Urban resident load power consumption risk identification method and system based on space-time coupling

Abstract

This invention discloses a method for identifying urban residential load electricity consumption risks based on spatiotemporal coupling. The method includes acquiring historical operational data of the target power system and preprocessing it to construct a spatiotemporal feature set; training a dynamic isolated forest model using the obtained spatiotemporal feature set to obtain an anomaly score set; acquiring current operational data of the target power system; obtaining the current anomaly score of the target power system using the trained dynamic isolated forest model; and identifying the urban residential load electricity consumption risks of the target power system based on spatiotemporal coupling based on the obtained current anomaly score. This invention also discloses a system for implementing the aforementioned method for identifying urban residential load electricity consumption risks based on spatiotemporal coupling. This invention not only enables the identification of urban residential load electricity consumption risks of the target power system based on spatiotemporal coupling, but also offers higher reliability and better accuracy.

Description

Technical Field

[0001] This invention belongs to the field of electrical automation and relates to a method and system for identifying the electricity consumption risk of urban residents based on spatiotemporal coupling. Background Technology

[0002] With economic and technological development and the improvement of people's living standards, electricity has become an indispensable secondary energy source in people's production and daily life, bringing endless convenience. Therefore, ensuring a stable and reliable supply of electricity has become one of the most important tasks of the power system.

[0003] Currently, the load on power distribution networks is growing rapidly, and the electricity consumption behaviors of users are becoming increasingly complex, making electricity safety issues in power distribution networks increasingly prominent. Therefore, identifying electricity safety risks is particularly important.

[0004] Traditional electricity safety risk identification schemes primarily employ fixed threshold rules or offline-trained static clustering models, aiming to identify abnormal states such as overload and overvoltage by monitoring time-series data such as voltage and current at individual nodes. Moreover, these schemes typically rely on manually set thresholds or batch training methods. However, with the increasing diversification of residential electricity consumption behaviors and the time-varying nature of electricity usage patterns, traditional schemes struggle to adapt to the non-stationary distribution of massive amounts of time-series data, and the need for frequent manual adjustments to fixed thresholds leads to increased false alarm and false negative rates. Furthermore, most existing methods neglect the electrical coupling and spatial correlation between distribution network nodes, resulting in insufficient ability to identify spatially propagated risks such as regional overloads and voltage dips. Summary of the Invention

[0005] One of the objectives of this invention is to provide a highly reliable and accurate method for identifying urban residential electricity consumption risks based on spatiotemporal coupling.

[0006] The second objective of this invention is to provide a system for implementing the aforementioned method for identifying urban residential electricity load risks based on spatiotemporal coupling.

[0007] The urban residential electricity load risk identification method based on spatiotemporal coupling provided by this invention includes the following steps:

[0008] S1. Obtain historical operating data information of the target power system;

[0009] S2. Preprocess the data information obtained in step S1 to construct a spatiotemporal feature set;

[0010] S3. Using the spatiotemporal feature set obtained in step S2, train the dynamic isolation forest model and obtain an anomaly score set;

[0011] S4. Obtain the current operating data information of the target power system;

[0012] S5. Using the trained dynamic isolated forest model, the current anomaly score of the target power system is obtained;

[0013] S6. Based on the current anomaly score obtained in step S5, complete the identification of urban residential load electricity consumption risks in the target power system based on spatiotemporal coupling.

[0014] Step S1, which involves obtaining historical operating data information of the target power system, specifically includes the following steps:

[0015] The voltage and current values ​​of each node in the target distribution network area are obtained at a set sampling period.

[0016] Step S2, which involves preprocessing the data obtained in step S1 to construct a spatiotemporal feature set, specifically includes the following steps:

[0017] For the voltage and current values ​​obtained in step S1, a period is set, and the average voltage value and average current value are calculated for each period. These values ​​are then normalized to obtain the normalized voltage value for each node. and normalized current value ;

[0018] Construct the spatial association matrix A between nodes as follows: , where N is the total number of nodes; This represents the electrical connection strength or spatial dependence between node i and node j, and it is set that if node i and node j are not directly connected, then... If node i and node j are the same node, then ;

[0019] The spatial features of node i are calculated using the following formula. :

[0020] In the formula Here is the normalized current value at node j; This is a set minimum value to prevent the denominator from being 0;

[0021] Construct the spatiotemporal feature vector of node i at time t for The spatiotemporal feature vectors of all nodes constitute the spatiotemporal feature set. for .

[0022] Step S3, which involves using the spatiotemporal feature set obtained in step S2 to train the dynamic isolation forest model and obtain an anomaly score set, specifically includes the following steps:

[0023] The spatiotemporal coupling factor of node i at time t is calculated. for ;

[0024] Set time decay weight for ,in The time decay parameter is set, where q represents the sample age;

[0025] The spatial augmentation coefficient of node i at time t was calculated. for ;in The set space enhancement parameters;

[0026] The comprehensive weight of node i at time t is calculated. for ;

[0027] Build a global queue , Each element in the set corresponds to the spatiotemporal feature set at time t. ; upon reaching a new time t+1, then in Add the spatiotemporal feature set at time t+1. and remove the spatiotemporal feature set. ; For global queue The total number of spatiotemporal feature sets in the data;

[0028] According to the overall weight In the current global queue Sampling without replacement was performed to obtain... Each sample is used as a training subset of an isolation tree; sampling is repeated T times; each tree is recursively partitioned, the recursive partitioning process including: randomly selecting feature dimensions. , The value can be 1, 2, or 3; the threshold is selected evenly between the set minimum and maximum values. Divide the sample into left and right child nodes until the number of samples is no greater than 1 or the height of the tree reaches a certain value. , The function is a rounding function; given the load data, the model outputs the age factor for each tree. for and the feature mean vector of the training subset for ;in, For training subset One of the samples; For the sample The corresponding overall weight; For the sample The age of the sample is the number of time steps since the current training moment; This is the training subset of the k-th isolation tree;

[0029] Calculate the global feature mean for time t. for The drift distance was calculated. for ,in The L2 norm is used; the adaptive correction factor is calculated. for , The first correction parameter is set. The second correction parameter is set. This represents the average drift distance.

[0030] For each node i and each tree k, obtain the original path length. , The number of edges from the root to the leaf is given; the corrected output path length is calculated. for ; Calculate the average corrected path length of node i for T represents the total number of isolated trees in the isolated forest; calculate the baseline path length. for , The Euler-Macheroni constant is used; the anomaly fraction of node i at time t is calculated. for ;

[0031] Anomaly scores for all nodes This constitutes a set of abnormal scores.

[0032] Step S4, which involves obtaining the current target power system's operational data, specifically includes the following steps:

[0033] The voltage and current values ​​of each node in the target distribution network area are obtained at a set sampling period and then preprocessed.

[0034] Step S5, which uses the trained dynamic isolation forest model to obtain the current anomaly score of the target power system, specifically includes the following steps:

[0035] The data obtained in step S4 is input into the trained dynamic isolated forest model to obtain the current anomaly score of each node n in the target power system. .

[0036] Step S6, which involves identifying the urban residential load electricity consumption risk of the target power system based on the current anomaly score obtained in step S5, specifically includes the following steps:

[0037] Take the set quantile , and , ;

[0038] Based on the set quantiles, the corresponding first threshold is obtained from the set of abnormal scores obtained in step S3. Second threshold and the third threshold , ;

[0039] Based on the current anomaly score of each node n in the target power system Determine the risk status of the target node n:

[0040] like If so, then the target node n is determined to be in a normal state;

[0041] like If so, then the target node n is determined to be in a state of interest;

[0042] like If so, then the target node n is determined to be in an overload state;

[0043] like If so, the target node n is determined to be in a severely abnormal state.

[0044] Step S6 further includes the following steps:

[0045] If node n in the target power system is in an overload state, then the following determination is made again:

[0046] like If so, then node n is determined to be in a current overload state;

[0047] like If so, then node n is determined to be in a voltage overload state;

[0048] Otherwise, determine that node n is in a composite overload state;

[0049] For the set current threshold, The set voltage threshold;

[0050] The frequency of use of feature dimensions in all split nodes of the current dynamic isolated forest model is statistically analyzed, and the dimension with the highest frequency of use is selected as the dominant cause of the anomaly of node n; the feature dimensions include current dimension, voltage dimension and spatial feature dimension.

[0051] The regional risk index of the region where node n is located is calculated. for N is the total number of nodes in the region where node n is located;

[0052] If the regional risk index If the value exceeds the set threshold, a regional risk warning will be issued.

[0053] This invention also provides a system for implementing the aforementioned method for identifying urban residential load electricity risks based on spatiotemporal coupling, comprising a data acquisition module, a data processing module, a model training module, a real-time acquisition module, a score calculation module, and a risk identification module; the data acquisition module, data processing module, model training module, real-time acquisition module, score calculation module, and risk identification module are connected in series; the data acquisition module is used to acquire historical operating data information of the target power system and upload the data information to the data processing module; the data processing module is used to preprocess the acquired data information according to the received data information to construct a spatiotemporal feature set and upload the data information to the model training module; the model training .... The received data is used to train the dynamic isolated forest model using the obtained spatiotemporal feature set, resulting in an anomaly score set. This data is then uploaded to the real-time acquisition module. The real-time acquisition module uses the received data to obtain the current operating data of the target power system and uploads it to the score calculation module. The score calculation module uses the trained dynamic isolated forest model to obtain the current anomaly score of the target power system and uploads it to the risk identification module. The risk identification module uses the received data and the obtained current anomaly score to identify the spatiotemporally coupled urban residential load electricity consumption risk of the target power system.

[0054] The method and system for identifying urban residential load electricity consumption risks based on spatiotemporal coupling provided by this invention, by acquiring and modeling data information of the target power system and comprehensively considering the spatiotemporal and abnormal characteristics of each node, can not only identify urban residential load electricity consumption risks of the target power system based on spatiotemporal coupling, but also has higher reliability and better accuracy. Attached Figure Description

[0055] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0056] Figure 2 This is a schematic diagram of the functional modules of the system of the present invention. Detailed Implementation

[0057] like Figure 1The diagram shown is a flowchart of the method of the present invention: The urban residential electricity load risk identification method based on spatiotemporal coupling disclosed in this invention includes the following steps:

[0058] S1. Obtain historical operating data of the target power system; specifically including the following steps:

[0059] The voltage and current values ​​of each node in the target distribution network area are obtained at a set sampling period.

[0060] S2. Preprocess the data information obtained in step S1 to construct a spatiotemporal feature set; specifically including the following steps:

[0061] For the voltage and current values ​​obtained in step S1, a period is set, and the average voltage value and average current value are calculated for each period. These values ​​are then normalized to obtain the normalized voltage value for each node. and normalized current value ;

[0062] Construct the spatial association matrix A between nodes as follows: , where N is the total number of nodes; This represents the electrical connection strength or spatial dependence between node i and node j, and it is set that if node i and node j are not directly connected, then... If node i and node j are the same node, then In specific implementation, When representing the electrical connection strength, its value can be normalized using the admittance magnitude between nodes as the coupling coefficient, i.e. ,in Let the admittance be the branch between nodes i and j (if there is no direct branch between the two nodes, then...). ); When representing the degree of spatial correlation, its value can be obtained using a Gaussian radial basis function based on geographic distance, i.e. ,in denoted as , representing the spatial distance (in km) between nodes i and j. This is the distance scale parameter (preferably 1km).

[0063] The spatial features of node i are calculated using the following formula. :

[0064] In the formula Here is the normalized current value at node j; The minimum value is set to prevent the denominator from being 0; this spatial feature is the weighted average of the currents of all neighboring nodes of node i, reflecting the overall current level of the local area where the node is located.

[0065] Construct the spatiotemporal feature vector of node i at time t for The spatiotemporal feature vectors of all nodes constitute the spatiotemporal feature set. for ;

[0066] S3. Using the spatiotemporal feature set obtained in step S2, train the dynamic isolation forest model and obtain an anomaly score set; specifically including the following steps:

[0067] The spatiotemporal coupling factor of node i at time t is calculated. for ;

[0068] Set time decay weight for ,in The set time decay parameter (preferably 0.02) represents the sample age;

[0069] The spatial augmentation coefficient of node i at time t was calculated. for ;in Set the space enhancement parameter (preferably 0.5);

[0070] The comprehensive weight of node i at time t is calculated. for ;

[0071] Build a global queue , Each element in the set corresponds to the spatiotemporal feature set at time t. ; upon reaching a new time t+1, then in Add the spatiotemporal feature set at time t+1. and remove the spatiotemporal feature set. ; For global queue The total number of spatiotemporal feature sets in the data (preferably 500);

[0072] According to the overall weight In the current global queue Sampling without replacement was performed to obtain... Each sample is used as a training subset of an isolation tree; sampling is repeated T times; each tree is recursively partitioned, the recursive partitioning process including: randomly selecting feature dimensions. , The value can be 1, 2, or 3; the threshold is selected evenly between the set minimum and maximum values. Divide the sample into left and right child nodes until the number of samples is no greater than 1 or the height of the tree reaches a certain value. , The function is a rounding function; given the load data, the model outputs the age factor for each tree. for and the feature mean vector of the training subset for Among them, among them, For training subset One of the samples; For the sample The corresponding overall weight; For the sample The age of the sample is the number of time steps since the current training moment; This is the training subset of the k-th isolation tree;

[0073] Calculate the global feature mean for time t. for The drift distance was calculated. for ,in The L2 norm is used; the adaptive correction factor is calculated. for , The first correction parameter is set (preferably 0.1). The second correction parameter is set (preferably with a value of 0.2). This represents the average drift distance.

[0074] For each node i and each tree k, obtain the original path length. , The number of edges from the root to the leaf is given; the corrected output path length is calculated. for ; Calculate the average corrected path length of node i for T represents the total number of isolated trees in the isolated forest; calculate the baseline path length. for , Let be the Euler-Macheroni constant (preferably 0.5772); calculate the anomaly score of node i at time t. for ;

[0075] Anomaly scores for all nodes This constitutes a set of abnormal scores;

[0076] S4. Obtain the current operating data information of the target power system; specifically including the following steps:

[0077] The voltage and current values ​​of each node in the target distribution network area are obtained at a set sampling period and preprocessed. The preprocessing process can be the preprocessing process described in step S2.

[0078] S5. Using the trained dynamic isolation forest model, obtain the current anomaly score of the target power system; specifically including the following steps:

[0079] The data obtained in step S4 is input into the trained dynamic isolated forest model to obtain the current anomaly score of each node n in the target power system. ;

[0080] S6. Based on the current anomaly score obtained in step S5, complete the identification of urban residential load electricity consumption risks in the target power system based on spatiotemporal coupling; specifically including the following steps:

[0081] Take the set quantile (The preferred value is 95%) (The preferred value is 98%) and (The preferred value is 99.5%) ;

[0082] Based on the set quantiles, the corresponding first threshold is obtained from the set of abnormal scores obtained in step S3. Second threshold and the third threshold , ;

[0083] Based on the current anomaly score of each node n in the target power system Determine the risk status of the target node n:

[0084] like If so, then the target node n is determined to be in a normal state;

[0085] like If so, then the target node n is determined to be in a state of interest;

[0086] like If so, then the target node n is determined to be in an overload state;

[0087] like If so, then the target node n is determined to be in a severely abnormal state;

[0088] In practice, the following steps may also be included:

[0089] If node n in the target power system is in an overload state, then the following determination is made again:

[0090] like If so, then node n is determined to be in a current overload state;

[0091] like If so, then node n is determined to be in a voltage overload state;

[0092] Otherwise, determine that node n is in a composite overload state;

[0093] The set current threshold (preferably 0.9) is used. The set voltage threshold (preferably 0.2);

[0094] The frequency of use of feature dimensions in all split nodes of the current dynamic isolated forest model is statistically analyzed, and the dimension with the highest frequency of use is selected as the dominant cause of the anomaly of node n; the feature dimensions include current dimension, voltage dimension and spatial feature dimension.

[0095] The regional risk index of the region where node n is located is calculated. for N is the total number of nodes in the region where node n is located;

[0096] If the regional risk index If the value exceeds the set value (preferably 0.7), a regional risk warning will be issued, indicating that the overall power consumption risk in the area where node n is located is relatively high.

[0097] This invention introduces spatiotemporal coupling features and a dynamic adaptive mechanism. First, it acquires voltage and current time-series data from multiple monitoring nodes in the distribution network. A spatial correlation matrix is ​​constructed based on the topology, and the spatial features of each node are obtained through weighted aggregation. These features are then fused with the voltage and current time features to form a three-dimensional spatiotemporal feature vector. An adaptive weighted sliding window isolated forest model is constructed, retaining recent samples with fixed-time decay weights through a sliding window. Spatial difference-enhanced sampling is introduced to make the model more attentive to local anomaly patterns. Multiple isolated trees are trained based on weighted sampling, recording the tree age factor and the mean of the training subset features for each tree. In the real-time identification stage, considering both the tree age and the current global distribution drift, an adaptive path length correction factor is designed to reduce the impact of older trees and trees with significant drift on the anomaly score, resulting in a dynamic anomaly score. Finally, based on the anomaly score distribution of historical normal samples, a percentile threshold is adaptively determined, classifying risks into normal, watch-out, overload, and severe anomaly levels, and outputting the anomaly attribution.

[0098] like Figure 2The diagram shows the functional modules of the system of this invention: The system disclosed in this invention for implementing the urban residential load electricity risk identification method based on spatiotemporal coupling includes a data acquisition module, a data processing module, a model training module, a real-time acquisition module, a score calculation module, and a risk identification module; the data acquisition module, data processing module, model training module, real-time acquisition module, score calculation module, and risk identification module are connected in series; the data acquisition module is used to acquire historical operating data information of the target power system and upload the data information to the data processing module; the data processing module is used to preprocess the acquired data information according to the received data information to construct a spatiotemporal feature set and upload the data information to the model training module; the model training module... The training module trains the dynamic isolated forest model using the received data and the obtained spatiotemporal feature set, generates an anomaly score set, and uploads the data to the real-time acquisition module. The real-time acquisition module acquires the current operating data of the target power system based on the received data and uploads the data to the score calculation module. The score calculation module uses the trained dynamic isolated forest model to obtain the current anomaly score of the target power system based on the received data and uploads the data to the risk identification module. The risk identification module identifies the spatiotemporally coupled urban residential load electricity consumption risk of the target power system based on the received data and the obtained current anomaly score.

Claims

1. A method for identifying urban residential electricity consumption risks based on spatiotemporal coupling, comprising the following steps: S1. Obtain historical operating data information of the target power system; S2. Preprocess the data information obtained in step S1 to construct a spatiotemporal feature set; S3. Using the spatiotemporal feature set obtained in step S2, train the dynamic isolation forest model and obtain an anomaly score set; S4. Obtain the current operating data information of the target power system; S5. Using the trained dynamic isolated forest model, the current anomaly score of the target power system is obtained; S6. Based on the current anomaly score obtained in step S5, complete the identification of urban residential load electricity consumption risks in the target power system based on spatiotemporal coupling.

2. The urban residential electricity consumption risk identification method based on spatiotemporal coupling according to claim 1, characterized in that... Step S1, which involves obtaining historical operating data information of the target power system, specifically includes the following steps: The voltage and current values ​​of each node in the target distribution network area are obtained at a set sampling period.

3. The urban residential load electricity consumption risk identification method based on spatiotemporal coupling according to claim 2, characterized in that... Step S2, which involves preprocessing the data obtained in step S1 to construct a spatiotemporal feature set, specifically includes the following steps: For the voltage and current values ​​obtained in step S1, a period is set, and the average voltage value and average current value are calculated for each period. These values ​​are then normalized to obtain the normalized voltage value for each node. and normalized current value ; Construct the spatial association matrix A between nodes as follows: , where N is the total number of nodes; This represents the electrical connection strength or spatial dependence between node i and node j, and it is set that if node i and node j are not directly connected, then... If node i and node j are the same node, then ; The spatial features of node i are calculated using the following formula. : In the formula Here is the normalized current value at node j; This is a set minimum value to prevent the denominator from being 0; Construct the spatiotemporal feature vector of node i at time t for The spatiotemporal feature vectors of all nodes constitute the spatiotemporal feature set. for .

4. The urban residential load electricity consumption risk identification method based on spatiotemporal coupling according to claim 3, characterized in that... Step S3, which involves using the spatiotemporal feature set obtained in step S2 to train the dynamic isolation forest model and obtain an anomaly score set, specifically includes the following steps: The spatiotemporal coupling factor of node i at time t is calculated. for ; Set time decay weight for ,in The time decay parameter is set, where q represents the sample age; The spatial augmentation coefficient of node i at time t was calculated. for ;in The set space enhancement parameters; The comprehensive weight of node i at time t is calculated. for ; Build a global queue , Each element in the set corresponds to the spatiotemporal feature set at time t. ; upon reaching a new time t+1, then in Add the spatiotemporal feature set at time t+1. and remove the spatiotemporal feature set. ; For global queue The total number of spatiotemporal feature sets in the data; According to the overall weight In the current global queue Sampling without replacement was performed to obtain... Each sample is used as a training subset of the isolation tree; Repeated sampling T times; Each tree is recursively partitioned. The recursive partitioning process includes: randomly selecting feature dimensions. , The value can be 1, 2, or 3; the threshold is selected evenly between the set minimum and maximum values. Divide the sample into left and right child nodes until the number of samples is no greater than 1 or the height of the tree reaches a certain value. , The function is a rounding function; given the load data, the model outputs the age factor for each tree. for and the feature mean vector of the training subset for ;in, For training subset One of the samples; For the sample The corresponding overall weight; For the sample Age; This is the training subset of the k-th isolation tree; Calculate the global feature mean for time t. for The drift distance was calculated. for ,in The L2 norm is used; the adaptive correction factor is calculated. for , The first correction parameter is set. The second correction parameter is set. This represents the average drift distance. For each node i and each tree k, obtain the original path length. , The number of edges from the root to the leaf is given; the corrected output path length is calculated. for ; Calculate the average corrected path length of node i for T represents the total number of isolated trees in the isolated forest; calculate the baseline path length. for , The Euler-Macheroni constant is used; the anomaly fraction of node i at time t is calculated. for ; Anomaly scores for all nodes This constitutes a set of abnormal scores.

5. The urban residential load electricity consumption risk identification method based on spatiotemporal coupling according to claim 4, characterized in that... Step S4, which involves obtaining the current target power system's operational data, specifically includes the following steps: The voltage and current values ​​of each node in the target distribution network area are obtained at a set sampling period and then preprocessed.

6. The urban residential load electricity consumption risk identification method based on spatiotemporal coupling according to claim 5, characterized in that... Step S5, which uses the trained dynamic isolation forest model to obtain the current anomaly score of the target power system, specifically includes the following steps: The data obtained in step S4 is input into the trained dynamic isolated forest model to obtain the current anomaly score of each node n in the target power system. .

7. The urban residential load electricity consumption risk identification method based on spatiotemporal coupling according to claim 6, characterized in that... Step S6, which involves identifying the urban residential load electricity consumption risk of the target power system based on the current anomaly score obtained in step S5, specifically includes the following steps: Take the set quantile , and , ; Based on the set quantiles, the corresponding first threshold is obtained from the set of abnormal scores obtained in step S3. Second threshold and the third threshold , ; Based on the current anomaly score of each node n in the target power system Determine the risk status of the target node n: like If so, then the target node n is determined to be in a normal state; like If so, then the target node n is determined to be in a state of interest; like If so, then the target node n is determined to be in an overload state; like If so, the target node n is determined to be in a severely abnormal state.

8. The urban residential load electricity consumption risk identification method based on spatiotemporal coupling according to claim 7, characterized in that... Step S6 further includes the following steps: If node n in the target power system is in an overload state, then the following determination is made again: like If so, then node n is determined to be in a current overload state; like If so, then node n is determined to be in a voltage overload state; Otherwise, determine that node n is in a composite overload state; For the set current threshold, The set voltage threshold; The frequency of use of feature dimensions in all split nodes of the current dynamic isolated forest model is statistically analyzed, and the dimension with the highest frequency of use is selected as the dominant cause of the anomaly of node n; the feature dimensions include current dimension, voltage dimension and spatial feature dimension. The regional risk index of the region where node n is located is calculated. for N is the total number of nodes in the region where node n is located; If the regional risk index If the value exceeds the set threshold, a regional risk warning will be issued.

9. A system for implementing the urban residential load electricity consumption risk identification method based on spatiotemporal coupling as described in any one of claims 1 to 8, characterized in that... It includes a data acquisition module, a data processing module, a model training module, a real-time acquisition module, a score calculation module, and a risk identification module; the data acquisition module, data processing module, model training module, real-time acquisition module, score calculation module, and risk identification module are connected in series; the data acquisition module is used to acquire historical operating data information of the target power system and upload the data information to the data processing module; The data processing module is used to preprocess the acquired data information based on the received data information in order to construct a spatiotemporal feature set and upload the data information to the model training module; The model training module is used to train the dynamic isolation forest model based on the received data information and the obtained spatiotemporal feature set, obtain the anomaly score set, and upload the data information to the real-time acquisition module. The real-time acquisition module is used to obtain the current operating data information of the target power system based on the received data information, and upload the data information to the score calculation module; The score calculation module is used to obtain the current anomaly score of the target power system based on the received data information and the trained dynamic isolated forest model, and then upload the data information to the risk identification module. The risk identification module is used to identify the urban residential load electricity consumption risk of the target power system based on the received data and the obtained current anomaly score.

Images