Intelligent monitoring system based on power dispatching data network

Abstract

The invention discloses an intelligent monitoring system based on an electric power dispatching data network, and the system comprises the steps: obtaining electric power dispatching data in the electric power dispatching data network through a data preprocessing module, carrying out the preprocessing of the electric power dispatching data, carrying out the integration of the preprocessed electric power dispatching data through a multi-dimensional information integration module, and carrying out the monitoring of the electric power dispatching data. The method comprises the steps of obtaining comprehensive operation state characteristics through a comprehensive operation module, carrying out safety evaluation on the comprehensive operation state characteristics through a safety evaluation module to obtain a safety evaluation result, carrying out power dispatching calculation through an analysis module according to the safety evaluation result to obtain an optimized dispatching scheme, and carrying out operability judgment on the optimized dispatching scheme through a judgment module. And according to the operability judgment result, obtaining an execution scheduling instruction of power scheduling, and carrying out load real-time monitoring and network node optimization on the power scheduling data network through a network analysis module to obtain final network optimization configuration.

Description

Technical Field

[0001] This invention belongs to the field of power monitoring technology, and in particular relates to an intelligent monitoring system based on a power dispatch data network. Background Technology

[0002] Power dispatching is a core area for ensuring the safe and stable operation of the power grid, directly impacting the reliability and economy of energy supply. With the expansion of power system scale and the widespread integration of new energy sources, the complexity and real-time requirements of dispatching decisions have significantly increased. Intelligent monitoring systems play a crucial role in optimizing dispatching operations and improving power grid operating efficiency. However, existing monitoring systems generally suffer from insufficient depth of decision analysis and weak dynamic adaptation capabilities to network resources when dealing with complex dispatching scenarios. These systems often struggle to comprehensively integrate multi-dimensional information, leading to inaccurate assessments of the rationality of dispatching decisions. Furthermore, they are slow to respond to dynamic load changes in the data network, making it difficult to effectively identify and optimize network bottlenecks.

[0003] In assessing the rationality of decisions, dispatchers need to comprehensively consider factors such as safety, economy, operability, and risk. However, existing systems often fail to balance real-time performance and accuracy when processing this multi-dimensional information. For example, during peak load periods, dispatchers may face pressure to quickly adjust power generation plans, but the system struggles to synthesize historical data, operating procedures, and real-time constraints to generate reliable decision evaluations in a short time, potentially leading to safety hazards or economic losses. Furthermore, this inadequacy in multi-dimensional information integration directly affects the dynamic perception and optimization of data network load. Because the dispatch data network carries massive amounts of real-time data, network nodes may experience delays or even failures due to high loads or abnormal traffic. However, existing systems lack accurate analysis of network heat distribution, making it difficult to adjust resource allocation in a timely manner, thus exacerbating the instability of decision support.

[0004] Therefore, the key issue of this research is how to build an intelligent monitoring system that can integrate multi-dimensional information in real time to evaluate the rationality of decision-making and optimize resource allocation by dynamically analyzing the heat distribution of data networks in complex and ever-changing power dispatch scenarios. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention proposes an intelligent monitoring system based on a power dispatch data network, thereby resolving the issues present in the prior art.

[0006] To achieve the above objectives, the present invention provides an intelligent monitoring system based on a power dispatch data network, comprising:

[0007] The system comprises a data preprocessing module, a multi-dimensional information integration module, a security assessment module, an analysis module, a judgment module, and a network analysis module, connected sequentially. The data preprocessing module acquires power dispatching data from the power dispatching data network and preprocesses it. The multi-dimensional information integration module integrates the preprocessed power dispatching data to obtain comprehensive operational status characteristics. The security assessment module performs a security assessment on the comprehensive operational status characteristics to obtain a security assessment result. The analysis module performs power dispatching calculations based on the security assessment result to obtain an optimized dispatching scheme. The judgment module assesses the operability of the optimized dispatching scheme and obtains the power dispatching execution command based on the operability assessment result. The network analysis module performs real-time load monitoring and network node optimization on the power dispatching data network to obtain the final optimized network configuration.

[0008] Optionally, in the data preprocessing module, the preprocessing of the power dispatch data includes missing value filling, outlier removal, and format and unit standardization.

[0009] Optionally, in the multi-dimensional information integration module, the process of integrating the preprocessed power dispatch data includes: performing dimensionality reduction processing on the preprocessed power dispatch data to obtain a low-dimensional feature set, wherein the power dispatch data includes real-time load data, power generation plan data, and historical dispatch data; performing dimensionality judgment on the low-dimensional feature set; clustering the low-dimensional feature set based on the judgment result to obtain preliminary operating status categories; optimizing the boundaries of the preliminary operating status types using machine learning algorithms based on the preliminary operating status categories to obtain precise operating status classifications; extracting key features from the precise operating status classifications and performing weighted processing to obtain comprehensive operating status features.

[0010] Optionally, in the security assessment module, the process of performing a security assessment on the comprehensive operational status characteristics includes:

[0011] The comprehensive operating status characteristics are assessed for safety thresholds to obtain compliance status characteristics. These compliance status characteristics are then classified using a machine learning model to obtain classified risk feature vectors. Power grid operation risks are calculated based on these classified risk feature vectors to obtain risk index values. These risk index values ​​are analyzed using a logistic regression algorithm to obtain a risk probability distribution. Based on this risk probability distribution, a decision tree algorithm is used to generate a safety assessment result, resulting in a power grid operation safety level. By comparing the safety level with a preset safety threshold, the stability of the power grid operation status is determined, and a safety assessment output is obtained.

[0012] Optionally, in the analysis module, the process of performing power dispatch calculations based on the security assessment results includes:

[0013] A comprehensive dataset is constructed based on cost data and security assessment results of power dispatching. The security assessment results in the comprehensive dataset are evaluated. Based on the evaluation results, an initial dispatching scheme is calculated using a linear programming algorithm to obtain a preliminary scheme set. Based on the preliminary scheme set, combined with cost data and corresponding constraints, a greedy algorithm is used to iteratively optimize the preliminary schemes to generate an optimized dispatching scheme set. The dispatching schemes in the optimized dispatching scheme set are evaluated for economic efficiency. Based on the economic efficiency scores, the dispatching schemes are evaluated to obtain candidate schemes. The candidate schemes are then validated to obtain the optimized dispatching scheme. The dispatching scheme includes power dispatching task priority and resource allocation information.

[0014] Optionally, the network analysis module includes a network heat analysis module, a resource allocation optimization module, a network bottleneck identification module, and a resource configuration strategy generation module connected in sequence. Specifically, the network heat analysis module acquires node data from the power dispatch data network, performs heat analysis on the node data to obtain the network load status, the resource allocation optimization module allocates network resources to the nodes according to the network load status to obtain a resource configuration strategy, the network bottleneck identification module identifies high-load areas according to the resource configuration strategy to obtain a bottleneck node list, and the resource configuration strategy generation module performs optimized resource allocation according to the bottleneck node list to obtain the final network optimization configuration scheme for the power dispatch data network.

[0015] Optionally, in the network popularity analysis module, the process of performing popularity analysis on node data includes:

[0016] The node data is preprocessed, including real-time traffic data and node load data. The preprocessed node data is classified and network hotspot features are analyzed to obtain a heat distribution vector. The heat distribution vector is judged to obtain a set of high-load nodes. The contribution rate of the set of high-load nodes is calculated, and the weighted processing is performed based on the contribution rate calculation results to obtain the network load status.

[0017] Optionally, in the resource allocation optimization module, the process of allocating network resources to nodes includes:

[0018] The network load status is analyzed to obtain traffic distribution patterns and network performance indicators. The traffic distribution patterns are judged, and based on the judgment results, load balancing requirements are obtained. Threshold judgments are made on the load balancing requirements. Based on the threshold judgment results, a resource allocation optimization scheme is calculated using a dynamic allocation algorithm according to the traffic distribution patterns. Network resources are adjusted according to the resource allocation optimization scheme to obtain an optimized resource allocation scheme. Resource scheduling efficiency is extracted based on the optimized resource allocation scheme, and the resource scheduling efficiency is evaluated and judged. Based on the judgment results, the resource allocation scheme is adjusted to obtain a resource configuration strategy.

[0019] Optionally, in the network bottleneck identification module, the process of identifying high-load areas according to the resource configuration strategy includes:

[0020] Resource allocation strategies are used to configure resources in the power dispatch data network. After configuration, the load status of the power dispatch data network is analyzed through a load detection mechanism. Based on the load status, a bottleneck node set is obtained. Based on the bottleneck node set, high-load areas are identified through network topology analysis. Based on the high-load areas, the bottleneck nodes are grouped using the k-means clustering algorithm to obtain a node group set. Based on the node group set, the load distribution characteristics of each group are calculated to obtain the load distribution vector. The variance of the load distribution vector is judged. Based on the judgment result, the shortest path between nodes is calculated using the Dijkstra algorithm to obtain the load redistribution path. Based on the load redistribution path, the network traffic distribution is adjusted to obtain a new load balance state. By monitoring the new load balance state, the bottleneck node set is updated and obtained.

[0021] Optionally, in the resource allocation strategy generation module, the process of optimizing resource allocation based on bottleneck nodes includes:

[0022] Obtain node analysis data from the bottleneck node list. Based on the node analysis data, use a linear programming algorithm to calculate the resource allocation ratio for the bottleneck nodes, resulting in a bandwidth adjustment scheme and a resource scheduling scheme. If the allocation ratio in the bandwidth adjustment scheme exceeds a preset threshold, dynamically divide the network bandwidth to generate a bandwidth allocation configuration. Based on the bandwidth allocation configuration, adjust the scheduling priority of the resource scheduling scheme for the bottleneck nodes to obtain a resource scheduling configuration. Through the resource scheduling configuration, use a simulated annealing algorithm to optimize the network topology and generate an optimized network configuration. Extract performance improvement indicators from the optimized network configuration and determine whether the performance improvement indicators meet the preset threshold. If not, return to the second step to recalculate the resource allocation ratio. Based on the performance improvement indicators, generate the final system configuration.

[0023] Compared with the prior art, the present invention has the following advantages and technical effects:

[0024] This invention discloses a comprehensive scheduling and management technology that integrates power grid operation and network resource optimization. Addressing the operational risks and inefficiencies caused by load fluctuations, uneven resource allocation, and potential bottlenecks in the coordinated operation of the power grid and data network, it achieves efficient and stable system operation through multi-dimensional information integration and dynamic optimization. First, this invention extracts real-time load, generation plans, and historical scheduling data from power grid operation data. Data preprocessing techniques are used to clean and standardize the data into a structured dataset. A multi-dimensional information integration algorithm is employed to fuse load, generation, and constraints, generating comprehensive operational status characteristics. If the characteristics meet a safety threshold, a safety assessment algorithm calculates risk indicators, and an economic analysis algorithm evaluates costs and constraints, generating optimized scheduling suggestions. Feasibility is verified through an operability judgment algorithm, and executable scheduling instructions are output. Simultaneously, this invention acquires traffic and node load data from the data network, generates load status characteristics through a network heat analysis algorithm, and dynamically adjusts resource allocation using a resource allocation optimization algorithm. If a bottleneck node is detected, a network bottleneck identification algorithm locates high-load areas and reallocates bandwidth and computing resources. Through the coordinated optimization of the power grid and network, this invention significantly improves operational stability and resource utilization efficiency, while reducing operational risks and costs. Attached Figure Description

[0025] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0026] Figure 1 This is a schematic diagram of the system structure according to an embodiment of the present invention. Detailed Implementation

[0027] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0028] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0029] like Figure 1 As shown, the present invention provides an intelligent monitoring system based on a power dispatch data network, comprising, in sequence:

[0030] The data preprocessing module is used to obtain real-time load, power generation plan and historical dispatch data from power grid operation data, and to clean and standardize multi-dimensional information through data preprocessing technology to obtain structured datasets;

[0031] The multidimensional information integration module is used to integrate the constraints in load, power generation and historical dispatch data based on the structured dataset using a multidimensional information integration algorithm to generate comprehensive operating status characteristics.

[0032] The safety assessment module is used to calculate the power grid operation risk index and obtain the safety assessment result by means of a safety assessment algorithm if the comprehensive operating status characteristics meet the preset safety threshold.

[0033] The economic analysis module is used to calculate the economic score of the scheduling scheme based on the safety assessment results, using an economic analysis algorithm combined with cost data and constraints, and to obtain the optimized scheduling scheme.

[0034] The operability judgment module is used to verify the feasibility of the scheduling scheme and determine the executable scheduling instruction if the operability score of the optimized scheduling scheme is higher than a preset threshold.

[0035] The network heat analysis module is used to obtain real-time traffic and node load data from the data network, calculate the network heat distribution characteristics through the network heat analysis algorithm, and obtain the network load status.

[0036] The resource allocation optimization module is used to dynamically adjust the network resource allocation scheme according to the network load status using a resource allocation optimization algorithm, and generate an optimized resource configuration strategy.

[0037] The network bottleneck identification module is used to locate high-load areas and obtain a list of bottleneck nodes if the network load status shows bottleneck nodes.

[0038] The resource allocation strategy generation module is used to reallocate network bandwidth and computing resources based on the bottleneck node list and using resource allocation optimization algorithms to generate the final optimized network configuration.

[0039] The present invention provides an intelligent monitoring system based on a power dispatch data network. The above modules are connected sequentially, and the data processing process of each module is as follows:

[0040] In the data preprocessing module, real-time load, power generation plan and historical dispatch data are obtained from the power grid operation data. The multi-dimensional information is cleaned and standardized by data preprocessing technology to obtain a structured dataset.

[0041] Real-time load data, generation plan data, and historical dispatch data from the power grid operation data network are acquired, and multi-dimensional data information is extracted through a data extraction process. If missing or outlier values ​​exist in the multi-dimensional data, data cleaning techniques are used to fill or remove them, resulting in a cleaned dataset. Data standardization is then applied to unify the format and normalize the dimensions of the cleaned dataset, generating a standardized dataset. Data integrity verification checks the completeness and consistency of the standardized dataset, resulting in a verified dataset. If format inconsistencies exist in the verified dataset, data format conversion is performed to generate a structured dataset.

[0042] For example, power grid operation data processing involves the extraction, cleaning, standardization, and verification of real-time load data, generation plan data, and historical dispatch data. The following analysis focuses on the power grid operation field through specific scenarios and multi-dimensional data processing examples.

[0043] For example, real-time load data reflects the current electricity demand of the power grid and typically includes dimensions such as time, load value, and region. Suppose a power grid system collects load data every 5 minutes. A record for a certain day might be: Time 2025-07-23 08:00, Load 1000MW, Region A; however, some records may be missing load values ​​or contain outliers such as negative values. The data extraction process requires separating multi-dimensional information from the database, such as timestamps, numerical values, and region identifiers.

[0044] In one possible implementation, data for a specified time period can be extracted from the original database via SQL queries, ensuring the integrity of multidimensional information. This includes cleaning up missing or outlier values.

[0045] Preferably, interpolation can be used to fill in missing values.

[0046] For example, if load data is missing at a certain point in time, it can be filled by averaging the load values ​​from the preceding and following time points. If 08:05 is missing, and 08:00 has a load of 1000MW, while 08:10 has a load of 1010MW, then the filled value would be 1005MW. Outliers, such as negative values, can be directly removed because they do not conform to physical meaning. After cleaning, the dataset retains valid records, improving data reliability. In data standardization, the cleaned dataset needs to have a unified format and normalized units.

[0047] For example, the unit of load data might be MW or kW, and it needs to be standardized to MW; the time format might be "YYYY-MM-DDHH:MM" or "YYYY / MM / DD", and it needs to be converted to a unified format. Dimensional normalization can use min-max normalization, mapping load values ​​to the 0-1 range. Assuming the load range is 500-1500MW, 1000MW is normalized to 0.5. This process ensures consistency in subsequent analyses and improves model compatibility. Data integrity verification checks the completeness and consistency of records.

[0048] For example, the validated dataset must ensure that each record contains fields such as time, load, and region, and that there are no null values. This can be achieved by counting the number of non-null values ​​in each column or checking the continuity of the time series. If data is missing in a certain region, it can be marked as incomplete and needs to be extracted or filled in again. For inconsistent formatting issues, such as inconsistent time formats, a script can be used to convert it to a standard format to generate a structured dataset.

[0049] Specifically, structured datasets can be used for power grid load forecasting or scheduling optimization.

[0050] For example, based on cleaned and standardized datasets, predictive models can more accurately analyze regional load trends, optimize power generation plans, and reduce resource waste. Validated data consistency ensures the accuracy of dispatch instructions and avoids grid risks caused by data errors.

[0051] In one embodiment, assuming that significant deviations are found between planned and actual power generation in historical dispatch data for a certain region, cleaning and standardization can identify the causes of these deviations, such as equipment failure or unreasonable planning, thereby optimizing future dispatch decisions. This process significantly improves the efficiency and stability of power grid operation.

[0052] For example, data cleaning and standardization can also support anomaly detection. Suppose that the load suddenly increases to 2000MW on a certain day, far exceeding the normal range. After cleaning, it can be identified as an anomaly. Combined with historical data analysis, the cause can be identified, such as a sudden increase in industrial electricity demand, thus providing a basis for grid expansion.

[0053] Understandably, the above processing steps, through multidimensional data extraction, cleaning, standardization, and verification, form a high-quality structured dataset, providing reliable support for power grid operation, optimizing resource allocation, reducing operational risks, and ensuring power supply stability.

[0054] In the multidimensional information integration module, based on the structured dataset, the multidimensional information integration algorithm is used to fuse the constraints in the load, power generation and historical dispatch data to generate comprehensive operating status characteristics.

[0055] Using a structured dataset containing constraints from real-time load data, generation plan data, and historical dispatch data, principal component analysis (PCA) is employed to reduce the dimensionality of the standardized dataset, resulting in a low-dimensional feature set. If the dimensionality of the low-dimensional feature set meets a preset threshold, k-means clustering is used to classify the feature set, generating preliminary operating state categories. Based on these preliminary operating state categories, support vector machine (SVM) is used to optimize the category boundaries, yielding precise operating state classifications. Key features are extracted from these precise operating state classifications and weighted according to the constraints to generate comprehensive operating state features.

[0056] In one possible implementation, the structured dataset of power grid operation data includes real-time load data, generation plan data, and historical dispatch data. This data typically possesses high-dimensional characteristics, such as load curves, generator output, and dispatch instructions. The implementation of principal component analysis (PCA) aims to reduce data redundancy and retain core information through dimensionality reduction.

[0057] For example, from a structured dataset containing 100 features, the principal components with the highest contribution rates can be extracted. Assuming the top three principal components can explain 80% of the variance, the data dimensionality can be reduced from 100 dimensions to 3 dimensions. Specifically, this can be achieved by calculating the covariance matrix of the dataset, extracting eigenvalues ​​and eigenvectors, and selecting the main eigenvectors to form a low-dimensional feature set.

[0058] It should be noted that during the dimensionality reduction process, it is necessary to ensure that the retained features can reflect key changes in the grid's operating status, such as load peak-valley differences or the stability of power generation plans.

[0059] For example, in the classification process of the k-means clustering algorithm, a low-dimensional feature set is used to identify different operating states of the power grid. Assuming that the low-dimensional feature set includes three dimensions: load fluctuation, power generation output ratio, and dispatch frequency, and the preset number of clusters k=3, the operating states of the power grid are divided into peak, valley, and stable states.

[0060] Specifically, the Euclidean distance from each data point to the cluster center can be calculated iteratively, and the data points can be assigned to the nearest cluster center to generate a preliminary running status category.

[0061] In one embodiment, if the load data of a certain regional power grid shows 2000 MW during peak hours, the power generation output ratio is 80%, and the dispatch frequency is high, it may be classified as a peak state.

[0062] It should be noted that the choice of k value needs to be combined with the business scenario. For example, the elbow rule can be used to determine the appropriate number of clusters to ensure that the classification results are consistent with the actual operating state.

[0063] In one possible implementation, the support vector machine algorithm is used to optimize the boundaries of the initial running state categories.

[0064] For example, peak and trough categories obtained based on k-means clustering may have blurred boundaries for some data points. For instance, when the load is close to 1500 MW, it's difficult to clearly distinguish between peak and stable states. Support Vector Machines (SVMs) improve classification accuracy by finding the maximum margin hyperplane to optimize category boundaries.

[0065] Specifically, a low-dimensional feature set can be input into a support vector machine model, and a kernel function can be used to process non-linearly separable data to generate an accurate operational state classification.

[0066] In one embodiment, assuming a data point has a load of 1450 MW and a power generation output ratio of 75%, it is clearly classified as a stationary state after optimization using a support vector machine. This precise classification helps to accurately identify the power grid operation mode.

[0067] For example, when extracting key features and weighting them in conjunction with constraints, features such as load peak, generation plan deviation, and dispatch response time can be extracted from the precise operating state classification. Constraints such as generator capacity limitations or grid stability requirements need to be weighted to reflect their importance.

[0068] Specifically, a weight of 0.5 can be assigned to the load peak, 0.3 to the generation plan deviation, and 0.2 to the dispatch response time, thereby generating comprehensive operating status characteristics.

[0069] In one embodiment, if a power grid has a peak load of 2200 MW during peak hours, a generation plan deviation of 5%, and a dispatch response time of 10 minutes, a weighted average of these values ​​can generate a characteristic value reflecting the overall operating status. This characteristic value can intuitively display the overall characteristics of the power grid operation, facilitating subsequent analysis and decision-making.

[0070] In one possible implementation, the generation of comprehensive operational status characteristics needs to be combined with the actual needs of the business scenario.

[0071] For example, for a power grid in a certain region, the sources of load pressure during peak hours can be identified by analyzing comprehensive characteristic values, such as an excessively high proportion of industrial electricity consumption, thereby optimizing power generation plans or dispatch strategies.

[0072] It should be noted that this method can effectively support dynamic monitoring of the power grid's operating status, improving dispatch efficiency and resource utilization.

[0073] In the safety assessment module, if the comprehensive operating status characteristics meet the preset safety threshold, the power grid operation risk index is calculated through the safety assessment algorithm to obtain the safety assessment result.

[0074] If the operational status dataset meets a preset safety threshold, data compliance is determined through threshold comparison logic to obtain compliance status features. A support vector machine algorithm is used to classify these compliance status features, resulting in a classified risk feature vector. Based on this risk feature vector, power grid operation risk indicators are calculated, yielding risk indicator values. Logistic regression is used to further analyze these risk indicator values, obtaining a risk probability distribution. Based on this risk probability distribution, a decision tree algorithm is used to generate a safety assessment result, resulting in a power grid operation safety level. By comparing the safety level with a preset safety threshold, the stability of the power grid operation status is determined, leading to the final safety assessment output.

[0075] For example, in the scenario of power grid operation status analysis, the operation status dataset typically contains multi-dimensional information such as load, generation, and dispatch constraints. Suppose a regional power grid dataset includes real-time load power, generator output, and historical dispatch constraints, and it needs to be determined whether it meets safety thresholds. Safety thresholds could be a load fluctuation rate below 5%, generation output stable within 80%-100% of rated power, and dispatch constraints satisfying grid frequency fluctuations of less than 0.2 Hz. The threshold comparison logic generates compliance status characteristics by checking these indicators item by item.

[0076] For example, if at a certain moment the load fluctuation rate is 3%, the power generation output is 90% of the rated power, and the frequency fluctuation is 0.1 Hz, the data is judged to be compliant, and a positive compliance status feature is generated.

[0077] In one possible implementation, support vector machine classification based on compliance status features can further distinguish potential risks.

[0078] For example, compliance status characteristics include load fluctuations, output deviations, and frequency fluctuations. Using a support vector machine (SVM) algorithm, these feature vectors are categorized into low-risk, medium-risk, and high-risk classes. Assuming a dataset has feature vectors of [3%, 10%, 0.1 Hz], after classification, it is assigned to the low-risk category, generating a corresponding risk feature vector. This process constructs a hyperplane to separate features at different risk levels, optimizing the classification boundary.

[0079] Specifically, when calculating power grid operation risk indicators based on risk feature vectors, a weighted summation method can be used.

[0080] For example, the weights of the low-risk feature vector are set to load fluctuation 0.4, output deviation 0.3, and frequency fluctuation 0.3, and the risk index value is calculated.

[0081] For example, a risk index value of 2.5 for a certain set of data indicates low risk. The logistic regression algorithm further analyzes this index value to generate a risk probability distribution.

[0082] For example, the probability distribution corresponding to a risk index value of 2.5 shows that the probability of a power grid failure is 10%, and the probability of normal operation is 90%.

[0083] In one possible implementation, the decision tree algorithm generates security assessment results based on the risk probability distribution.

[0084] For example, nodes with a failure probability below 15% and an index value less than 3 in the probability distribution are classified as high-safety level. Assuming a power grid has a failure probability of 10% and an index value of 2.5, it is ultimately classified as Level 1 safety. This process, through multi-layered conditional branching, comprehensively considers both probability and index to generate a clear assessment result.

[0085] For example, the stability of the power grid's operating status can be determined by comparing the security level with a preset security threshold.

[0086] For example, if the preset safety threshold requires a Level 1 safety rating and an indicator value below 3, and the aforementioned power grid meets these conditions, it is determined to be in a stable operating state. The final safety assessment output is "stable," accompanied by detailed indicator values ​​and probability distributions as reference. This assessment method, through multi-level analysis, gradually refines risk characteristics, ultimately forming an operable judgment of operating status, providing a reliable basis for power grid dispatch.

[0087] In the economic analysis module, based on the safety assessment results, an economic analysis algorithm is used to calculate the economic score of the scheduling scheme by combining cost data and constraints, thereby obtaining the optimized scheduling scheme.

[0088] Acquire security assessment data and cost data, extract constraints from preset business constraints, and construct a comprehensive dataset. If the security assessment data in the comprehensive dataset meets preset thresholds, a linear programming algorithm is used to calculate an initial scheduling scheme, resulting in a preliminary scheme set. Based on the preliminary scheme set, combined with cost data and constraints, a greedy algorithm is used to iteratively optimize the schemes, generating an optimized scheduling scheme set. Economic score features are extracted from the optimized scheduling scheme set, and a weighted summation method is used to calculate the economic score of each scheme, determining the score ranking. If the scheme with the highest economic score in the ranking meets the constraints, it is marked as a candidate optimized scheme, and a candidate scheme description is generated. Based on the candidate scheme description, combined with business constraints and cost data, a secondary verification is performed to determine whether all constraints are met, resulting in the final optimized scheduling scheme. Key parameters are extracted from the final optimized scheduling scheme to generate scheduling execution instructions, outputting an executable scheduling scheme, which includes task priority and resource allocation information.

[0089] For example, in power grid operation optimization and scheduling scenarios, obtaining security assessment data and cost data is fundamental to constructing a comprehensive dataset. Security assessment data may include indicators such as grid load, voltage stability, and failure rate, while cost data involves generation costs, transmission loss costs, etc. Suppose that the security assessment data for a certain regional power grid indicates a load rate below 80%, meeting a preset threshold; the cost data shows that thermal power generation costs 0.5 yuan per kilowatt-hour, and renewable energy generation costs 0.3 yuan per kilowatt-hour. Conditions are extracted from business constraints, such as the maximum load not exceeding 90% and the renewable energy share needing to reach 30%. These data are integrated into a comprehensive dataset, providing a basis for subsequent optimization.

[0090] In one possible implementation, a linear programming algorithm is used to compute the initial scheduling scheme. Linear programming generates a preliminary set of schemes through an objective function (such as minimizing total cost) and constraints (such as load balancing).

[0091] For example, given the grid demand during a certain hour, the algorithm may generate three solutions: Solution A, which relies primarily on thermal power, costs 1 million yuan; Solution B, which increases the proportion of renewable energy, costs 950,000 yuan; and Solution C, which relies entirely on renewable energy, costs 1.1 million yuan. All these solutions meet the load constraints, but their costs and renewable energy proportions differ.

[0092] Specifically, based on the initial set of solutions, and combined with cost data and constraints, a greedy algorithm is used for iterative optimization. The greedy algorithm prioritizes the power generation unit with the lowest cost while ensuring the constraints are met. Assuming that, based on solution B, the algorithm reduces the cost to 920,000 yuan by adjusting the ratio of renewable energy to thermal power, while simultaneously meeting the requirement of renewable energy accounting for 30%, thus forming an optimized dispatch solution set.

[0093] Preferably, economic scoring features, such as power generation cost and transmission efficiency, are extracted, and the economic score is calculated by weighted summation. Assuming a cost weight of 0.6 and an efficiency weight of 0.4, Option B scores 92, Option A scores 88, and Option C scores 85.

[0094] In one embodiment, the highest-scoring scheme B, if it satisfies all constraints, such as load factor and renewable energy ratio, is marked as a candidate optimization scheme. Its description includes information such as power generation unit allocation and scheduling time. The secondary verification stage, combined with business constraints, checks whether the scheme meets dynamic load requirements, such as whether it remains economical when the load decreases at night. Assuming that scheme B passes the verification check, it becomes the final optimized scheduling scheme.

[0095] For example, key parameters are extracted from the final scheme, such as the operating power of thermal power units being 500 MW and that of renewable energy units being 300 MW, to generate dispatch execution instructions. These instructions clearly define task priorities, such as prioritizing the dispatch of renewable energy units and allocating resources, such as transmission line capacity. The final executable dispatch scheme ensures stable grid operation and optimized costs, and the clear instructions can be directly applied to actual dispatching.

[0096] In the operability judgment module, if the operability score of the optimized scheduling scheme is higher than the preset threshold, the operability judgment algorithm is used to verify the feasibility of the scheduling scheme and determine the executable scheduling instruction.

[0097] The raw data of the scheduling scheme, including task priority and resource allocation information, is acquired through the data acquisition module to obtain structured scheduling suggestions. If the structured scheduling suggestions contain complete task information, an operability score is calculated using an operability assessment model to obtain the score result. The operability score is compared with a preset threshold by the score comparison module to determine whether the scheduling suggestions meet the execution conditions, and a comparison result is obtained. If the comparison result shows that the operability score is higher than the preset threshold, a feasibility analysis of the scheduling scheme is performed using an operability verification algorithm to obtain a feasibility confirmation result. Based on the feasibility confirmation result, the feasible scheduling scheme is converted into specific execution instructions using the instruction generation module to obtain executable scheduling instructions. The executable scheduling instructions are transmitted to the target execution unit through the instruction distribution system to obtain feedback information from the execution unit and obtain the instruction execution status. Based on the instruction execution status, the execution effect of the scheduling instructions is evaluated by the status analysis module to generate optimized scheduling suggestion data.

[0098] For example, during the initial data acquisition process for a scheduling scheme, a data acquisition module can obtain task priority and resource allocation information from multiple sources. Assuming a power grid dispatching scenario, the data acquisition module obtains real-time data from the power dispatching data network, such as the power dispatching data required by different regions, line and resource type availability, and dispatching time windows. The raw data may include priority scores (range 0 to 10) for the power dispatching data required by different regions, as well as line planning and required available time periods. This data is cleaned and integrated to form structured dispatching recommendations, containing the task priorities and resource allocation for different regions. Structured dispatching recommendations ensure data consistency and facilitate subsequent analysis. This is also relevant in the application of operability assessment models.

[0099] Understandably, the evaluation model will calculate an operability score based on task priority, resource availability, and time constraints.

[0100] For example, the model might consider the urgency of power dispatch, resource types such as wind and thermal power, and time. If an order has a priority of 8, and the allocated resources (wind and thermal power) have line connections and are plentiful, the score will be high, such as 0.9 (out of 1.0). Conversely, if there are no direct line connections or few resources, the score may drop to 0.6. The score reflects the feasibility of the dispatch recommendations and provides a basis for subsequent screening.

[0101] Specifically, the score comparison module compares the operability score with a preset threshold (such as 0.8) to determine whether the scheduling suggestion meets the execution conditions.

[0102] For example, a scheduling suggestion with a score of 0.9, above the threshold, indicates strong executability; another suggestion with a score of 0.7 needs optimization or elimination. Comparison results help filter out high-quality scheduling suggestions, ensuring execution efficiency. In the feasibility analysis phase, the operability verification algorithm further examines the constraints of the scheduling scheme.

[0103] For example, the algorithm verifies whether power resource scheduling can be completed within a specified time window and whether the power scheduled for different resource types meets the demand. If a plan schedules power for different resource types in two regions, and the verification results show that all data scheduling is completed within the time window and there are still resources remaining, then the plan is confirmed to be feasible. The feasibility confirmation results provide a reliable basis for subsequent instruction generation.

[0104] For example, the instruction generation module converts feasible scheduling schemes into specific execution instructions. Suppose a feasible scheme involves scheduling power generation equipment at 3 locations to supply power to 5 demand locations. The instructions might include, "Power generation equipment A will be scheduled at 10:00 to supply power to demand locations 1 and 2, expected to be completed at 12:00." These instructions clearly define task allocation and timing, facilitating operation by the execution unit. In instruction distribution system applications, executable scheduling instructions are transmitted via network to the target execution unit, such as the vehicle's dispatch terminal.

[0105] For example, an instruction is distributed to the terminal of power generation equipment A, the terminal confirms receipt and begins execution. The execution unit provides feedback information, such as "Electricity demand 1 has been met," forming the instruction execution status and reflecting the execution progress.

[0106] Specifically, the status analysis module evaluates the effectiveness of scheduling instructions based on the execution status.

[0107] For example, if feedback shows that 90% of the power dispatch demand has been met, the module may generate optimization suggestions, such as adjusting power supply routes to shorten transmission time. The optimized dispatch suggestion data can be used in the next round of dispatching to improve overall efficiency and resource utilization.

[0108] In the network heat analysis module, real-time traffic and node load data are obtained from the data network, and the network heat distribution characteristics are calculated through the network heat analysis algorithm to obtain the network load status.

[0109] Real-time traffic and node load data are acquired, and timestamped traffic records and node operating parameters are extracted from the power data network system to obtain the raw dataset. The raw dataset is cleaned using a pre-defined standardization method to remove outliers and normalize the traffic and load data, resulting in a standardized dataset. The standardized dataset is then classified using a k-means clustering algorithm. Network heat features are extracted based on the distribution characteristics of traffic and load, yielding a heat distribution vector. If the variance of the heat distribution vector exceeds a pre-defined threshold, a locally high-load region is identified, and a set of high-load nodes is determined. Based on the high-load node set, the load contribution rate of each node is calculated, and a weighted average method is used to generate a network load status index, thus obtaining the network load status.

[0110] For example, in the field of power grid traffic scheduling, acquiring real-time traffic data and node load data is a core component in building an efficient scheduling system.

[0111] Understandably, data network systems typically include distributed traffic monitoring devices and node status sensors, which periodically record traffic throughput and node operating parameters such as CPU utilization and memory usage.

[0112] For example, in a data center network, suppose a traffic monitoring device records inbound and outbound traffic every 5 seconds, generating records containing timestamps, source IPs, destination IPs, and traffic volume; while node sensors collect node load data every 10 seconds, such as CPU utilization of 70% and memory usage of 60%. This data forms the raw dataset, providing the foundation for subsequent analysis.

[0113] In one embodiment, standardization is crucial for cleaning the original dataset. Outliers may originate from network attacks or device malfunctions, such as a node experiencing a sudden surge in traffic to 10Gbps, far exceeding the normal range of 2-3Gbps. The cleaning process removes these outliers and normalizes the traffic and load data to the 0-1 range, facilitating subsequent algorithm processing.

[0114] For example, traffic data can be linearly normalized, mapping 1Gbps to 0.1 and 10Gbps to 1.0; load data, such as CPU utilization at 60%, can be normalized to 0.6. This standardization ensures data consistency and helps improve the accuracy of algorithms.

[0115] For example, the k-means clustering algorithm is used to classify standardized datasets and extract network heatmap features. The clustering process categorizes nodes into high, medium, and low load groups based on traffic and load distribution. Assuming a network has 100 nodes, clustering might identify 20 nodes with significantly higher traffic and load than the others, indicating the existence of local hotspots. The heatmap distribution vector is generated by statistically analyzing the number of nodes and the mean load for each category; for example, the mean load for high load is 0.8, medium load is 0.5, and low load is 0.2. If the vector variance is 0.15, exceeding the preset threshold of 0.1, it indicates uneven network load distribution and the presence of a set of high-load nodes. This analysis helps to quickly locate network bottlenecks.

[0116] In one embodiment, after identifying the set of high-load nodes, calculating the load contribution rate is a key step.

[0117] For example, for 20 nodes under high load, the proportion of traffic and CPU utilization for each node to the total is calculated. Assuming a certain node contributes 30%, it indicates that it is the main bottleneck. The weighted average method can combine traffic and load data to generate network load status indicators.

[0118] For example, assigning a weight of 0.6 to traffic and a weight of 0.4 to load, the weighted average yields an index of 0.75, reflecting the overall high-load state of the network. This index provides a quantitative basis for subsequent scheduling optimization.

[0119] Understandably, the above method ensures real-time monitoring and accurate analysis of network load through a complete process from data collection to status assessment. Each step is closely aligned with scheduling optimization goals: data cleaning improves data quality, cluster analysis accurately identifies problem areas, and load contribution rate and status indicators provide reliable data for scheduling decisions. These steps support each other, collectively improving the efficiency and response speed of network resource allocation.

[0120] In the resource allocation optimization module, the network resource allocation scheme is dynamically adjusted according to the network load status using a resource allocation optimization algorithm to generate an optimized resource configuration strategy.

[0121] Real-time monitoring data is acquired, network load status is analyzed, and the current traffic distribution pattern and network performance indicators are determined. By analyzing the real-time monitoring data, a load balancing mechanism is employed to determine whether the traffic distribution pattern meets preset network performance indicator thresholds, thus obtaining the load balancing requirements. If the load balancing requirements do not meet the preset thresholds, a dynamic allocation algorithm is used to calculate an optimized resource allocation scheme for the traffic distribution pattern, determining the initial resource configuration. Based on the initial resource configuration and resource utilization, a linear programming algorithm is used to adjust the network resource configuration, resulting in an optimized resource allocation scheme. Resource scheduling efficiency is extracted from the optimized resource allocation scheme, and it is determined whether it reaches a preset resource scheduling efficiency threshold, obtaining a scheduling efficiency evaluation result. If the scheduling efficiency evaluation result does not reach the preset threshold, resources are reallocated through a dynamic adjustment mechanism to generate the final optimized allocation strategy.

[0122] For example, in network load status analysis, the process of acquiring real-time monitoring data involves extracting timestamped traffic records and node performance parameters from network device logs. Assuming a data center network contains 100 nodes, generating traffic and CPU utilization data per second, real-time data acquisition tools, such as the SNMP protocol, are used to obtain the inflow and outflow traffic and resource utilization of each node. The collected data includes node A with 500 Mbps traffic and 70% CPU utilization, and node B with 300 Mbps traffic and 50% CPU utilization. This data constitutes the raw monitoring dataset for subsequent analysis. When analyzing traffic distribution patterns, statistical analysis methods can be used to determine areas of concentrated network traffic.

[0123] For example, time series analysis is used to observe traffic fluctuations over 24 hours. Suppose that during the evening peak hours, the traffic at node A surges to 800 Mbps, while the traffic at node B remains stable. This distribution pattern suggests that node A may be a bottleneck.

[0124] It should be noted that the traffic distribution pattern is determined based on preset performance thresholds, such as peak traffic not exceeding 600Mbps or node CPU utilization not exceeding 80%. If node A's traffic exceeds the threshold, load balancing is triggered. For the load balancing mechanism, dynamic traffic scheduling can be implemented using software-defined networking technology.

[0125] For example, if node A is detected to be overloaded, the load balancer will offload some traffic to node B. Assuming node B has 400Mbps of remaining bandwidth, the traffic allocation algorithm will migrate 200Mbps of traffic from node A to node B, ensuring that the traffic at node A drops below 600Mbps. This dynamic allocation improves network stability. In resource allocation optimization schemes, linear programming algorithms can be used to adjust resource configuration.

[0126] For example, the resource allocation ratio is calculated based on the bandwidth utilization and CPU utilization of nodes A and B. Assuming node A has a bandwidth utilization of 80% and node B has 50%, linear programming is used to preferably migrate some computational tasks from node A to node B, balancing the utilization of both nodes to approximately 65%. When extracting scheduling efficiency from the optimized resource allocation scheme, task completion time and resource occupancy can be statistically analyzed.

[0127] For example, after the adjustment, the task completion time of node A is shortened from 5 seconds to 3 seconds, indicating an improvement in scheduling efficiency. If the scheduling efficiency does not meet the standard, resources can be reallocated through a dynamic adjustment mechanism.

[0128] For example, if the response time of node B is still detected to be higher than the threshold of 2 seconds, then the low-priority task is further migrated to node C, assuming that the utilization rate of node C is 40%. The final optimized allocation strategy ensures that the response time of all nodes is lower than 2 seconds. This dynamic adjustment mechanism significantly improves the overall performance and resource utilization efficiency of the network.

[0129] In the network bottleneck identification module, if the network load status displays a bottleneck node, the high-load area is located through the network bottleneck identification algorithm to obtain a list of bottleneck nodes.

[0130] The resource allocation strategy configures the network. After configuration, a load detection mechanism analyzes the network load status to obtain a bottleneck node set. If the bottleneck node set is not empty, high-load areas are identified through network topology analysis. Based on the high-load areas, the bottleneck nodes are grouped using a k-means clustering algorithm, resulting in node group sets. The load distribution characteristics of each group are calculated to obtain a load distribution vector. If the variance of the load distribution vector exceeds a preset threshold, the shortest path between nodes is calculated using Dijkstra's algorithm to determine the load redistribution path. Based on the load redistribution path, network traffic allocation is adjusted to obtain a new load balancing state. The bottleneck node set is updated by monitoring the new load balancing state.

[0131] For example, in network load monitoring scenarios, load monitoring mechanisms identify bottleneck nodes by collecting performance data from network devices in real time. Assuming an enterprise network contains multiple routers and switches, the load monitoring mechanism monitors the CPU utilization, memory usage, and bandwidth consumption of each device.

[0132] For example, if a router's bandwidth utilization reaches 90%, far exceeding the normal threshold of 50%, then that router is marked as a bottleneck node. The formation of a bottleneck node set relies on the dynamic analysis of this data to ensure the timely detection of high-load points in the network.

[0133] In one possible implementation, when identifying high-load areas through network topology analysis, the network structure can be mapped using a topology graph.

[0134] For example, an enterprise network comprises a core layer, an aggregation layer, and an access layer. Topology analysis reveals that bottleneck nodes are primarily concentrated in a single switch cluster at the aggregation layer. By analyzing the inflow and outflow patterns in this area, high-load regions are identified as the data center entry points connecting multiple departments. This analysis is based on statistical data of traffic paths.

[0135] For example, the inbound traffic of the aggregation layer switch is 800Mbps and the outbound traffic is 600Mbps, which is outside the normal range.

[0136] Specifically, when using the k-means clustering algorithm to group bottleneck nodes, the nodes are classified according to their load characteristics (such as bandwidth usage and latency).

[0137] For example, the 10 bottleneck nodes are divided into three groups: a high-latency group, a high-bandwidth-occupancy group, and a mixed-problem group. The load distribution characteristics of each group are generated by calculating the peak and average traffic of the nodes to create a load distribution vector. Assuming the vector for the high-bandwidth-occupancy group shows a peak traffic of 950 Mbps, a mean of 700 Mbps, and a variance of 200, exceeding the preset threshold of 150, it indicates that further optimization is needed.

[0138] For example, when calculating load redistribution paths using Dijkstra's algorithm, the shortest path can be found based on the network topology. Suppose there are multiple paths from switch A in a high-load area to switch B in a low-load area; Dijkstra's algorithm selects the path with the lowest latency and sufficient bandwidth.

[0139] For example, path ACB has a total latency of 10ms and a bandwidth margin of 500Mbps. Based on this path, some traffic is offloaded from switch A to switch B, reducing the load on switch A to 60%.

[0140] In one possible implementation, after adjusting network traffic distribution, the new load balancing status is verified through real-time monitoring.

[0141] For example, after traffic offloading, switch A's bandwidth usage dropped to 55%, and latency decreased from 20ms to 8ms. The updated bottleneck node set showed that the original 10 bottleneck nodes were reduced to 3. This dynamic adjustment mechanism, through continuous monitoring and optimization, ensures more reasonable allocation of network resources and improves the stability of overall network performance.

[0142] It should be noted that the implementation of each of the above steps revolves around the core objective of network load optimization, and is closely integrated with real-time data analysis and dynamic adjustments.

[0143] For example, variance analysis of load distribution vectors helps identify unbalanced traffic patterns, while the application of Dijkstra's algorithm provides efficient traffic offloading paths. These methods support each other, forming a complete network resource optimization scheme. The implementation of each technical topic is based on a real network environment, ensuring the relevance and practicality of the solution.

[0144] In the resource allocation strategy generation module, based on the bottleneck node list, a resource allocation optimization algorithm is used to reallocate network bandwidth and computing resources to generate the final optimized network configuration.

[0145] Node analysis data is obtained from the bottleneck node list to identify performance bottlenecks. Using this data, a linear programming algorithm is employed to calculate resource allocation ratios, resulting in bandwidth adjustment and resource scheduling schemes. If the allocation ratio in the bandwidth adjustment scheme exceeds a preset threshold, network bandwidth is dynamically partitioned to generate a bandwidth allocation configuration. Based on the bandwidth allocation configuration, the scheduling priority of computing resources is adjusted to obtain a resource scheduling configuration. Using this configuration, a simulated annealing algorithm is used to optimize the network topology, generating an optimized network configuration. Performance improvement metrics are extracted from the optimized configuration, and it is determined whether these metrics meet preset thresholds. If not, the process returns to step two to recalculate the resource allocation ratios. Finally, the final system configuration is generated based on these performance improvement metrics.

[0146] In one possible implementation, when obtaining node analysis data from the bottleneck node list, network monitoring tools can be used to collect node performance metrics, such as CPU utilization, memory usage, and bandwidth consumption.

[0147] For example, suppose there are 10 nodes in a network. Monitoring reveals that node A's bandwidth utilization reaches 95%, far exceeding the average of 60%, indicating that it is a performance bottleneck. The collected data includes the node's latency, packet loss rate, and throughput, forming a multidimensional performance dataset. The advantage of this approach is that it can quickly locate resource bottlenecks, providing a precise basis for subsequent optimization.

[0148] Specifically, when using linear programming algorithms to calculate resource allocation ratios, objective functions can be set based on node analysis data, such as minimizing network latency or maximizing throughput.

[0149] For example, given the high load on node A, linear programming can be used to adjust its bandwidth allocation from the current 30% to 50%, while simultaneously reducing the allocation to other low-load nodes. This method ensures reasonable resource allocation through mathematical optimization, avoiding resource waste or insufficiency.

[0150] In one possible implementation, dynamic segmentation is triggered if the allocation ratio in the bandwidth adjustment scheme exceeds a preset threshold, such as 70%.

[0151] For example, if node A has an 80% bandwidth allocation, its bandwidth can be divided into high-priority and low-priority parts. High-priority bandwidth is used for critical business traffic, and low-priority bandwidth is used for non-critical traffic. This partitioning method improves network efficiency through fine-grained management.

[0152] For example, when adjusting the scheduling priority of computing resources, different scheduling weights can be assigned to nodes based on bandwidth allocation. Suppose node A's bandwidth allocation increases, its scheduling priority can be set to the highest, ensuring that critical tasks are processed first. This approach improves resource utilization efficiency through priority adjustment.

[0153] Specifically, when using simulated annealing to optimize network topology, the global optimal solution can be found by randomly adjusting the node connection relationships.

[0154] For example, if node A in the initial topology is connected to the core network through node B, we can try connecting it directly to the core node to reduce one-hop latency. Simulated annealing finds the topology configuration with the lowest latency through multiple iterations. The advantage of this method is that it avoids getting trapped in local optima and improves the overall network performance.

[0155] In one possible implementation, when extracting performance improvement metrics from network optimization configuration, attention can be paid to changes in network latency and throughput.

[0156] For example, after optimization, node A's latency decreased from 100ms to 50ms, and its throughput increased from 500Mbps to 800Mbps. If the metrics do not reach the preset thresholds, such as latency needing to be below 40ms, the resource allocation ratio is recalculated. This iterative approach ensures that the optimization effect gradually approaches the target.

[0157] For example, when generating the final system configuration, the results of bandwidth allocation, resource scheduling, and topology optimization can be integrated to form a complete configuration file.

[0158] For example, node A is allocated 50% of its bandwidth, has a scheduling priority of 1, and its connection path is adjusted to directly connect to the core node. This configuration improves network stability and efficiency through multi-dimensional optimization.

[0159] In one possible implementation, the combination of the above methods can be applied to data center network optimization.

[0160] For example, high-load nodes within a data center can be gradually optimized through the steps described above, ultimately achieving low latency and high throughput. This approach ensures a comprehensive improvement in network performance through multi-layered collaborative optimization.

[0161] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An intelligent monitoring system based on a power dispatch data network, characterized in that, include: The system comprises a data preprocessing module, a multi-dimensional information integration module, a security assessment module, an analysis module, a judgment module, and a network analysis module, connected sequentially. The data preprocessing module acquires power dispatching data from the power dispatching data network and preprocesses it. The multi-dimensional information integration module integrates the preprocessed power dispatching data to obtain comprehensive operational status characteristics. The security assessment module performs a security assessment on the comprehensive operational status characteristics to obtain a security assessment result. The analysis module performs power dispatching calculations based on the security assessment result to obtain an optimized dispatching scheme. The judgment module assesses the operability of the optimized dispatching scheme and obtains the power dispatching execution command based on the operability assessment result. The network analysis module performs real-time load monitoring and network node optimization on the power dispatching data network to obtain the final optimized network configuration.

2. The system according to claim 1, characterized in that, In the data preprocessing module, the preprocessing of the power dispatch data includes missing value filling, outlier removal, and format and unit standardization.

3. The system according to claim 1, characterized in that, In the multi-dimensional information integration module, the process of integrating the preprocessed power dispatch data includes: performing dimensionality reduction processing on the preprocessed power dispatch data to obtain a low-dimensional feature set, wherein the power dispatch data includes real-time load data, power generation plan data, and historical dispatch data; performing dimensionality judgment on the low-dimensional feature set; clustering the low-dimensional feature set based on the judgment result to obtain preliminary operating status categories; optimizing the boundaries of the preliminary operating status types according to the preliminary operating status categories using machine learning algorithms to obtain precise operating status classifications; extracting key features from the precise operating status classifications and performing weighted processing to obtain comprehensive operating status features.

4. The system according to claim 1, characterized in that, In the security assessment module, the process of conducting a security assessment on the comprehensive operational status characteristics includes: The comprehensive operating status characteristics are assessed for safety thresholds to obtain compliance status characteristics. These compliance status characteristics are then classified using a machine learning model to obtain classified risk feature vectors. Power grid operation risks are calculated based on these classified risk feature vectors to obtain risk index values. These risk index values ​​are analyzed using a logistic regression algorithm to obtain a risk probability distribution. Based on this risk probability distribution, a decision tree algorithm is used to generate a safety assessment result, resulting in a power grid operation safety level. By comparing the safety level with a preset safety threshold, the stability of the power grid operation status is determined, and a safety assessment output is obtained.

5. The system according to claim 1, characterized in that, In the analysis module, the process of performing power dispatch calculations based on the security assessment results includes: A comprehensive dataset is constructed based on cost data and security assessment results of power dispatching. The security assessment results in the comprehensive dataset are evaluated. Based on the evaluation results, an initial dispatching scheme is calculated using a linear programming algorithm to obtain a preliminary scheme set. Based on the preliminary scheme set, combined with cost data and corresponding constraints, a greedy algorithm is used to iteratively optimize the preliminary schemes to generate an optimized dispatching scheme set. The dispatching schemes in the optimized dispatching scheme set are evaluated for economic efficiency. Based on the economic efficiency scores, the dispatching schemes are evaluated to obtain candidate schemes. The candidate schemes are then validated to obtain the optimized dispatching scheme. The dispatching scheme includes power dispatching task priority and resource allocation information.

6. The system according to claim 1, characterized in that, The network analysis module includes a network heat analysis module, a resource allocation optimization module, a network bottleneck identification module, and a resource configuration strategy generation module connected in sequence. Specifically, the network heat analysis module acquires node data from the power dispatch data network, performs heat analysis on the node data to obtain the network load status, the resource allocation optimization module allocates network resources to the nodes according to the network load status to obtain a resource configuration strategy, the network bottleneck identification module identifies high-load areas according to the resource configuration strategy to obtain a bottleneck node list, and the resource configuration strategy generation module performs optimized resource allocation according to the bottleneck node list to obtain the final network optimization configuration scheme for the power dispatch data network.

7. The system according to claim 1, characterized in that, In the network popularity analysis module, the process of performing popularity analysis on node data includes: The node data is preprocessed, including real-time traffic data and node load data. The preprocessed node data is classified and network hotspot features are analyzed to obtain a heat distribution vector. The heat distribution vector is judged to obtain a set of high-load nodes. The contribution rate of the set of high-load nodes is calculated, and the weighted processing is performed based on the contribution rate calculation results to obtain the network load status.

8. The system according to claim 1, characterized in that, In the resource allocation optimization module, the process of allocating network resources to nodes includes: The network load status is analyzed to obtain traffic distribution patterns and network performance indicators. The traffic distribution patterns are judged, and based on the judgment results, load balancing requirements are obtained. Threshold judgments are made on the load balancing requirements. Based on the threshold judgment results, a resource allocation optimization scheme is calculated using a dynamic allocation algorithm according to the traffic distribution patterns. Network resources are adjusted according to the resource allocation optimization scheme to obtain an optimized resource allocation scheme. Resource scheduling efficiency is extracted based on the optimized resource allocation scheme, and the resource scheduling efficiency is evaluated and judged. Based on the judgment results, the resource allocation scheme is adjusted to obtain a resource configuration strategy.

9. The system according to claim 1, characterized in that, In the network bottleneck identification module, the process of identifying high-load areas based on the resource configuration strategy includes: Resource allocation strategies are used to configure resources in the power dispatch data network. After configuration, the load status of the power dispatch data network is analyzed through a load detection mechanism. Based on the load status, a bottleneck node set is obtained. Based on the bottleneck node set, high-load areas are identified through network topology analysis. Based on the high-load areas, the bottleneck nodes are grouped using the k-means clustering algorithm to obtain a node group set. Based on the node group set, the load distribution characteristics of each group are calculated to obtain the load distribution vector. The variance of the load distribution vector is judged. Based on the judgment result, the shortest path between nodes is calculated using the Dijkstra algorithm to obtain the load redistribution path. Based on the load redistribution path, the network traffic distribution is adjusted to obtain a new load balance state. By monitoring the new load balance state, the bottleneck node set is updated and obtained.

10. The system according to claim 1, characterized in that, In the resource allocation strategy generation module, the process of optimizing resource allocation based on bottleneck nodes includes: Obtain node analysis data from the bottleneck node list. Based on the node analysis data, use a linear programming algorithm to calculate the resource allocation ratio for the bottleneck nodes, resulting in a bandwidth adjustment scheme and a resource scheduling scheme. If the allocation ratio in the bandwidth adjustment scheme exceeds a preset threshold, dynamically divide the network bandwidth to generate a bandwidth allocation configuration. Based on the bandwidth allocation configuration, adjust the scheduling priority of the resource scheduling scheme for the bottleneck nodes to obtain a resource scheduling configuration. Through the resource scheduling configuration, use a simulated annealing algorithm to optimize the network topology and generate an optimized network configuration. Extract performance improvement indicators from the optimized network configuration and determine whether the performance improvement indicators meet the preset threshold. If not, return to the second step to recalculate the resource allocation ratio. Based on the performance improvement indicators, generate the final system configuration.

Images