Regional electricity utilization transaction monitoring method and system based on space-time gridding

Abstract

The invention discloses a regional power utilization transaction monitoring method and system based on space-time gridding, and relates to the related field of smart power grids, and the method comprises the steps: setting a space division rule and a time slicing rule, and constructing a multi-level space-time grid of a target monitoring region; for each space-time grid, calculating a power utilization characteristic vector in a corresponding time slice, the power utilization characteristic vector comprising a load total quantity characteristic, a load curve form characteristic and a power utilization volatility characteristic; performing multi-dimensional comparison on the power utilization characteristic vector of the target space-time grid to obtain a multi-dimensional deviation degree; and performing power utilization transaction index analysis according to the multi-dimensional deviation degree, generating a grading transaction result according to a preset threshold value, and performing transaction grading early warning. According to the invention, the technical problem of insufficient accuracy of existing regional power utilization transaction monitoring is solved, and the technical effect of improving the transaction monitoring accuracy is achieved.

Description

Technical Field

[0001] This application relates to the field of smart grids, and in particular to a method and system for monitoring regional power consumption anomalies based on spatiotemporal grids. Background Technology

[0002] Accurate monitoring and anomaly identification of regional electricity consumption behavior are crucial for ensuring the safe operation of the power grid and improving load management efficiency. Currently, methods mainly rely on statistical analysis of overall regional electricity consumption data or threshold alarms based on fixed zones and time windows. Existing methods struggle to precisely capture local differences in electricity consumption patterns within a given space and their dynamic evolution over time, resulting in insensitivity to concealed, regional, and time-varying electricity consumption anomalies, and a high likelihood of missed or false alarms.

[0003] Currently, the relevant technologies for monitoring regional power consumption anomalies suffer from insufficient accuracy. Summary of the Invention

[0004] This application provides a method and system for monitoring regional power consumption anomalies based on spatiotemporal gridding. It constructs a multi-level spatiotemporal grid for the target monitoring area by setting spatial division and time slicing rules. For each spatiotemporal grid, it calculates a power consumption feature vector containing total load, load curve shape, and power consumption fluctuation within the corresponding time slice. The power consumption feature vectors of the target spatiotemporal grid are compared in multiple dimensions to obtain multi-dimensional deviation. Based on the multi-dimensional deviation, power consumption anomaly index analysis is performed. Combined with preset thresholds, graded anomaly results are generated, and graded anomaly early warning is implemented. These technical means solve the technical problem of insufficient accuracy in existing regional power consumption anomaly monitoring, achieving the technical effect of improving the accuracy of anomaly monitoring.

[0005] This application provides a method for monitoring regional power consumption anomalies based on spatiotemporal gridding, including: setting spatial division rules and time slicing rules to construct a multi-level spatiotemporal grid for the target monitoring area; for each spatiotemporal grid, calculating the power consumption feature vector within the corresponding time slice, wherein the power consumption feature vector includes total load characteristics, load curve shape characteristics, and power consumption fluctuation characteristics; performing multi-dimensional comparison of the power consumption feature vector of the target spatiotemporal grid to obtain multi-dimensional deviation; performing power consumption anomaly index analysis based on the multi-dimensional deviation, generating graded anomaly results based on a preset threshold, and performing graded anomaly warning.

[0006] In a possible implementation, the electricity consumption feature vector of the target spatiotemporal grid is compared in multiple dimensions to obtain a multidimensional deviation. The following processing is performed: based on a multi-level spatiotemporal grid, the electricity consumption feature vector is compared longitudinally with the historical baseline of the target spatiotemporal grid to obtain a longitudinal deviation; the spatiotemporal network group features of the same attribute of the target spatiotemporal grid are compared laterally to obtain a lateral deviation; the target spatiotemporal grid is compared with topologically adjacent or electrically associated spatiotemporal grids to obtain a neighborhood deviation; the longitudinal deviation, lateral deviation, and neighborhood deviation are integrated to obtain the multidimensional deviation.

[0007] In a possible implementation, the characteristics of similar spatiotemporal network groups of the target spatiotemporal grid are compared laterally to obtain the lateral deviation. The following processing is then performed: grid attributes are analyzed based on the industry classification, power supply voltage level, and user scale of the spatiotemporal grid to construct a grid attribute feature profile; using the grid attribute feature profile and electricity consumption behavior features, grid features are clustered to establish a similar spatiotemporal grid group; the degree of deviation between the electricity consumption feature vector of the target spatiotemporal grid and the feature center of the similar spatiotemporal grid group is calculated to obtain the lateral deviation.

[0008] In a possible implementation, a multi-level spatiotemporal grid of the target monitoring area is constructed, and the following processing is performed: a first-level grid is established by dividing the geographical location grid according to administrative boundaries and physical stations; the first-level grid is aggregated or split based on electricity density and user attribute similarity to construct the multi-level spatiotemporal grid.

[0009] In possible implementations, time slicing rules are set, and the following processing is performed: at least two different time windows are set, including a first time window for capturing real-time time and a second time window for analyzing periodic patterns; or the time window scale is dynamically set according to the type of spatiotemporal grid or the needs of external early warning events.

[0010] In a possible implementation, the target spatiotemporal grid is compared with topologically adjacent or electrically associated spatiotemporal grids to obtain the neighborhood deviation. The following processing is then performed: Based on the power grid topology connection relationship, the electrical neighborhood range of the target spatiotemporal grid is determined. The electrical neighborhood includes all spatiotemporal grids powered by the same distribution transformer or located under the same circuit. The degree of deviation between the power consumption feature vector of the target spatiotemporal grid and the statistical center value of the power consumption feature vector of all grids in the electrical neighborhood is calculated to obtain the neighborhood deviation.

[0011] In a possible implementation, the target spatiotemporal grid is compared with topologically adjacent or electrically associated spatiotemporal grids to obtain the neighborhood deviation. The following processing is also performed: when an anomaly is detected in the target spatiotemporal grid, it is determined whether the electrically upstream neighboring grid of the target spatiotemporal grid has experienced the same or associated type of anomaly within a preset time window; if so, it is determined that there is abnormal propagation, and a neighborhood deviation including the propagation direction and root cause location inference is generated.

[0012] In a possible implementation, an electricity consumption anomaly index analysis is performed based on the multidimensional deviation, and a graded anomaly result is generated based on a preset threshold. The following processing is then performed: weight coefficients for the vertical deviation, horizontal deviation, and neighborhood deviation are configured based on season, grid type, or historical feedback data from anomaly investigation; a weighted fusion method is used to weight and fuse the multidimensional deviation based on the weight coefficients to obtain an electricity consumption anomaly index; the electricity consumption anomaly index is compared with a multi-level judgment threshold, and the graded anomaly result is generated based on the comparison result.

[0013] In a possible implementation, the following processing is also performed: the graded anomaly results are associated and matched with external business events, wherein the external business event data includes planned power outage information, orderly power consumption plans, holiday information and major weather warnings; when the graded anomaly results are successfully matched with external business events, the anomaly grade warning is downgraded, labeled or blocked.

[0014] This application also provides a regional power consumption anomaly monitoring system based on spatiotemporal gridding, including: a multi-level spatiotemporal grid construction module, used to set spatial division rules and time slicing rules to construct a multi-level spatiotemporal grid for the target monitoring area; a power consumption feature vector calculation module, used to calculate the power consumption feature vector in the corresponding time slice for each spatiotemporal grid, wherein the power consumption feature vector includes total load characteristics, load curve shape characteristics, and power consumption fluctuation characteristics; a multi-dimensional comparison module, used to perform multi-dimensional comparison of the power consumption feature vector of the target spatiotemporal grid to obtain multi-dimensional deviation; and a power consumption anomaly analysis module, used to perform power consumption anomaly index analysis based on the multi-dimensional deviation, generate graded anomaly results based on preset thresholds, and perform graded anomaly warning.

[0015] The proposed method and system for monitoring regional power consumption anomalies based on spatiotemporal gridding, as described in this application, first establishes spatial division rules and time slicing rules to construct a multi-level spatiotemporal grid for the target monitoring area. Then, for each spatiotemporal grid, a power consumption characteristic vector is calculated within the corresponding time slice. This characteristic vector includes total load characteristics, load curve morphology characteristics, and power consumption fluctuation characteristics. Next, the power consumption characteristic vectors of the target spatiotemporal grid are compared in multiple dimensions to obtain multi-dimensional deviations. Finally, power consumption anomaly index analysis is performed based on the multi-dimensional deviations, and graded anomaly results are generated according to preset thresholds, with graded anomaly warnings issued. Through the above process, the method and system proposed in this application achieve the technical effect of improving the accuracy of anomaly monitoring. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 This is a flowchart illustrating the regional power consumption anomaly monitoring method based on spatiotemporal gridding provided in an embodiment of this application.

[0018] Figure 2 A schematic diagram of the structure of a regional power consumption anomaly monitoring system based on spatiotemporal gridding provided in an embodiment of this application.

[0019] Figure labeling: Multi-level spatiotemporal grid construction module 10, electricity consumption feature vector calculation module 20, multi-dimensional comparison module 30, electricity consumption anomaly analysis module 40. Detailed Implementation

[0020] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0021] This application provides a method for monitoring regional power consumption anomalies based on spatiotemporal gridding, such as... Figure 1 As shown, the method includes: Step S100: Set spatial division rules and time slicing rules to construct a multi-level spatiotemporal grid for the target monitoring area.

[0022] Specifically, a two-tiered spatial and temporal grid monitoring framework is established. First, the spatial division criteria and temporal segmentation criteria for the region are determined. Then, the two are combined to construct a multi-level spatiotemporal grid that can adapt to different monitoring granularities. The spatial grid's hierarchy can be adjusted according to actual monitoring needs, and the temporal slice scale can take into account both real-time monitoring and periodic pattern analysis, providing a unified analysis unit for subsequent electricity consumption characteristic calculations.

[0023] In one possible implementation, a multi-level spatiotemporal grid of the target monitoring area is constructed. Step S100 further includes step S110, which involves dividing the geographic location grid according to administrative boundaries and physical distribution areas to establish a first-level grid. Specifically, using geographic information system (GIS) spatial partitioning technology, the administrative boundary vector data of the target monitoring area, such as city, district, street, and community boundaries, and the geographic topology data of the power distribution network physical distribution areas, such as the power supply range vector map of distribution transformers, are loaded. Through the overlay analysis function of the GIS, the spatial intersection operation of the administrative boundaries and physical distribution area boundaries is performed to generate a non-overlapping, fully covered first-level grid.

[0024] For example, taking the streets of a city as the basis for administrative division, and superimposing the power supply area of ​​all distribution transformers within that street, if a certain area spans two streets, then according to the power supply line route and user affiliation of the area, the area is divided into the corresponding street grid, ensuring that users in each primary grid belong to the same administrative region and a clear physical power supply unit.

[0025] Step S120: Based on electricity density and user attribute similarity, the first-level grid is aggregated or split to construct the multi-level spatiotemporal grid. Specifically, the electricity density of each first-level grid is calculated by dividing the total electricity load within the grid by the geographical area of ​​the grid. Simultaneously, the attribute features of users within the grid are extracted, including industry type, electricity capacity, and voltage level. A clustering algorithm, such as K-means clustering, is used to cluster the first-level grids. Adjacent first-level grids with high electricity density and high user attribute similarity are aggregated to form a higher-level macro-grid for overall regional electricity monitoring. First-level grids with large differences in electricity density and complex user attributes are split into multiple finer-grained sub-grids for precise monitoring of local electricity consumption anomalies.

[0026] For example, if a certain level of grid contains both large industrial users and densely populated residential users, and their electricity consumption density and characteristics differ significantly, then the grid is split into industrial user sub-grids and residential user sub-grids for separate analysis.

[0027] In one possible implementation, time slicing rules are set, and step S100 further includes step S130, setting at least two different time windows, including a first time window for capturing real-time time and a second time window for analyzing periodic patterns; or dynamically setting the time window scale according to the type of spatiotemporal grid or the needs of external early warning events.

[0028] Specifically, time series segmentation technology is employed, with two preset configuration modes: fixed time window and dynamic time window. In the fixed time window mode, the first time window is set to 5 minutes, capturing real-time electricity load data at 5-minute intervals to identify changes in electricity consumption within the grid. The second time window is set to 24 hours, 7 days, and 30 days, respectively, to analyze daily, weekly, and monthly electricity consumption patterns. In the dynamic time window mode, a rule base is established to associate the time window scale with grid type and external events. For example, for industrial user grids, when a factory shutdown warning event is received, the time window is adjusted to 1 hour to shorten the monitoring cycle and quickly capture electricity consumption anomalies caused by shutdowns; for residential user grids, during holidays, the time window is adjusted to 12 hours to adapt to residential holiday electricity consumption patterns.

[0029] Step S200: For each spatiotemporal grid, calculate the electricity consumption feature vector within the corresponding time slice. The electricity consumption feature vector includes total load characteristics, load curve shape characteristics, and electricity consumption fluctuation characteristics.

[0030] Specifically, feature extraction of electricity consumption data is performed on each spatiotemporal grid cell, transforming the raw electricity load data into feature vectors suitable for comparative analysis. Among these, the total load feature reflects the overall electricity consumption scale within the grid, the load curve shape feature reflects the temporal distribution pattern of electricity consumption, and the electricity fluctuation feature reflects the stability of the electricity load. These three features together constitute the core indicators describing the grid's electricity consumption status, providing a data foundation for deviation calculation. The specific calculation method is as follows: the total load feature is obtained by summing the real-time load data of all users within the grid within a statistical time slice. For example, for a 5-minute time slice, 300 1-second load sampling points are collected from each user during this period, and the cumulative sum of the load values ​​of all user sampling points is calculated. Simultaneously, the average load, maximum load, and minimum load within this time slice are calculated as supplementary total load features. The load curve morphology characteristics are achieved by calculating the fitting coefficient and morphological indices of the load curve. First, the load sequence within the time slice is normalized, and then Fourier transform is used to extract the frequency components of the curve. Morphological parameters such as the proportion of dominant frequency, the mean slope of the curve, and the number of peaks and valleys are calculated. For example, the load curve of the residential user grid will show obvious peaks during the evening period, and its number of peaks and valleys and the proportion of dominant frequency will be significantly higher than those of the industrial user grid. The electricity consumption volatility characteristics are quantified by calculating the dispersion of the load sequence, specifically using three indicators: standard deviation, coefficient of variation, and load volatility. Among them, the standard deviation reflects the dispersion of the load data, the coefficient of variation is the ratio of the standard deviation to the average load, and the load volatility is the ratio of the difference between the maximum and minimum loads within the time slice to the average load. The final electricity consumption volatility characteristic value is obtained by weighted summing of these three indicators, thus comprehensively reflecting the stability of the electricity load within the grid.

[0031] Step S300: Perform multi-dimensional comparison of the electricity consumption feature vector of the target spatiotemporal grid to obtain the multi-dimensional deviation.

[0032] Specifically, through multi-dimensional comparative analysis, the degree of difference between the electricity consumption characteristics of the target spatiotemporal grid and the baseline state is quantified. Specifically, comparisons are made from three dimensions: longitudinal historical contemporaneous period, horizontal similar grids, and neighboring related grids. The corresponding deviation is calculated for each dimension, and then the deviations from the three dimensions are integrated to obtain a multi-dimensional deviation that comprehensively reflects the degree of electricity consumption anomalies, thus solving the problem of one-sided results from single-dimensional comparisons.

[0033] In one possible implementation, the electricity consumption feature vector of the target spatiotemporal grid is compared in multiple dimensions to obtain multidimensional deviation. Step S300 further includes step S310, which, based on a multi-level spatiotemporal grid, compares the electricity consumption feature vector with the historical baseline of the target spatiotemporal grid to obtain the longitudinal deviation. Specifically, the historical baseline of the target spatiotemporal grid is constructed by selecting electricity consumption feature vector data from the same period in the past three years, such as a specific date and time, and calculating the mean and standard deviation of each feature dimension to form the historical baseline feature vector. The Euclidean distance algorithm is used to calculate the degree of deviation between the current electricity consumption feature vector and the historical baseline feature vector. The formula for calculating the longitudinal deviation is: the longitudinal deviation equals the Euclidean distance between the current feature vector and the baseline feature vector divided by the standard deviation of the baseline feature vector.

[0034] For example, to calculate the longitudinal deviation of a residential grid from 7 PM to 8 PM on January 1, 2025, the average total load, average load curve shape coefficient, and average electricity fluctuation coefficient from 7 PM to 8 PM on January 1, 2022-2024 are selected as the baseline. The Euclidean distance between the current feature vector and the baseline vector is calculated, and then divided by the corresponding standard deviation to obtain the longitudinal deviation for that period.

[0035] Step S320 involves performing a lateral comparison of the spatiotemporal network group features of the target spatiotemporal grid with similar attributes to obtain the lateral deviation. Specifically, a group of grids with similar attributes is constructed based on the grid's attribute features and electricity consumption behavior features, and the degree of deviation between the target grid and the feature center of the group of grids with similar attributes is calculated. Specifically, a clustering algorithm is used to divide the grids with similar attributes, and then a cosine similarity algorithm is used to calculate the similarity between the feature vector of the target grid and the feature center vector of the group. The lateral deviation is equal to 1 minus the cosine similarity value.

[0036] For example, all residential user grids are divided into a similar group, and the mean of the total load, curve shape, and volatility characteristics of the group are calculated to form the group feature center. Then, the cosine similarity between the feature vector of the target residential grid and the vector of the center is calculated to obtain the lateral deviation.

[0037] Step S330: Perform a neighborhood comparison between the target spatiotemporal grid and topologically adjacent or electrically associated spatiotemporal grids to obtain the neighborhood deviation. Specifically, determine the neighborhood range of the target grid based on the distribution network topology, including geographically adjacent grids and electrically associated grids, such as grids powered by the same distribution transformer. Calculate the statistical center value (e.g., mean) of the electricity consumption feature vectors of all grids within the neighborhood, and use the Manhattan distance algorithm to calculate the distance between the target grid's feature vector and this statistical center value; this distance is the neighborhood deviation.

[0038] For example, a target grid is powered by distribution transformer A, and its neighborhood grid includes all other grids that are also powered by distribution transformer A. Calculate the mean of the power consumption characteristics of these neighborhood grids, and then calculate the Manhattan distance between the target grid's feature vector and this mean to obtain the neighborhood deviation.

[0039] Step S340: Integrate the vertical deviation, horizontal deviation, and neighborhood deviation to obtain the multidimensional deviation. Specifically, a weighted summation method is used to integrate the deviations of the three dimensions. First, the weight coefficients of each deviation are determined using the analytic hierarchy process (AHP). For example, the weight of the vertical deviation is 0.4, the weight of the horizontal deviation is 0.3, and the weight of the neighborhood deviation is 0.3. Then, the multidimensional deviation is calculated using the following formula: Multidimensional deviation equals vertical deviation multiplied by its weight, plus horizontal deviation multiplied by its weight, plus neighborhood deviation multiplied by its weight.

[0040] In one possible implementation, the spatiotemporal network group characteristics of the same attribute of the target spatiotemporal grid are compared laterally to obtain the lateral deviation. Step S320 further includes step S321, which analyzes the grid attributes based on the industry classification, power supply voltage level, and user scale of the spatiotemporal grid to construct a grid attribute feature profile. Specifically, a grid attribute feature index library is established, and the indexes include industry classification, power supply voltage level, and user scale. Among them, industry classification is divided into categories such as industrial, commercial, residential, and agricultural, and is represented by numerical codes, such as 1 for industrial and 2 for commercial; power supply voltage level is divided into low voltage 220V, 380V, high voltage 10kV, etc., and is represented by voltage values; user scale is divided into large users, medium users, and small users, based on the user's power consumption capacity, such as a power consumption capacity greater than 100kVA being a large user. The above attribute data of each spatiotemporal grid is extracted and standardized according to the specifications of the index library to form a one-dimensional feature vector composed of industry code, voltage level, and user scale level. This vector is the grid attribute feature profile.

[0041] For example, if all users in a certain grid are large industrial users and the power supply voltage is 10 kV, then its attribute feature profile is [1, 10000, 1].

[0042] Step S322: Using the grid attribute feature profile and electricity consumption behavior features, grid features are clustered to establish similar spatiotemporal grid groups. Specifically, an improved K-means clustering algorithm is used to fuse the grid attribute feature profile and electricity consumption behavior features, including total load, load curve shape, and electricity consumption fluctuation, to form a fused feature vector. The number of clusters is set; for example, based on the attribute differences of grids within the region, the number of clusters is set to 5, corresponding to categories such as industrial large user grids, commercial medium-sized user grids, and residential small user grids. By iteratively calculating the distance from each fused feature vector to the cluster center, the grids with the closest distance are grouped into the same category, ultimately forming multiple similar spatiotemporal grid groups. For example, grids with an industry code of 1 and a user scale level of 1 in all fused feature vectors are grouped into an industrial large user grid group.

[0043] Step S323: The deviation degree between the electricity consumption feature vector of the target spatiotemporal grid and the feature center of the same type of spatiotemporal grid group is calculated to obtain the lateral deviation degree. Specifically, the feature center of the same type of spatiotemporal grid group is calculated, that is, the mean of each dimension of the electricity consumption feature vector of all grids in the group, including total load, curve shape, and volatility, is calculated to form the group feature center vector. Then, the cosine similarity algorithm is used to calculate the similarity between the electricity consumption feature vector of the target grid and the feature center vector of the group. The formula for calculating the lateral deviation degree is: lateral deviation degree = 1 - cosine similarity.

[0044] In one possible implementation, the target spatiotemporal grid is compared with topologically adjacent or electrically associated spatiotemporal grids to obtain the neighborhood deviation. Step S330 further includes step S331, which determines the electrical neighborhood range of the target spatiotemporal grid based on the power grid topology connection relationship. The electrical neighborhood includes all spatiotemporal grids powered by the same distribution transformer or located under the same circuit. Specifically, based on the distribution network GIS topology system, the power distribution equipment identifiers corresponding to the target spatiotemporal grid, such as distribution transformer numbers and line numbers, are extracted. Through equipment association queries, all spatiotemporal grids with the same distribution transformer number as the target grid, as well as all spatiotemporal grids under the same line number, are filtered out. These grids are merged and duplicates are removed, which is the electrical neighborhood range of the target grid.

[0045] For example, if the distribution transformer of the target grid is numbered B-001 and the line number is L-005, then the electrical neighborhood includes all grids powered by B-001 and all grids belonging to the L-005 line.

[0046] Step S332: Calculate the deviation of the electricity consumption feature vector of the target spatiotemporal grid from the statistical center value of the electricity consumption feature vectors of all grids within its electrical neighborhood, obtaining the neighborhood deviation. Specifically, calculate the statistical center value of the electricity consumption feature vectors of all grids within the electrical neighborhood, and calculate the arithmetic mean for the three dimensions of total load, load curve shape, and electricity consumption fluctuation, forming the neighborhood statistical center vector. Then, use the Manhattan distance algorithm to calculate the distance between the target grid's electricity consumption feature vector and the neighborhood statistical center vector. This distance is the neighborhood deviation, calculated using the formula: the neighborhood deviation equals the sum of the absolute values ​​of the differences in the feature values ​​of each dimension.

[0047] For example, if the target grid feature vector is [80, 0.6, 0.3] and the neighborhood statistical center vector is [75, 0.62, 0.28], then the neighborhood deviation is |80-75|+|0.6-0.62|+|0.3-0.28|=5.04.

[0048] In one possible implementation, the target spatiotemporal grid is compared with topologically adjacent or electrically associated spatiotemporal grids to obtain a neighborhood deviation. Step S330 further includes step S333: when an anomaly is detected in the target spatiotemporal grid, it is determined whether the electrically upstream neighboring grid of the target spatiotemporal grid has experienced the same or associated type of anomaly within a preset time window; if so, it is determined that there is abnormal propagation, and a neighborhood deviation including the propagation direction and root cause location inference is generated.

[0049] Specifically, a rule base for anomaly types is established within a preset time window, such as 30 minutes. This base includes associations for types like power outage anomalies, overload anomalies, and load drop anomalies. For example, the association type for a power outage anomaly is a load drop anomaly in the downstream grid. When an anomaly is detected in the target grid, the anomaly records of the upstream neighboring grids, such as the upstream line grid corresponding to the distribution transformer, are retrieved over the past 30 minutes to determine if there are any identical or related anomalies. If so, the anomaly propagation direction is determined through topology tracing: upstream line grid → target grid, suggesting the root cause is a fault point in the upstream line. Then, labels for the propagation direction and root cause location are added to the neighborhood deviation calculation results to generate a neighborhood deviation score containing additional information.

[0050] Step S400: Perform electricity fluctuation index analysis based on the multidimensional deviation, generate graded fluctuation results based on preset thresholds, and issue a graded fluctuation warning.

[0051] Specifically, the integrated multidimensional deviation is transformed into anomaly levels that can be directly used for early warning. By assigning reasonable weights to the deviations of different dimensions, a comprehensive electricity consumption anomaly index is calculated. This index is then compared with preset multi-level thresholds to classify different anomaly levels and trigger corresponding early warnings, thereby achieving accurate identification and graded handling of regional electricity consumption anomalies.

[0052] In one possible implementation, electricity consumption anomaly index analysis is performed based on the multidimensional deviation, and graded anomaly results are generated based on preset thresholds. Step S400 further includes step S410, configuring the weight coefficients of the vertical deviation, horizontal deviation, and neighborhood deviation based on season, grid type, or historical feedback data from anomaly investigation. Specifically, a dynamic configuration model for weight coefficients is established, with input parameters being season, grid type, and historical feedback data. Seasons are divided into summer, winter, and spring / autumn; grid types are divided into industrial, commercial, and residential; and historical feedback data includes the accuracy rate of each dimension of deviation for anomaly warning over the past year. The model is trained using a multiple linear regression algorithm to obtain a weight coefficient configuration table under different input parameters. For example, historical data from the summer residential grid shows the highest accuracy rate for vertical deviation warnings, so the vertical deviation weight is configured as 0.5, the horizontal deviation weight as 0.2, and the neighborhood deviation weight as 0.3; historical data from the winter industrial grid shows the highest accuracy rate for neighborhood deviation warnings, so the neighborhood deviation weight is configured as 0.5, the vertical deviation weight as 0.3, and the horizontal deviation weight as 0.2.

[0053] Step S420: A weighted fusion method is used to weight and fuse the multidimensional deviations based on the weight coefficients to obtain the electricity consumption anomaly index. Specifically, the electricity consumption anomaly index is calculated using a weighted summation formula: Electricity Consumption Anomaly Index = Vertical Deviation × Vertical Weight Coefficient + Horizontal Deviation × Horizontal Weight Coefficient + Neighborhood Deviation × Neighborhood Weight Coefficient. The weight coefficients are obtained from the configuration table in step S410. During the calculation process, the deviation and weight coefficient of each dimension are multiplied, and the product results are summed to obtain the final electricity consumption anomaly index.

[0054] Step S430: The electricity consumption anomaly index is compared with a multi-level judgment threshold, and the graded anomaly result is generated based on the comparison result. Specifically, a multi-level judgment threshold system is established. For example, the electricity consumption anomaly index is divided into four levels, with corresponding thresholds as follows: Level 1 anomaly (no abnormality): 0-0.2; Level 2 anomaly (mild abnormality): 0.2-0.5; Level 3 anomaly (moderate abnormality): 0.5-0.8; Level 4 anomaly (severe abnormality): 0.8-1.0. The calculated electricity consumption anomaly index is compared with this threshold system to determine the corresponding anomaly level.

[0055] In one possible implementation, the method further includes step S500, which involves associating and matching the graded anomaly results with external business events. The external business event data includes planned power outage information, orderly power consumption plans, holiday information, and major weather warnings. Specifically, an external business event database is established to store planned power outage information, orderly power consumption plans, holiday information, and major weather warnings. Planned power outage information includes outage time, outage area, and outage reason; orderly power consumption plans include power-restricted users, power-restricted periods, and power-restricted loads; holiday information includes the dates and periods of statutory holidays and adjusted workdays; and major weather warnings include the warning areas and warning periods for typhoons, rainstorms, and cold waves. A keyword matching algorithm is used to match the grid location and anomaly time in the graded anomaly results with the event location and event time in the external business event database. The matching condition is that the grid locations coincide and the time intervals overlap.

[0056] For example, if a grid experiences a Level 3 anomaly between 9:00 and 12:00 on July 10, 2025, and a query of the external event database reveals a planned power outage event in the same grid during the same period, then the two events are considered a successful match.

[0057] Step S600: When the graded anomaly result successfully matches an external business event, the anomaly graded warning is downgraded, labeled, or blocked. Specifically, an anomaly warning processing rule base is established, with the following rules: 1. If a planned event such as a planned power outage or orderly power consumption is matched, and the anomaly level is level two or below, the warning is directly blocked; 2. If a planned event is matched, and the anomaly level is level three or above, the warning level is downgraded by one level, and a planned event is labeled as causing the anomaly; 3. If an unplanned event such as a major weather warning is matched, a weather factor is labeled as causing the anomaly, and the warning level remains unchanged. For example, if the graded anomaly result of a certain grid is level three and a planned power outage event is matched, the warning level is downgraded to level two, and a label indicating that a planned power outage caused the anomaly is added to the warning information; if a level two anomaly of a certain grid matches a rainstorm warning event, the warning is blocked.

[0058] This application's embodiments employ set spatial division and time slicing rules to construct a multi-level spatiotemporal grid for the target monitoring area. For each spatiotemporal grid, a power consumption feature vector containing the total load, load curve shape, and power consumption fluctuations within the corresponding time slice is calculated. The power consumption feature vectors of the target spatiotemporal grid are compared in multiple dimensions to obtain multi-dimensional deviation. Based on the multi-dimensional deviation, power consumption anomaly index analysis is performed. Combined with preset thresholds, graded anomaly results are generated, and graded anomaly early warning is implemented. These technical means solve the technical problem of insufficient accuracy in existing regional power consumption anomaly monitoring and achieve the technical effect of improving the accuracy of anomaly monitoring.

[0059] In the above text, refer to Figure 1A method for monitoring regional power consumption anomalies based on spatiotemporal gridding according to an embodiment of the present invention is described in detail. Next, reference will be made to... Figure 2 A spatiotemporal gridded regional power consumption anomaly monitoring system according to an embodiment of the present invention is described.

[0060] The spatiotemporal grid-based regional power consumption anomaly monitoring system according to embodiments of the present invention addresses the technical problem of insufficient accuracy in existing regional power consumption anomaly monitoring systems, thereby improving the accuracy of anomaly monitoring. The spatiotemporal grid-based regional power consumption anomaly monitoring system includes: a multi-level spatiotemporal grid construction module 10, a power consumption feature vector calculation module 20, a multi-dimensional comparison module 30, and a power consumption anomaly analysis module 40.

[0061] The multi-level spatiotemporal grid construction module 10 is used to set spatial division rules and time slicing rules to construct a multi-level spatiotemporal grid for the target monitoring area; the electricity consumption feature vector calculation module 20 is used to calculate the electricity consumption feature vector in the corresponding time slice for each spatiotemporal grid, wherein the electricity consumption feature vector includes total load characteristics, load curve shape characteristics, and electricity consumption fluctuation characteristics; the multi-dimensional comparison module 30 is used to perform multi-dimensional comparison of the electricity consumption feature vector of the target spatiotemporal grid to obtain multi-dimensional deviation; the electricity consumption anomaly analysis module 40 is used to perform electricity consumption anomaly index analysis based on the multi-dimensional deviation, generate graded anomaly results based on preset thresholds, and perform graded anomaly warning.

[0062] The detailed description of the specific configuration of the multi-dimensional comparison module 30 is explained as follows: As mentioned above, the electricity consumption feature vector of the target spatiotemporal grid is compared in multiple dimensions to obtain multi-dimensional deviation. The multi-dimensional comparison module 30 may further include: a longitudinal comparison unit for longitudinally comparing the electricity consumption feature vector with the historical baseline of the target spatiotemporal grid based on a multi-level spatiotemporal grid to obtain longitudinal deviation; a lateral comparison unit for laterally comparing the spatiotemporal network group features of the same attribute of the target spatiotemporal grid to obtain lateral deviation; a neighborhood comparison unit for performing neighborhood comparison between the target spatiotemporal grid and topologically adjacent or electrically associated spatiotemporal grids to obtain neighborhood deviation; and a deviation integration unit for integrating the longitudinal deviation, lateral deviation, and neighborhood deviation to obtain the multi-dimensional deviation.

[0063] Specifically, the horizontal comparison of the spatiotemporal network group features of the target spatiotemporal grid with similar attributes is performed to obtain the horizontal deviation. The horizontal comparison unit may further include: a grid attribute parsing subunit for parsing grid attributes based on the industry classification, power supply voltage level, and user scale of the spatiotemporal grid to construct a grid attribute feature profile; a grid clustering subunit for using the grid attribute feature profile and electricity consumption behavior features to perform grid clustering on the grid features and establish a spatiotemporal grid group with similar attributes; and a deviation degree calculation subunit for calculating the deviation degree between the electricity consumption feature vector of the target spatiotemporal grid and the feature center of the spatiotemporal grid group with similar attributes to obtain the horizontal deviation.

[0064] The detailed description of the specific configuration of the multi-level spatiotemporal grid construction module 10 is explained as follows: As mentioned above, to construct a multi-level spatiotemporal grid for the target monitoring area, the multi-level spatiotemporal grid construction module 10 may further include: a geographic location grid division unit for dividing the geographic location grid according to administrative boundaries and physical stations to establish a first-level grid; and an aggregation / splitting unit for aggregating or splitting the first-level grid based on electricity consumption density and user attribute similarity to construct the multi-level spatiotemporal grid.

[0065] The multi-level spatiotemporal grid construction module 10, which sets time slicing rules, may further include: a time window setting unit for setting at least two different time windows, including a first time window for capturing real-time time and a second time window for analyzing periodic patterns; or dynamically setting the time window scale according to the type of spatiotemporal grid or the needs of external early warning events.

[0066] Specifically, the target spatiotemporal grid is compared with topologically adjacent or electrically associated spatiotemporal grids to obtain the neighborhood deviation. The neighborhood comparison unit may further include: an electrical neighborhood range determination subunit for determining the electrical neighborhood range of the target spatiotemporal grid based on the power grid topology connection relationship, wherein the electrical neighborhood includes all spatiotemporal grids powered by the same distribution transformer or located under the same circuit; and a deviation degree calculation subunit for calculating the deviation degree of the electricity consumption feature vector of the target spatiotemporal grid from the statistical center value of the electricity consumption feature vector of all grids in the electrical neighborhood to obtain the neighborhood deviation.

[0067] The target spatiotemporal grid is compared with topologically adjacent or electrically associated spatiotemporal grids to obtain a neighborhood deviation. The neighborhood comparison unit may further include: a judgment subunit used to determine whether, within a preset time window, the electrically upstream neighboring grid of the target spatiotemporal grid has experienced the same or associated type of anomaly when an anomaly is detected in the target spatiotemporal grid; if so, it is determined that there is abnormal propagation, and a neighborhood deviation including the propagation direction and root cause location inference is generated.

[0068] The detailed description of the specific configuration of the electricity consumption anomaly analysis module 40 is explained as follows: As mentioned above, the electricity consumption anomaly index is analyzed based on the multidimensional deviation, and a graded anomaly result is generated based on a preset threshold. The electricity consumption anomaly analysis module 40 may further include: a weight coefficient configuration unit for configuring the weight coefficients of the longitudinal deviation, the lateral deviation, and the neighborhood deviation based on the season, grid type, or historical feedback data of anomaly investigation; a weighted fusion unit for using a weighted fusion method to weight and fuse the multidimensional deviation based on the weight coefficients to obtain the electricity consumption anomaly index; and a graded anomaly result generation unit for comparing the electricity consumption anomaly index with a multi-level judgment threshold and generating the graded anomaly result based on the comparison result.

[0069] The method may further include: an association matching module for associating and matching the graded anomaly results with external business events, wherein the external business event data includes planned power outage information, orderly power consumption plans, holiday information and major weather warnings; and an anomaly graded warning processing module for downgrading, labeling or blocking the anomaly graded warning when the graded anomaly results are successfully matched with external business events.

[0070] The spatiotemporal grid-based regional power consumption anomaly monitoring system provided in this invention can execute the spatiotemporal grid-based regional power consumption anomaly monitoring method provided in any embodiment of this invention, and has the corresponding functional modules and beneficial effects of the method.

[0071] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0072] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for monitoring regional power consumption anomalies based on spatiotemporal gridding, characterized in that, include: Set spatial division rules and time slicing rules to construct a multi-level spatiotemporal grid for the target monitoring area; For each spatiotemporal grid, calculate the electricity consumption feature vector within the corresponding time slice. The electricity consumption feature vector includes the total load characteristics, the load curve shape characteristics, and the electricity consumption fluctuation characteristics. The electricity consumption feature vector of the target spatiotemporal grid is compared in multiple dimensions to obtain the multidimensional deviation. Based on the multidimensional deviation, an electricity consumption anomaly index analysis is performed, and graded anomaly results are generated according to preset thresholds, and anomaly graded early warning is issued.

2. The method for monitoring regional power consumption anomalies based on spatiotemporal gridding according to claim 1, characterized in that, A multi-dimensional comparison is performed on the electricity consumption feature vector of the target spatiotemporal grid to obtain multi-dimensional deviation, including: Based on a multi-level spatiotemporal grid, the longitudinal deviation is obtained by longitudinally comparing the electricity consumption feature vector with the historical baseline of the target spatiotemporal grid. The lateral deviation is obtained by comparing the characteristics of spatiotemporal network groups with similar attributes of the target spatiotemporal grid. The target spatiotemporal grid is compared with topologically adjacent or electrically associated spatiotemporal grids to obtain the neighborhood deviation. The longitudinal deviation, lateral deviation, and neighborhood deviation are integrated to obtain the multidimensional deviation.

3. The method for monitoring regional power consumption anomalies based on spatiotemporal gridding according to claim 2, characterized in that, A lateral comparison is performed on the spatiotemporal network group characteristics of the target spatiotemporal grid with similar attributes to obtain the lateral deviation, including: Based on the industry classification, power supply voltage level, and user scale of the spatiotemporal grid, the grid attribute is analyzed to construct a grid attribute feature profile; Using the aforementioned grid attribute feature profile and electricity consumption behavior features, grid clustering is performed on the grid features to establish a spatiotemporal grid group of the same type; The deviation degree between the electricity consumption feature vector of the target spatiotemporal grid and the feature center of the same type of spatiotemporal grid group is calculated to obtain the lateral deviation degree.

4. The method for monitoring regional power consumption anomalies based on spatiotemporal gridding according to claim 1, characterized in that, Constructing a multi-level spatiotemporal grid for the target monitoring area, including: Geographical grids are divided according to administrative boundaries and physical distribution areas to establish a primary grid. Based on electricity consumption density and user attribute similarity, the first-level grid is aggregated or split to construct the multi-level spatiotemporal grid.

5. The method for monitoring regional power consumption anomalies based on spatiotemporal gridding according to claim 1, characterized in that, Configure time slicing rules, including: Set at least two different time windows, including a first time window for capturing real-time time and a second time window for analyzing periodic patterns; Alternatively, the time window scale can be dynamically set according to the type of spatiotemporal grid or the needs of external early warning events.

6. The method for monitoring regional power consumption anomalies based on spatiotemporal gridding according to claim 2, characterized in that, The target spatiotemporal grid is compared with topologically adjacent or electrically associated spatiotemporal grids to obtain the neighborhood deviation, including: Based on the power grid topology, the electrical neighborhood of the target spatiotemporal grid is determined. The electrical neighborhood includes all spatiotemporal grids powered by the same distribution transformer or located under the same circuit. The deviation of the power consumption feature vector of the target spatiotemporal grid from the statistical center value of the power consumption feature vector of all grids in its electrical neighborhood is calculated to obtain the neighborhood deviation.

7. The method for monitoring regional power consumption anomalies based on spatiotemporal gridding according to claim 6, characterized in that, The method further includes performing a neighborhood comparison between the target spatiotemporal grid and topologically adjacent or electrically associated spatiotemporal grids to obtain the neighborhood deviation, and also includes: When an anomaly is detected in the target spatiotemporal grid, it is determined whether the electrical upstream neighboring grid of the target spatiotemporal grid has experienced the same or related type of anomaly within a preset time window; If so, it is determined that there is abnormal propagation, and a neighborhood deviation degree containing the propagation direction and the inferred root cause location is generated.

8. The method for monitoring regional power consumption anomalies based on spatiotemporal gridding according to claim 2, characterized in that, Based on the multidimensional deviation, an electricity consumption anomaly index analysis is performed, and graded anomaly results are generated according to a preset threshold, including: Configure the weighting coefficients for the vertical deviation, horizontal deviation, and neighborhood deviation based on the season, grid type, or historical feedback data from anomaly investigation; A weighted fusion method is used to weight and fuse the multidimensional deviation based on the weight coefficients to obtain the electricity consumption anomaly index. The power consumption anomaly index is compared with a multi-level judgment threshold, and the graded anomaly result is generated based on the comparison result.

9. The method for monitoring regional power consumption anomalies based on spatiotemporal gridding according to claim 1, characterized in that, Also includes: The graded anomaly results are correlated and matched with external business events, where external business event data includes planned power outage information, orderly power consumption plans, holiday information, and major weather warnings; When the graded anomaly result is successfully matched with an external business event, the anomaly graded warning is downgraded, labeled, or blocked.

10. A regional power consumption anomaly monitoring system based on spatiotemporal gridding, characterized in that, The system is used to implement the regional power consumption anomaly monitoring method based on spatiotemporal gridding as described in any one of claims 1-9, the system comprising: The multi-level spatiotemporal grid construction module is used to set spatial division rules and time slicing rules to construct a multi-level spatiotemporal grid for the target monitoring area; The electricity consumption feature vector calculation module is used to calculate the electricity consumption feature vector within the corresponding time slice for each spatiotemporal grid. The electricity consumption feature vector includes the total load feature, the load curve shape feature, and the electricity consumption fluctuation feature. The multi-dimensional comparison module is used to perform multi-dimensional comparison of the electricity consumption feature vector of the target spatiotemporal grid to obtain the multi-dimensional deviation. The power consumption anomaly analysis module is used to perform power consumption anomaly index analysis based on the multidimensional deviation, generate graded anomaly results based on preset thresholds, and provide graded anomaly warnings.

Images