A method, system, equipment, and medium for classifying and managing multi-source power demand data.

Abstract

This invention relates to the field of power data technology and discloses a method, system, device, and medium for classifying and managing multi-source power demand data. The method includes: constructing a multi-layer graph structure based on the acquired multi-source power demand dataset; constructing a search sample set based on the multi-layer graph structure, where each search sample includes a color vector, a state vector, and a fitness state. The color vector is determined by semantic features, and the fitness state is set to reflect the semantic connectivity of the semantic graph, the state fluctuation amplitude of the state graph, and the adjacency density of the spatial adjacency graph; iteratively updating the search sample set using an improved peacock optimization algorithm to obtain the optimal fitness state of each search sample, thereby determining the classification label corresponding to the search sample. The improved peacock optimization algorithm is configured to iteratively update each search sample based on the perturbation guidance vector and the direction of the gravitational gradient. This method improves the accuracy and adaptability of power demand sorting.

Description

Technical Field

[0001] This invention relates to the field of power data management technology, and in particular to a method, system, device and medium for classifying and managing multi-source power demand data. Background Technology

[0002] Power companies need to handle a massive number of user requests every day, covering issues such as abnormal electricity bills, equipment failures, power outage repairs, and policy inquiries. Accurately classifying and assigning these requests to the appropriate departments is key to improving service efficiency.

[0003] Current mainstream methods for classifying and managing electricity complaints rely on keyword matching or simple rules. For example, they preset keywords like "overcharged electricity bill" or "no power at home" for keyword matching, and set rules like "anything containing 'fault' is classified as an equipment problem" for simple rule matching. However, electricity users' expressions are flexible and diverse, and the same complaint can be worded differently. Fixed keywords and simple rules are prone to missing information or misjudging, leading to processing delays and / or repeated user feedback. To address these issues, an algorithmic model has been introduced to optimize the electricity complaint classification process. However, this approach has the following drawbacks: the algorithm's process and parameters are fixed, making it unable to adapt to abnormal samples or sudden data changes, resulting in decreased classification accuracy; and it has low sensitivity to changes in user behavior, making it difficult for the model to adjust its strategy in real time when user electricity usage and complaint preferences change over time, resulting in insufficient adaptability.

[0004] Therefore, improving the accuracy and adaptability of electricity demand classification has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] This invention provides a method, system, device, and medium for classifying and managing multi-source power demand data, in order to solve the technical problem of improving the accuracy and adaptability of power demand classification, and to achieve the effect of improving the accuracy and adaptability of power demand sorting.

[0006] In a first aspect, the present invention provides a method for classifying and managing multi-source electricity demand data, the method comprising:

[0007] Based on the acquired multi-source power demand dataset, a multi-layer graph structure is constructed, which includes a semantic graph, a state graph, and a spatial adjacency graph.

[0008] Based on the multi-layer graph structure, a search sample set is constructed. Each search sample in the search sample set includes a color vector, a state vector, and a fitness state. The color vector is determined by the semantic features of the semantic graph, the state vector is determined by the state data of the state graph, and the fitness state is set to reflect the semantic connectivity of the semantic graph, the state fluctuation amplitude of the state graph, and the adjacency density of the spatial adjacency graph.

[0009] Based on the fitness state, a perturbation guidance vector is obtained. Based on the perturbation guidance vector, the color vector, and the preset target color vector, the gravitational gradient direction of the search sample is constructed.

[0010] An improved peacock optimization algorithm is used to iteratively update the search sample set to obtain the optimal fitness state of each search sample, and the classification label of each search sample is determined according to the optimal fitness state. The improved peacock optimization algorithm is set to execute the perturbation caused by the fitness state according to the perturbation guidance vector, and drive the movement direction of the color vector and the state vector according to the gravity gradient direction.

[0011] Preferably, the step of constructing a multi-layer graph structure based on the acquired multi-source power demand dataset includes:

[0012] Obtain a multi-source power demand dataset composed of multi-source power demand data from various power users, wherein the multi-source power demand data includes customer demand text, identity information, equipment status information, and geographical location information;

[0013] Using each of the multi-source power demand data as nodes, and the similarity between the semantic feature vectors of each of the customer demand texts as the first edge, a semantic graph is constructed. Using the time transition relationship of the state data composed of each of the identity information, the device status information and the geographical location information as the second edge, a state graph is constructed. Using the spatial distance between the geographical locations displayed by each of the geographical location information as the third edge, a spatial adjacency graph is constructed.

[0014] A multi-layer graph structure is constructed based on the semantic graph, the state graph, and the spatial adjacency graph.

[0015] Preferably, constructing the search sample set based on the multi-layer graph structure includes:

[0016] Based on the semantic feature vectors of each node, a color vector is obtained; based on the identity information, device status information, and geographical location information of each node, a status vector for each multi-source power demand data is constructed.

[0017] Based on the semantic graph, the semantic connectivity between each node is obtained; based on the state graph, the state fluctuation amplitude between each node is obtained; based on the spatial adjacency graph, the adjacency density between each node is obtained; and based on the semantic connectivity, the state fluctuation amplitude, and the adjacency density, the fitness state of each multi-source power demand data is constructed.

[0018] A search sample set is constructed based on the color vector, the state vector, and the fitness state of each of the multi-source power demand data.

[0019] Preferably, obtaining the perturbation guidance vector based on the fitness state includes:

[0020] Based on the semantic connectivity of each node, the connectivity difference between each node and its neighboring nodes is obtained, and the mean and standard deviation of the connectivity difference of each node are used as the structural distribution index of the search sample corresponding to the node.

[0021] Based on the state fluctuation amplitude, the adjacency density, and the structure distribution index, a disturbance guidance vector is constructed.

[0022] Preferably, constructing the gravitational gradient direction of the search sample based on the perturbation guidance vector, the color vector, and the preset target color vector includes:

[0023] Calculate the channel difference between the color vector and the preset target color vector, and normalize the channel difference to obtain the attraction direction vector;

[0024] Based on the disturbance guidance vector and the attraction direction vector, a position state driving vector is obtained, and the state vector of the search sample is updated based on the position state driving vector.

[0025] Search samples with fitness states higher than the average are selected from the search sample set as gravitational centers, and anomaly gravitational potential functions are constructed based on the gravitational centers.

[0026] Based on the anomalous gravitational potential function, the gradient of the updated state vector of the search sample is calculated to obtain the position state gradient vector.

[0027] The position state gradient vector is normalized to obtain the gravitational gradient direction of the search sample.

[0028] Preferably, the step of iteratively updating the search sample set using the improved peacock optimization algorithm to obtain the optimal fitness state of each search sample includes:

[0029] Based on the color vector and the preset target color vector, the semantic tag matching degree is obtained, and based on the position state gradient vector, the gravitational response degree is obtained.

[0030] A multi-objective fitness function is constructed based on the semantic tag matching degree, the gravitational response degree, the semantic connectivity degree, the state fluctuation amplitude, and the adjacency density.

[0031] Based on the iteration sequence of each search sample in the iterative process of the improved peacock optimization algorithm, a historical behavior trajectory tensor is constructed, and the historical behavior trajectory tensor and the graph state information tensor of the multi-layer graph structure are input into a pre-constructed gated neural network structure to obtain the peacock behavior gating factor.

[0032] Based on the perturbation guidance vector, the behavior of a peacock exploring a new region is simulated to perform the perturbation caused by the fitness state. Based on the gravity gradient direction, the movement direction of the color vector and the state vector is guided. Based on the peacock behavior gating factor, the peacock behavior is controlled to iteratively update each search sample.

[0033] Based on the multi-objective fitness function, each of the search samples after iterative updates is judged to determine the optimal fitness state of each search sample.

[0034] Preferably, the step of inputting the historical behavior trajectory tensor and the graph state information tensor of the multi-layer graph structure into a pre-constructed gated neural network structure to obtain the peacock behavior gating factor includes:

[0035] The historical behavior trajectory tensor and the graph state information tensor of the multi-layer graph structure are concatenated to obtain a joint feature tensor;

[0036] The joint feature tensor is input into a gated neural network structure to obtain peacock behavior gating factors, which include peacock behavior gating factors, jumping behavior gating factors, and trajectory splitting behavior gating factors.

[0037] Secondly, the present invention also provides a multi-source power demand data classification and management system to implement the multi-source power demand data classification and management method described above. The system includes: a multi-layer graph structure construction module, a search sample set construction module, a parameter construction module, and an iterative optimization and update module.

[0038] The multi-layer graph structure construction module is used to construct a multi-layer graph structure based on the acquired multi-source power demand dataset. The multi-layer graph structure includes a semantic graph, a state graph, and a spatial adjacency graph.

[0039] The search sample set construction module is used to construct a search sample set according to the multi-layer graph structure. Each search sample in the search sample set includes a color vector, a state vector, and a fitness state. The color vector is determined by the semantic features of the semantic graph, the state vector is determined by the state data of the state graph, and the fitness state is set to reflect the semantic connectivity of the semantic graph, the state fluctuation amplitude of the state graph, and the adjacency density of the spatial adjacency graph.

[0040] The parameter construction module is used to obtain a perturbation guidance vector based on the fitness state, and to construct the gravitational gradient direction of the search sample based on the perturbation guidance vector, the color vector, and a preset target color vector.

[0041] The iterative optimization update module is used to iteratively update the search sample set using an improved peacock optimization algorithm to obtain the optimal fitness state of each search sample, and to determine the classification label of each search sample based on the optimal fitness state. The improved peacock optimization algorithm is configured to execute the perturbation caused by the fitness state according to the perturbation guidance vector, and to drive the movement direction of the color vector and the state vector according to the gravity gradient direction.

[0042] Thirdly, the present invention also provides a computer device, the computer device including a memory, a processor and a transceiver, which are connected to each other via a bus; the memory is used to store a set of computer program instructions and data, and to transmit the stored data to the processor, the processor executing the computer program instructions stored in the memory to execute the above-described multi-source power demand data classification and management method.

[0043] Fourthly, the present invention also provides a computer-readable storage medium storing a computer program that, when the computer program is run, implements the above-described method for classifying and managing multi-source power demand data.

[0044] This application provides a method, system, device, and medium for classifying and managing multi-source power demand data. Compared with the prior art, the beneficial effects of the embodiments of this application are as follows:

[0045] By integrating heterogeneous graph structures such as semantic graphs, state graphs, and spatial adjacency graphs, the algorithm accurately expresses the semantic, state, and spatial coupling relationships between customer demands, enhancing modeling capabilities in complex customer contexts and providing structural support for multi-dimensional feature fusion and reasoning. Combining key indicators such as semantic connectivity, state fluctuation, and adjacency density, it achieves a quantitative expression of the semantic consistency, state stability, and neighborhood structure of power customer demands. A perturbation guidance mechanism and a gravitational potential field function are constructed, dynamically generating gravitational gradient directions using the differences between semantic features and spatial adjacency relationships to guide the movement of search samples in the feature space. Simultaneously, a gated neural network is introduced based on historical behavioral trajectories and graph state information to achieve adaptive control of search sample boasting, jumping, and trajectory splitting behaviors, improving flexibility and responsiveness when handling complex customer state changes. The improved Peacock Optimization algorithm uses color vectors, state vectors, and fitness states as evolutionary variables, integrating three types of driving factors: perturbation guidance, gravitational attraction, and behavioral control. This achieves precise screening and evolutionary convergence of key samples, effectively improving the accuracy of demand sorting and the stability of anomaly identification. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the steps of a multi-source power demand data classification and management method provided in a preferred embodiment of the present invention;

[0047] Figure 2 This is a schematic diagram of the structure of a multi-source power demand data classification and management system provided in a preferred embodiment of the present invention;

[0048] Figure 3 This is an internal structural diagram of the computer device in an embodiment of the present invention;

[0049] Figure label:

[0050] 1- Multi-layer graph structure construction module, 2- Search sample set construction module, 3- Parameter construction module, 4- Iterative optimization and update module. Detailed Implementation

[0051] The embodiments of the present invention are described in detail below with reference to the accompanying drawings. The embodiments are provided for illustrative purposes only and should not be construed as limiting the scope of the invention. The accompanying drawings are for reference and illustration only and do not constitute a limitation on the scope of protection of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of this invention.

[0052] In the description of this invention, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0053] Please see Figure 1 The diagram illustrates the steps of a multi-source power demand data classification and management method. In an embodiment of the present invention, a multi-source power demand data classification and management method is provided, the method comprising:

[0054] S1. Based on the acquired multi-source power demand dataset, a multi-layer graph structure is constructed, which includes a semantic graph, a state graph, and a spatial adjacency graph. In a preferred embodiment of this application, the collected multi-source power demand dataset includes multi-source power demand data from various power users. The multi-source power demand data includes customer demand text, identity information, equipment status information, and geographical location information. The identity information includes attributes such as customer category (e.g., residential users and industrial users), contract number, electricity consumption level, payment records, historical demand information, and the distribution area to which the customer belongs. These attributes are used to represent the customer's identity characteristics in the power service system in the state graph. Through joint modeling with equipment status information and geographical location information, the system can distinguish the semantic differences and risk sensitivities of different customer types in the subsequent multi-objective fitness function calculation. For example, residential customers and industrial customers have different urgency under the same semantic expression. The system can assign differentiated weights when filtering and retaining search samples based on identity information, thereby improving the accuracy of sorting and early warning. Equipment status information includes voltage, current, power factor, and tripping records of electricity meters; operating status, temperature, and load levels of power distribution equipment; alarm information and load changes of user-end equipment; and online status and communication quality of data acquisition terminals. This information is used to characterize the operating status of customer-used equipment and power distribution facilities in the status diagram. By integrating this information with identity information, customer request text, and geographic location information, subjective requests and objective status can be comprehensively considered during subsequent fitness state updates and anomaly gravitational potential function construction, thereby improving the accuracy and robustness of customer request sorting and anomaly identification. Geographic location information includes the latitude and longitude coordinates of the electricity address, the corresponding distribution substation number, and the location of the power transformer, reflecting the spatial distribution of equipment within the power network. The multi-layer graph structure includes a semantic graph, a state graph, and a spatial adjacency graph. All three graphs use various multi-source power demand data as nodes. The semantic graph uses the semantic feature vectors of customer demand texts as node content, and the similarity of the semantic feature vectors of customer demand texts between nodes is used as the first edge connection. A connection is formed when the similarity exceeds a preset similarity threshold; otherwise, no connection is formed. The state graph uses state data formed by the combination of identity information, equipment status information, and geographical location information as node content, and the time transition relationship of the state data between nodes is used as the second edge connection. The spatial adjacency graph uses geographical location information as node content, constructs spatial location vectors based on the geographical location information, and then calculates the spatial distance between nodes based on the spatial location vectors of each node. The spatial distance between nodes represents the third edge connection. A connection is established when the spatial distance between any two nodes is less than a preset first spatial distance threshold; otherwise, no connection is established. The edge weight of the spatial adjacency graph is represented by the reciprocal of the spatial distance.

[0055] In the preferred embodiment of this application, by integrating heterogeneous graph structures such as semantic graphs, state graphs and spatial adjacency graphs, the coupling relationship between customer demands in semantics, state and space is accurately expressed, the modeling capability in complex customer contexts is improved, and structural support is provided for multi-dimensional feature fusion and reasoning.

[0056] S2. Based on the multi-layer graph structure, a search sample set is constructed. Each search sample in the search sample set includes a color vector, a state vector, and a fitness state. The color vector is determined by the semantic features of the semantic graph, the state vector is determined by the state data of the state graph, and the fitness state is set to reflect the semantic connectivity of the semantic graph, the state fluctuation amplitude of the state graph, and the adjacency density of the spatial adjacency graph. In a preferred embodiment of this application, the color vector of each search sample includes at least the semantic features of the customer request text of the search sample, determined by the semantic features of the semantic graph. The state vector of each search sample includes at least the identity information, device status information, and geographical location information of the search sample. The state vector is a vectorized representation obtained after encoding and normalization based on the identity information, device status information, and geographical location information. The fitness state of each search sample includes at least the semantic connectivity between the node of the search sample and the nodes of other search samples in the semantic graph, the state fluctuation amplitude between the nodes of other search samples in the state graph, and the adjacency density between the nodes of other search samples in the spatial adjacency graph. The semantic connectivity between each node is calculated based on the semantic graph. The formula for calculating semantic connectivity is:

[0057]

[0058] in, Indicates the search sample With search samples Semantic connectivity between them For search samples In the Normalized vector values ​​in each semantic feature dimension Indicates the search sample In the Normalized vector values ​​in each semantic feature dimension This represents the total number of dimensions of the semantic feature vectors.

[0059] Based on the state diagram, the state fluctuation amplitude between each node at different time series is obtained. The Euclidean distance between the state vectors of each search sample at adjacent time steps is used as the metric. The formula for calculating the state fluctuation amplitude is:

[0060]

[0061] in, Indicates the search sample exist The amplitude of state fluctuation at any given moment Indicates the search sample The state vector at time t, Indicates the search sample The state vector at time t-1 This represents the timing length.

[0062] Based on the spatial adjacency graph, the adjacency density between each node is obtained. Specifically, based on the spatial adjacency graph, the number of adjacent samples whose spatial distance between the search sample of each node and the search samples of its neighboring nodes is less than a preset second spatial distance threshold is counted to form the adjacency density. The formula for calculating the adjacency density is as follows:

[0063]

[0064] in, Indicates the search sample adjacency density, Indicates the search sample Spatial position vector, Indicates the search sample Spatial position vector, Indicates the search sample and search samples Spatial distance between them This represents the second spatial distance threshold. This represents an indicator function, which is 1 if the condition is true and 0 otherwise.

[0065] Furthermore, based on semantic connectivity, state fluctuation amplitude, and adjacency density, the fitness state of each multi-source power demand data is constructed. Based on the color vector, state vector, and fitness state of each multi-source power demand data, search samples for the corresponding multi-source power demand data are constructed. A search sample set is constructed from all search samples.

[0066] In the preferred embodiment of this application, a structured search sample is constructed, and key indicators such as semantic connectivity, state fluctuation and adjacency density are combined to realize the quantitative expression of semantic consistency, state stability and neighborhood structure of power customer demands, providing basic data for subsequent improvement of the fitness state evaluation and behavior guidance of the peacock optimization algorithm.

[0067] S3. Based on the fitness state, a perturbation guidance vector is obtained. Based on the perturbation guidance vector, the color vector, and the preset target color vector, the gravitational gradient direction of the search sample is constructed. In a preferred embodiment of this application, an improved peacock optimization algorithm is used to iteratively update the search sample set. The peacock optimization algorithm is a swarm intelligence optimization algorithm inspired by the "tail display" and "group interaction" behavior of peacocks during courtship. Its core idea is to simulate the natural selection mechanism of peacocks competing for mates by displaying their tail feathers, transforming "tail feather attraction" into "fitness evaluation" in the algorithm. It finds the optimal solution through competition and cooperation among individuals within the group, possessing advantages such as high search accuracy, fast convergence speed, strong robustness, and simple parameters. In this improved peacock optimization algorithm, in each iteration, the perturbation caused by the fitness state is adjusted according to the perturbation guidance vector, and the gravitational gradient direction is used to drive the color vector and state vector.

[0068] For the construction of the perturbation guidance vector, based on each semantic connectivity, the connectivity difference between each node and its neighboring nodes is calculated, and the mean and standard deviation of the connectivity difference of each node are used as the structural distribution index of the search samples of that node.

[0069] Furthermore, based on the state fluctuation amplitude, adjacency density, and structural distribution indicators, a disturbance guidance vector is constructed. Specifically, a behavioral fluctuation score is generated based on the state fluctuation amplitude, and a score amplification mapping is performed using a monotonically increasing exponential function to obtain a risk score. After normalizing the risk score, adjacency density, and structural distribution indicators, a disturbance guidance vector is constructed according to the weights. The disturbance guidance vector is expressed as:

[0070]

[0071] in, Indicates the search sample The perturbation guiding vector. , , The weighting coefficients are for the direction of the disturbance. Indicates the search sample Structural distribution indicators Indicates the search sample Risk score, Indicates the search sample adjacency density, This represents the minimum-maximum normalization function.

[0072] For constructing the gravitational gradient direction, an attraction direction vector is built based on the color vector and the preset target color vector. The target color vector is the semantic label prototype vector of each sorting label. The semantic label prototype vector is obtained by normalizing the semantic feature vectors of the historical customer request text under that sorting label, calculating the mean, and then normalizing the mean again. Based on the channel difference between the color vector of the current search sample and the target color vector, an attraction direction vector is constructed. Furthermore, the normalized perturbation guidance vector and the attraction direction vector are weighted and superimposed to generate a position state driving vector. The position vector of the search sample is updated based on the position state driving vector. The position vector of the search sample consists of the color vector and state vector corresponding to the search sample. The update of the position vector is to iteratively adjust the current position vector of the search sample guided by the position state driving vector. That is, in each iteration, the current position vector of the search sample is added to its corresponding position state driving vector, and the update amplitude is controlled by a set step size factor to obtain a new position vector. This allows the search sample to gradually move to a region with higher fitness under the combined effect of the perturbation guidance vector and the attraction direction vector. It effectively combines the differences in semantic features of the appeal text and the structural risk perturbation information to improve the pertinence and accuracy of the dynamic position adjustment process.

[0073] In a preferred embodiment of this application, search samples with fitness states higher than the average are selected from the search sample set as gravitational centers, and anomaly gravitational potential functions are constructed based on the gravitational centers. The anomaly gravitational potential function is expressed as:

[0074]

[0075] in, Indicates the search sample The anomalous gravitational potential function value, Indicates the number of gravitational centers. Represents the center of gravity fitness status, Indicates the first A gravitational center, Indicates the search sample With the The shortest path distance between each center of gravity in the semantic graph Indicates the search sample and the center of gravity Euclidean distance in a spatial adjacency graph A coefficient representing the degree to which semantic distance and spatial distance are balanced.

[0076] Furthermore, based on the anomalous gravitational potential function, the gradient of the updated state vector of the search sample is calculated to obtain the position-state gradient vector, which is expressed as:

[0077]

[0078] in, Indicates the search sample The position state gradient vector, Represents the center of gravity The spatial location vector.

[0079] Furthermore, the position state gradient vector is normalized to generate the gravitational gradient direction.

[0080] In a preferred embodiment of this application, a fitness state-driven anomalous gravitational potential modeling mechanism is introduced to quantify the intensity and direction of gravitational action in both spatial location and semantic features. The positional state gradient vector is constructed by combining the shortest semantic path and spatial distance, providing a physical perception basis and algorithmic support for guiding the aggregation of search samples.

[0081] S4. The improved peacock optimization algorithm is used to iteratively update the search sample set to obtain the optimal fitness state of each search sample, and the classification label of each search sample is determined according to the optimal fitness state. The improved peacock optimization algorithm is set to execute the perturbation caused by the fitness state according to the perturbation guidance vector, and drive the movement direction of the color vector and the state vector according to the gravity gradient direction. In a preferred embodiment of this application, the semantic label matching degree is obtained according to the color vector of the search sample and the target color vector, and the semantic label matching degree is expressed as:

[0082]

[0083] in, Indicates the first The semantic label matching degree of each search sample. express and cosine similarity, Indicates the first The color vector of each search sample, This represents the target color vector.

[0084] Based on the position-state gradient vector, the gravitational responsivity is obtained, which is expressed as:

[0085]

[0086] in, Indicates the first Gravitational response of each search sample, express The 2-norm, This represents the Sigmoid function. This represents the slope adjustment coefficient. .

[0087] Based on semantic label matching degree, gravitational response degree, semantic connectivity, state fluctuation amplitude, and adjacency density, a multi-objective fitness function is constructed. In practical applications, the fitness state value of the improved peacock optimization algorithm is calculated according to the multi-objective fitness function. The multi-objective fitness function is expressed as:

[0088]

[0089] in, Represents the semantic graph and the search sample A set of adjacent samples that are connected. express The base number, This indicates that a minimum-maximum normalization mapping is performed on the current set of samples in the iteration, mapping the corresponding quantity to... interval, This represents a non-negative weighting coefficient used to balance the contributions of semantic label matching, gravitational response, semantic connectivity, state fluctuation amplitude, and adjacency density to the fitness state, and satisfies the following conditions: .

[0090] In a preferred embodiment of this application, the multi-objective fitness function integrates semantic consistency and spatial responsiveness, and expresses nonlinear adjustment relationships through the response function, thereby enhancing the discriminative ability and adaptability of search samples in high-dimensional space.

[0091] During the iterative update of the search sample set using the improved peacock optimization algorithm, the peacock behavior is controlled by a peacock behavior gating factor. For the peacock behavior gating factor, a graph state information tensor is constructed based on the graph node states corresponding to the search samples in the semantic graph, state graph, and spatial adjacency graph of the multi-layer graph structure. A historical behavior trajectory tensor is constructed based on the iterative sequence of each search sample in the improved peacock optimization algorithm. The historical behavior trajectory tensor and the graph state information tensor are concatenated to obtain a joint feature vector. Then, a pre-constructed gated neural network structure is used to process the joint feature vector to obtain the peacock behavior gating factor. The peacock behavior gating factor includes a show-off gating factor, a jump gating factor, and a trajectory splitting gating factor, which are represented as follows:

[0092]

[0093]

[0094]

[0095] in, This indicates a gating factor for ostentatious behavior. This represents the gating factor for jump behavior. The gating factor represents the trajectory splitting behavior. This represents the joint feature tensor obtained by concatenating the graph state information tensor and the historical behavior trajectory tensor. It is the Sigmoid activation function. This represents the weight matrix corresponding to the boasting behavior. This represents the weight matrix corresponding to the jump behavior. The weight matrix represents the trajectory splitting behavior. This represents the bias vector corresponding to the boasting behavior. This represents the bias vector corresponding to the jump behavior. This represents the bias vector corresponding to the trajectory splitting behavior.

[0096] In a preferred embodiment of this application, a behavior control mechanism that integrates graph state information and historical behavior trajectories is constructed to output a behavior gating factor from a gated neural network. This effectively controls the jumping behavior, show-off behavior, and trajectory splitting behavior of search samples in the subsequent improved peacock optimization process, thereby enhancing the plasticity and convergence stability of the improved peacock optimization algorithm.

[0097] Furthermore, color vector, state vector, and fitness state are set as evolutionary variables for the search samples. The search sample set is initialized, and an upper limit for the number of iterations and a fitness state convergence threshold are set. In each iteration, the perturbation guidance vector generated by the fitness state, the position state gradient vector calculated by the anomalous gravitational potential function, and the peacock behavior gating factor generated by the graph state information tensor and the historical behavior trajectory tensor through a gated neural network are input. Perturbation adjustment, gravitational attraction, and behavior control operations are performed to update the color vector, state vector, and fitness state of each search sample. After the update, the multi-objective fitness function is called to calculate the fitness state value of each search sample and record the current best sample. Based on the output value of the peacock behavior gating factor and the preset... The comparison results of the gating factor thresholds determine and control the behavior of the search sample. When the boasting gating factor exceeds the preset boasting gating factor threshold, the position state of the search sample is adjusted to the historical best position. When the jump gating factor exceeds the preset jump gating factor threshold, a random perturbation vector of preset amplitude is superimposed on the current position to update the position state. When the trajectory splitting gating factor exceeds the preset trajectory splitting gating factor threshold, the current search sample is copied and different small perturbations are added to achieve multi-path exploration. After the position state adjustment is completed, the fitness state is updated in real time based on the new position state and related features. When the termination condition is met, the final search sample set is output for generating sorting labels and anomaly identification results.

[0098] Specifically, it determines whether the preset number of iterations has been reached or whether the fitness state value is greater than the preset fitness state threshold. If not, it indicates that the iteration termination condition has not been met, and the next round of iteration continues. If yes, it indicates that the iteration termination condition has been met, the iteration process terminates, and the final search sample set is retained. Specifically, it first calculates the color vector similarity between the color vector of the target customer's final search sample and the color vectors of each sorting label, and the state vector similarity between the target customer's final search sample and the state vectors of each sorting label. Then, it performs a weighted sum calculation of the color vector similarity, state vector similarity, and the fitness state of the final search sample to obtain the target customer's comprehensive score on each sorting label, thereby determining the customer's sorting label. Sorting labels include main sorting labels and auxiliary sorting labels.

[0099] Furthermore, based on the weighted sum of the abnormal gravitational potential function value, state fluctuation amplitude, adjacency density, and structural distribution index of the final search sample of the target customer, an anomaly score is obtained. Based on the relationship between the anomaly score and the preset anomaly score threshold, abnormal samples are identified, and the anomaly identification results of the abnormal samples are output.

[0100] In a preferred embodiment of this application, the improved peacock optimization algorithm uses color vector, state vector and fitness state as evolutionary variables, integrates three types of driving factors: perturbation guidance, attraction and behavior control, and performs a global-local joint evolutionary process to effectively achieve sorting optimization and anomaly detection for customer demands.

[0101] In a preferred embodiment of this application, the multi-source power demand data classification and management method of this application is applied to the customer demand processing flow of a power service unit. A comparative analysis is conducted using actual operational data. The unit's original demand processing mainly relied on keyword screening and manual preliminary review. When faced with complex or vaguely worded demands, it often failed to promptly determine the problem type, especially when there were equipment malfunctions, high location concentration, or / and customer status anomalies. The manual method often resulted in delays or misjudgments. Fifteen days of historical multi-source power demand data from the power service unit were collected, including customer demand text, identity information, equipment status information, and geographical location information. The multi-source power demand data classification and management method of this application was used to generate sorting labels and early warning results, which were then compared with the results of manual processing. The comparison results are shown in Table 1.

[0102] Table 1

[0103]

[0104] As shown in Table 1, the multi-source power demand data classification and management method of this application has significantly improved response time and sorting efficiency. Furthermore, this application can perform more in-depth modeling of power customer intent, reducing judgment errors caused by ambiguous expressions or failure to promptly report equipment malfunctions. The increase in the number of malfunction warning hits also reflects that the multi-source power demand data classification and management method of this application has a strong perception capability in capturing abnormal behavior patterns.

[0105] Specifically analyzing certain abnormal samples, such as when a customer does not explicitly state that their equipment is malfunctioning, but the background status graph shows that their electricity meter has tripped multiple times and the status information of nearby users' equipment also fluctuates significantly, this application can identify such potential risks through gravity function modeling and risk scoring mechanisms, while traditional manual methods fail to detect them in a timely manner due to information fragmentation.

[0106] Based on the above test results, the multi-source power demand data classification and management method of this application demonstrates better sorting response speed and higher anomaly detection rate in real data streams. It is particularly suitable for large-scale, information-complex power customer service scenarios and has good adaptability and practicality.

[0107] In a preferred embodiment of the present invention, a multi-layer graph structure is constructed based on the acquired multi-source power demand dataset. The multi-layer graph structure includes a semantic graph, a state graph, and a spatial adjacency graph. Based on the multi-layer graph structure, a search sample set is constructed. Each search sample in the search sample set includes a color vector, a state vector, and a fitness state. The color vector is composed of the semantic features of the semantic graph, the state vector is composed of the state data of the state graph, and the fitness state is set to reflect the semantic connectivity of the semantic graph, the state fluctuation amplitude of the state graph, and the adjacency density of the spatial adjacency graph. Based on the fitness state, a perturbation guidance vector is obtained, and based on the perturbation guidance vector, the gravitational gradient direction of the search sample is constructed. An improved peacock optimization algorithm is used to iteratively update the search sample set to obtain the optimal fitness state of each search sample. The classification label of each search sample is determined based on the optimal fitness state. The improved peacock optimization algorithm is set to execute the perturbation caused by the fitness state according to the perturbation guidance vector and drive the color vector and state vector according to the gravitational gradient direction. This application discloses a multi-source electricity demand data classification and management method. By integrating heterogeneous graph structures such as semantic graphs, state graphs, and spatial adjacency graphs, it accurately expresses the semantic, state, and spatial coupling relationships between customer demands, improving modeling capabilities in complex customer contexts and providing structural support for multi-dimensional feature fusion and reasoning. Combining key indicators such as semantic connectivity, state fluctuation, and adjacency density, it achieves a quantitative expression of the semantic consistency, state stability, and neighborhood structure of electricity customer demands. It constructs a perturbation guidance mechanism and a gravitational potential field function, and dynamically generates gravitational gradient directions using the differences between semantic features and spatial adjacency relationships to guide the movement of search samples in the feature space. At the same time, based on historical behavior trajectories and graph state information, it introduces a gated neural network to achieve adaptive control of the boasting, jumping, and trajectory splitting behaviors of search samples, improving the flexibility and responsiveness when handling complex customer state changes. The improved peacock optimization algorithm uses color vectors, state vectors, and fitness states as evolutionary variables, integrating three types of driving factors: perturbation guidance, gravitational attraction, and behavior control, to achieve accurate screening and evolutionary convergence of key samples, effectively improving the accuracy of demand sorting and the stability of anomaly identification.

[0108] Accordingly, such as Figure 2 The diagram shown illustrates the structure of a multi-source power demand data classification and management system. Based on a multi-source power demand data classification and management method, this embodiment of the invention also provides a multi-source power demand data classification and management system to implement the multi-source power demand data classification and management method disclosed in this embodiment. The system includes: a multi-layer graph structure construction module 1, a search sample set construction module 2, a parameter construction module 3, and an iterative optimization and update module 4.

[0109] The multi-layer graph structure construction module 1 is used to construct a multi-layer graph structure based on the acquired multi-source power demand dataset. The multi-layer graph structure includes a semantic graph, a state graph, and a spatial adjacency graph.

[0110] The search sample set construction module 2 is used to construct a search sample set according to the multi-layer graph structure. Each search sample in the search sample set includes a color vector, a state vector, and a fitness state. The color vector is determined by the semantic features of the semantic graph, the state vector is determined by the state data of the state graph, and the fitness state is set to reflect the semantic connectivity of the semantic graph, the state fluctuation amplitude of the state graph, and the adjacency density of the spatial adjacency graph.

[0111] The parameter construction module 3 is used to obtain a perturbation guidance vector based on the fitness state, and to construct the gravitational gradient direction of the search sample based on the perturbation guidance vector, the color vector, and the preset target color vector.

[0112] The iterative optimization update module 4 is used to iteratively update the search sample set using an improved peacock optimization algorithm to obtain the optimal fitness state of each search sample, and to determine the classification label of each search sample based on the optimal fitness state. The improved peacock optimization algorithm is configured to execute the perturbation caused by the fitness state according to the perturbation guidance vector, and to drive the movement direction of the color vector and the state vector according to the gravity gradient direction.

[0113] For specific limitations regarding the multi-source power demand data classification and management system, please refer to the above-described limitations regarding the multi-source power demand data classification and management method, which will not be repeated here. Those skilled in the art will recognize that the various modules and steps described in conjunction with the embodiments disclosed in this invention can be implemented in hardware, software, or a combination of both. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.

[0114] like Figure 3 The diagram shows the internal structure of a computer device. An embodiment of the present invention provides a computer device including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the steps described in the embodiment of the multi-source power demand data classification and management method, for example... Figure 1 Steps S1 to S4 as described above.

[0115] Those skilled in the art will understand that the illustrations Figure 3This is merely an example of a computer device and does not constitute a limitation on the computer device. It may include more or fewer components than shown, or combine certain components, or different components. For example, the computer device may also include input / output devices, network access devices, buses, etc.

[0116] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the computer device, connecting various parts of the computer device via various interfaces and lines.

[0117] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the computer device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0118] If the modules integrated into the computer device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0119] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0120] Accordingly, embodiments of the present invention provide a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the steps described in the embodiments of the multi-source power demand data classification and management method, for example... Figure 1 Steps S1 to S4 as described above.

[0121] In summary, the embodiments of this application provide a method, system, device, and medium for classifying and managing multi-source electricity demand data, solving the technical problem of improving the accuracy and adaptability of electricity demand classification. The method includes: constructing a multi-layer graph structure based on the acquired multi-source electricity demand dataset, the multi-layer graph structure including a semantic graph, a state graph, and a spatial adjacency graph; constructing a search sample set based on the multi-layer graph structure, each search sample in the search sample set including a color vector, a state vector, and a fitness state, the color vector being determined by the semantic features of the semantic graph, the state vector being determined by the state data of the state graph, and the fitness state being set to reflect the semantic graph. Semantic connectivity, state fluctuation amplitude of the state graph, and adjacency density of the spatial adjacency graph are considered. Based on the fitness state, a perturbation guidance vector is obtained. The gravitational gradient direction of the search sample is constructed based on the perturbation guidance vector's color vector and a preset target color vector. An improved peacock optimization algorithm is used to iteratively update the search sample set, obtaining the optimal fitness state for each search sample. The classification label for each search sample is determined based on the optimal fitness state. The improved peacock optimization algorithm is configured to execute perturbations caused by the fitness state based on the perturbation guidance vector, and drive the movement direction of the color vector and state vector based on the gravitational gradient direction. This application discloses a multi-source electricity demand data classification and management method. By integrating heterogeneous graph structures such as semantic graphs, state graphs, and spatial adjacency graphs, it accurately expresses the semantic, state, and spatial coupling relationships between customer demands, improving modeling capabilities in complex customer contexts and providing structural support for multi-dimensional feature fusion and reasoning. Combining key indicators such as semantic connectivity, state fluctuation, and adjacency density, it achieves a quantitative expression of the semantic consistency, state stability, and neighborhood structure of electricity customer demands. It constructs a perturbation guidance mechanism and a gravitational potential field function, and dynamically generates gravitational gradient directions using the differences between semantic features and spatial adjacency relationships to guide the movement of search samples in the feature space. At the same time, based on historical behavior trajectories and graph state information, it introduces a gated neural network to achieve adaptive control of the boasting, jumping, and trajectory splitting behaviors of search samples, improving the flexibility and responsiveness when handling complex customer state changes. The improved peacock optimization algorithm uses color vectors, state vectors, and fitness states as evolutionary variables, integrating three types of driving factors: perturbation guidance, gravitational attraction, and behavior control, to achieve accurate screening and evolutionary convergence of key samples, effectively improving the accuracy of demand sorting and the stability of anomaly identification.

[0122] The various embodiments in this specification are described in a progressive manner. For directly identical or similar parts of the embodiments, refer to each other. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0123] The embodiments described above are merely preferred embodiments of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the technical principles of this application, and these improvements and substitutions should also be considered within the scope of protection of this application. Therefore, the scope of protection of this patent application should be determined by the scope of the claims.

Claims

1. A multi-source power appeal data classification management method, characterized in that, The method comprises: According to the obtained multi-source power appeal data set, a multi-layer graph structure is constructed, the multi-layer graph structure comprising a semantic graph, a state graph and a spatial adjacency graph; According to the multi-layer graph structure, a search sample set is constructed, each search sample of the search sample set comprising a color vector, a state vector and a fitness state, the color vector being determined by semantic features of the semantic graph, the state vector being determined by state data of the state graph, and the fitness state being set to reflect semantic connectivity of the semantic graph, state fluctuation amplitude of the state graph and adjacency density of the spatial adjacency graph; According to the fitness state, a perturbation guide vector is obtained, comprising: according to each semantic connectivity, a connectivity difference value between each node and a neighborhood node is obtained, and a mean value and a standard deviation corresponding to the connectivity difference value of each node are taken as a structure distribution index of the search sample corresponding to the node; according to the state fluctuation amplitude, the adjacency density and the structure distribution index, a perturbation guide vector is constructed, wherein each multi-source power appeal data of a power user is taken as a node; according to the perturbation guide vector, the color vector and a preset target color vector, a gravitational gradient direction of the search sample is constructed; An improved peacock optimization algorithm is used to iteratively update the search sample set, to obtain an optimal fitness state of each search sample, and a classification label of each search sample is determined according to the optimal fitness state, the improved peacock optimization algorithm being set to execute a perturbation caused by the fitness state according to the perturbation guide vector, and to drive a moving direction of the color vector and the state vector according to the gravitational gradient direction.

2. The multi-source power solicitation data classification management method of claim 1, wherein, According to the obtained multi-source power appeal data set, a multi-layer graph structure is constructed, comprising: A multi-source power appeal data set composed of multi-source power appeal data of each power user is obtained, the multi-source power appeal data comprising customer appeal text, identity information, equipment state information and geographic location information; A semantic graph is constructed with similarity between semantic feature vectors of each customer appeal text as a first edge, a state graph is constructed with a time transfer relationship of state data composed of each identity information, equipment state information and geographic location information as a second edge, and a spatial adjacency graph is constructed with a spatial distance between geographic locations displayed by each geographic location information as a third edge; According to the semantic graph, the state graph and the spatial adjacency graph, a multi-layer graph structure is constructed.

3. The multi-source power solicitation data classification management method of claim 2, wherein, According to the multi-layer graph structure, a search sample set is constructed, comprising: According to the semantic feature vector of each node, a color vector is obtained, and according to the identity information, the equipment state information and the geographic location information of each node, a state vector of each multi-source power appeal data is constructed; According to the semantic graph, semantic connectivity between each of the nodes is obtained, according to the state graph, state fluctuation amplitude between each of the nodes is obtained, according to the spatial adjacency graph, adjacency density between each of the nodes is obtained, and fitness states of each of the multi-source power appeal data are constructed according to the semantic connectivity, the state fluctuation amplitude and the adjacency density; According to the color vector, the state vector and the fitness state of each of the multi-source power appeal data, a search sample set is constructed.

4. The multi-source power solicitation data classification management method of claim 3, wherein, The construction of the gravitational gradient direction of the search sample according to the perturbation guide vector, the color vector and a preset target color vector includes: A channel difference between the color vector and the preset target color vector is calculated, and the channel difference is normalized to obtain an attraction direction vector; According to the perturbation guide vector and the attraction direction vector, a position state driving vector is obtained, and a position of the search sample is updated according to the position state driving vector; The search sample with a fitness state higher than an average value is selected as a gravitational center from the search sample set, and an abnormal gravitational potential function is constructed according to the gravitational center; Based on the abnormal gravitational potential function, a gradient calculation is performed on the state vector after the position update of the search sample to obtain a position state gradient vector; The position state gradient vector is normalized to obtain the gravitational gradient direction of the search sample.

5. The multi-source power solicitation data classification management method of claim 4, wherein, The iterative update of the search sample set by using the improved peacock optimization algorithm to obtain an optimal fitness state of each of the search samples includes: According to the color vector and the preset target color vector, a semantic label matching degree is obtained, and according to the position state gradient vector, a gravitational response degree is obtained; According to the semantic label matching degree, the gravitational response degree, the semantic connectivity, the state fluctuation amplitude and the adjacency density, a multi-objective fitness function is constructed; According to an iteration sequence of each of the search samples in the iteration process of the improved peacock optimization algorithm, a historical behavior trajectory tensor is constructed, and the historical behavior trajectory tensor and a graph state information tensor of the multi-layer graph structure are input into a pre-constructed gated neural network structure to obtain a peacock behavior gating factor; According to the perturbation guide vector, a behavior of peacock exploring a new area is simulated to perform the perturbation caused by the fitness state, according to the gravitational gradient direction, a moving direction of the color vector and the state vector is guided, according to the peacock behavior gating factor, a peacock behavior is controlled to iteratively update each of the search samples; Based on the multi-objective fitness function, each of the search samples after the iterative update is judged to determine the optimal fitness state of each of the search samples.

6. The multi-source power solicitation data classification management method of claim 5, wherein, The input of the historical behavior trajectory tensor and the graph state information tensor of the multi-layer graph structure into the pre-constructed gated neural network structure to obtain the peacock behavior gating factor includes: The historical behavior trajectory tensor and the graph state information tensor of the multi-layer graph structure are spliced to obtain a joint feature tensor; The joint feature tensor is input into a gated neural network structure to obtain peacock behavior gating factors, including a show-off behavior gating factor, a jumping behavior gating factor, and a trajectory splitting behavior gating factor.

7. A multi-source power demand data classification management system for implementing the multi-source power demand data classification management method according to any one of claims 1 to 6, characterized by, The system comprises a multi-layer graph structure construction module, a search sample set construction module, a parameter construction module, and an iterative optimization update module. The multi-layer graph structure construction module is configured to construct a multi-layer graph structure according to the obtained multi-source power appeal data set, wherein the multi-layer graph structure comprises a semantic graph, a state graph, and a spatial adjacency graph. The search sample set construction module is configured to construct a search sample set according to the multi-layer graph structure, wherein each search sample of the search sample set comprises a color vector, a state vector, and a fitness state, the color vector is determined by semantic features of the semantic graph, the state vector is determined by state data of the state graph, and the fitness state is set to reflect semantic connectivity of the semantic graph, state fluctuation amplitude of the state graph, and adjacency density of the spatial adjacency graph. The parameter construction module is configured to obtain a perturbation guide vector according to the fitness state, comprising: obtaining connectivity difference values between each node and adjacent nodes according to each semantic connectivity, and taking a mean value and a standard deviation of the connectivity difference value of each node as a structure distribution index of the search sample corresponding to the node; constructing a perturbation guide vector according to the state fluctuation amplitude, the adjacency density, and the structure distribution index, wherein each multi-source power appeal data of a power user is taken as a node; and constructing a gravitational gradient direction of the search sample according to the perturbation guide vector, the color vector, and a preset target color vector. The iterative optimization update module is configured to perform iterative update on the search sample set by using an improved peacock optimization algorithm to obtain an optimal fitness state of each search sample, and determine a classification label of each search sample according to the optimal fitness state, wherein the improved peacock optimization algorithm is set to perform a perturbation caused by the fitness state according to the perturbation guide vector, and drive a moving direction of the color vector and the state vector according to the gravitational gradient direction.

8. A computer device, comprising: The computer device comprises a memory, a processor, and a transceiver connected through a bus, the memory is configured to store a set of computer program instructions and data, and transmit the stored data to the processor, the processor executes the computer program instructions stored in the memory to perform the multi-source power appeal data classification management method in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer readable storage medium stores a computer program, when the computer program is executed, the multi-source power appeal data classification management method in any one of claims 1 to 6 is implemented.

Images