A cross-departmental government big data business collaborative processing method based on computing power platform and data fusion

Abstract

This invention relates to the technical field of power government administration and proposes a cross-departmental government big data collaborative processing method based on computing power platforms and data fusion. The method comprises the following steps: S1. Deploying a heterogeneous computing power cluster to establish a data processing platform for dynamic scheduling of computing resources, including CPUs, GPUs, FPGAs, and edge computing nodes; S2. Acquiring multidimensional data on power grid equipment, loads, industry electricity consumption, and government policies; and S3. Designing a cross-departmental data access system based on a zero-trust architecture to develop multidimensional models for power grid risk prediction, industry energy consumption analysis, and carbon emission accounting in power management. By applying a spatiotemporal joint probability prediction and conflict priority sorting algorithm to power-government emergency response, this method achieves intelligent and precise cross-departmental collaboration, allowing for the rapid resolution of critical power outages based on priority, thereby reducing the harmfulness and unpredictability of power outages.

Description

Technical Field

[0001] The present invention relates to the field of power government administration technology, and in particular to a cross-departmental government big data business collaborative processing method based on computing power platform and data fusion. Background Art

[0002] With the deep integration of digital government and the energy internet, cross-departmental collaboration has become a key requirement for improving public service efficiency and social governance capabilities. As a core component of national infrastructure, the operational status of the power system is closely linked to the livelihood security and policy implementation of government services. For example, government platforms require real-time electricity usage data to support public services, and the power system must optimize resource allocation based on government policy guidance. However, existing technologies face the following core challenges in cross-departmental collaboration.

[0003] First, traditional government affairs and power systems each deploy independent computing facilities, and computing resources such as CPUs and GPUs cannot be dynamically allocated on demand, resulting in an imbalance in computing power utilization in high-concurrency scenarios, making it difficult to support the real-time processing needs of big data. At the same time, multi-source data such as power equipment operation data, grid load data, and government policy documents have problems such as format heterogeneity and semantic ambiguity. Cross-departmental data sharing only stays at the basic field exchange level, lacking in-depth correlation analysis of equipment status, load characteristics, and policy orientation, forming "data islands" and resulting in a lack of comprehensiveness and timeliness in collaborative decision-making.

[0004] Secondly, existing cross-departmental data access mostly uses static permission control, which cannot cope with dynamically changing security threats. At the same time, the power-government collaboration scenario relies on manual experience to formulate plans, lacks data-driven intelligent prediction and optimization models, and is difficult to prioritize fault handling in a timely manner according to the severity of the consequences caused by the fault, resulting in high-risk and unpredictable power failures. Summary of the Invention

[0005] In response to the problems existing in the prior art, the purpose of the present invention is to provide a cross-departmental government big data business collaborative processing method based on computing power platform and data fusion to solve the problems raised by the above background technology.

[0006] To achieve the above objectives, the present invention provides a cross-departmental government big data business collaborative processing method based on computing power platform and data fusion, comprising the following steps:

[0007] S1. Deploy heterogeneous computing clusters and establish a data processing platform for dynamic scheduling of computing resources including CPU, GPU, FPGA, and edge computing nodes;

[0008] S2. Acquire multidimensional data on power grid equipment, grid load, industry power consumption, and government policy data, perform semantic annotation on the data, build a fusion model for power and government data, establish a related data structure for equipment operating status, grid load status, industry energy consumption characteristics, and policy-oriented information, and process all acquired data.

[0009] S3. Design a cross-departmental data access system based on a zero-trust architecture, develop a multi-dimensional model for power grid risk prediction, industry energy consumption analysis, and carbon emission accounting in power management, and design cross-departmental business collaboration content for power distribution planning and emergency response.

[0010] S4. Based on knowledge graphs, historical data, and data analysis models, build intelligent optimization recommendations for different decision-making scenarios and develop intelligent decision support systems;

[0011] S5. Establish a business collaboration effectiveness evaluation indicator system, optimize system parameters, and continuously optimize and analyze system models.

[0012] Preferably, in step S2, processing all acquired data includes the following steps:

[0013] S21. Establish a data integrity, accuracy, and consistency evaluation indicator system to standardize the acquired data;

[0014] S22. Design ontology-based semantic annotation rules to automatically map data objects to standard semantic annotations, and build a semantic annotation platform that supports human-computer collaboration and provides automatic annotation and manual verification functions;

[0015] S23. Automatically identify entities related to power equipment, grid load, industry policies, and policies from unstructured text, extract relationships, and establish a relationship model between entities;

[0016] S24. Fill in data attribute values ​​through data mapping and knowledge reasoning, map entities and relationships into low-dimensional vector space, and set up a query index mechanism;

[0017] S25. Model the time dimension of power grid operation data and industry electricity consumption data;

[0018] S26. Spatial modeling of power grid equipment, industry locations, and government planning based on spatiotemporal proximity and spatiotemporal co-occurrence relationships, and mining spatiotemporal correlation patterns between data;

[0019] S27. Make spatiotemporal forecasts of grid load, equipment status, industry electricity consumption, and government planning, and calculate the joint probability of events in the spatiotemporal domain.

[0020] Preferably, in step S3, cross-departmental business collaboration includes the following steps:

[0021] S31. Encrypt data in a hierarchical manner and set zero-trust access control permissions;

[0022] S32 collects and processes historical data, real-time monitoring data, and environmental data of power grid equipment and loads, and constructs a power grid risk prediction model and an industry energy consumption analysis model. The formula is: , where is the predicted value of power grid risk, is the observation vector, is the environmental feature vector, and are all eigenvectors, is the activation function, is the bias term, For prediction Energy consumption value of the industry at all times, The weight matrix of the fully connected layer, is the attention weight vector, is element-wise and multiplication, is a gated recurrent unit, is the latent space representation vector, For history The energy consumption sequence at each moment, is the bias vector;

[0023] S33. Based on the prediction results of the power grid risk prediction model and the industry energy consumption analysis model, and taking into account the power grid transmission loss, a carbon emission accounting model is constructed. The formula is: , where is the total carbon emissions from actual power generation, For the moment Industry electricity consumption, For the moment The power grid from The proportion of electricity generation from this type of energy, For the Carbon intensity of energy sources, is the power transmission loss rate of the power grid;

[0024] S34. Based on the prediction results of the multidimensional model and combined with government planning and industrial planning, detect conflicts between power grid planning and its related policies, evaluate conflicts, and prioritize conflict resolution. The formula is: , where Calculate the value for the priority, The impact of the conflict on scale, intensity and duration, Planning solutions for power grids The irreplaceability of For conflict resolution costs, To prevent the denominator from being zero, the value is ;

[0025] S35. Based on the conflict resolution priority ranking results, build an emergency response plan for the power grid planning scheme, and coordinate command across departments to solve problems in power grid dispatching.

[0026] Preferably, in step S31, access permission control includes the following steps:

[0027] S311. Classify the power-government data according to the sensitivity of the data, divide it into security levels, and encrypt the data in a hierarchical manner;

[0028] S312. Build a dynamic authorization model based on the multi-factor identity access system, monitor abnormal access in real time, and calculate the multi-factor access risk value;

[0029] S313. Dynamically adjust access permissions based on risk assessment results.

[0030] Preferably, in step S32, processing the collected data includes the following steps:

[0031] S321. Clean the collected data, process outliers, missing values, and duplicate values ​​in the data, unify the data format and dimension, and divide the processed data into a training set, a prediction set, and a validation set;

[0032] S322. Extract risk-related features from the data, evaluate the importance of each feature to risk prediction, and combine and transform the original features to generate new features;

[0033] S323, perform dimensionality reduction processing on high-order features and dynamically update feature engineering strategies;

[0034] S324. Based on the characteristics of power grid risk prediction, a neural network correlation model between characteristics and risks is constructed. The formula is: , where Based on the fractional The predicted value of , that is, the value of the correlation between the quantitative characteristics and the risk, is the input feature, is the true value, To treat the optimization variables Take the minimum parameter value of the loss function, that is, solve the optimal fractional prediction value through optimization.

[0035] Preferably, in step S35, solving the problem of power grid dispatching includes the following steps:

[0036] S351. Call the power grid GIS system to locate the conflict location, confirm the priority of conflict resolution, generate multiple solutions based on the priority of conflict resolution, and establish a cost-benefit matrix for the solutions. Then, evaluate the solutions and select the optimal solution.

[0037] S352. Based on the selected plan, establish a cross-departmental joint command center and assign power grid departments, government departments, and external experts to formulate emergency repair, material deployment, and backup management plans to achieve information sharing and coordinated command.

[0038] S353. Based on the carbon emission accounting model, carbon emissions are updated and calculated in real time, and carbon quotas are allocated according to the government's carbon emission reduction targets and actual conditions;

[0039] S354. Regularly evaluate the carbon emission reduction effects of the industry, verify the effectiveness of the plan, and formulate policy incentives or constraints based on the industry's carbon emission reduction situation, and share and publicize data to improve transparency.

[0040] Preferably, in step S4, the knowledge graph construction includes the following steps:

[0041] S41. Collect and annotate training data for entity recognition, construct a training dataset, establish a training model for the power-government sector, and fine-tune the data;

[0042] S42. Define the relationship types between entities in the power-government sector, construct a training dataset for relationship extraction, and establish a relationship extraction model. , where Fixed head entity feature and tail entity features When , the predicted relationship type is The probability distribution of is the weight matrix of the relation classification layer, is the bias vector, is the splicing symbol;

[0043] S43. Design an entity and relationship attribute system in the power-government sector. Extract attribute values ​​from entities and relationships, unify the extracted attribute values ​​into a unified data format, process outliers, and populate the processed attribute values ​​into the knowledge graph. Establish a dynamic update mechanism for attribute values ​​to ensure the timeliness of attribute values.

[0044] S44. Use test data to evaluate the performance of the entity recognition model, perform relational reasoning based on existing relationships, complete missing relationships in the knowledge graph, and perform manual verification.

[0045] Preferably, in step S5, continuously optimizing and analyzing the system model includes the following steps:

[0046] S51. Clarify the scenarios and core objectives of power-government collaboration, divide it into four dimensions: data, process, service, and resources, and form an evaluation system after preliminary selection of indicators through screening and weighting.

[0047] S52. Develop data interaction and process scheduling parameter strategies by category, design AB testing solutions for pilot verification, and establish risk control mechanisms;

[0048] S53. Continuously optimize the model through sensitivity analysis, develop intelligent platform solidification capabilities, and provide supporting training and systems to ensure long-term operation.

[0049] A cross-departmental government big data business collaborative processing system based on computing power platform and data fusion, applied to any one of the above-mentioned cross-departmental government big data business collaborative processing methods based on computing power platform and data fusion, comprising:

[0050] Heterogeneous computing power scheduling module, used to integrate CPU, GPU, FPGA and edge computing nodes to build a dynamic scheduling platform;

[0051] A multi-dimensional data fusion module is used to aggregate data on power grid equipment, loads, industry electricity consumption, and government policies, and build a cross-domain data fusion structure through semantic annotation and association modeling;

[0052] The security collaboration modeling module designs a cross-departmental data security access system based on a zero-trust architecture, develops multi-dimensional models for power grid risk prediction, energy consumption analysis, and carbon emission accounting, and plans cross-departmental collaborative business logic for power distribution and emergency response.

[0053] The intelligent decision support module combines knowledge graphs, historical data, and analytical models to build intelligent optimization recommendation solutions for different scenarios;

[0054] The performance evaluation and optimization module is used to establish a business collaboration performance evaluation indicator system, optimize system parameters through data monitoring and analysis, and continuously iterate and improve the model.

[0055] The present invention provides a cross-departmental government big data collaborative processing method based on computing power platform and data fusion, which has the following beneficial effects:

[0056] 1. By integrating CPU, GPU, FPGA and edge computing nodes to build a dynamic computing power platform, and introducing semantic annotation, spatiotemporal modeling and knowledge graph technology, we can achieve deep correlation and structured processing of multi-source data such as power equipment, loads, and policies, breaking through the traditional cross-departmental data silos and improving the efficiency of heterogeneous data fusion and analysis accuracy.

[0057] 2. Design a cross-departmental data access control mechanism based on a zero-trust architecture, combine multi-dimensional models such as power grid risk prediction, energy consumption analysis, and carbon emission accounting, build a "data security-business collaboration-intelligent decision-making" closed loop, and apply the spatiotemporal joint probability prediction and conflict priority sorting algorithm to power-government emergency response to achieve intelligent and precise cross-departmental business collaboration. In the event of a power network failure, it can quickly generate a collaborative processing plan for multiple departments and quickly handle power failures with serious consequences according to priority levels, which is conducive to reducing the harmfulness and unpredictability of power failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0059] Figure 1 A flowchart of the steps of a cross-departmental government big data business collaborative processing method based on computing power platform and data fusion provided in this application;

[0060] Figure 2 A schematic diagram of the system modules of a cross-departmental government big data business collaborative processing method based on computing power platform and data fusion provided for this application. DETAILED DESCRIPTION

[0061] The following embodiments of the present invention are described in further detail in conjunction with the accompanying drawings and examples. The following embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention.

[0062] like Figure 1-Figure 2 As shown, this embodiment proposes a cross-departmental government big data business collaborative processing method based on computing power platform and data fusion, including the following steps:

[0063] S1. Deploy heterogeneous computing clusters and establish a data processing platform for dynamic scheduling of computing resources including CPU, GPU, FPGA, and edge computing nodes;

[0064] S2. Acquire multidimensional data on power grid equipment, grid load, industry power consumption, and government policy data, perform semantic annotation on the data, build a fusion model for power and government data, establish a related data structure for equipment operating status, grid load status, industry energy consumption characteristics, and policy-oriented information, and process all acquired data.

[0065] S3. Design a cross-departmental data access system based on a zero-trust architecture, develop a multi-dimensional model for power grid risk prediction, industry energy consumption analysis, and carbon emission accounting in power management, and design cross-departmental business collaboration content for power distribution planning and emergency response.

[0066] S4. Based on knowledge graphs, historical data, and data analysis models, build intelligent optimization recommendations for different decision-making scenarios and develop intelligent decision support systems;

[0067] S5. Establish a business collaboration effectiveness evaluation indicator system, optimize system parameters, and continuously optimize and analyze system models.

[0068] This invention builds a dynamic computing power platform by integrating CPU, GPU, FPGA and edge computing nodes, and introduces semantic annotation, spatiotemporal modeling and knowledge graph technology to achieve deep correlation and structured processing of multi-source data such as power equipment, load, and policies, breaking through the traditional cross-departmental data island problem and improving the efficiency of heterogeneous data fusion and analysis accuracy.

[0069] In this embodiment, in step S2, processing all acquired data includes the following steps:

[0070] S21. Establish a data integrity, accuracy, and consistency evaluation indicator system to standardize the acquired data;

[0071] S22. Design ontology-based semantic annotation rules to automatically map data objects to standard semantic annotations, and build a semantic annotation platform that supports human-computer collaboration and provides automatic annotation and manual verification functions;

[0072] S23. Automatically identify entities related to power equipment, grid load, industry policies, and policies from unstructured text, extract relationships, and establish a relationship model between entities. The formula is: , where is the label sequence predicted by the model, For the set of all possible label sequences In the above example, we select the sequence that maximizes the product of subsequent probabilities. is the number of all label sequences, For the Characteristics of the location and the previous label Under the condition that the current label The conditional probability of

[0073] S24. Fill in data attribute values ​​through data mapping and knowledge reasoning, map entities and relationships into low-dimensional vector space, and set up a query index mechanism;

[0074] S25. Model the time dimension of the power grid operation data and industry electricity consumption data using the following formula: , where For data variables at time The observed value of is the intercept term, is the number of data variables, For the The smooth function corresponding to the independent variable, For the independent variable in time The observed value of For time The random error term at ;

[0075] S26. Spatial modeling of power grid equipment, industry locations, and government planning is performed using the following formula: , where and Respectively Layer and Nodes in a layer neural network The eigenvector of For nodes The set of spatial neighbors of Belong to the spatial neighbor set Variables , and Respectively The layer's weight matrix and bias terms, is the activation function, based on the relationship between spatiotemporal proximity and spatiotemporal co-occurrence, and mining the spatiotemporal correlation pattern between data. The formula is: , where For spatial location In time The predicted value of 、 and are the spatial and temporal latent factor matrices and the spatial-latent factor association tensor, respectively. 、 as well as They are spatial location attributes, spatial characteristics of temporal regularity government planning;

[0076] S27. Perform spatiotemporal forecasts on grid load, equipment status, industry electricity consumption, and government planning, and calculate the joint probability of events in the spatiotemporal domain. The formula is: , where is the joint probability value, is the spatial region attribute, Time window properties.

[0077] Specifically, by establishing a data quality assessment system and a semantic annotation platform, the power-government data is standardized and entity relationship modeled. Combined with mathematical modeling of the spatiotemporal dimensions, the spatiotemporal correlation patterns and joint probability distributions between data are mined, providing a highly consistent and highly available structured data foundation for subsequent power grid risk prediction, industry energy consumption analysis and other models, ensuring the accuracy and timeliness of cross-departmental collaborative analysis.

[0078] In this embodiment, in step S3, cross-departmental business collaboration includes the following steps:

[0079] S31. Encrypt data in a hierarchical manner and set zero-trust access control permissions;

[0080] S32 collects and processes historical data, real-time monitoring data, and environmental data of power grid equipment and loads, and constructs a power grid risk prediction model and an industry energy consumption analysis model. The formula is: , where is the predicted value of power grid risk, is the observation vector, is the environmental feature vector, and are all eigenvectors, is the activation function, is the bias term, For prediction Energy consumption value of the industry at all times, The weight matrix of the fully connected layer, is the attention weight vector, is element-wise and multiplication, is a gated recurrent unit, is the latent space representation vector, For history The energy consumption sequence at each moment, is the bias vector;

[0081] S33. Based on the prediction results of the power grid risk prediction model and the industry energy consumption analysis model, and taking into account the power grid transmission loss, a carbon emission accounting model is constructed. The formula is: , where is the total carbon emissions from actual power generation, For the moment Industry electricity consumption, For the moment The power grid from The proportion of electricity generation from this type of energy, For the Carbon intensity of energy sources, is the power transmission loss rate of the power grid;

[0082] S34. Based on the prediction results of the multidimensional model and combined with government planning and industrial planning, detect conflicts between power grid planning and its related policies, evaluate conflicts, and prioritize conflict resolution. The formula is: , where Calculate the value for the priority, The impact of the conflict on scale, intensity and duration, Planning solutions for power grids The irreplaceability of For conflict resolution costs, To prevent the denominator from being zero, the value is ;

[0083] S35. Based on the conflict resolution priority ranking results, build an emergency response plan for the power grid planning scheme, and coordinate command across departments to solve problems in power grid dispatching.

[0084] Specifically, based on the zero-trust architecture, data is hierarchically encrypted and access rights are dynamically controlled. Through the power grid risk prediction model, industry energy consumption analysis model and carbon emission accounting model, real-time monitoring of the power grid operation status and cross-departmental collaborative planning are achieved. At the same time, through conflict detection and priority sorting algorithms, the consistency of power grid planning and government policies is guaranteed, and the scientific nature and collaborative efficiency of emergency response are improved.

[0085] In this embodiment, in step S31, access permission control includes the following steps:

[0086] S311. Classify the power-government data according to the sensitivity of the data, divide it into security levels, and encrypt the data in a hierarchical manner;

[0087] S312. Based on the multi-factor identity access system, a dynamic authorization model is constructed, abnormal access is monitored in real time, and the multi-factor access risk value is calculated using the following formula: , where is the comprehensive risk value, and Respectively The weight coefficient and risk score of the class factors, and Respectively The penalty coefficient and Boolean value of the risk-like rule, and are the number of factors and risk rules, respectively;

[0088] S313. Dynamically adjust access rights based on risk assessment results. The formula is: , where For the overall environmental security and credibility, is the total number of environmental factors, For the Normalized safety value of an environmental factor.

[0089] Specifically, by classifying the sensitivity of power-government data and encrypting them in a hierarchical manner, combining multi-factor authentication and dynamic authorization models, calculating access risk values ​​in real time and adjusting permissions, a security protection chain of "data classification-risk assessment-dynamic adjustment of permissions" is formed, effectively preventing risks such as unauthorized access and data leakage in cross-departmental data sharing, and ensuring the compliant use and transmission security of sensitive data in collaborative scenarios.

[0090] In this embodiment, in step S35, solving the problem of power grid scheduling includes the following steps:

[0091] S351. Call the power grid GIS system to locate the conflict location, confirm the priority of conflict resolution, generate multiple solutions based on the priority of conflict resolution, and establish a cost-benefit matrix for the solutions. Then, evaluate the solutions and select the optimal solution.

[0092] S352. Based on the selected plan, establish a cross-departmental joint command center and assign power grid departments, government departments, and external experts to formulate emergency repair, material deployment, and backup management plans to achieve information sharing and coordinated command.

[0093] S353. Based on the carbon emission accounting model, carbon emissions are updated and calculated in real time. Carbon quotas are allocated according to the government's carbon emission reduction targets and actual conditions. The formula is: , where For the The annual carbon quota for each industry, is the annual output value of the industry, Adjust the average carbon emission intensity of the industry in the previous three years. is the adjustment factor, and , The annual emission reduction rate target is issued based on the baseline;

[0094] S354. Regularly evaluate the carbon emission reduction effects of the industry, verify the effectiveness of the plan, and formulate policy incentives or constraints based on the industry's carbon emission reduction situation, and share and publicize data to improve transparency.

[0095] Specifically, the power grid GIS system is used to locate conflicting locations and generate multiple solutions. The optimal strategy is selected based on the cost-benefit matrix. Resource allocation and information sharing are achieved through the cross-departmental joint command center. At the same time, the carbon emission accounting model is used to dynamically update carbon quotas. The power grid dispatch is combined with the government's carbon emission reduction targets. This not only ensures the stability of power grid operation, but also promotes the green transformation of the industry through data sharing and policy incentives, and enhances the environmental benefits and social transparency of cross-departmental collaboration.

[0096] In this embodiment, in step S4, the knowledge graph construction includes the following steps:

[0097] S41. Collect and label training data for entity recognition, build a training dataset, and establish a training model for the power-government sector. Fine-tune the data using the formula: , where is the adjusted parameter value, and are the number of samples and categories of government policies or power failures, and The samples Belong to category The true label and predicted logits value, is the exponential of the prediction score, is the regularization parameter, is the model parameter The squared norm of , that is, the sum of the squares of all parameters;

[0098] S42. Define the relationship types between entities in the power-government sector, construct a training dataset for relationship extraction, and establish a relationship extraction model. , where Fixed head entity feature and tail entity features When , the predicted relationship type is The probability distribution of is the weight matrix of the relation classification layer, is the bias vector, is the splicing symbol;

[0099] S43. Design an entity and relationship attribute system in the power-government sector. Extract attribute values ​​from entities and relationships, unify the extracted attribute values ​​into a unified data format, process outliers, and populate the processed attribute values ​​into the knowledge graph. Establish a dynamic update mechanism for attribute values ​​to ensure the timeliness of attribute values.

[0100] S44. Use test data to evaluate the performance of the entity recognition model, perform relational reasoning based on existing relationships, complete missing relationships in the knowledge graph, and perform manual verification.

[0101] Specifically, through entity recognition model training and relationship extraction algorithm, a knowledge graph in the power-government affairs field is constructed, integrating multi-dimensional entity relationships such as equipment, policies, and energy consumption. Combined with the dynamic update of attribute values ​​and the missing relationship reasoning and completion mechanism, a structured knowledge network is formed to provide domain knowledge support for the intelligent decision support system, enabling the system to generate optimized recommendation plans for different scenarios based on historical data and graph correlation relationships, thereby improving the accuracy and explainability of decisions.

[0102] In this embodiment, in step S32, processing the collected data includes the following steps:

[0103] S321. Clean the collected data, process outliers, missing values, and duplicate values ​​in the data, unify the data format and dimension, and divide the processed data into a training set, a prediction set, and a validation set;

[0104] S322. Extract risk-related features from the data, evaluate the importance of each feature to risk prediction, and combine and transform the original features to generate new features;

[0105] S323, perform dimensionality reduction processing on high-order features and dynamically update feature engineering strategies;

[0106] S324. Based on the characteristics of power grid risk prediction, a neural network correlation model between characteristics and risks is constructed. The formula is: , where Based on the fractional The predicted value of , that is, the value of the correlation between the quantitative characteristics and the risk, is the input feature, is the true value, To treat the optimization variables Take the minimum parameter value of the loss function, that is, solve the optimal fractional prediction value through optimization.

[0107] Specifically, the accuracy and robustness of power grid risk prediction and industry energy consumption analysis models are improved by cleaning, feature engineering and neural network modeling of collected data. Data cleaning ensures the data quality of the input model, feature engineering mines key influencing factors and reduces dimensional redundancy, and the neural network model achieves dynamic prediction of power grid operation status and energy consumption trend analysis by quantifying the correlation between features and risks, providing a scientific basis for cross-departmental collaboration in formulating power distribution plans and emergency response strategies, and enhancing the foresight and accuracy of power management.

[0108] In this embodiment, in step S5, continuously optimizing and analyzing the system model includes the following steps:

[0109] S51. Clarify the scenarios and core objectives of power-government collaboration, divide it into four dimensions: data, process, service, and resources, and form an evaluation system after preliminary selection of indicators through screening and weighting.

[0110] S52. Develop data interaction and process scheduling parameter strategies by category, design AB testing solutions for pilot verification, and establish risk control mechanisms;

[0111] S53. Continuously optimize the model through sensitivity analysis, develop intelligent platform solidification capabilities, and provide supporting training and systems to ensure long-term operation.

[0112] Specifically, by establishing a four-dimensional evaluation system covering data, processes, services, and resources, quantifying collaborative efficiency and identifying bottleneck indicators, and combining AB testing and sensitivity analysis to optimize system parameters, while relying on intelligent platforms to solidify optimization strategies, and supporting training and assessment systems to ensure long-term operation, a closed-loop management mechanism of "evaluation-optimization-verification-iteration" is formed to continuously improve the performance and user experience of the power-government collaborative system.

[0113] A cross-departmental government big data business collaborative processing system based on computing power platform and data fusion, applied to any one of the above-mentioned cross-departmental government big data business collaborative processing methods based on computing power platform and data fusion, comprising:

[0114] Heterogeneous computing power scheduling module, used to integrate CPU, GPU, FPGA and edge computing nodes to build a dynamic scheduling platform;

[0115] A multi-dimensional data fusion module is used to aggregate data on power grid equipment, loads, industry electricity consumption, and government policies, and build a cross-domain data fusion structure through semantic annotation and association modeling;

[0116] The security collaboration modeling module designs a cross-departmental data security access system based on a zero-trust architecture, develops multi-dimensional models for power grid risk prediction, energy consumption analysis, and carbon emission accounting, and plans cross-departmental collaborative business logic for power distribution and emergency response.

[0117] The intelligent decision support module combines knowledge graphs, historical data, and analytical models to build intelligent optimization recommendation solutions for different scenarios;

[0118] The performance evaluation and optimization module is used to establish a business collaboration performance evaluation indicator system, optimize system parameters through data monitoring and analysis, and continuously iterate and improve the model.

[0119] The present invention designs a cross-departmental data access control mechanism based on a zero-trust architecture, combines multi-dimensional models such as power grid risk prediction, energy consumption analysis, and carbon emission accounting, and constructs a "data security-business collaboration-intelligent decision-making" closed loop. The spatiotemporal joint probability prediction and conflict priority sorting algorithm are applied to power-government emergency response, realizing the intelligent and precise cross-departmental business collaboration. In the event of a power network failure, it can quickly generate a collaborative processing plan for multiple departments and quickly handle power failures with serious consequences according to priority levels, which is conducive to reducing the harmfulness and unpredictability of power failures.

[0120] The above embodiments are intended to illustrate the present invention only and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the embodiments, it should be understood by those skilled in the art that various combinations, modifications, or equivalent substitutions of the technical solutions of the present invention do not depart from the spirit and scope of the technical solutions of the present invention and should be encompassed by the scope of the claims of the present invention.

Claims

1. A cross-departmental government big data business collaborative processing method based on computing power platform and data fusion, characterized by: The following steps are involved: S1. Deploy heterogeneous computing clusters and establish a data processing platform for dynamic scheduling of computing resources including CPUs, GPUs, FPGAs, and edge computing nodes; S2. Acquire multidimensional data on power grid equipment, grid load, industry power consumption, and government policy data, perform semantic annotation on the data, build a fusion model for power and government data, establish a related data structure for equipment operating status, grid load status, industry energy consumption characteristics, and policy-oriented information, and process all acquired data. S3. Design a cross-departmental data access system based on a zero-trust architecture, develop a multi-dimensional model for power grid risk prediction, industry energy consumption analysis, and carbon emission accounting in power management, and design cross-departmental business collaboration content for power distribution planning and emergency response. In step S3, cross-departmental business collaboration includes the following steps: S31. Encrypt data in a hierarchical manner and set zero-trust access control permissions; S32 collects and processes historical data, real-time monitoring data, and environmental data of power grid equipment and loads, and constructs a power grid risk prediction model and an industry energy consumption analysis model. The formula is: , where is the predicted value of power grid risk, is the observation vector, is the environmental feature vector, and are all eigenvectors, is the activation function, is the bias term, For prediction Energy consumption value of the industry at all times, The weight matrix of the fully connected layer, is the attention weight vector, is element-wise and multiplication, is a gated recurrent unit, is the latent space representation vector, For history The energy consumption sequence at each moment, is the bias vector; S33. Based on the prediction results of the power grid risk prediction model and the industry energy consumption analysis model, and taking into account the power grid transmission loss, a carbon emission accounting model is constructed. The formula is: , where is the total carbon emissions from actual power generation, For the moment Industry electricity consumption, For the moment The power grid from The proportion of electricity generation from this type of energy, For the Carbon intensity of energy sources, is the power transmission loss rate of the power grid; S34. Based on the prediction results of the multidimensional model and combined with government planning and industrial planning, detect conflicts between power grid planning and its related planning, evaluate conflicts, and rank conflict resolution priorities. The formula is: , where Calculate the value for the priority, The impact of the conflict on scale, intensity and duration, Planning solutions for power grids The irreplaceability of For conflict resolution costs, To prevent the denominator from being zero, the value is ; S35. Based on the conflict resolution priority ranking results, construct an emergency response plan for the power grid planning scheme, and coordinate command across departments to solve problems in power grid dispatching; S4. Based on knowledge graphs, historical data, and data analysis models, build intelligent optimization recommendations for different decision-making scenarios and develop intelligent decision support systems; S5. Establish a business collaboration effectiveness evaluation indicator system, optimize system parameters, and continuously optimize and analyze system models.

2. The cross-departmental government affairs big data business collaborative processing method based on computing power platform and data fusion according to claim 1 is characterized in that: In step S2, processing all acquired data includes the following steps: S21. Establish a data integrity, accuracy, and consistency evaluation indicator system to standardize the acquired data; S22. Design ontology-based semantic annotation rules to automatically map data objects to standard semantic annotations, and build a semantic annotation platform that supports human-computer collaboration and provides automatic annotation and manual verification functions; S23. Automatically identify entities related to power equipment, grid load, industry policies, and policies from unstructured text, extract relationships, and establish a relationship model between entities; S24. Fill in data attribute values ​​through data mapping and knowledge reasoning, map entities and relationships into low-dimensional vector space, and set up a query index mechanism; S25. Model the time dimension of power grid operation data and industry electricity consumption data; S26. Spatial modeling of power grid equipment, industry locations, and government planning based on spatiotemporal proximity and spatiotemporal co-occurrence relationships, and mining spatiotemporal correlation patterns between data; S27. Make spatiotemporal forecasts of grid load, equipment status, industry electricity consumption, and government planning, and calculate the joint probability of events in the spatiotemporal domain.

3. The cross-departmental government affairs big data business collaborative processing method based on computing power platform and data fusion according to claim 1 is characterized in that: In step S31, access permission control includes the following steps: S311. Classify the power-government data according to the sensitivity of the data, divide it into security levels, and encrypt the data in a hierarchical manner; S312. Build a dynamic authorization model based on the multi-factor identity access system, monitor abnormal access in real time, and calculate the multi-factor access risk value; S313. Dynamically adjust access permissions based on risk assessment results.

4. The cross-departmental government affairs big data business collaborative processing method based on computing power platform and data fusion according to claim 1 is characterized in that: In step S32, processing the collected data includes the following steps: S321. Clean the collected data, process outliers, missing values, and duplicate values ​​in the data, unify the data format and dimension, and divide the processed data into a training set, a prediction set, and a validation set; S322. Extract risk-related features from the data, evaluate the importance of each feature to risk prediction, and combine and transform the original features to generate new features; S323, perform dimensionality reduction processing on high-order features and dynamically update feature engineering strategies; S324. Based on the characteristics of power grid risk prediction, a neural network correlation model between characteristics and risks is constructed. The formula is: , where Based on the fractional The predicted value of , that is, the value of the correlation between the quantitative characteristics and the risk, is the input feature, is the true value, To treat the optimization variables Take the minimum parameter value of the loss function, that is, solve the optimal fractional prediction value through optimization.

5. The cross-departmental government affairs big data business collaborative processing method based on computing power platform and data fusion according to claim 1 is characterized in that: In step S35, solving the problem of power grid dispatching includes the following steps: S351. Call the power grid GIS system to locate the conflict location, confirm the priority of conflict resolution, generate multiple solutions based on the priority of conflict resolution, and establish a cost-benefit matrix for the solutions. Then, evaluate the solutions and select the optimal solution. S352. Based on the selected plan, establish a cross-departmental joint command center and assign power grid departments, government departments, and external experts to formulate emergency repair, material deployment, and backup management plans to achieve information sharing and coordinated command; S353. Based on the carbon emission accounting model, carbon emissions are updated and calculated in real time, and carbon quotas are allocated according to the government's carbon emission reduction targets and actual conditions; S354. Regularly evaluate the carbon emission reduction effects of the industry, verify the effectiveness of the plan, and formulate policy incentives or constraints based on the industry's carbon emission reduction situation, and share and publicize data to improve transparency.

6. The cross-departmental government affairs big data collaborative processing method based on computing power platform and data fusion according to claim 1 is characterized by: In step S4, the knowledge graph construction includes the following steps: S41. Collect and annotate training data for entity recognition, construct a training dataset, establish a training model for the power-government sector, and fine-tune the data; S42. Define the relationship types between entities in the power-government sector, construct a training dataset for relationship extraction, and establish a relationship extraction model. , where Fixed head entity feature and tail entity features When , the predicted relationship type is The probability distribution of is the weight matrix of the relation classification layer, is the bias vector, is the splicing symbol; S43. Design an entity and relationship attribute system in the power-government sector. Extract attribute values ​​from entities and relationships, unify the extracted attribute values ​​into a unified data format, process outliers, and populate the processed attribute values ​​into the knowledge graph. Establish a dynamic update mechanism for attribute values ​​to ensure the timeliness of attribute values. S44. Use test data to evaluate the performance of the entity recognition model, perform relational reasoning based on existing relationships, complete missing relationships in the knowledge graph, and perform manual verification.

7. The cross-departmental government affairs big data business collaborative processing method based on computing power platform and data fusion according to claim 1 is characterized in that: In step S5, continuously optimizing and analyzing the system model includes the following steps: S51. Clarify the scenarios and core objectives of power-government collaboration, divide it into four dimensions: data, process, service, and resources, and form an evaluation system after preliminary selection of indicators through screening and weighting. S52. Develop data interaction and process scheduling parameter strategies by category, design AB testing solutions for pilot verification, and establish risk control mechanisms; S53. Continuously optimize the model through sensitivity analysis, develop intelligent platform solidification capabilities, and provide supporting training and systems to ensure long-term operation.

8. A cross-departmental government affairs big data business collaborative processing system based on computing power platform and data fusion, applied to a cross-departmental government affairs big data business collaborative processing method based on computing power platform and data fusion as described in any one of claims 1-7, characterized in that: include: Heterogeneous computing power scheduling module, used to integrate CPU, GPU, FPGA and edge computing nodes to build a dynamic scheduling platform; A multi-dimensional data fusion module is used to aggregate data on power grid equipment, loads, industry electricity consumption, and government policies, and build a cross-domain data fusion structure through semantic annotation and association modeling; The security collaboration modeling module designs a cross-departmental data security access system based on a zero-trust architecture, develops multi-dimensional models for power grid risk prediction, energy consumption analysis, and carbon emission accounting, and plans cross-departmental collaborative business logic for power distribution and emergency response. The intelligent decision support module combines knowledge graphs, historical data, and analytical models to build intelligent optimization recommendation solutions for different scenarios; The performance evaluation and optimization module is used to establish a business collaboration performance evaluation indicator system, optimize system parameters through data monitoring and analysis, and continuously iterate and improve the model.

Images