The present invention will be described in detail below in conjunction with the drawings and specific embodiments. At the same time, it should be noted here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments. For some well-known technologies, those skilled in the art can also adopt other alternative methods for implementation; and the drawings are only for more specifically describing the embodiments, and are not intended to specifically limit the present invention.

[0063] It should be noted that when referring to "an embodiment", "embodiments", "exemplary embodiments", "some embodiments", etc. in the specification, it indicates that the described embodiments may include specific features, structures, or characteristics, but not necessarily every embodiment includes such specific features, structures, or characteristics. Additionally, when combining embodiments to describe specific features, structures, or characteristics, implementing such features, structures, or characteristics in combination with other embodiments (whether explicitly described or not) should be within the knowledge of those skilled in the relevant art.

[0064] Generally, terms can be understood at least in part from their use in context. For example, depending at least in part on the context, the term "one or more" as used herein can be used to describe any feature, structure, or property in a singular sense, or can be used to describe a combination of features, structures, or properties in a plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey a set of exclusive factors, but rather can alternatively, depending at least in part on the context, allow for the existence of other factors that are not necessarily explicitly described.

[0065] As Figure 1 - Figure 2 shown, an artificial intelligence-based electricity quantity accounting management system includes:

[0066] Dynamic electricity quantity accounting module: Based on the historical operation data of user equipment, a dynamic electricity quantity calculation model is constructed, and the actual voltage waveform, current harmonic component, and power pulse sequence of the current power-consuming equipment are collected in real time, and the theoretical electricity quantity value is output through the dynamic electricity quantity calculation model;

[0067] Multi-dimensional accounting difference analysis module: Conducts three-dimensional difference analysis on the theoretical electricity quantity value and the measured electricity quantity value of the metering terminal. The three-dimensional difference analysis includes time-domain cumulative deviation, frequency-domain energy offset, and pulse sequence similarity, and generates a difference coefficient matrix at the device level;

[0068] Abnormal area correlation and location module: Inputs the difference coefficient matrix into the spatial topology analysis network, combines the grid node impedance parameters with the user equipment location information, generates a heat map of the abnormal probability of the abnormal current propagation path, and locates the abnormal electricity consumption physical area;

[0069] Adaptive verification strategy generation module: Configures verification parameters according to the heat map of the abnormal probability, including:

[0070] Enables the high-frequency power quality monitoring mode for high-probability abnormal areas;

[0071] Applies impedance perturbation tests to adjacent node devices.

[0072] The dynamic electricity quantity accounting module specifically includes:

[0073] 1. Historical data feature mining: Conducts multi-dimensional decomposition processing on the historical operation data of user equipment, including:

[0074] 1.1 Steady-state and transient energy ratio extraction: Adopts an adaptive wavelet packet decomposition algorithm to extract the steady-state component of the voltage waveform and the transient component , and calculates their energy ratios:

[0075] , is the energy proportion of the steady-state component in the voltage waveform, reflecting the energy ratio of the periodic and continuously stable part in the voltage signal. is the energy proportion of the transient component in the voltage waveform, reflecting the energy proportion of the non-stationary components generated due to switching, disturbances, etc.

[0076] 1.2. Construction of harmonic-load rate correlation matrix: Using the harmonic coupling analysis algorithm, establish the correlation matrix of the current harmonic components and the equipment load rate : : , represents the th harmonic current component, is the equipment load rate;

[0077] 1.3. Dynamic alignment of pulse sequences: Use the dynamic time warping (DTW) algorithm to align the power pulse sequences under different working conditions : , where respectively represent the power pulse sequences under the 𝑖-th and 𝑗-th historical working conditions.

[0078] 2. Construction of the electricity quantity calculation model: Input the processed historical data into the hybrid prediction network, including:

[0079] 2.1. Time series prediction unit: An LSTM network established based on the equipment aging curve, input the historical feature vector , and output the basic electricity quantity prediction value : ;

[0080] 2.2. Frequency domain attention correction module: A fully connected network embedded with the frequency domain attention mechanism, correct , and the output after correction is: , FC-Attn represents the fully connected network with the frequency domain attention mechanism (Fully Connected + Attention);

[0081] 2.3. Pulse feature compensator: Calculate the compensation factor according to the currently collected power pulse sequence , and adjust the output: .

[0082] 3. Real-time data fusion: Synchronously capture the following features through the edge-side heterogeneous data acquisition module:

[0083] The zero-crossing distortion rate of the voltage waveform and the phase jitter parameter ;

[0084] 3 - 39th sub - content of current harmonics and phase - angle distribution , is the phase - angle of the th harmonic;

[0085] Amplitude mutation gradient of power pulse and time - interval entropy value .

[0086] 4. Theoretical value calculation: Input the above - mentioned real - time acquisition features into the dynamic power calculation model and execute:

[0087] 4.1. Adaptive filtering: Apply an adaptive filter according to the current operation stage and environmental conditions (temperature and humidity) of the device to filter out noise interference: , represents the environmental - noise adaptive filter;

[0088] 4.2. Spatiotemporal feature fusion calculation: Use the fusion function to integrate features such as voltage, current, and power pulse to generate a fused feature vector: ;

[0089] 4.3. Output the theoretical power value: Based on the above - mentioned fused feature vector, output the theoretical power prediction value including the dynamic fluctuation range : , represents the hybrid prediction network.

[0090] Predictor receives the comprehensive feature vector obtained through the fusion function , calls the constructed dynamic power model structure, and the output result combines the dynamic power theoretical values of time - series, frequency - domain, and pulse features, that is , and the specific detailed process is as follows:

[0091] 4.3.1. Feature access stage: Receive the fused feature vector integrated by , and this vector integrates: Voltage, current, and power - pulse signals after adaptive filtering;

[0092] Auxiliary features such as spatial phase jitter, harmonic phase distribution, and pulse - interval entropy;

[0093] Context information characterizing the device aging degree and operation stage.

[0094]

[0095] 4.3.2. Main process of model prediction: Call the constructed hybrid prediction network model, mainly including:

[0096] ​a) LSTM Time Series Prediction Sub - module:

[0097] Utilize historical feature data and current fusion features ;

[0098] Output the predicted value of the basic power consumption at the current stage of the device , reflecting the aging trend and periodic load characteristics.

[0099] b) Frequency - domain Attention Correction Sub - module:

[0100] Input and the harmonic load matrix ;

[0101] Weight and correct the error caused by harmonic distortion, and output .

[0102] c) Pulse Feature Compensator:

[0103] Model the current power pulse features (such as mutation gradient, interval entropy), and output the pulse compensation term , correcting the fluctuations caused by sudden loads.

[0104] 4.3.3. Final Theoretical Power Output: Integrate the results of the above three parts to generate the final theoretical power prediction value : .

[0105] The multi - dimensional accounting difference analysis module specifically includes:

[0106] 1. Time - domain Cumulative Deviation Calculation:

[0107] 1.1. Compare the theoretical power value with the measured power value using a sliding window for comparison, where is the length of the sliding time window;

[0108] 1.2. Align the asynchronous - sampled power sequences using dynamic time warping (DTW) to obtain the aligned sequence: ;

[0109] 1.3. Calculate the cumulative deviation amount within each window :

[0110] , represents the time index within the sliding window, is the cumulative deviation amount corresponding to time 𝑡;

[0111] 1.4. Combine the deviation amounts of all windows into a time - domain deviation vector : ; where, is the theoretical power value, is the measured power value, is the power sequence aligned by DTW.

[0112] 2. Frequency-domain energy offset detection:

[0113] 2.1. Perform wavelet packet decomposition on the and current harmonic components respectively, and extract the frequency range kHz, 15kHz];

[0114] 2.2. Calculate the energy spectral density ratio in this frequency band : , is the energy spectral density of the measured harmonic current signal in the frequency band 𝑓, represents the energy spectral density of the theoretical harmonic current signal in the frequency band 𝑓;

[0115] Extract the energy offset index of each harmonic order , and form the frequency-domain offset vector :

[0116] , where, is the th harmonic energy spectral density in the frequency band , are the theoretical and measured harmonic current signals, represents the th harmonic frequency-domain energy offset index.

[0117] 3. Pulse sequence similarity evaluation:

[0118] 3.1. Define the theoretical pulse sequence and the measured pulse sequence ;

[0119] 3.2. Use the Hausdorff distance to match the set of pulse amplitude mutation points in the theoretical pulse sequence with the set of pulse amplitude mutation points in the measured pulse sequence :

[0120] , represents the Hausdorff distance, which is used to measure the maximum and minimum distance difference between two point sets (the theoretical and measured pulse mutation point sets), is the set of mutation points extracted from the theoretical pulse sequence, is the set of mutation points extracted from the measured pulse sequence, is any mutation point in the theoretical sequence, is any mutation point in the measured sequence, represents the point and the point The Euclidean distance between them. sup is the supremum (i.e., "maximum value"), and the maximum value is taken in this set. inf is the infimum (i.e., "minimum value"), and the minimum value is taken in this set. represents taking the larger of the maximum and minimum distances in two directions as the final symmetry distance metric.

[0121] 3.3. Calculate the phase synchronization error of mutation points , First, extract the position points where the amplitude mutates from the theoretical pulse sequence and the measured pulse sequence respectively. These position points represent typical electrical events such as equipment start / stop, current impact, or state switching. The extraction method can be based on the power change rate or completed, and the output is two sets of time points, corresponding to the mutation moments in the theoretical sequence and the measured sequence respectively. Then, establish a one-to-one correspondence between these two sets of mutation points. Adopt a matching strategy based on the minimum time distance, that is, pair each mutation point in the theoretical sequence with the mutation point with the closest time in the measured sequence. After completing the pairing of mutation points, calculate the time difference between each pair of mutation points. To convert the time difference into a periodic and comparable phase quantity, introduce the average period of the pulse sequence as the normalization scale. That is, divide the time difference of each pair of mutation points by the period length and multiply by the angular range corresponding to a complete period to map it to an angular quantity, so as to obtain the relative phase offset value of each pair of mutation points. Finally, statistically summarize the phase offset amounts of all mutation point pairs and take the average value as the overall phase synchronization error to reflect the synchronization degree between the mutation behavior predicted theoretically and the actual observation in the time dimension.

[0122] 3.4. Statistically analyze the pulse interval probability distributions of theory and measurement , calculate the KL divergence:

[0123] , is the Kullback-Leibler divergence between the two;

[0124] 3.5. Construct the pulse similarity vector : , where, represents the sets of theoretical and measured pulse mutation points, represents the phase synchronization error, is the KL divergence of the pulse interval distribution.

[0125] 4. Generation of the coefficient of variation matrix:

[0126] 4.1. Concatenate the above three types of features into a combined feature tensor: ;

[0127] 4.2. Normalize and weight the tensor (using feature importance as the weight :

[0128] , represents the \(i\)-th type of differential feature (such as time domain, frequency domain, pulse), is the weighting factor of the \(i\)-th type of differential feature, is the weighted differential feature tensor after normalization;

[0129] 4.3. Output the final differential coefficient third-order matrix : ;

[0130] where, is the device number index, \(t\) is the timestamp index, \(c\) is the differential type (time domain, frequency domain, pulse), is the concatenated differential tensor, is the feature weight factor, is the normalized differential tensor.

[0131] The abnormal area correlation positioning module specifically includes:

[0132] 1. Power grid topology modeling:

[0133] 1.1. Construct an electrical topology diagram: Based on the geographical location information of user equipment and the distribution structure, establish a topological diagram of node connection relationships: , where, represents the node set (equipment, transformer, access point), represents the electrical connection edge;

[0134] 1.2. Line parameter annotation: Each edge is associated with an impedance value , and an impedance matrix is constructed: , represents the line impedance from node to ;

[0135] 1.3. Reverse current constraint modeling: For the access point nodes with distributed power sources , define the reverse current boundary condition: , represents the node set with distributed power sources, represents the reverse current of node , represents the node Reverse current threshold;

[0136] 1.4. Admittance matrix generation: Derive the nodal admittance matrix from the topological structure and impedance relationship , forming an electrified topological model: .

[0137] 2. Abnormal propagation simulation:

[0138] 2.1. Differential data mapping: Map the differential coefficient matrix to the power grid nodes: , this expression represents extracting the differential tensor slice of the 𝑣-th node under all times and all differential types, represents the input feature vector of node 𝑣 in the power grid topological graph, which is used for abnormal propagation simulation in the subsequent graph neural network. It integrates all differential features related to this node, and Agg() represents the "aggregation operation" for compressing the three-dimensional tensor into a one-dimensional vector;

[0139] 2.2. Graph neural network propagation modeling: Based on the graph neural network propagation mechanism, calculate the abnormal state vector of each node , and the propagation update rule is as follows:

[0140] i. Attenuation factor calculation: , the attenuation factor reflects the propagation attenuation degree of abnormal current between nodes, represents the propagation attenuation factor of node , represents the impedance attenuation coefficient;

[0141] ii. Multi-head attention fusion:

[0142] ; where, is the updated feature representation of node in the -th layer of the graph neural network (i.e., the state vector after this round of propagation), is the feature vector of the adjacent node in the -th layer, representing its previous round of state information, is the set of adjacent nodes of node , representing the power grid nodes directly connected to it (determined according to the topological structure), is the attention weight of the -th attention head for the edge , reflecting the information influence degree of node on node under this channel, obtained through softmax normalization, is the The linear transformation matrix corresponding to each attention head, which is used to map the input features to the attention space of this channel. The number of attention heads in the multi-head attention mechanism. Each attention head learns a different way of weighting neighbor features, improving the model's ability to model various relationships in complex graph structures. Indicates to concatenate all the features output by the attention heads to form the final feature representation of the node .

[0143] iii. Abnormal diffusion simulation: Repeat the above propagation process times, introduce the Monte Carlo sampling strategy, simulate each initial perturbation, and record the possible propagation paths of abnormal current. Among them, is the input feature vector of node (derived from the coefficient of variation matrix), represents the node state of the th layer, represents the set of adjacent nodes of node , is the simulation round.

[0144] 3. Probability distribution generation:

[0145] 3.1. Propagation frequency statistics: For each edge in the times of simulation, count the frequency of abnormal propagation and perform normalization statistics: ;

[0146] 3.2. Residence probability calculation: Introduce the line impedance weight and calculate the abnormal residence probability: ;

[0147] 3.3. Heat distribution generation: Aggregate the residence probabilities of all lines to form an abnormal probability heat map , and screen the set of suspicious lines that exceed the abnormal residence probability threshold : , where is the frequency of abnormal propagation of line , is the propagation frequency, is the propagation residence probability, is the impedance correction coefficient, represents the abnormal residence probability threshold, with a value of 65% - 70%, is the abnormal probability heat map, is the set of suspicious lines.

[0148] 4. Physical area positioning:

[0149] 4.1. Spatial clustering analysis: Based on the distribution of suspicious lines in the abnormal heat map, perform clustering division on its topological structure to identify regional subgraphs with abnormal central tendency: , denote the topological subgraph of the th clustering region;

[0150] 4.2. Boundary delimitation and equipment mapping: Combine the grid connection relationship and the geographical location of user equipment to map the abnormal area to the specific physical coordinate space and approximate the area boundary with a polygon;

[0151] 4.3. Output of location report: Output a comprehensive report including the following content:

[0152] Set of suspicious device identifiers ;

[0153] Main path sequence of abnormal propagation .

[0154] The abnormal area associated location module also includes outputting suspicious device identifiers and the main path of abnormal propagation, where:

[0155] The suspicious device identifiers are based on the grid nodes connected by the set of suspicious lines, bind the grid nodes to the actual user equipment, form a set of suspicious device identifiers, and indicate potential abnormal power sources or affected terminals;

[0156] The main path of abnormal propagation records the node paths and their orders experienced in each round of propagation during the abnormal diffusion process of Monte Carlo simulation. Among all the simulation paths, count the occurrence frequency of each path, and select the path sequence with the highest cumulative occurrence frequency as the main path of abnormal propagation. The output main path sequence of abnormal propagation is a structured and ordered list of nodes, reflecting the main propagation trajectory of abnormal information in the power grid.

[0157] The adaptive verification strategy generation module specifically includes:

[0158] When the abnormal probability value in a certain area of the abnormal probability heat distribution map exceeds the preset abnormal residence probability threshold, send a monitoring mode switching instruction to the metering terminal belonging to this area and execute:

[0159] i. Increase the voltage / current sampling frequency to 5 - 10 times the original frequency;

[0160] ii. Synchronously enable the harmonic component real-time tracking mode to capture the phase mutation of 3 - 39 times of sub-harmonics;

[0161] iii. Deploy a transient event detector on the edge side to record voltage sag / swell waveform segments;

[0162] Impedance perturbation test execution: Apply multi - band impedance perturbations to the adjacent node devices with the largest change in the anomaly probability gradient in the anomaly probability heat distribution map.

[0163] The multi - band impedance perturbations include:

[0164] i. Inject a 0.1 - 2 kHz swept - frequency test signal through a controllable current source;

[0165] ii. Theoretical impedance response calculation: Based on the topological model and impedance matrix , calculate the theoretical response impedance spectrum:

[0166] ;

[0167] Obtain the measured impedance response: Record the voltage of each node after injecting the test signal to obtain the measured response impedance spectrum: ;

[0168] iii. Anomaly offset determination: Calculate the Euclidean distance between the theoretical response impedance spectrum and the measured response impedance spectrum;

[0169] iv. Compare the Euclidean distance between the measured impedance spectrum and the theoretical impedance spectrum, calculate the anomaly node offset percentage. When the anomaly node offset percentage exceeds the second threshold (12% - 18%), mark it as an "impedance anomaly - related device". The impedance offset is the relative deviation of the measured impedance value Zmeas from the theoretically expected impedance Zref.

[0170] Through the implementation of refined and enhanced on - line verification operations on the metering terminals and key node devices in high - risk areas, the accurate identification and technical closed - loop verification of suspected abnormal electricity consumption behaviors are achieved.

[0171] The high - frequency monitoring part is used to improve the time and frequency resolution of anomaly capture, so as to identify micro - perturbations and harmonic phase mutations that cannot be detected under traditional sampling;

[0172] The impedance perturbation test part realizes the accurate identification of the deviation of the physical - layer electrical properties of suspicious nodes by comparing the offset between the injection - response characteristics and the theoretical model.

[0173] Through the improvement of the sampling frequency and harmonic tracking, the detection ability of abnormal characteristics such as transient perturbations, voltage distortion, and harmonic phase mutations is enhanced, enabling the system to have the ability of rapid perception and diagnosis of sudden and high - frequency anomalies. Through multi - band perturbation injection and impedance response comparison, it can accurately judge whether there are electrical characteristic anomalies or physical access deviations in the equipment, avoiding misjudgment based solely on statistical indicators and enhancing the physical credibility of positioning.

[0174] The present invention covers any alternatives, modifications, equivalent methods and solutions made to the essence and scope of the present invention. For the public to have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, and those skilled in the art can fully understand the present invention without the description of these details. In addition, well-known methods, processes, procedures, components and circuits, etc. are not described in detail to avoid unnecessary confusion to the essence of the present invention.

[0175] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. An electric power accounting and management system based on artificial intelligence, characterized in that: include: Dynamic power accounting module: builds a dynamic power calculation model based on the historical operation data of the user equipment, collects the actual voltage waveform, current harmonic components and power pulse sequence of the current power-consuming equipment in real time, and outputs the theoretical power value through the dynamic power calculation model, specifically including: Historical data feature mining: Multi-dimensional decomposition and processing of historical operation data of user equipment, including: using adaptive wavelet packet decomposition to extract the energy proportion of steady-state components and transient components of the voltage waveform, establishing the correlation matrix between current harmonic components and equipment load rate through harmonic coupling analysis, and using dynamic time warping algorithm to align power pulse sequence patterns under different working conditions; Dynamic power calculation model construction: The processed historical operation data is input into the hybrid prediction network, which includes: The LSTM timing prediction unit based on the equipment aging curve is used to generate the basic power prediction value. The fully connected network embedded with the frequency domain attention mechanism corrects the prediction deviation caused by harmonic distortion. Based on the pulse feature compensator, the prediction weight is dynamically adjusted according to the power pulse sequence collected in real time. The heterogeneous data acquisition units deployed on the edge synchronously capture the zero-crossing distortion rate and phase jitter parameters of the voltage waveform, the 3rd to 39th order content and phase angle distribution of the current harmonics, the amplitude mutation gradient and time interval entropy of the power pulse sequence; Theoretical value calculation: input the real-time collected data into the dynamic power calculation model to obtain the theoretical power value; Multi-dimensional accounting difference analysis module: performs three-dimensional difference analysis on the theoretical power value and the actual power value measured by the metering terminal, wherein the three-dimensional difference analysis includes time domain cumulative deviation, frequency domain energy offset and pulse sequence similarity, and generates a device-level difference coefficient matrix; Abnormal area association positioning module: input the difference coefficient matrix into the spatial topology analysis network, combine the grid node impedance parameters and the user equipment location information, generate the abnormal probability thermal distribution map of the abnormal current propagation path, and locate the abnormal power consumption physical area; Adaptive verification strategy generation module: configures verification parameters according to the abnormal probability thermal distribution map, including: Enable high-frequency power quality monitoring mode for high-probability abnormal areas; Apply impedance disturbance test to adjacent node devices.

2. The power accounting and management system based on artificial intelligence according to claim 1 is characterized in that: The theoretical power value calculation includes executing: Adaptive filtering based on the equipment operation phase can eliminate the measurement noise caused by ambient temperature and humidity; Integrate the correlation features of voltage, current and pulse through spatiotemporal feature fusion algorithm; The output includes the theoretical power value within the normal operating fluctuation range, and the theoretical power value is dynamically updated as the equipment ages.

3. The power accounting and management system based on artificial intelligence according to claim 1 is characterized in that: The multi-dimensional accounting difference analysis module specifically includes: Time domain cumulative deviation calculation: The theoretical power value and the measured power value are compared in a sliding manner in a preset time window, and the dynamic time warping algorithm is used to align the asynchronously sampled power sequence, calculate the cumulative deviation in each window, and generate a time domain deviation vector; Frequency domain energy offset detection: Decompose the current harmonic components of the theoretical value and the measured value using the frequency domain decomposition algorithm, calculate the energy spectrum density ratio, extract the energy offset index of each harmonic, and construct the frequency domain offset vector; Pulse sequence similarity evaluation: Based on the structure matching algorithm, the distance distribution between the theoretical value and the measured pulse sequence is matched, the phase synchronization error of the pulse amplitude mutation point is calculated, the KL divergence of the pulse interval distribution is quantified, and the pulse similarity vector is generated; Generation of difference coefficient matrix: The time domain deviation vector, frequency domain offset vector, and pulse similarity vector are tensor-concatenated, and dimension differences are eliminated through normalization processing weighted by feature importance, and a third-order difference coefficient matrix with the output dimension of [device number × timestamp × difference type] is output.

4. The power accounting and management system based on artificial intelligence according to claim 3 is characterized in that: The frequency domain energy offset detection specifically includes: In the frequency domain analysis stage, the corresponding current harmonic components in the theoretical and measured electrical quantity sequences are first extracted, and multi-scale frequency band analysis is performed on each group of harmonic signals using wavelet packet decomposition to extract the energy distribution characteristics within a preset sensitive frequency band. The sensitive frequency band is selected to cover the operating frequency range of typical power electronic equipment. After the extraction is completed, the energy density of the theoretical and measured data within the sensitive frequency band is quantitatively calculated, and the offset index of each order of harmonics is extracted based on the relative offset degree of the two. The energy offset results of all orders are summarized to construct a frequency domain offset vector.

5. The power accounting and management system based on artificial intelligence according to claim 3 is characterized in that: In the pulse sequence similarity evaluation, the structure matching algorithm adopts the Hausdorff distance algorithm to perform position matching on the mutation point sets in the two sequences and identify the phase synchronization error.

6. The power accounting and management system based on artificial intelligence according to claim 1 is characterized in that: The abnormal area association positioning module specifically includes: Grid topology modeling: Build a grid node connection relationship topology map based on user equipment location information, annotate the line impedance parameters between nodes, superimpose the reverse current constraint conditions of the distributed power access point on the topology map, and generate an electrification topology model including the impedance matrix and the node admittance matrix; Abnormal propagation simulation: Map the difference coefficient matrix to the corresponding nodes of the power grid topology model; perform abnormal propagation deduction based on the graph neural network, and the calculation of abnormal propagation deduction includes: i. Calculate the attenuation factor of abnormal current according to the node impedance parameters; ii. Capture cross-region abnormal correlation features through multi-head attention mechanism; iii. Monte Carlo method is used to simulate the diffusion path of abnormal current in the topological network; Probability distribution generation: Count the frequency of abnormal currents in each line in the simulated propagation, calculate the abnormal current residence probability value based on the line impedance parameters, generate an abnormal probability thermal distribution map covering the entire network, and mark the suspicious line set whose probability value exceeds the preset abnormal residence probability threshold; Physical area positioning: perform spatial clustering analysis on the abnormal probability thermal distribution map to identify abnormal probability clustering areas; define the physical boundaries of abnormal power consumption based on the user equipment location and the grid topology connection relationship.

7. The power accounting and management system based on artificial intelligence according to claim 6 is characterized in that: The abnormal area association positioning module also includes outputs including suspicious device identification and abnormal propagation main path, wherein: The suspicious device identification is based on the grid node connected to the suspicious line set, and the grid node is bound to the actual user device to form a suspicious device identification set, indicating a potential abnormal power source head or affected terminal; The main path of abnormal propagation is achieved by recording the node paths and their order experienced in each round of propagation during the abnormal diffusion process of Monte Carlo simulation. Among all simulated paths, the frequency of occurrence of each path is counted, and the path sequence with the highest cumulative occurrence frequency is selected as the main path of abnormal propagation. The output main path sequence of abnormal propagation is a structured and ordered node list, which reflects the main propagation trajectory of abnormal information in the power grid.

8. The power accounting and management system based on artificial intelligence according to claim 1 is characterized in that: The adaptive verification strategy generation module specifically includes: When the abnormal probability value of a certain area in the abnormal probability thermal distribution map exceeds the preset abnormal residence probability threshold, a monitoring mode switching instruction is sent to the metering terminal to which the area belongs to execute: i. Increase the voltage / current sampling frequency to 5-10 times the original frequency; ii. Synchronously enable the real-time tracking mode of harmonic components to capture the 3rd to 39th order harmonic phase mutations; iii. Deploy transient event detectors on the edge side to record voltage sag / swell waveform segments; Impedance perturbation test execution: Apply multi-band impedance perturbation to the adjacent node devices with the largest abnormal probability gradient change in the abnormal probability thermal distribution diagram.

9. The power accounting and management system based on artificial intelligence according to claim 8 is characterized in that: The multi-band impedance disturbance includes: i. Inject 0.1-2kHz swept frequency test signal through controllable current source; ii. Theoretical impedance response calculation: based on topological model and impedance matrix , calculate the theoretical response impedance spectrum; Acquisition of measured impedance response: Record the voltage of each node after the test signal is injected to obtain the measured response impedance spectrum; iii. Abnormal offset determination: Calculate the Euclidean distance between the theoretical response impedance spectrum and the measured response impedance spectrum; iv. Compare the Euclidean distance between the measured impedance spectrum and the theoretical impedance spectrum, calculate the abnormal node offset percentage, and when the abnormal node offset percentage exceeds the second threshold, mark it as "abnormal impedance associated device".

Images