Regional power load real-time allocation system and method based on artificial intelligence

Abstract

The invention discloses a regional power load real-time allocation system and method based on artificial intelligence, and belongs to the technical field of power distribution. According to the method, after an electric load data set and an alarm condition data set are integrated, a multi-dimensional feature interaction matrix is decomposed based on a factorization machine, a high-order interaction relation between electric load features and alarm condition features is extracted, hidden variable modeling is introduced, accurate division of risk levels of different areas is further achieved, and the accuracy of the risk levels is improved. Compared with a traditional single-domain feature analysis method, the method has the advantages that complex nonlinear relations can be captured, potential influence rules of regional power anomalies on alarm risks can be mined, and experimental verification shows that the method is high in accuracy and high in reliability. According to the modeling method, the recognition accuracy of the high-correlation region is improved by more than 15%, and a more scientific basis is provided for resource allocation decisions.

Description

Technical Field

[0001] The present invention belongs to the technical field of power distribution, and particularly relates to a real-time allocation system and method for regional power loads based on artificial intelligence. Background Art

[0002] With the development of artificial intelligence and big data technologies, the construction of smart cities has gradually become an important direction for urban management and governance. The power load management and police dispatch systems in most cities operate independently. The power department usually relies on smart grid technology to monitor and regulate loads, mainly applied in the fields of power anomaly detection, load prediction, and equipment fault warning. The police system relies on historical police situation data, real-time alarm information, and manual judgment to conduct public security risk assessment and police force deployment. Although both play a certain role in their respective fields, the separated management method in the smart city scenario has obvious defects: on the one hand, power load anomalies often have potential associations with public security events in specific areas. For example, large-scale gatherings and emergencies may cause a sharp increase in local area loads or equipment failures, but existing technologies are difficult to establish such associative relationships; on the other hand, the police system lacks support information on power resources when deploying police forces, resulting in a possible reduction in emergency response efficiency due to power outages when dealing with emergencies.

[0003] In recent years, some single-field optimization methods based on artificial intelligence have begun to appear. For example, using deep learning models for power load prediction or police situation trend analysis has exposed the following problems in practical applications: (1) Power load data and police situation data belong to different management departments, with inconsistent data formats and storage standards, making it difficult to achieve cross-field data fusion and sharing.

[0004] (2) Traditional methods lack in-depth exploration of the complex associations between power loads and police situation dynamics, and it is difficult to identify the dynamic relationships between abnormal loads and potential public security risks within a region.

[0005] (3) The power and police systems cannot cooperate efficiently when facing emergencies. For example, the power department cannot give priority to ensuring power supply in key public security areas, and the police department cannot optimize the police force deployment plan based on power failure information.

[0006] Therefore, the response speed and dynamic adjustment ability of single-field optimization technologies in emergencies are relatively low, and there are significant deficiencies in data fusion, correlation analysis, cross-departmental cooperation, and real-time regulation, resulting in limitations in the comprehensive governance ability of smart cities in power load and police situation allocation, and there are relatively large limitations in the real-time monitoring, allocation, and cooperation ability of power loads and police situations, making it difficult to meet the demand for optimized resource scheduling in complex urban environments. Summary of the Invention

[0007] The objective of the present invention is to overcome the disadvantages of the above-mentioned existing technologies, and provide a real-time regional power load allocation system and method based on artificial intelligence, so as to solve the problem that there are significant limitations in the real-time monitoring, allocation, and coordination capabilities of power loads and police situations in the existing technologies, and it is difficult to meet the demand for optimized resource scheduling in complex urban environments.

[0008] To achieve the above objective, the present invention adopts the following technical solutions: A real-time regional power load allocation method based on artificial intelligence, comprising the following steps: Obtain the real-time power load data set and police situation data set of the region, and integrate them into a multi-dimensional power data set; Extract features from the multi-dimensional power data set to obtain a power load feature vector and a police situation feature vector, and combine the power load feature vector and the police situation feature vector to construct a multi-dimensional feature interaction matrix; Decompose the multi-dimensional feature interaction matrix through a factorization machine to obtain latent variables, where the latent variables are the high-order interaction relationships between power load features and police situation features; perform cluster analysis on the latent variable representation to divide the correlation levels of power load and police situation risks; Based on the latent variables and the correlation levels of power load and police situation risks, combined with the historical police situation risk records and power load abnormality records of the region, calculate the risk correlation score set of each region, and obtain a list of high-correlation regions; Based on the list of high-correlation regions, sort the risk correlation scores, and dynamically adjust the recommendation results of high-priority regions in combination with the real-time multi-dimensional power data set to generate a recommendation report; When a police situation or a major power load abnormality occurs in a high-priority region, push a collaborative control scheme in combination with power load data, police situation data, and the recommendation report.

[0009] A further improvement of the present invention lies in: Preferably, the multi-dimensional power data set is a multi-dimensional power data set constructed based on timestamps.

[0010] Preferably, the power load data set includes current, voltage, power factor, and peak power consumption information; the police situation data set includes real-time police situation alarm information, historical public security event records, and regional population density information; the multi-dimensional feature interaction matrix is:

[0011] Wherein, Represents the feature cross-operation, Represents the police situation feature vector at time t, Represents the power load feature vector at time t.

[0012] Preferably, before extracting features from the multi-dimensional power dataset, it also includes a process of cleaning and normalizing the accessed multi-dimensional power dataset and normalizing the data.

[0013] Preferably, the specific process of decomposing the multi-dimensional feature interaction matrix through a factorization machine to obtain latent variables is as follows: S31, factorize the multi-dimensional feature interaction matrix, extract the high-order interaction relationship between the power load feature and the police situation feature, and construct a factorization model; S32, generate latent variables according to the factorization result.

[0014] Preferably, the process of calculating the risk correlation score for each region is as follows: S41, calculate the influence coefficient of the power load anomaly on the police situation risk through the latent variable, and construct a regional influence factor set; S42, combine the historical police situation risk records and the power load anomaly records to calculate the risk occurrence probability set for each region; S43, combine the regional influence factor set and the risk occurrence probability set to calculate the risk correlation score set for each region.

[0015] Preferably, the recommended report includes the region number, risk level, abnormal type of power load feature, and recommended disposal measures. The specific process of generating the recommended report is as follows: S51, sort the risk correlation scores through a sorting algorithm to obtain a regional priority list; S52, according to the regional priority list, filter out the regions where the risk correlation score is greater than the preset threshold to obtain a high-priority region set; S53, dynamically adjust the recommended results of the high-priority regions in combination with the real-time multi-dimensional power dataset; S54, generate a recommended report according to the risk level of the adjusted high-priority regions.

[0016] Preferably, the formula for dynamically adjusting the recommended results of the high-priority regions in combination with the real-time multi-dimensional power dataset is:

[0017] Among them, represents the dynamically adjusted risk level of high-priority region j; and are the adjustment weights of the risk score and the real-time data, is the risk correlation score of region j, is the regional risk dynamic update function calculated based on the real-time multi-dimensional power dataset.

[0018] Preferably, the process of pushing the coordinated control plan is: S61. Monitor the power load data and alarm data of high-priority areas in real time, and determine the event types of alarms or major power load anomalies; S62. Obtain a list of key response areas based on the risk level information and characteristics of emergencies in the high-priority areas in the recommendation report; S63. Determine the power load adjustment strategy based on the real-time power load data and event types; S64. Generate police deployment suggestions according to the nature and impact scope of the alarm or major power load; S65. Integrate the power load adjustment strategy and police deployment suggestions to generate a collaborative control plan.

[0019] A real-time regional power load allocation system based on artificial intelligence, comprising: A data acquisition module, configured to obtain the real-time power load data set and alarm data set of the region, and integrate them into a multi-dimensional power data set; A feature extraction module, configured to extract features from the multi-dimensional power data set, obtain a power load feature vector and an alarm feature vector, and combine the power load feature vector and the alarm feature vector to construct a multi-dimensional feature interaction matrix; A factor decomposition modeling module, configured to decompose the multi-dimensional feature interaction matrix through a factorization machine to obtain latent variables, where the latent variables are the high-order interaction relationships between the power load features and the alarm features; the latent variables are used for clustering analysis to divide the correlation levels of the power load and the alarm risks; A risk assessment module, configured to calculate a set of risk correlation scores for each region based on the latent variables and the correlation levels of the power load and the alarm risks, and combine the historical alarm risk records and power load anomaly records of the region to obtain a list of high-correlation regions; A sorting and recommendation module, configured to sort the risk correlation scores based on the list of high-correlation regions, and dynamically adjust the recommendation results of the high-priority regions in combination with the real-time multi-dimensional power data set to generate a recommendation report; A collaborative control module, configured to push a collaborative control plan when an alarm or major power load anomaly occurs in a high-priority region, in combination with the power load data, alarm data and recommendation report.

[0020] Compared with the prior art, the present invention has the following beneficial effects: The present invention discloses a real-time allocation method for regional power loads based on artificial intelligence. This method decomposes the multi-dimensional feature interaction matrix based on the factorization machine, extracts the high-order interaction relationship between power load features and police situation features. Compared with the traditional single-domain feature analysis method, this method can capture complex non-linear relationships, explore the potential influence law of regional power anomalies on police situation risks, and further realizes the accurate classification of risk levels in different regions by introducing latent variable modeling. Experimental verification shows that the modeling method of the present invention has an accuracy improvement of more than 15% in the identification of high-correlation regions, providing a more scientific basis for resource allocation decisions. By combining real-time multi-dimensional power data sets and police situation data, and using sorting algorithms to dynamically adjust the recommendation results of high-priority regions, the system designs a linkage optimization algorithm for load adjustment strategies and police force deployment plans, ensuring that resources can quickly tilt towards high-risk regions in case of sudden police situations or major power load anomalies. Compared with traditional static rule allocation, this method significantly improves the response speed of emergencies, with a deployment efficiency improvement of more than 20%, while reducing resource occupancy in non-critical regions and enhancing the overall allocation ability of the system. Through the feedback optimization mechanism, the system can dynamically monitor and adjust the control strategy to adapt to the changes of emergencies and the dynamic state of resources. Compared with traditional systems, the present invention can maintain the effectiveness of the control scheme under complex multi-variable conditions, with a control accuracy improvement of 18%, showing stronger robustness and adaptability in dealing with complex scenarios of cross-departmental linkages, providing more efficient technical support for smart city governance. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flowchart of latent variable modeling based on the factorization machine and regional risk correlation analysis in the present invention. Figure 2 It is an implementation flowchart of a real-time allocation method for power loads in key regions based on artificial intelligence in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] Hereinafter, the terms "first", "second", "third", and "fourth" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first", "second", "third", and "fourth" may explicitly or implicitly include one or more of such features.

[0023] The co - shooting method provided by the embodiments of this application can be applied to terminal devices such as mobile phones, tablet computers, wearable devices, vehicle - mounted devices, augmented reality (AR) / virtual reality (VR) devices, laptop computers, ultra - mobile personal computers (UMPCs), netbooks, personal digital assistants (PDAs), etc. The embodiments of this application do not impose any restrictions on the specific types of terminal devices.

[0024] It should be noted that the terms "first", "second", etc. in the specification and drawings of the present invention are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non - exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0025] Figure 1 It is the flowchart of the implementation of a real - time power load allocation method for key areas based on artificial intelligence in the embodiments of the present invention; Please refer to Figure 1 , which is the flowchart of a real - time power load allocation method for key areas based on artificial intelligence in the embodiments of the present invention. The method includes the following steps: S1, Obtain the real - time power load data set and the police situation data set of the area, and integrate them into a multi - dimensional power data set; S2, Extract features from the multi - dimensional power data set to obtain a power load feature vector and a police situation feature vector, and combine the power load feature vector and the police situation feature vector to construct a multi - dimensional feature interaction matrix; S3, Decompose the multi - dimensional feature interaction matrix through a factorization machine to obtain latent variables. The latent variables are the high - order interaction relationships between the power load features and the police situation features; perform clustering analysis on the latent variable representation to divide the correlation levels of regional power loads and police situation risks; S4, Based on the latent variables and the correlation levels of power loads and police situation risks, combined with the historical police situation risk records and power load anomaly records of the area, calculate the risk correlation score set of each area to obtain a list of high - correlation areas; S5. Based on the high - correlation region list, sort the risk - correlation scores, and dynamically adjust the recommendation results of high - priority regions in combination with the real - time multi - dimensional power dataset to generate a recommendation report. S6. When an alarm or a major power load anomaly occurs in a high - priority region, push a collaborative control scheme in combination with power load data, alarm data, and the recommendation report.

[0026] In this embodiment, the power load dataset in S1 includes current, voltage, power factor, and peak power consumption information, and the alarm dataset includes real - time alarm information, historical public security event records, and regional population density information. Specifically, S1 includes the following steps: S11. Obtain the real - time power load data of key regions from power monitoring terminals and construct a power load dataset:

[0027] Among them, represents the current at time t, represents the voltage at time t, represents the power factor at time t, and its value range is [0, 1]; represents the peak power consumption at time t, and T represents the total data collection time. S12. Obtain the alarm data of key regions from the police system of the public security department and construct an alarm dataset:

[0028] Among them, represents the alarm level at time t, and its value range is {0, 1, 2, 3}, corresponding to no alarm, minor alarm, medium alarm, and major alarm respectively; represents the historical public security event record at time t, is the statistical value of the k - th historical public security event; K represents the total number of types of public security events; represents the regional population density at time t. S13. Perform time synchronization and correlation matching on the power load dataset and the alarm dataset, and construct a multi - dimensional power dataset through a unified time stamp t: .

[0029] In this embodiment, between S1 and S2, there is also a process of data cleaning and standardization for the accessed multi - dimensional power dataset, removing outliers and redundant information, and normalizing the data, which specifically includes the following steps: S11. Detect and remove outliers in the multi - dimensional power dataset to construct a cleaned multi - dimensional power dataset ; S12. Remove redundant information in the cleaned multi-dimensional power dataset to construct a redundant-information-removed multi-dimensional power dataset ; S13. Normalize the redundant-information-removed dataset to construct a normalized multi-dimensional power dataset .

[0030] In this embodiment, the features in the multi-dimensional feature interaction matrix in S2 include area numbers, power load features, police situation features, and auxiliary features. The multi-dimensional feature interaction matrix is used to represent the potential correlation between power loads and police situation risks in each area. The construction of the multi-dimensional feature interaction matrix specifically includes the following steps: S21. Extract power load feature vectors from the normalized multi-dimensional power dataset to construct a power load feature set:

[0031] wherein, represents the power load feature vector at time t, is a power load feature extraction function. By using the feature extraction function, the feature representations of the power load at different time points are obtained, which are used to characterize the operating state and load change pattern of the power system; S22. Extract police situation feature vectors from the normalized multi-dimensional power dataset to construct a police situation feature set:

[0032] wherein, represents the police situation feature vector at time t, is a police situation feature extraction function. By using the feature extraction function, the feature representations of the police situation at different time points are obtained, which are used to characterize the public security situation and risk level in the area; S23. Combine the power load feature vectors and the police situation feature vectors to construct a multi-dimensional feature interaction matrix:

[0033] wherein, represents a feature cross operation. By interacting the power load feature vectors and the police situation feature vectors, high-dimensional features reflecting the potential correlation between power loads and police situation risks are generated.

[0034] In this embodiment, in S3, through clustering, regions with similar latent variable representations can be divided into several "power load - police situation risk" correlation levels, which are displayed in the form of clusters, forming a grouping based on high - order interaction relationships. There are often significant differences in the coupling degree of power load anomalies and police situation risks in regions of different clusters. The results of clustering analysis are used for differential processing in subsequent scoring or algorithms (such as setting the weights of the impact of power load anomalies and the probability balance coefficients of police situation risks).

[0035] In this embodiment, in the above - mentioned S3, the latent variable representation models the feature combinations of the multi - dimensional feature interaction matrix to explore the potential influence law of regional power load anomalies on police situation risks, including the following steps: S31. Factorize the multi - dimensional feature interaction matrix to extract the high - order interaction relationship between power load features and police situation features, and construct a factorization model:

[0036] Among them, is the latent variable matrix, represents the interaction intensity between time t and latent variable k, is the feature factor matrix, represents the relationship intensity between the l - th feature and latent variable k, is the regularization parameter, is the interaction weight, represents the trace operation of the matrix; S32. Generate latent variables according to the factorization results, and construct a region for capturing the dynamic interaction relationship between power load and police situation features in the time series, the latent variable set

[0037]

[0038] Among them, represents the latent variable representation value at time t. The k - th latent variable extracts the high - order relationship between power load and police situation features, is the multi - dimensional feature interaction matrix whose element value represents the l - th feature value at time t; is the factor value in the feature factor matrix; is the bias term of the latent variable, used to correct the feature interaction result; is the activation function.

[0039] In this embodiment, in S4, the risk correlation score is calculated by weighted summation of the influence coefficient of power load anomalies on police situation risks and the regional risk occurrence probability, including the following steps: S41. Use the latent variable set Calculate the influence coefficient of abnormal power load on the risk of police alerts, and construct a set of regional influence factors:

[0040] Among them, represents the set of risk influence factors for different regions at each time point within the time series, is the influence factor of the j-th region at time t, is the specific risk weight of the latent variable and region j, used to describe the contribution intensity of the latent variable k to the risk of region j. M is the total number of regions; in this step, the results of cluster analysis affect and .

[0041] S42. Calculate the set of risk occurrence probabilities for each region by combining historical police alert risk records and abnormal power load records:

[0042] Among them, represents the set of risk occurrence probabilities for each region, is the risk occurrence probability of the j-th region, is the coefficient for adjusting the influence weight of abnormal power load, represents the police alert risk value of region j at the l-th historical time point, represents the abnormal power load value of region j at the l-th historical time point. P is the total number of historical time points; S43. Combine the set of regional influence factors and the set of risk occurrence probabilities to calculate the set of risk correlation scores for each region:

[0043] Among them, represents the set of risk correlation scores for each region, is the risk correlation score of the j-th region, is the weight balance coefficient between the influence factor and the risk occurrence probability, represents the total sum of comprehensive influence factors of region j at all time points within the time series, represents the average influence factor of region j.

[0044] In this embodiment, in S5, according to the list of high-correlation regions, use a sorting algorithm to sort the risk correlation scores, give priority to screening high-risk regions, and dynamically adjust the recommendation results of high-priority regions in combination with real-time multi-dimensional power data sets to generate a recommendation report. The recommendation report includes region numbers, risk levels, abnormal types of power load characteristics, and recommended disposal measures; specifically, it includes the following steps: S51. Score the risk correlation set Use a sorting algorithm to sort in descending order to generate a list of regional priorities:

[0045] wherein, represents a list of regional numbers sorted from high to low according to the risk correlation score, represents the regional number whose score is higher than or equal to the score of the regional number ; S52. According to the sorting result, preferentially screen high-risk regions and construct a set of high-priority regions:

[0046] wherein, represents the set of high-priority regions, including all regional numbers with risk correlation scores higher than the preset threshold , is the score threshold for screening high-priority regions; S53. Combine the real-time multi-dimensional power dataset to dynamically adjust the recommended results of high-priority regions:

[0047] wherein, represents the dynamically adjusted risk level of high-priority region j; and are the adjustment weights of the risk score and real-time data, is the risk correlation score of region j, is the regional risk dynamic update function calculated based on the real-time multi-dimensional power dataset, used to adjust the real-time risk level of the region; S54. Generate a recommendation report according to the adjusted risk level of high-priority regions. The recommendation report includes the following content:

[0048] wherein, represents the generated recommendation report; j is the regional number; is the risk category label corresponding to the risk level; is the multi-dimensional feature interaction matrix of the region; is the recommended disposal measure, including the power load adjustment strategy and the police deployment plan.

[0049] In this embodiment, when a sudden police emergency or a major abnormal power load occurs in the high-priority area, the S6 combines the recommendation report and pushes the collaborative control plan to the power department and the public security department in real time, including the power load adjustment strategy and the police deployment suggestion within the area. The specific steps are as follows: S61. Monitor the power load data and police emergency data in the high-priority area in real time, judge the event type by analyzing the characteristics and influence scope of the sudden event, including equipment overload, power outage, large-scale assembly or public security event, generate event description information, and record the event impact area and the preliminary severity assessment.

[0050] S62. Dynamically evaluate the response priority of each area by combining the risk level information and the characteristics of the sudden event in the recommendation report for the high-priority area, calculate the list of key response areas according to the priority allocation, and make the resources tilt towards the area with the largest influence scope and the highest risk level. S63. Develop a power load adjustment strategy based on the real-time power load data and the sudden event type, including load transfer, load shedding or emergency power supply restoration measures, to ensure the power supply stability in the key area, and at the same time combine the grid operation safety and the load balance requirements of other areas. S64. Generate police deployment suggestions according to the nature and influence scope of the sudden event, including the number of deployed police, deployment locations and response time limits, and at the same time optimize the collaborative cooperation strategy between the police and the power department, giving priority to protecting power supply equipment or assisting in evacuating the crowd. S65. Integrate the power load adjustment strategy and the police deployment suggestions to generate a collaborative control plan. The collaborative control plan includes event description, response priority, load adjustment strategy, police deployment suggestions and expected implementation effects, and is pushed to the decision-making systems of the power department and the public security department in real time through the communication interface.

[0051] A real-time power load allocation system for key areas based on artificial intelligence, the system includes the following modules: A data acquisition module, which is used to obtain the real-time power load data set and police emergency data set of the area and integrate them into a multi-dimensional power data set. A feature extraction module, which is used to extract features from the multi-dimensional power data set, obtain the power load feature vector and the police emergency feature vector, and combine the power load feature vector and the police emergency feature vector to construct a multi-dimensional feature interaction matrix. A factorization modeling module, which is used to decompose the multi-dimensional feature interaction matrix through a factorization machine to obtain latent variables. The latent variables are the high-order interaction relationships between the power load features and the police emergency features, and are used to describe the potential correlation between the power load and the police emergency risk. A risk assessment module, which is used to calculate a set of risk correlation scores for each region based on latent variables, combined with the historical police situation risk records and power load anomaly records of the region, to obtain a list of regions with high correlation; the risk assessment module evaluates the risk level of each region by weighted summing the influence coefficient of power load anomalies on police situation risks and the probability of regional risk occurrence.

[0052] A sorting and recommendation module, which is used to sort the risk correlation scores based on the list of regions with high correlation, dynamically adjust the recommendation results of high-priority regions in combination with real-time multi-dimensional power data sets, and generate a recommendation report; A collaborative regulation module, which is used to push a collaborative regulation plan when a police situation or a major power load anomaly occurs in a high-priority region. The collaborative regulation module synthesizes power load adjustment strategies and police deployment suggestions, and pushes the regulation plan to the decision-making systems of the power department and the public security department through a communication interface.

[0053] In some embodiments of the present invention, the data acquisition module can also support data cleaning and standardization processing, be able to remove outliers and redundant information, and perform normalization operations on the data.

[0054] In some embodiments of the present invention, a feedback optimization module is further included, which is used to perform real-time monitoring and feedback analysis on the execution process of the regulation plan, dynamically adjust the power load and police situation risk regulation strategies, and update the model parameters to optimize the overall performance of the system.

[0055] A data storage and analysis module, which is used to store all the collected original data, feature extraction results, latent variable modeling results, and regulation execution records, and support post-event analysis and model optimization.

[0056] The following is further illustrated with specific embodiments: Embodiment 1 In the core business district of a certain extra-large city, an emergency is simulated: due to a sudden increase in a large-scale public gathering activity, the power load in the region rapidly climbs to the peak. At the same time, due to the excessive gathering of people, multiple public security alarm events occur. At this time, the traditional power load monitoring and police situation handling systems operate independently and fail to timely discover the potential association between power load anomalies and police situation events, resulting in low resource allocation efficiency and insufficient rapid response ability to events.

[0057] The present invention provides an artificial intelligence-based real-time power load allocation system and method for key regions, which can quickly identify regional risks in the above scenarios and achieve cross-departmental collaborative regulation to ensure stable power supply and public safety: At around 3 p.m. on a certain date in a certain month, in the CBD Square in the core business district of City A, due to the temporary holding of a public gathering event with approximately 5,000 participants, the power load in the area surged. Real-time monitoring data showed that the power load of the power supply equipment rapidly climbed from the usual 600 kW to 1,100 kW, exceeding 90% of the rated load of the equipment. At the same time, the police system received multiple alarm messages regarding crowd congestion, lost property, and physical conflicts. Due to the failure of the traditional system to promptly associate these data, the power supply equipment tripped due to overload at 15:20, and the power supply interruption lasted for 45 minutes, causing a significant social impact.

[0058] With the support of the method of the present invention, after the event occurred, the system immediately obtained real-time data through the power data acquisition module and the police situation data acquisition module: Power load data: At 15:05, the current reached 400 A (usually 250 A), the voltage fluctuation dropped to 208 V (normal value is 220 V), and the power factor was 0.78 (lower than the normal value of 0.95); Police situation data: At 15:08, the number of alarms reached 7 (usually 2 per hour), and the types of alarms included 2 for congestion, 3 for conflicts, and 2 for theft; the regional population density monitoring was 3,500 people per square kilometer (usually 1,200 people).

[0059] After the data preprocessing, the system uses the factorization machine to decompose the multi-dimensional feature interaction matrix to extract latent variables, and mines the potential correlation between the power load and the police situation risk. The analysis results show that the occurrence of abnormal power in the area has a high correlation with the high population density and the surge in police situations. The risk level is classified as "major risk" and requires priority response. According to the results calculated by the risk assessment module, the system generated the following control plan: Power load adjustment strategy: Divert the load to the power supply equipment in the neighboring area (within 1 km), and a total of 200 kW of power is adjusted. Start the standby generator in the CBD Square to provide an additional 100 kW of power supply.

[0060] Police deployment suggestions: Deploy 50 police officers to enter the key area, and divide them into groups to be responsible for evacuating the crowd, maintaining order, and patrolling the scene. Prioritize handling congestion alarms to ensure that the crowd can evacuate in an orderly manner. Dispatch additional security vehicles to strengthen patrols to prevent further security incidents.

[0061] The control plan was pushed to the power department and the public security department through the system at 15:10 and was quickly implemented. By 15:25, the power supply had returned to the normal level, and all the police officers had arrived at the designated locations and started to perform their tasks.

[0062] To verify the effectiveness of the method of the present invention, a comparative simulation test was conducted on the performance of the method of the present invention and the traditional method in similar scenarios. The following are the comparative data: Table 1 Comparative implementation data of the method of the present invention and traditional methods

[0063] The method of the present invention uses approximately 5,000 sets of historical data of power loads and police alerts as training samples, covering various scenarios such as daily power consumption fluctuations, emergencies, and public security alarms. By dynamically adjusting the weight parameters of the factorization machine, the model has higher adaptability in dealing with risks in different regions. Experiments show that the method of the present invention is superior to traditional methods in terms of data response speed, risk assessment accuracy, and regulation efficiency.

[0064] In summary, this embodiment demonstrates the application process and effect verification of the method of the present invention in actual scenarios, fully proving its innovation and practicality in the field of real-time allocation of power loads and police alerts, and providing an effective solution for urban public safety and power stability.

[0065] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A real-time allocation method for regional power loads based on artificial intelligence, characterized in that, Including the following steps: Obtain the real-time power load dataset and police situation dataset of the region, and integrate them into a multi-dimensional power dataset; Extract features from the multi-dimensional power dataset to obtain a power load feature vector and a police situation feature vector, and combine the power load feature vector and the police situation feature vector to construct a multi-dimensional feature interaction matrix; Decompose the multi-dimensional feature interaction matrix through a factorization machine to obtain latent variables, where the latent variables are the high-order interaction relationships between power load features and police situation features; Perform clustering analysis on the latent variable representation to divide the correlation level of power load and police situation risks; Based on the latent variables and the correlation level of power load and police situation risks, combined with the historical police situation risk records and power load anomaly records of the region, calculate the risk correlation score set of each region to obtain a list of high-correlation regions; Based on the list of high-correlation regions, sort the risk correlation scores, and dynamically adjust the recommendation results of high-priority regions in combination with the real-time multi-dimensional power dataset to generate a recommendation report; When a police situation or a major power load anomaly occurs in a high-priority region, push a collaborative control plan in combination with power load data, police situation data, and the recommendation report.

2. The real-time allocation method of regional power load based on artificial intelligence according to claim 1, characterized in that The multi-dimensional power dataset is a multi-dimensional power dataset constructed based on timestamps.

3. A real-time allocation method for regional power loads based on artificial intelligence according to claim 1, characterized in that, The power load dataset includes current, voltage, power factor, and peak power consumption information; the police situation dataset includes real-time police situation alarm information, historical public security event records, and regional population density information; the multi-dimensional feature interaction matrix is: Among them, represents the feature cross operation, represents the alarm feature vector at time t, represents the power load feature vector at time t.

4. A real-time allocation method for regional power loads based on artificial intelligence according to claim 1, characterized in that, Before extracting features from the multi-dimensional power dataset, it also includes a process of cleaning and standardizing the accessed multi-dimensional power dataset and normalizing the data.

5. A real-time allocation method for regional power load based on artificial intelligence according to claim 1, characterized in that, The specific process of decomposing the multi-dimensional feature interaction matrix through a factorization machine to obtain latent variables is: S31, Factorize the multi-dimensional feature interaction matrix, extract the high-order interaction relationship between power load features and police situation features, and construct a factorization model; S32, Generate latent variables according to the factorization result.

6. The real-time allocation method of regional power load based on artificial intelligence according to claim 1, characterized in that The process of calculating the risk correlation score of each region is: S41, Through latent variables, calculate the influence coefficient of power load anomalies on police situation risks, and construct a set of regional influence factors; S42, Combine historical police situation risk records and power load anomaly records to calculate the risk occurrence probability set of each region; S43, Combine the set of regional influence factors and the risk occurrence probability set to calculate the risk correlation score set of each region.

7. A real-time allocation method for regional power loads based on artificial intelligence according to claim 1, characterized in that The recommendation report includes region number, risk level, power load feature anomaly type, and recommended disposal measures; the specific process of generating the recommendation report is: S51, Sort the risk correlation scores through a sorting algorithm to obtain a region priority list; S52, According to the region priority list, filter out regions with risk correlation scores greater than a preset threshold to obtain a set of high-priority regions; S53, Dynamically adjust the recommendation results of high-priority regions in combination with the real-time multi-dimensional power dataset; S54, Generate a recommendation report according to the risk levels of the adjusted high-priority regions.

8. A real-time allocation method for regional power loads based on artificial intelligence according to claim 7, characterized in that The formula for dynamically adjusting the recommendation results of high-priority regions in combination with the real-time multi-dimensional power dataset is: Among them, represents the dynamically adjusted risk level of the high-priority area j; and are the adjustment weights of the risk score and real-time data, is the risk correlation score of area j, is the dynamic update function of the regional risk calculated based on the real-time multi-dimensional power dataset.

9. A real-time allocation method for regional power loads based on artificial intelligence according to claim 1, characterized in that The process of pushing the coordinated control plan is: S61. Monitor the power load data and police situation data in high-priority areas in real time, and judge the event types of police situations or major power load anomalies; S62. Obtain a list of key response areas according to the risk level information and characteristics of emergencies in high-priority areas in the recommendation report; S63. Determine the power load adjustment strategy based on real-time power load data and event types; S64. Generate police deployment suggestions according to the nature and impact scope of police situations or major power loads; S65. Integrate the power load adjustment strategy and police deployment suggestions to generate a collaborative control plan.

10. A real-time regional power load allocation system based on artificial intelligence, characterized in that, Including: A data acquisition module for obtaining the real-time power load data set and police situation data set of the area and integrating them into a multi-dimensional power data set; A feature extraction module for extracting features from the multi-dimensional power data set to obtain a power load feature vector and a police situation feature vector, and combining the power load feature vector and the police situation feature vector to construct a multi-dimensional feature interaction matrix; A factorization modeling module for decomposing the multi-dimensional feature interaction matrix through a factorization machine to obtain latent variables, where the latent variables are the high-order interaction relationships between power load features and police situation features; Conduct cluster analysis on the latent variable representation to divide the correlation level of power load and police situation risks; A risk assessment module for calculating the risk correlation score set of each area based on the latent variables and the correlation level of power load and police situation risks, and combining the historical police situation risk records and power load anomaly records of the area to obtain a list of high-correlation areas; A sorting and recommendation module for sorting the risk correlation scores based on the list of high-correlation areas, and dynamically adjusting the recommendation results of high-priority areas in combination with the real-time multi-dimensional power data set to generate a recommendation report; A collaborative control module for pushing a collaborative control plan in combination with power load data, police situation data and a recommendation report when a police situation or major power load anomaly occurs in a high-priority area.

Images