Power grid load scheduling method based on power big data

Abstract

The application relates to the technical field of power big data intelligent scheduling, and discloses a power grid load scheduling method based on power big data. The method comprises the following steps: constructing a high-density data stream by deploying distributed sensing nodes in the whole domain of a power grid, and generating a power grid load situation map after fusion processing. According to the situation map, a time-space aggregation area and a power hollow area of the load are identified, and the dynamic correlation strength thereof is calculated. Then, a plurality of virtual load migration paths connecting the two areas are deduced in a digital space by taking the strength as a link, and the feasibility and cost of each path are evaluated. A load dynamic migration strategy is formulated based on the evaluation results, and an instruction is generated to drive the load to migrate along the selected path in the physical power grid. The method realizes accurate description and active guidance of the time-space distribution of the load, and makes the power grid scheduling change from passive response to local overload to active optimization of global load distribution, thereby improving the flexibility and economy of operation.

Description

Technical Field

[0001] This invention relates to the field of intelligent dispatching technology for power big data, specifically a power grid load dispatching method based on power big data. Background Technology

[0002] Current power grid load dispatch primarily relies on classical optimization algorithms within energy management systems. These methods typically use relatively fixed power grid topology models and predicted load curves, employing economic efficiency and safety as objective functions to solve for generation plans or load shedding schemes. Another common approach is automatic generation control based on regional control deviations, or reliance on operator experience to activate pre-set load control plans when line overloads or voltage exceedances are detected. These technologies form the core of existing dispatch automation systems.

[0003] The aforementioned existing technical solutions have shortcomings. Traditional optimization models heavily rely on accurate predictions, are slow to respond to rapid spatiotemporal changes in load and renewable energy output, and struggle to handle massive amounts of distributed node data. Their scheduling commands are essentially power increase / decrease commands oriented towards "nodes" or "regions," failing to treat the power grid as a "spatiotemporal field" of dynamic load distribution and migration. This results in a lack of foresight and flexibility in scheduling strategies, often resorting to crude load shedding to address local emergencies, and failing to proactively guide load distribution on a global spatiotemporal scale to prevent congestion and fully utilize distributed resources.

[0004] The challenge lies in accurately depicting the aggregation and dispersion of loads across a wide spatial and temporal area, and proactively planning load flow paths accordingly. This requires elevating dispatching decisions from "addressing instantaneous power imbalances" to "optimizing load spatial and temporal distribution." The core challenge is identifying load "sources" and "sinks" with migration potential from massive amounts of data, and quantitatively assessing the various feasible "paths" connecting them and their costs under physical grid constraints, thereby transforming passive control into proactive guidance. Summary of the Invention

[0005] The purpose of this invention is to provide a power grid load scheduling method based on power big data to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a power grid load dispatching method based on power big data, the method comprising:

[0007] Deploy distributed sensing nodes across the entire power grid to construct a multi-dimensional, high-density power operation data stream;

[0008] Establish a fusion channel for multi-source heterogeneous power operation data, perform feature extraction and state analysis on the power operation data stream, and generate a power grid load status map with spatiotemporal correlation;

[0009] Based on the power grid load situation diagram, identify the spatiotemporal clustering areas and power void areas of the load in the power grid, and calculate the dynamic correlation strength between the spatiotemporal clustering areas and the power void areas;

[0010] A load migration simulation model is constructed. The load migration simulation model uses the dynamic correlation strength as a link to deduce multiple virtual load migration paths between the spatiotemporal clustering region and the power void region, and performs feasibility assessment and cost calculation for each virtual load migration path.

[0011] Based on the output of the load migration simulation model, a dynamic load migration strategy is formulated, which includes time-scheduled resource binding and execution timing orchestration for the selected virtual load migration path.

[0012] The load dynamic migration strategy is executed, and a sequence of coordinated scheduling instructions is sent to the corresponding control units in the power grid to drive the load to migrate along the path specified by the strategy in the physical power grid.

[0013] Preferably, the deployment of distributed sensing nodes across the entire power grid to construct a multi-dimensional, high-density power operation data stream specifically includes:

[0014] Based on the power grid topology and regional functional characteristics, distributed sensing nodes with multi-parameter sensing capabilities are configured at key line nodes, load access points, and power conversion hubs.

[0015] The distributed sensing nodes synchronously collect voltage, current, phase angle, power, and equipment operating status information to form raw sampling data packets;

[0016] Each of the distributed sensing nodes adds spatial coordinate identifiers and timestamps to its own generated raw sampling data packets according to a unified time base to generate standard data frames;

[0017] The standard data frames are aggregated to the data hub center through a dedicated power communication network and a wireless sensor network to form a continuous, high-density power operation data stream.

[0018] Preferably, the establishment of a fusion channel for multi-source heterogeneous power operation data, and the feature extraction and state analysis of the power operation data stream specifically include:

[0019] Establish a fusion channel for multi-source heterogeneous power operation data in the data hub center, and perform data cleaning, format alignment and time synchronization correction on the power operation data stream;

[0020] Using the fusion channel of the multi-source heterogeneous power operation data, multi-dimensional feature extraction is performed on the corrected data. The extracted features include load change rate features, power fluctuation cycle features, regional power consumption pattern features, and equipment operation correlation features.

[0021] The extracted multidimensional features are input into the state parsing engine, which maps the multidimensional features to a predefined set of states to generate a semantic description of the current operating state for each monitored power grid unit.

[0022] The semantic descriptions are spatially spliced ​​according to the power grid topology and bound to the corresponding timestamps to generate the power grid load status map. The power grid load status map visually reflects the load distribution and operating status of the entire power grid at a specific time in the form of a graph.

[0023] Preferably, identifying the spatiotemporal clusters and power voids of load in the power grid based on the power grid load situation diagram specifically includes:

[0024] The power grid load situation map is divided into a spatiotemporal grid, dividing the entire power grid into a series of grid units with spatiotemporal boundaries;

[0025] Calculate the load density index and power margin for each grid cell. The load density index reflects the degree of load concentration per unit area or unit line capacity, and the power margin reflects the grid cell's ability to accommodate additional loads after meeting existing loads.

[0026] Grid cell clusters with load density index exceeding the high density threshold are identified as the spatiotemporal clustering region, and grid cell clusters with power margin exceeding the high margin threshold and load density index below the low density threshold are identified as the power void region.

[0027] On the power grid load situation map, the spatiotemporal clustering area and the power void area are highlighted, and the topological distance and electrical coupling degree between each spatiotemporal clustering area and each power void area are calculated. The dynamic correlation strength is jointly determined by the topological distance and the electrical coupling degree.

[0028] Preferably, the construction of the load migration simulation model specifically includes:

[0029] Starting from the spatiotemporal clustering region and ending from the power void region, a load migration simulation model is constructed on the virtual network formed by the power grid load situation diagram.

[0030] The load migration simulation model first generates all possible physical connection paths connecting a spatiotemporal cluster to a power void region based on the impedance parameters of the power grid lines, the transformer tap position status, and the protection setting constraints.

[0031] For each of the physical connection paths, the load migration simulation model simulates the power flow changes under different migration power amounts, calculates the load rate changes, network loss increments, and voltage offsets of each device in the path, and uses these as the core indicators for evaluating the feasibility of the virtual load migration path.

[0032] The load migration simulation model calculates the migration cost for each physical connection path and its corresponding simulated power quantity. The migration cost is obtained by weighted summation of network loss cost, equipment lifespan depreciation cost, and scheduling operation complexity cost.

[0033] The load migration simulation model outputs a simulation report containing multiple virtual load migration paths, their feasibility assessment results, and migration costs.

[0034] Preferably, the step of formulating a dynamic load migration strategy based on the output of the load migration simulation model specifically includes:

[0035] Receive the simulation report output by the load migration simulation model, and filter out all virtual load migration paths whose feasibility assessment results are feasible;

[0036] Based on the current global optimization objectives of the power grid, one or more target migration paths are selected from the screened virtual load migration paths. The global optimization objectives include minimizing total network loss, achieving the most balanced load on key sections, or minimizing overall migration cost.

[0037] For each selected target migration path, determine its specific migrateable load, migration start time, migration duration, and migration rate curve, and complete the time-schedule resource binding. The time-schedule resources include the transmission capacity reservation of relevant lines, the relevant transformer tap adjustment plan, and the setting coordination of relevant protection devices.

[0038] The execution times of multiple target migration paths are conflict detected and coordinated, the execution sequence is arranged, and a load dynamic migration strategy containing a series of ordered scheduling instructions is generated.

[0039] Preferably, the step of executing the load dynamic migration strategy and issuing a sequence of coordinated dispatch instructions to the corresponding control units in the power grid specifically includes:

[0040] The load dynamic migration strategy is parsed into a series of atomic-level power grid operation commands, including circuit breaker opening and closing commands, transformer tap adjustment commands, capacitor bank switching commands, and interruptible load control commands.

[0041] According to the execution timing, the atomic-level power grid operation instructions are packaged into several coordinated scheduling instruction sequences with specific execution time windows;

[0042] The coordinated scheduling instruction sequence is sent to the control unit of the corresponding area through the scheduling automation system;

[0043] During the issuance of the coordinated scheduling instruction sequence, the status feedback of the control unit receiving the instructions is monitored in real time to ensure the smooth flow and executability of the instruction channel.

[0044] Preferably, after the driving load migrates along the path specified by the strategy in the physical power grid, the process further includes:

[0045] During the load migration process, the power grid response data is collected in real time to form the feedback part of the power operation data stream;

[0046] The power operation data stream of the feedback section is input into the fusion channel of the multi-source heterogeneous power operation data to update the power grid load status map and generate a real-time power grid load status map.

[0047] The real-time power grid load situation map is compared with the situation map predicted before the load dynamic migration strategy is implemented, and the deviation between the actual migration trajectory and the path specified by the strategy is calculated.

[0048] If the deviation exceeds the allowable range, the strategy correction process is initiated, the dynamic correlation strength is recalculated based on the new real-time power grid load situation map, and a local or global load migration strategy replanning is triggered.

[0049] Preferably, the startup strategy correction process specifically includes:

[0050] Based on the deviation, the causes of the deviation are analyzed, including load forecast deviation, abnormal equipment response, or new disturbance events in the power grid.

[0051] Pause the unexecuted portions of the currently running load migration strategy;

[0052] Using the current real-time power grid load situation map as the initial state, the load migration simulation model is invoked, and within the remaining schedulable time window, the simulation calculation is re-performed to generate a new simulation report.

[0053] Based on the new simulation report, a revised load migration strategy was quickly formulated and immediately implemented to ensure that the load migration process converges toward the optimization objective under controlled conditions.

[0054] Preferably, the method further includes closed-loop learning of the entire scheduling process:

[0055] Completely record the entire process of each load dynamic migration strategy from formulation, execution to correction, including the initial power grid load status diagram, simulation report, load dynamic migration strategy, coordinated dispatch instruction sequence, real-time power grid load status diagram, deviation and corrected strategy;

[0056] Using big data processing technology, the data of the entire process is mined and analyzed to extract successful scheduling patterns and scheduling risk characteristics;

[0057] The successful scheduling mode and scheduling risk characteristics are fed back to the parameter library and rule library of the load migration simulation model to optimize the accuracy of subsequent simulation calculations and the calculation precision of the dynamic correlation strength, so that the entire scheduling method has adaptive evolution capability.

[0058] Compared with the prior art, the beneficial effects of the present invention are:

[0059] This technology achieves a deep analysis of the power grid state by identifying "spatiotemporal clusters" and "power voids" from the power grid load status map and calculating their dynamic correlation strength. It transforms dispersed load data into entities with spatiotemporal location and potential energy attributes, accurately depicting the dynamic evolution trend of load imbalance. The dispatching system can thus proactively locate the load "sources" that need to be alleviated and the power "lowlands" that can accommodate the load, providing clear spatiotemporal targets and quantitative correlations for intervention measures, fundamentally improving the pertinence and predictability of decision-making.

[0060] This technology upgrades traditional scheduling commands into path planning and comparison in a digital space by dynamically extrapolating multiple virtual load migration paths based on correlation strength and conducting quantitative evaluation. Each potential load movement scheme undergoes feasibility verification and cost calculation, transforming strategy formulation from choosing a single scheme to selecting the best among multiple pre-verified paths. The decision-making process shifts from relying on experience-based judgment to selecting based on simulation data, enhancing the scientific rigor and adaptability of the strategy and reducing uncertainty and operational risks in actual execution. Attached Figure Description

[0061] Figure 1 This is a schematic diagram illustrating the working principle of the power grid load dispatching method based on power big data as described in this invention.

[0062] Figure 2 A flowchart illustrating the deployment of distributed sensing nodes and the construction of power operation data streams;

[0063] Figure 3 A flowchart for constructing and evaluating a load migration simulation model;

[0064] Figure 4 A dynamic comparison curve of multipath load migration rates;

[0065] Figure 5 This is a dynamic correlation intensity heat map of the spatiotemporal load concentration area and the power void area of ​​the power grid. Detailed Implementation

[0066] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] Please see Figure 1 This invention provides a power grid load scheduling method based on power big data. The method includes: optimizing power grid load scheduling by integrating sensing, analysis, decision-making, and execution processes. Distributed sensing nodes are deployed throughout the power grid to construct a multi-dimensional, high-density power operation data stream. A fusion channel for multi-source heterogeneous power operation data is established, and feature extraction and state analysis are performed on the power operation data stream to generate a power grid load status map with spatiotemporal correlation. Based on the power grid load status map, spatiotemporal clusters and power voids in the power grid are identified, and the dynamic correlation strength between these clusters and voids is calculated. A load migration simulation model is constructed, which uses the aforementioned dynamic correlation strength as a link to deduce multiple virtual load migration paths between the spatiotemporal clusters and power voids, and performs feasibility assessment and cost calculation for each virtual load migration path. Based on the output of the load migration simulation model, a dynamic load migration strategy is formulated, which includes time-space scheduling resource binding and execution timing arrangement for the selected virtual load migration paths. The system executes the established load dynamic migration strategy, issues a sequence of coordinated dispatch instructions to the corresponding control units in the power grid, and drives the load to migrate along the path specified by the strategy in the physical power grid.

[0068] Example 1: See Figure 2In specific implementation, taking the dispatching system of a regional power grid as an example, the power grid topology includes multi-voltage level transmission and distribution networks and various power load access points. Based on the power grid topology and regional functional characteristics, distributed sensing nodes with multi-parameter sensing capabilities are configured at key line nodes such as transmission corridor intersections, load access points such as large industrial parks, and power conversion hubs such as substation main transformers. The distributed sensing nodes synchronously collect voltage amplitude, current phase, phase angle difference, active power, reactive power, and equipment operating status information such as circuit breaker position signals to form raw sampling data packets. Each distributed sensing node adds spatial coordinate identifiers such as latitude and longitude encoding and timestamps such as millisecond accuracy to its own generated raw sampling data packets according to a unified time reference such as the time synchronization protocol based on the Global Positioning System, generating standard data frames. Through the backbone transmission network of the power dedicated communication network and the low-power wide area network of the wireless sensor network, the standard data frames are aggregated to the data hub center to form a continuous, high-density power operation data stream.

[0069] In some embodiments, a fusion channel for multi-source heterogeneous power operation data is established in a data hub center. This fusion channel cleanses the power operation data stream to remove outliers, aligns formats to standardize data encoding, and performs time synchronization correction to compensate for transmission delays. The fusion channel then extracts multi-dimensional features from the corrected data, including load change rate features, power fluctuation cycle features, regional electricity consumption pattern features, and equipment operation correlation features. The load change rate feature is obtained through power difference calculation between adjacent time points. This calculation is based on power data synchronously collected by distributed sensing nodes, and is achieved by performing differential calculations on power measurements at adjacent sampling time points. Specifically, the fusion channel processes the corrected power time series, sequentially calculating the power difference between each sampling point and its preceding sampling point, thus directly reflecting the instantaneous load change trend. This differential calculation relies on a high-density data stream under a unified time base, ensuring that the change rate feature accurately captures dynamic load fluctuations and provides crucial input for generating a power grid load situation map. Power fluctuation cycle characteristics are obtained by identifying the dominant oscillation mode through spectrum analysis. Regional electricity consumption pattern characteristics are obtained by clustering historical data into typical daily curves. Equipment operation correlation characteristics are obtained by quantifying the degree of equipment status linkage through statistical methods. The extracted multidimensional features are input into the status analysis engine, which maps the multidimensional features to a predefined status set, including overload, light load, normal, and alarm states. A semantic description of the current operating status is generated for each monitored power grid unit. The semantic description is spatially stitched according to the power grid topology and bound to the corresponding timestamp to finally generate a power grid load status map. The power grid load status map intuitively reflects the load distribution and operating status of the entire power grid at a specific time in a graphical form.

[0070] Optionally, during the multidimensional feature extraction process, the calculation of device operation-related features uses the following formula:

[0071]

[0072] in: This represents the operational correlation characteristic value between device x and device y. This represents the state measurement value of device x at the k-th sampling point. This represents the average value of the device x state measurement over N sampling points. This represents the state measurement value of device y at the k-th sampling point. This represents the average value of the device's y-state measurement across N sampling points. This indicates the total number of sampling points used in feature extraction.

[0073] It is understandable that the deployment location of distributed sensing nodes is dynamically adjusted according to the regional functional characteristics. The sampling frequency of distributed sensing nodes at power conversion hubs is higher than that of ordinary load access points. The original sampling data packets of distributed sensing nodes are stored using a circular caching mechanism to ensure data integrity. The fusion channel of multi-source heterogeneous power operation data adopts a parallel computing architecture to achieve high throughput processing. The state parsing engine realizes the mapping from features to states based on the rule engine and fuzzy logic. The power grid load status map is superimposed on the power grid topology map in the form of a heat map to achieve dynamic display.

[0074] Example 2: In a specific implementation, taking the specific processing flow of a provincial power grid dispatch center as an example, the power grid load status map is stored in the computing memory in the form of a graph structure containing topological connection relationships, real-time load values ​​and node power injection data. The power grid load status map is divided into a series of grid units with spatiotemporal boundaries by spatiotemporal grid segmentation. The grid units of the spatiotemporal boundaries are divided in geographical space according to the physical corridor of the line, and divided in the time dimension according to the preset fixed time slice length. Each grid unit of the spatiotemporal boundary has a unique spatial code and time index.

[0075] In some embodiments, the load density index and power margin of each spatiotemporal boundary grid cell are calculated. The load density index is calculated as the ratio of the sum of the loads carried by all lines within the spatiotemporal boundary grid cell to the rated total capacity of the lines within the cell. The power margin is calculated as the sum of the maximum output of all power nodes within the spatiotemporal boundary grid cell, the sum of the power injection capabilities of adjacent cells, minus the real-time total load value within the cell. The load density index reflects the degree of load concentration per unit area or unit line capacity, while the power margin reflects the ability of the spatiotemporal boundary grid cell to accept additional loads after meeting the existing load. Grid cell clusters at spatiotemporal boundaries with load density indices exceeding a high-density threshold are identified as spatiotemporal clusters. The high-density threshold is statistically derived from historical power grid operation data. Grid cell clusters at spatiotemporal boundaries with power margins exceeding a high-density threshold and load density indices below a low-density threshold are identified as power voids. The low-density threshold is also determined based on historical data. On the power grid load situation map, spatiotemporal clusters and power voids are highlighted, and the topological distance and electrical coupling degree between each spatiotemporal cluster and each power void are calculated. The topological distance is calculated based on the minimum number of nodes traversed in the path connecting the two regions on the power grid topology map, and the electrical coupling degree is calculated based on the impedance parameters of the transmission channel between the two regions, the correlation of historical power flow data, and the loop relationship of the network structure. The dynamic correlation strength is jointly determined by the topological distance and the electrical coupling degree.

[0076] Optional, load density index The calculation uses the following formula:

[0077]

[0078] in: Indicates the load density index. A grid cell representing a spacetime boundary. Indicates belonging to a grid cell One of the routes, Indicates the line The real-time active load carried at the calculation moment. Indicates belonging to a grid cell Line or transformer capacity unit, Represents capacity unit The maximum apparent power capacity rating. It can be understood that the summation operation in the numerator of the formula applies to the mesh element. The real-time load of all lines within the grid, and the summation operation of the denominator for each grid cell. The capacity of all lines or transformers within the system.

[0079] It is understandable that the granularity of the grid cells at the spatiotemporal boundary can be adjusted according to the requirements of scheduling precision. The calculation cycle of the load density index and power margin is synchronized with the grid data refresh cycle. The setting of high-density threshold and low-density threshold can be dynamically fine-tuned according to the system safety margin strategy. The calculation of topology distance can further introduce the line length weight factor. The calculation of electrical coupling degree can be quantified by the numerical relationship of the corresponding elements in the admittance matrix. On the grid load status map, the spatiotemporal cluster area and the power void area are highlighted and distinguished by red blocks and blue blocks. The calculation of dynamic correlation strength adopts the method of weighted summation of the normalized topology distance and electrical coupling degree.

[0080] Example 3: See Figure 3 In specific implementation, taking a regional power grid containing multiple 220 kV substations and 110 kV distribution networks as an example, the virtual network composed of the power grid load situation map includes topology nodes, lines, and marked spatiotemporal clusters and power voids. The load migration simulation model is constructed on the virtual network composed of the power grid load situation map, starting from the spatiotemporal clusters and ending at the power voids. The load migration simulation model first generates all possible physical connection paths connecting a spatiotemporal cluster to a power void based on the impedance parameters of the power grid lines, transformer tap status, and protection setting constraints. The search for physical connection paths is based on the power grid topology, starting from the power nodes or hub nodes in the spatiotemporal clusters, traversing the network until reaching the load nodes or access nodes in the power void. During the search process, paths that do not meet the line transmission limits, transformer overload capacity, and protection device action setting constraints are filtered out, generating a set of all feasible physical connection paths.

[0081] In some embodiments, the load migration simulation model for each physical connection path simulates power flow changes under different migration power levels. The simulation process is performed by calling the built-in power flow calculation module to calculate the load rate change, network loss increment, and voltage offset of each device in the path. These are used as core indicators to evaluate the feasibility of the virtual load migration path. The load rate change of a device is the change in the ratio of the actual power of the device to its rated capacity under the simulated migration power level. The network loss increment is the difference between the total active power loss on the entire physical connection path under the simulated migration power level and the total active power loss in the initial state. The voltage offset is the change in the voltage offset of all nodes in the physical connection path under the simulated migration power level. The maximum deviation between the voltage amplitude under load and the rated voltage amplitude is calculated. The load migration simulation model calculates the migration cost for each physical connection path and its corresponding simulated power. The migration cost is obtained by weighted summation of network loss cost, equipment life loss cost, and scheduling operation complexity cost. The network loss cost is calculated based on the product of the simulated energy loss increment and the unit electricity price. The equipment life loss cost is estimated based on the functional relationship between the life impact factor corresponding to the change in equipment load rate and the equipment asset value. The scheduling operation complexity cost is calculated based on the number of switch and transformer tap adjustment operations and the complexity rating required to complete the load migration of the physical connection path.

[0082] Optional, migration cost The calculation uses the following formula:

[0083]

[0084] in: This represents the total migration cost for a physical connection path under a specific simulated migration power level. , , These represent the weighting coefficients for network loss, equipment lifespan degradation, and operational complexity, respectively. This indicates the physical connection path during a preset migration duration under simulated migration power. The increase in total functional energy loss generated internally. This indicates the cost price per unit of electricity. This represents the set of all major electrical equipment included in the physical connection path. Represents a set One of the devices. Indicates equipment The asset value, Indicates the change in equipment load rate The relevant equipment lifespan reduction factor, This indicates the number of scheduling operation instructions required to complete the load migration for this path. This represents the complexity cost per unit of scheduling operation. As can be understood, the weighting coefficients... , , The value is preset by the dispatch strategy maker based on the economic and security preferences of the power grid operation, and the migration duration is... It is a fixed value set according to the scheduling plan.

[0085] It is understandable that the load migration simulation model generates a feasibility assessment result for each physical connection path. The feasibility assessment result is based on a comprehensive judgment of whether the load rate change of the equipment exceeds the limit, whether the voltage deviation is within the allowable range, and whether the protection setting constraints are violated. The load migration simulation model outputs a simulation report containing multiple virtual load migration paths and their feasibility assessment results and migration costs. The virtual load migration path is defined by the physical connection path, the suggested migration power, and the migration rate range. The simulation report is presented in the form of a structured data list. Each virtual load migration path records the corresponding physical connection path identifier, the maximum feasible migration power, key constraints, feasibility assessment results, and the calculated migration cost value.

[0086] Example 4: In a specific implementation, taking a regional power grid dispatch center receiving a simulation report containing five virtual load migration paths with feasible feasibility assessment results as an example, the simulation report output by the load migration simulation model is stored in the form of database records. The dispatch decision system receives the simulation report output by the load migration simulation model, and the dispatch decision system runs a screening program to select all virtual load migration paths with feasible feasibility assessment results from the simulation report. These virtual load migration paths constitute a candidate path set.

[0087] In some embodiments, one or more target migration paths are selected from the selected virtual load migration paths according to the current global optimization objectives of the power grid. The global optimization objectives include minimizing total network loss, achieving the most balanced load on key sections, or minimizing overall migration cost. The global optimization objectives are selected by the dispatcher from preset options or set by custom weight combinations on the decision interface. For each selected target migration path, the specific migrateable load, migration start time, migration duration, and migration rate curve are determined, and the time-space scheduling resources are bound. The migrateable load is determined based on the maximum migration power recommended in the feasibility assessment report of the target migration path and the current actual power deficit. The migration start time is determined based on the power grid load forecast curve and the trend of system safety margin changes. The migration duration is determined based on the migrateable load and the power change rate acceptable to the power grid. The migration rate curve is used to define a smooth transition function of power change over time during the migration duration. The time-space scheduling resources include the transmission capacity reservation of relevant lines, the relevant transformer tap adjustment plan, and the setting coordination of relevant protection devices. The execution time of multiple target migration paths is conflict detected and coordinated to arrange the execution sequence. Conflict detection checks whether there is time overlap and functional mutual exclusion in the operation of different target migration paths on the same line, transformer or switchgear. Coordination and arrangement generate a conflict-free operation sequence based on the priority of the target migration path, operation duration and system security constraints, and generate a load dynamic migration strategy containing a series of ordered scheduling instructions.

[0088] Optionally, the logical judgment formula for conflict detection in execution timing orchestration is as follows:

[0089]

[0090] in: This indicates a conflicting logical judgment result. This represents any two distinct target migration paths. and Perform a union operation. Indicates the target migration path The total set of physical equipment resources required. Indicates the target migration path The total set of physical equipment resources required. This represents the intersection of the physical device resources occupied by two paths. Indicates the target migration path The planned time window set, Indicates the target migration path The planned time window set, This represents the intersection of the time windows for the two planned routes. Let represent the empty set. It can be understood that the formula determines a conflict exists if and only if there are two paths that both occupy shared physical equipment resources and whose planned time windows overlap.

[0091] Referring to Table 1, the load dynamic migration strategy sends a sequence of coordinated dispatch instructions to the corresponding control units in the power grid. The load dynamic migration strategy is parsed into a series of atomic-level power grid operation instructions, including circuit breaker opening and closing instructions, transformer tap adjustment instructions, capacitor bank switching instructions, and interruptible load control instructions. Based on the execution sequence, these atomic-level power grid operation instructions are packaged into several coordinated dispatch instruction sequences with specific execution time windows. Each coordinated dispatch instruction sequence contains a set of instructions that need to be issued to multiple control units within the same region or substation within the same time window. The coordinated dispatch instruction sequence is then sent to the control units in the corresponding regions through the dispatch automation system. The dispatch automation system establishes a connection with the control units and transmits instructions via a standard communication protocol. During the issuance of the coordinated dispatch instruction sequence, the status feedback of the receiving control units is monitored in real time. Status feedback includes instruction reception confirmation signals and equipment readiness signals.

[0092] Table 1: Cooperative Scheduling Instruction Sequence Table

[0093]

[0094] It is understandable that atomic-level power grid operation instructions are generated strictly in accordance with power system operation procedures and contain complete operation objects, target values ​​and security check codes. The planned execution time windows of the coordinated scheduling instruction sequence have strict order and time intervals to ensure the orderliness of power grid operation. The dispatch automation system has instruction simulation and pre-play functions and security error prevention and verification functions. The status feedback of the control unit is recorded in the dispatch log in real time as traceable data of the strategy execution process.

[0095] See Figure 4During the load migration execution phase of power grid load dispatching, the dynamic variation characteristics of the migration rate (unit: MW / min) of three virtual load migration paths with migration time (unit: minutes) were observed. Specifically, path 1 (blue curve) exhibited the largest rate fluctuation, reaching a peak of approximately 12.2 MW / min around the 20-minute migration time, followed by multiple rounds of rises and falls, reflecting its dynamic adjustment characteristics in power migration. Path 3 (green curve) maintained a rate consistently around 10 MW / min, with relatively smooth overall fluctuations, reflecting the stability of its migration process. Path 2 (orange curve) showed a trend of gradually decreasing rate followed by a slight rebound, with a minimum rate of approximately 6.4 MW / min, demonstrating its adaptability to low-power migration. The rate differences among the three paths correspond to the feasibility assessment (e.g., equipment load rate, voltage deviation) and cost calculation (e.g., network loss, operational complexity) results of different paths in the load migration simulation model, providing an intuitive dynamic characteristic reference for selecting the migration rate curve in the dynamic load migration strategy.

[0096] Example 5: In a specific implementation, taking the execution of a load dynamic migration strategy involving the transfer of power across multiple lines in a regional power grid as an example, during the load migration process, real-time response data of the power grid is collected to form the feedback part of the power operation data stream. The response data originates from voltage, current, power, and switch status change information uploaded by distributed sensing nodes. The power operation data stream of the feedback part is input into the fusion channel of multi-source heterogeneous power operation data to update the power grid load situation map and generate a real-time power grid load situation map. The real-time power grid load situation map reflects the latest operating status and load distribution of the power grid after the execution of the load dynamic migration strategy. The real-time power grid load situation map is compared with the situation map predicted before the execution of the load dynamic migration strategy to calculate the deviation between the actual migration trajectory and the path specified by the strategy. The comparison is performed on the same spatiotemporal grid cell, and for each node or line affected by the migration, its actual power value is compared with the expected power value specified by the strategy.

[0097] In some embodiments, deviation The calculation uses the following formula:

[0098]

[0099] in: Indicator representing overall deviation. This indicates the total number of key monitoring points participating in the comparison. Indicates the first Key monitoring points Indicates key monitoring points The actual active power value in the real-time power grid load status diagram. This indicates a key monitoring point in the load dynamic migration strategy. The specified expected active power value, The reference power value used for normalization is typically taken from key monitoring points. The rated capacity of the equipment. (Understandable, deviation.) It is a dimensionless scalar; the larger its value, the more significant the deviation between the actual execution and the expected strategy. If the deviation exceeds the allowable range, a strategy correction process is initiated. The allowable range is preset to a threshold by the scheduling procedure. Based on the new real-time power grid load situation map, the dynamic correlation strength is recalculated, and local or global load migration strategy replanning is triggered.

[0100] The strategy correction process begins by analyzing the deviation to determine its causes. These causes include load forecasting errors, abnormal equipment response, or new disturbances in the power grid. The analysis involves checking whether the power curves of specific nodes in the real-time power grid load situation diagram deviate from the predicted trajectory, whether the relevant switch states match the instructions, and whether any protection action signals have been generated. The unexecuted portions of the currently executing load migration strategy are paused, and the dispatch automation system sends instructions to relevant control units to suspend or halt subsequent operations. Using the current real-time power grid load situation diagram as the initial state, the load migration simulation model is invoked. Within the remaining schedulable time window, a new simulation report is generated. The schedulable time window is determined by the overall completion time of the original strategy and the time already consumed. Based on the new simulation report, a revised load migration strategy is quickly formulated and immediately implemented to ensure that the load migration process converges towards the optimization objective under controlled conditions.

[0101] Optionally, this also includes closed-loop learning of the entire scheduling process, fully recording the data of each load dynamic migration strategy from formulation, execution to correction. This full-process data includes the initial grid load situation diagram, simulation report, load dynamic migration strategy, coordinated dispatch command sequence, real-time grid load situation diagram, deviation degree, and corrected strategy. This data is stored in a historical case database indexed by timestamps. Big data processing techniques are used to mine and analyze the full-process data to extract successful scheduling patterns and scheduling risk characteristics. These techniques include cluster analysis, association rule mining, and sequence pattern analysis. Successful scheduling patterns are characterized by a combination of strategy features with small deviation degrees, a smooth migration process, and a high degree of optimization goal achievement. Scheduling risk characteristics are patterns that are prone to large deviations or operational failures under specific grid structures, load types, or operational sequences. Successful scheduling patterns and scheduling risk characteristics are fed back into the parameter and rule bases of the load migration simulation model to optimize the accuracy of subsequent simulation calculations and the precision of dynamic correlation strength calculations, enabling the entire scheduling method to have adaptive evolution capabilities.

[0102] It is understandable that the update frequency of the real-time power grid load situation map is higher than the refresh frequency of the situation map on which the strategy is based. The deviation can be calculated and comprehensively evaluated for multiple electrical quantities such as voltage and phase angle. The triggering of the strategy correction process can be automatic or initiated after manual confirmation by the dispatcher. The remaining schedulable time window is the total time of the original strategy minus the time elapsed from the start of execution to the triggering of correction. Data mining and analysis during the closed-loop learning process are performed offline on a regular basis. The successful scheduling patterns and scheduling risk characteristics mined are injected into the decision knowledge base of the load migration simulation model in the form of knowledge entries. In subsequent simulation calculations, the load migration simulation model will prioritize the use of parameter settings that match the successful scheduling patterns and impose stricter constraints on scenarios that match the scheduling risk characteristics.

[0103] See Figure 5 In the stage of identifying spatiotemporal clusters and power voids in power grid load dispatching, heatmaps quantified the dynamic correlation strength between different spatiotemporal clusters and power voids. Specifically, using spatiotemporal clusters (clusters A, B, and C) as the vertical dimension and power voids (voids 1, 2, 3, and 4) as the horizontal dimension, correlation strength values ​​in the range of 0.50-0.90 were represented by different color depths. Cluster B and void 2 showed the highest correlation strength (0.88), reflecting a strong correlation in topological distance and electrical coupling. Cluster A and void 4 showed the lowest correlation strength (0.58), corresponding to a weaker spatiotemporal coupling relationship. The quantification of dynamic correlation strength provided a crucial link for the subsequent deduction of load migration paths. In the parameter configuration, the correlation strength was calculated by weighting topological distance and electrical coupling, and its numerical distribution intuitively reflected the potential closeness of the correlation between load migration in different regions.

[0104] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0105] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A power grid load dispatching method based on power big data, characterized in that, It includes the following processing stages: Deploy distributed sensing nodes across the entire power grid to construct a multi-dimensional, high-density power operation data stream; Establish a fusion channel for multi-source heterogeneous power operation data, perform feature extraction and state analysis on the power operation data stream, and generate a power grid load status map with spatiotemporal correlation; Based on the power grid load situation diagram, identify the spatiotemporal clustering areas and power void areas of the load in the power grid, and calculate the dynamic correlation strength between the spatiotemporal clustering areas and the power void areas; A load migration simulation model is constructed. The load migration simulation model uses the dynamic correlation strength as a link to deduce multiple virtual load migration paths between the spatiotemporal clustering region and the power void region, and performs feasibility assessment and cost calculation for each virtual load migration path. Based on the output of the load migration simulation model, a dynamic load migration strategy is formulated, which includes time-scheduled resource binding and execution timing orchestration for the selected virtual load migration path. The load dynamic migration strategy is executed, and a sequence of coordinated scheduling instructions is sent to the corresponding control units in the power grid to drive the load to migrate along the path specified by the strategy in the physical power grid.

2. The power grid load dispatching method based on power big data according to claim 1, characterized in that, The deployment of distributed sensing nodes across the entire power grid to construct a multi-dimensional, high-density power operation data stream specifically includes: Based on the power grid topology and regional functional characteristics, distributed sensing nodes with multi-parameter sensing capabilities are configured at key line nodes, load access points, and power conversion hubs. The distributed sensing nodes synchronously collect voltage, current, phase angle, power, and equipment operating status information to form raw sampling data packets; Each of the distributed sensing nodes adds spatial coordinate identifiers and timestamps to its own generated raw sampling data packets according to a unified time base to generate standard data frames; The standard data frames are aggregated to the data hub center through a dedicated power communication network and a wireless sensor network to form a continuous, high-density power operation data stream.

3. The power grid load dispatching method based on power big data according to claim 2, characterized in that, The establishment of a fusion channel for multi-source heterogeneous power operation data, and the feature extraction and state analysis of the power operation data stream specifically include: Establish a fusion channel for multi-source heterogeneous power operation data in the data hub center, and perform data cleaning, format alignment and time synchronization correction on the power operation data stream; Using the fusion channel of the multi-source heterogeneous power operation data, multi-dimensional feature extraction is performed on the corrected data. The extracted features include load change rate features, power fluctuation cycle features, regional power consumption pattern features, and equipment operation correlation features. The extracted multidimensional features are input into the state parsing engine, which maps the multidimensional features to a predefined set of states to generate a semantic description of the current operating state for each monitored power grid unit. The semantic descriptions are spatially spliced ​​according to the power grid topology and bound to the corresponding timestamps to generate the power grid load status map. The power grid load status map visually reflects the load distribution and operating status of the entire power grid at a specific time in the form of a graph.

4. The power grid load dispatching method based on power big data according to claim 3, characterized in that, The process of identifying spatiotemporal load clusters and power voids in the power grid based on the power grid load situation diagram specifically includes: The power grid load situation map is divided into a spatiotemporal grid, dividing the entire power grid into a series of grid units with spatiotemporal boundaries; Calculate the load density index and power margin for each grid cell. The load density index reflects the degree of load concentration per unit area or unit line capacity, and the power margin reflects the grid cell's ability to accommodate additional loads after meeting existing loads. Grid cell clusters with load density index exceeding the high density threshold are identified as the spatiotemporal clustering region, and grid cell clusters with power margin exceeding the high margin threshold and load density index below the low density threshold are identified as the power void region. On the power grid load situation map, the spatiotemporal clustering area and the power void area are highlighted, and the topological distance and electrical coupling degree between each spatiotemporal clustering area and each power void area are calculated. The dynamic correlation strength is jointly determined by the topological distance and the electrical coupling degree.

5. The power grid load dispatching method based on power big data according to claim 4, characterized in that, The construction of the load migration simulation model specifically includes: Starting from the spatiotemporal clustering region and ending from the power void region, a load migration simulation model is constructed on the virtual network formed by the power grid load situation diagram. The load migration simulation model first generates all possible physical connection paths connecting a spatiotemporal cluster to a power void region based on the impedance parameters of the power grid lines, the transformer tap position status, and the protection setting constraints. For each of the physical connection paths, the load migration simulation model simulates the power flow changes under different migration power amounts, calculates the load rate changes, network loss increments, and voltage offsets of each device in the path, and uses these as the core indicators for evaluating the feasibility of the virtual load migration path. The load migration simulation model calculates the migration cost for each physical connection path and its corresponding simulated power quantity. The migration cost is obtained by weighted summation of network loss cost, equipment lifespan depreciation cost, and scheduling operation complexity cost. The load migration simulation model outputs a simulation report containing multiple virtual load migration paths, their feasibility assessment results, and migration costs.

6. The power grid load dispatching method based on power big data according to claim 5, characterized in that, The specific steps of formulating a dynamic load migration strategy based on the output of the load migration simulation model include: Receive the simulation report output by the load migration simulation model, and filter out all virtual load migration paths whose feasibility assessment results are feasible; Based on the current global optimization objectives of the power grid, one or more target migration paths are selected from the screened virtual load migration paths. The global optimization objectives include minimizing total network loss, achieving the most balanced load on key sections, or minimizing overall migration cost. For each selected target migration path, determine its specific migrateable load, migration start time, migration duration, and migration rate curve, and complete the time-schedule resource binding. The time-schedule resources include the transmission capacity reservation of relevant lines, the relevant transformer tap adjustment plan, and the setting coordination of relevant protection devices. The execution times of multiple target migration paths are conflict detected and coordinated, the execution sequence is arranged, and a load dynamic migration strategy containing a series of ordered scheduling instructions is generated.

7. A power grid load dispatching method based on power big data according to claim 6, characterized in that, The execution of the load dynamic migration strategy, specifically including issuing a sequence of coordinated dispatch instructions to the corresponding control units in the power grid, includes: The load dynamic migration strategy is parsed into a series of atomic-level power grid operation commands, including circuit breaker opening and closing commands, transformer tap adjustment commands, capacitor bank switching commands, and interruptible load control commands. According to the execution timing, the atomic-level power grid operation instructions are packaged into several coordinated scheduling instruction sequences with specific execution time windows; The coordinated scheduling instruction sequence is sent to the control unit of the corresponding area through the scheduling automation system; During the issuance of the coordinated scheduling instruction sequence, the status feedback of the control unit receiving the instructions is monitored in real time to ensure the smooth flow and executability of the instruction channel.

8. A power grid load dispatching method based on power big data according to claim 7, characterized in that, After the driving load migrates along the path specified by the strategy in the physical power grid, the process also includes: During the load migration process, the power grid response data is collected in real time to form the feedback part of the power operation data stream; The power operation data stream of the feedback section is input into the fusion channel of the multi-source heterogeneous power operation data to update the power grid load status map and generate a real-time power grid load status map. The real-time power grid load situation map is compared with the situation map predicted before the load dynamic migration strategy is implemented, and the deviation between the actual migration trajectory and the path specified by the strategy is calculated. If the deviation exceeds the allowable range, the strategy correction process is initiated, the dynamic correlation strength is recalculated based on the new real-time power grid load situation map, and a local or global load migration strategy replanning is triggered.

9. A power grid load dispatching method based on power big data according to claim 8, characterized in that, The startup strategy correction process specifically includes: Based on the deviation, the causes of the deviation are analyzed, including load forecast deviation, abnormal equipment response, or new disturbance events in the power grid. Pause the unexecuted portions of the currently running load migration strategy; Using the current real-time power grid load situation map as the initial state, the load migration simulation model is invoked, and within the remaining schedulable time window, the simulation calculation is re-performed to generate a new simulation report. Based on the new simulation report, a revised load migration strategy was quickly formulated and immediately implemented to ensure that the load migration process converges toward the optimization objective under controlled conditions.

10. A power grid load dispatching method based on power big data according to claim 9, characterized in that, It also includes closed-loop learning of the entire scheduling process: Completely record the entire process of each load dynamic migration strategy from formulation, execution to correction, including the initial power grid load status diagram, simulation report, load dynamic migration strategy, coordinated dispatch instruction sequence, real-time power grid load status diagram, deviation and corrected strategy; Using big data processing technology, the data of the entire process is mined and analyzed to extract successful scheduling patterns and scheduling risk characteristics; The successful scheduling mode and scheduling risk characteristics are fed back to the parameter library and rule library of the load migration simulation model to optimize the accuracy of subsequent simulation calculations and the calculation precision of the dynamic correlation strength, so that the entire scheduling method has adaptive evolution capability.

Images