A Smart Power Grid Dispatch Decision System and Method Based on Multi-Source Heterogeneous Data Fusion

Abstract

This invention discloses a power grid intelligent dispatch decision-making system and method based on multi-source heterogeneous data fusion, belonging to the field of power grid intelligent dispatch technology. The specific steps of the power grid intelligent dispatch decision-making method based on multi-source heterogeneous data fusion include: Step 1, collecting multi-source heterogeneous data, constructing a meteorological feature vector set, predicting power grid load changes, and outputting load forecast values; Step 2, fusing load forecast values ​​and equipment health coefficients to construct a dual-risk collaborative analysis matrix, generating risk entropy values ​​and mapping them to standard risk models; Step 3, according to the standard risk models, performing generator output pre-adjustment, energy storage charging and discharging control, and interruptible load management in a tiered manner, and coordinating microgrid resources through blockchain. This invention avoids the one-sidedness of single-risk factor decision-making by constructing a dual-dimensional risk collaborative decision-making architecture, orthogonally decomposing demand-side forecast risk and equipment health risk into multiple state modes.

Description

Technical Field

[0001] This invention belongs to the field of intelligent power grid dispatching technology, specifically relating to an intelligent power grid dispatching decision-making system and method based on multi-source heterogeneous data fusion. Background Technology

[0002] As the power system accelerates its evolution towards "dual high" characteristics (high proportion of new energy and high proportion of power electronic equipment), the sources of risks faced by the power grid operation are becoming increasingly diversified, and there are complex nonlinear coupling effects between different risks. It is easy to overlook the coupling effect between demand-side and equipment-side risks, leading to one-sided dispatch decisions. Traditional single risk analysis methods are no longer able to meet the dual requirements of system safety and efficiency.

[0003] Extreme weather (such as high temperatures, rainstorms, and typhoons) not only directly leads to a surge in load demand (such as summer high-temperature scenarios where air conditioning load accounts for more than 40%), but also indirectly causes equipment failure risks through equipment performance degradation (such as accelerated aging of transformer insulation and transmission line galloping). For example, in high-temperature environments, the probability of failure may increase by 20%-30% for every 10% increase in equipment load rate, and the superposition effect of load fluctuations and equipment failures may further amplify the vulnerability of the system.

[0004] The intermittency of new energy power generation (such as the daily fluctuation rate of photovoltaic output exceeding 50%) leads to a widening of the load peak-valley difference. Frequent start-ups and shutdowns or deep peak shaving of equipment may accelerate mechanical wear and thermal stress damage. The grid structure, equipment aging degree and meteorological sensitivity vary significantly in different geographical regions. Changes in equipment health status (such as insulation aging and metal fatigue) usually take several years to accumulate, while meteorological events (such as continuous rainstorms) may trigger failures instantly.

[0005] Therefore, there is an urgent need to propose a smart grid dispatching decision-making method based on the fusion of multi-source heterogeneous data to solve the above problems. Summary of the Invention

[0006] The purpose of this invention is to provide a smart grid dispatching decision system and method based on multi-source heterogeneous data fusion, which solves the technical problem in the prior art that easily overlooks the coupling effect of demand-side and equipment-side risks, leading to one-sided dispatching decisions.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A smart grid dispatching decision-making system and method based on multi-source heterogeneous data fusion, including:

[0009] Step 1: Collect multi-source heterogeneous data, construct a meteorological feature vector set, predict power grid load changes, and output load forecast values;

[0010] Step 2: Integrate load forecast values ​​and equipment health coefficients to construct a dual-risk collaborative analysis matrix, generate risk entropy values, and map them to standard risk patterns;

[0011] Step 3: Based on the standard risk model, perform generator output pre-adjustment, energy storage charging and discharging control, and interruptible load management in stages, and coordinate microgrid resources through blockchain.

[0012] Furthermore, multi-source heterogeneous data are collected to construct a meteorological feature vector set. The specific method is as follows:

[0013] The system collects power grid operating parameters, including current, voltage, and active power data, in real time through the SCADA system at preset time intervals. At the same time, it calls the meteorological department's API interface to obtain regional temperature, humidity, wind speed, and air pressure data. The multi-source heterogeneous data is mapped to a specific range through Z-Score standardization. The meteorological data is up-processed using cubic spline interpolation technology to align it with the power grid data timestamp. A feature vector set containing temperature change rate, humidity gradient, and wind speed fluctuation values ​​is constructed and a graded early warning is triggered.

[0014] Furthermore, a feature vector set containing temperature change rate, humidity gradient, and wind speed fluctuation values ​​is constructed and a tiered early warning system is triggered. The specific method is as follows:

[0015] The temperature change rate is obtained by calculating the ratio of the temperature difference between adjacent time points in the monitoring area to the time interval. When the temperature change rate is greater than or equal to A1, a high temperature warning monitoring is triggered and the monitoring continues for a duration of a1. When the average temperature change rate within the monitoring duration is greater than or equal to x1, a high temperature red warning is issued; otherwise, the monitoring returns to normal.

[0016] The humidity gradient is obtained by calculating the difference in humidity between adjacent time points in the monitoring area to determine the monitoring area range. When the difference in humidity gradient between adjacent time points in the monitoring area (current time point minus previous time point) is greater than or equal to A2, humidity early warning monitoring is triggered and the monitoring continues for a duration of a2. When the humidity in the monitoring area is continuously greater than or equal to x2 within the duration of a2, a continuous rainy warning is issued; otherwise, a regular warning is returned.

[0017] The wind speed fluctuation value is obtained by statistically analyzing the standard deviation of wind speed data in the monitoring area at adjacent time points. When the standard deviation of wind speed is greater than or equal to A3, typhoon warning monitoring is triggered. Monitoring continues for a duration of a3. When the average rate of air pressure drop in the monitoring area is greater than or equal to x3, an orange typhoon warning is issued. Otherwise, it returns to a regular warning.

[0018] Furthermore, the specific methods for predicting changes in power grid load are as follows:

[0019] Collect historical multi-source heterogeneous data for N time periods, dynamically adjust the feature weights according to the real-time weather type, increase the temperature feature weight to n1 times the original value during the high-temperature warning period, increase the humidity feature weight to n2 times the original value during the continuous rainy period, and increase the wind speed feature weight to n3 times the original value during the typhoon warning period;

[0020] Divide the power grid coverage area into M meteorological grids, input the historical feature vectors into the gradient boosting tree prediction model, the weight of each grid is equal to the ratio of the load of the grid to the total load of the monitoring area, adopt a recursive prediction strategy, output the load change values of each grid for the future N1 time period, and then perform residual correction on the prediction results through a long short-term memory network to form the load prediction value ΔPd of the final power grid monitoring area.

[0021] Furthermore, fuse the load prediction value and the equipment health coefficient to construct a dual-risk collaborative analysis matrix. The specific method is as follows: conduct a fusion analysis on the load prediction value ΔPd and the equipment health coefficient Rd, calculate the demand-side risk coefficient, and its value is the percentage of the predicted load value to the regional benchmark load. Set three-level risk thresholds: when the demand risk coefficient > M1, it is a low risk; when M1 ≤ demand risk coefficient ≤ M2, it is a medium risk; when the demand risk coefficient > M2, it is a high risk. The equipment risk level follows the preset threshold division method. When the equipment health coefficient > m2, it is a low risk; when m1 ≤ equipment health coefficient ≤ m2, it is a medium risk; when the equipment health coefficient < m1, it is a high risk;

[0022] Create a dual-risk collaborative analysis matrix, forming a total of nine basic risk combination units of 3×3. The row dimension of the matrix is the demand-side risk level, and the column dimension of the matrix is the equipment risk level.

[0023] Furthermore, the equipment health coefficient specifically includes:

[0024] Continuously monitor the average load rate of the equipment within the preset time period. When it exceeds the set threshold, it is marked as a high-load state. Determine the total load duration of the power grid equipment. Determine the equipment health coefficient by comprehensively considering the total load duration and the high-load duration of the power grid equipment, and use the formula to represent, where x represents the xth equipment, Zh(x) represents the total load duration of the power grid equipment, and zh(x) represents the high-load duration of the power grid equipment.

[0025] Furthermore, generate the risk entropy value and map it to the standard risk mode. The specific method is as follows:

[0026] For each risk combination unit, the raw values ​​of the demand-side risk percentage and the equipment Rd value are extracted and weighted according to a preset weight ratio. The dispersion of the weighted result is calculated as the risk entropy value. The risk entropy value is generated by quantifying the coupling effect of different risk combinations. When the risk entropy value exceeds the dynamic threshold of the corresponding risk combination unit, the system automatically enters the standard risk mode and performs generator output pre-adjustment, energy storage charging and discharging control, and interruptible load management. A pattern-policy mapping knowledge base is established on the digital twin platform to match the standard mode corresponding to the current risk combination in real time. When the entropy value suddenly exceeds the historical average W%, pattern rematching is triggered.

[0027] Furthermore, based on the standard risk model, generator output pre-adjustment, energy storage charging and discharging control, and interruptible load management are implemented in stages, and microgrid resources are coordinated through blockchain. The specific methods are as follows:

[0028] At the generator control level, a pre-adjustment mechanism is activated when demand risk reaches medium to high risk. In load growth scenarios, output is increased by H1% of the predicted deficit, and in load decline scenarios, output is reduced by H2% of the predicted surplus. The adjustment process is constrained by the equipment risk status. Output increases are prohibited in high-risk equipment areas, and output adjustments in medium-risk equipment areas are limited to no more than h% of the unit's rated capacity.

[0029] At the energy storage system control level, when the load forecast increases, the discharge mode is activated to supplement the power shortage, and when the load forecast decreases, the charging mode is switched to absorb the excess energy. The charging and discharging power adopts the maximum capacity operation mode in the low-risk area of ​​the equipment, and is derated to H3% of the rated power in the medium- and high-risk areas of the equipment.

[0030] At the load management level, the interruptible load management procedure is activated only when the conditions of high demand risk, low equipment risk and forecast error of less than H4% for three consecutive time periods are met simultaneously. Commercial loads are interrupted first, while critical loads are kept powered.

[0031] This invention also provides a power grid intelligent dispatching decision system based on multi-source heterogeneous data fusion, applied to a power grid intelligent dispatching decision method based on multi-source heterogeneous data fusion, comprising:

[0032] The data acquisition and load forecasting module collects multi-source heterogeneous data, constructs a meteorological feature vector set, predicts power grid load changes, and outputs load forecast values.

[0033] The risk entropy generation module integrates load forecast values ​​and equipment health coefficients to construct a dual-risk collaborative analysis matrix, generate risk entropy values, and map them to standard risk patterns.

[0034] The intelligent dispatch decision-making module performs generator output pre-adjustment, energy storage charging and discharging control, and interruptible load management in a tiered manner according to the standard risk model, and coordinates microgrid resources through blockchain.

[0035] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0036] 1. This invention quantifies the dynamic impact of meteorological elements on power grid load by constructing a set of meteorological feature vectors (temperature change rate, humidity gradient, wind speed fluctuation value). It triggers differentiated dispatching measures through graded early warning (high temperature red warning, continuous rain warning, typhoon orange warning). Based on real-time load forecast values ​​and meteorological feature weight adjustments, it dynamically optimizes power generation plans, improves dispatching accuracy and real-time performance, adapts to extreme weather and new energy fluctuations, and strengthens risk prevention and control capabilities.

[0037] 2. This invention achieves dual quantification of demand-side load fluctuation risk and equipment operation health risk through independent modeling and fusion analysis of load forecast values ​​and equipment health coefficients, forming multiple basic risk combination units. Based on weighted fusion calculation of risk entropy values, it quantifies the risk coupling effect, quantifies system stability by calculating risk entropy values, and sets dynamic thresholds to improve the real-time and adaptability of risk warnings. It maps risk entropy values ​​to standard risk patterns and establishes a pattern-strategy mapping knowledge base through a digital twin platform to achieve rapid response to risk status.

[0038] 3. This invention achieves dynamic balance of generator output through demand risk classification and equipment risk constraints, avoiding equipment overload or failure caused by blindly increasing output. Through dynamic charging and discharging control of the energy storage system, load fluctuations are smoothed, reducing the need for frequent generator start-stop or deep peak shaving, and improving power generation efficiency. Through blockchain smart contracts, a power support protocol between the main network and the microgrid is implemented, ensuring the transparency, reliability and real-time nature of resource allocation. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 A flowchart of a power grid intelligent dispatch decision-making method based on multi-source heterogeneous data fusion is shown.

[0041] Figure 2 The diagram illustrates the steps involved in constructing a feature vector set containing temperature change rate, humidity gradient, and wind speed fluctuation values, and triggering a graded early warning system.

[0042] Figure 3 A block diagram of a smart grid dispatching decision system based on multi-source heterogeneous data fusion is shown. Detailed Implementation

[0043] 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.

[0044] Example 1, such as Figure 1 , Figure 2 The power grid intelligent dispatch decision-making method based on multi-source heterogeneous data fusion shown includes the following steps:

[0045] Step 1: Collect multi-source heterogeneous data, construct a meteorological feature vector set, predict power grid load changes, and output load forecast values.

[0046] At preset time intervals, the SCADA system collects real-time power grid operating parameters, including current, voltage, and active power data. At the same time, it calls the meteorological department's API interface to obtain regional temperature, humidity, wind speed, and air pressure data. Through Z-Score standardization, the multi-source heterogeneous data is mapped to a specific range. Cubic spline interpolation technology is used to up-process the meteorological data to align it with the power grid data timestamp. A feature vector set containing temperature change rate, humidity gradient, and wind speed fluctuation values ​​is constructed and a graded early warning is triggered.

[0047] The temperature change rate is obtained by calculating the ratio of the temperature difference between adjacent time points in the monitoring area to the time interval. When the temperature change rate is greater than or equal to A1, a high temperature warning monitoring is triggered and the monitoring continues for a duration of a1. When the average temperature change rate within the monitoring duration is greater than or equal to x1, a high temperature red warning is issued; otherwise, the monitoring returns to normal.

[0048] The humidity gradient is obtained by calculating the difference in humidity between adjacent time points in the monitoring area to determine the monitoring area range. When the difference in humidity gradient between adjacent time points in the monitoring area (current time point minus previous time point) is greater than or equal to A2, humidity early warning monitoring is triggered and the monitoring continues for a duration of a2. When the humidity in the monitoring area is continuously greater than or equal to x2 within the duration of a2, a continuous rainy warning is issued; otherwise, a regular warning is returned.

[0049] The wind speed fluctuation value is obtained by statistically analyzing the standard deviation of wind speed data in the monitoring area at adjacent time points. When the standard deviation of wind speed is greater than or equal to A3, typhoon warning monitoring is triggered. The monitoring continues for a duration of a3. When the average rate of air pressure drop in the monitoring area is greater than or equal to x3, an orange typhoon warning is issued. Otherwise, it returns to a regular warning.

[0050] Collect historical multi-source heterogeneous data for N duration, analyze the variation law of the power grid load in the monitoring area under different meteorological environments, establish an association analysis library between meteorological elements and historical load changes, dynamically adjust the feature weights according to the real-time weather type, increase the temperature feature weight to n1 times the original value during the high temperature warning period, increase the humidity feature weight to n2 times the original value during the continuous rainy period, and increase the wind speed feature weight to n3 times the original value during the typhoon warning period. The weight coefficient is updated hourly through an online learning model to ensure that the prediction model adapts to the climate mutation scenario;

[0051] Divide the power grid coverage area into M meteorological grids, input the historical feature vector into the gradient boosting tree prediction model. The weight of each grid is equal to the ratio of the load of the grid to the total load of the monitoring area. Adopt a recursive prediction strategy to output the load change value of each grid for the future N1 duration, and then perform residual correction on the prediction result through a long short-term memory network to form the load prediction value ΔPd of the final power grid monitoring area.

[0052] Step 2: Integrate the load prediction value and the equipment health coefficient to construct a dual-risk collaborative analysis matrix, generate the risk entropy value and map it to a standard risk mode.

[0053] Based on the power grid equipment as the basic grid unit, each equipment corresponds to only one basic grid. Continuously monitor the average load rate of the equipment corresponding to the basic grid within the preset duration. When it exceeds the set threshold, it is marked as a high-load state. Determine the total load duration of the power grid equipment. Combine the total load duration and the high-load duration of the power grid equipment to determine the equipment health coefficient, using the formula It is expressed as, where x represents the xth equipment, Zh(x) represents the total load duration of the power grid equipment, and zh(x) represents the high-load duration of the power grid equipment.

[0054] Perform a fusion analysis on the load prediction value ΔPd and the equipment health coefficient Rd, calculate the independent demand risk coefficient of each grid, and its value is the percentage of the predicted load value to the regional reference load. Set three-level risk thresholds: when the demand risk coefficient > M1, it is a low risk; when M1 ≤ demand risk coefficient ≤ M2, it is a medium risk; when the demand risk coefficient > M2, it is a high risk. The equipment risk level follows the preset threshold division method. When the equipment health coefficient > m2, it is a low risk; when m1 ≤ equipment health coefficient ≤ m2, it is a medium risk; when the equipment health coefficient < m1, it is a high risk.

[0055] Identify high-risk aggregation areas through a spatial autocorrelation algorithm (such as Moran's I index). If more than 3 adjacent grids simultaneously meet the high demand risk or high equipment risk, trigger a regional cascade risk warning and shorten the monitoring time interval of the surrounding adjacent grids (shorten the duration corresponding to the original time interval by half);

[0056] A dual-risk collaborative analysis matrix is ​​created, forming nine basic risk combination units in a 3×3 matrix. The matrix row dimension represents the demand-side risk level (low / medium / high), and the matrix column dimension represents the equipment risk level (low / medium / high). For each risk combination unit, the raw values ​​of the demand-side risk percentage and the equipment Rd value are extracted and weighted according to a preset weight ratio (e.g., when demand risk is high and equipment risk is low, y1 times the demand value + y2 times the equipment value is used). The dispersion of the weighted result is calculated as the risk entropy value (the higher the entropy value, the more unstable the system state). The risk entropy value is generated by quantifying the coupling effect of different risk combinations. When the risk entropy value exceeds the dynamic threshold of the corresponding risk combination unit, the system automatically enters the standard risk mode and performs generator output pre-adjustment, energy storage charging and discharging control, and interruptible load management.

[0057] The dynamic threshold of the i-th risk combination unit is calculated using the formula... It is expressed as, where t represents a time point. Let represent the basic threshold of the i-th risk combination unit. k1(t), k2(t), and k3(t) represent the graded early warning condition functions for time t. When there is a red high temperature warning at time t, k1(t) equals f1, where f1 is the high temperature weight; otherwise, k1(t) equals 0. When there is a continuous rain warning at time t, k2(t) equals f2, where f2 is the rain weight; otherwise, k2(t) equals 0. When there is a typhoon orange warning at time t, k3(t) equals f3, where f3 is the typhoon weight; otherwise, k3(t) equals 0.

[0058] Establish a pattern-strategy mapping knowledge base on the digital twin platform to match the standard pattern corresponding to the current risk portfolio in real time. When the entropy value suddenly exceeds the historical average W%, pattern rematch is triggered.

[0059] Step 3: Based on different basic risk combinations, implement generator output pre-adjustment, energy storage charging and discharging control, and interruptible load management in stages, and coordinate microgrid resources through blockchain.

[0060] Based on the risk model, hierarchical scheduling decisions are implemented. At the generator control level: when the demand risk reaches medium to high risk, the pre-adjustment mechanism is activated. In the load growth scenario, the output is increased by H1% of the predicted deficit value, and in the load decline scenario, the output is reduced by H2% of the predicted surplus value. This adjustment process is constrained by the equipment risk status - the output increase operation is prohibited in the high-risk area of ​​the equipment, and the output adjustment range in the medium-risk area of ​​the equipment is limited to not exceeding h of the unit's rated capacity.

[0061] At the energy storage system control level: when the load forecast increases, the discharge mode is activated to supplement the power shortage; when the load forecast decreases, the charging mode is switched to absorb the excess energy. The charging and discharging power adopts the maximum capacity operation mode in the low-risk area of ​​the equipment, and is reduced to H3% of the rated power in the medium- and high-risk areas of the equipment to ensure the safety margin of the equipment.

[0062] When the main grid's regulation capacity is insufficient, a resource coordination request is sent to the grid-connected microgrid. The microgrid controller aggregates local photovoltaic, energy storage, and controllable load resources, and reaches a power support agreement through a blockchain smart contract. The main grid dispatch center integrates the contribution value of distributed resources and dynamically adjusts the energy storage compensation power demand.

[0063] At the load management level: The interruptible load management procedure is activated only when the conditions of high demand risk, low equipment risk and forecast error of less than H4% for three consecutive time periods are met simultaneously. Commercial loads are interrupted first, while power supply to critical loads (such as medical and transportation loads) is maintained, so as to achieve a balance between power supply reliability and system security.

[0064] Example 2, as follows Figure 3 The power grid intelligent dispatching and decision-making system based on multi-source heterogeneous data fusion shown includes the following:

[0065] The data acquisition and load forecasting module collects grid operation parameters in real time through the SCADA system at preset time intervals, including current, voltage, and active power data. At the same time, it calls the meteorological department's API interface to obtain regional temperature, humidity, wind speed, and air pressure data. Through Z-Score standardization, the multi-source heterogeneous data is mapped to a specific range. Cubic spline interpolation technology is used to up-process the meteorological data to align it with the grid data timestamp. A feature vector set containing temperature change rate, humidity gradient, and wind speed fluctuation values ​​is constructed and a graded early warning is triggered.

[0066] The temperature change rate is obtained by calculating the ratio of the temperature difference between adjacent time points in the monitoring area to the time interval. When the temperature change rate is greater than or equal to A1, a high temperature warning monitoring is triggered and the monitoring continues for a duration of a1. When the average temperature change rate within the monitoring duration is greater than or equal to x1, a high temperature red warning is issued; otherwise, the monitoring returns to normal.

[0067] The humidity gradient is obtained by calculating the difference in humidity between adjacent time points in the monitoring area to determine the monitoring area range. When the difference in humidity gradient between adjacent time points in the monitoring area (current time point minus previous time point) is greater than or equal to A2, humidity early warning monitoring is triggered and the monitoring continues for a duration of a2. When the humidity in the monitoring area is continuously greater than or equal to x2 within the duration of a2, a continuous rainy warning is issued; otherwise, a regular warning is returned.

[0068] The wind speed fluctuation value is obtained by statistically calculating the standard deviation of the wind speed data in the monitoring area at adjacent time points. When the wind speed standard deviation is greater than or equal to A3, typhoon warning monitoring is triggered and monitored for a duration of a3. When the average value of the air pressure drop rate in the monitoring area is greater than or equal to x3, an orange typhoon warning is issued; otherwise, a regular warning is returned.

[0069] Collect historical multi-source heterogeneous data for N duration, analyze the variation law of the power grid load in the monitoring area under different meteorological environments, establish an association analysis library of meteorological elements and historical load changes, dynamically adjust the feature weights according to the real-time weather type, increase the temperature feature weight to n1 times the original value during the high-temperature warning period, increase the humidity feature weight to n2 times the original value during the continuous rainy period, and increase the wind speed feature weight to n3 times the original value during the typhoon warning period. The weight coefficients are updated hourly through an online learning model to ensure that the prediction model adapts to climate mutation scenarios.

[0070] Divide the power grid coverage area into M meteorological grids, input the historical feature vectors into the gradient boosting tree prediction model. The weight of each grid is equal to the ratio of the load of the grid to the total load of the monitoring area. Adopt a recursive prediction strategy to output the load change values of each grid for the future N1 duration, and then perform residual correction on the prediction results through a long short-term memory network to form the final load prediction value ΔPd of the power grid monitoring area.

[0071] The risk entropy value generation module takes power grid equipment as the basic grid unit. Each equipment corresponds to only one basic grid. Continuously monitor the average load rate of the equipment corresponding to the basic grid within the preset duration. When it exceeds the set threshold, it is marked as a high-load state. Determine the total load duration of the power grid equipment, and comprehensively determine the equipment health coefficient based on the total load duration and high-load duration of the power grid equipment. It is expressed by the formula, where x represents the xth equipment, Zh(x) represents the total load duration of the power grid equipment, and zh(x) represents the high-load duration of the power grid equipment.

[0072] Conduct a fusion analysis of the load prediction value ΔPd and the equipment health coefficient Rd, calculate the independent demand risk coefficient for each grid, and its value is the percentage of the predicted load value to the regional benchmark load. Set three-level risk thresholds: when the demand risk coefficient > M1, it is a low risk; when M1 ≤ demand risk coefficient < M2, it is a medium risk; when the demand risk coefficient > M2, it is a high risk. The equipment risk level follows the preset threshold division method. When the equipment health coefficient > m2, it is a low risk; when m1 ≤ equipment health coefficient < m2, it is a medium risk; when the equipment health coefficient < m1, it is a high risk.

[0073] High-risk clusters are identified by spatial autocorrelation algorithms (such as Moran's I index). If three or more adjacent grids simultaneously meet the requirements for high risk or equipment risk, a regional cascade risk warning is triggered, and the monitoring time interval of the surrounding adjacent grids is shortened (the original time interval is shortened by half).

[0074] A dual-risk collaborative analysis matrix is ​​created, forming nine basic risk combination units in a 3×3 matrix. The matrix row dimension represents the demand-side risk level (low / medium / high), and the matrix column dimension represents the equipment risk level (low / medium / high). For each risk combination unit, the raw values ​​of the demand-side risk percentage and the equipment Rd value are extracted and weighted according to a preset weight ratio (e.g., when demand risk is high and equipment risk is low, y1 times the demand value + y2 times the equipment value is used). The dispersion of the weighted result is calculated as the risk entropy value (the higher the entropy value, the more unstable the system state). The risk entropy value is generated by quantifying the coupling effect of different risk combinations. When the risk entropy value exceeds the dynamic threshold of the corresponding risk combination unit, the system automatically enters the standard risk mode and performs generator output pre-adjustment, energy storage charging and discharging control, and interruptible load management.

[0075] The dynamic threshold of the i-th risk combination unit is expressed by the formula, where t represents the time point, f represents the basic threshold of the i-th risk combination unit, and k1(t) and f represent the graded early warning condition functions for time t. When there is a high temperature red warning at time t, k1(t) equals f1, where f1 is the high temperature weight; otherwise, k1(t) equals 0. When there is a continuous rain warning at time t, k2(t) equals f2, where f2 is the rain weight; otherwise, k2(t) equals 0. When there is a typhoon orange warning at time t, k3(t) equals f3, where f3 is the typhoon weight; otherwise, k3(t) equals 0.

[0076] Establish a pattern-strategy mapping knowledge base on the digital twin platform to match the standard pattern corresponding to the current risk portfolio in real time. When the entropy value suddenly exceeds the historical average W%, pattern rematch is triggered.

[0077] The intelligent scheduling decision module executes hierarchical scheduling decisions based on risk patterns. At the generator control level, when the demand risk reaches medium to high risk, a pre-adjustment mechanism is activated. In load growth scenarios, output is increased by H1% of the predicted deficit value, and in load decline scenarios, output is reduced by H2% of the predicted surplus value. This adjustment process is constrained by the equipment risk status—output increase is prohibited in high-risk areas of the equipment, and the output adjustment range is limited to no more than h of the unit's rated capacity in medium-risk areas of the equipment.

[0078] At the energy storage system control level: when the load forecast increases, the discharge mode is activated to supplement the power shortage; when the load forecast decreases, the charging mode is switched to absorb the excess energy. The charging and discharging power adopts the maximum capacity operation mode in the low-risk area of ​​the equipment, and is reduced to H3% of the rated power in the medium- and high-risk areas of the equipment to ensure the safety margin of the equipment.

[0079] When the main grid's regulation capacity is insufficient, a resource coordination request is sent to the grid-connected microgrid. The microgrid controller aggregates local photovoltaic, energy storage, and controllable load resources, and reaches a power support agreement through a blockchain smart contract. The main grid dispatch center integrates the contribution value of distributed resources and dynamically adjusts the energy storage compensation power demand.

[0080] At the load management level: The interruptible load management procedure is activated only when the conditions of high demand risk, low equipment risk and forecast error of less than H4% for three consecutive time periods are met simultaneously. Commercial loads are interrupted first, while power supply to critical loads (such as medical and transportation loads) is maintained, so as to achieve a balance between power supply reliability and system security.

[0081] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

[0082] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to specific implementations. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A power grid intelligent dispatching decision-making method based on multi-source heterogeneous data fusion, characterized in that, It includes: Step 1: Collect multi-source heterogeneous data, construct a meteorological feature vector set, predict the change of power grid load, and output the load prediction value; Step 2: Integrate the load prediction value and the equipment health coefficient to construct a dual-risk collaborative analysis matrix, generate a risk entropy value, and map it to a standard risk mode; Conduct a comprehensive analysis by integrating the load prediction value ΔPd and the equipment health coefficient Rd, calculate the demand-side risk coefficient, whose value is the percentage of the predicted load value to the regional benchmark load. Set three-level risk thresholds: when the demand risk coefficient > M1, it is a low risk; when M1 ≤ demand risk coefficient ≤ M2, it is a medium risk; when the demand risk coefficient > M2, it is a high risk. The equipment risk level follows the preset threshold division method. When the equipment health coefficient > m2, it is a low risk; when m1 ≤ equipment health coefficient ≤ m2, it is a medium risk; when the equipment health coefficient < m1, it is a high risk; Create a dual-risk collaborative analysis matrix, forming a total of nine basic risk combination units of 3×3. The row dimension of the matrix is the demand-side risk level, and the column dimension of the matrix is the equipment risk level; Extract the original values of the demand-side risk percentage value and the equipment Rd value for each risk combination unit, perform weighted integration according to the preset weight ratio, calculate the dispersion degree of the weighted result as the risk entropy value, generate the risk entropy value by quantifying the coupling effect of different risk combinations. When the risk entropy value exceeds the dynamic threshold of the corresponding risk combination unit, it automatically enters the standard risk mode, executes the pre-adjustment of generator output, energy storage charge and discharge control, and interruptible load management. Establish a mode-strategy mapping knowledge base on the digital twin platform, and real-time match the standard mode corresponding to the current risk combination. When the entropy value mutation exceeds W% of the historical average, trigger mode re-matching; Step 3: According to the standard risk mode, hierarchically execute the pre-adjustment of generator output, energy storage charge and discharge control, and interruptible load management, and coordinate the microgrid resources through the blockchain.

2. The power grid intelligent dispatching decision-making method based on multi-source heterogeneous data fusion according to claim 1, characterized in that, Collect multi-source heterogeneous data and construct a meteorological feature vector set. The specific method is as follows: Real-time collect the power grid operation parameters, including current, voltage, and active power data, through the SCADA system at a preset time interval. At the same time, call the API interface of the meteorological department to obtain the regional temperature, humidity, wind speed, and air pressure data. Map the multi-source heterogeneous data to a specific range through Z-Score standardization, and use the cubic spline interpolation technology to perform upsampling processing on the meteorological data to align its time stamp with the power grid data. Construct a feature vector set containing the temperature change rate, humidity gradient, and wind speed fluctuation value and trigger hierarchical early warnings.

3. The power grid intelligent dispatching decision-making method based on multi-source heterogeneous data fusion according to claim 2, characterized in that, Construct a feature vector set containing the temperature change rate, humidity gradient, and wind speed fluctuation value and trigger hierarchical early warnings. The specific method is as follows: The temperature change rate is obtained by calculating the ratio of the temperature difference in the monitored area between adjacent time points to the time interval. When the temperature change rate is greater than or equal to A1, trigger high-temperature warning monitoring and continuously monitor for a1 duration. When the average value of the temperature change rate during the monitoring duration is greater than or equal to x1, issue a high-temperature red warning; otherwise, return to normal monitoring; The humidity gradient is obtained by calculating the difference in humidity between adjacent time points in the monitoring area to determine the monitoring area range. When the difference in humidity gradient between adjacent time points in the monitoring area is greater than or equal to A2, humidity early warning monitoring is triggered and the monitoring continues for a duration of a2. When the humidity in the monitoring area is continuously greater than or equal to x2 within the duration of a2, a continuous rainy warning is issued; otherwise, a regular warning is issued. The wind speed fluctuation value is obtained by statistically analyzing the standard deviation of wind speed data in the monitoring area at adjacent time points. When the standard deviation of wind speed is greater than or equal to A3, typhoon warning monitoring is triggered. Monitoring continues for a duration of a3. When the average rate of air pressure drop in the monitoring area is greater than or equal to x3, an orange typhoon warning is issued. Otherwise, it returns to a regular warning.

4. The power grid intelligent dispatching decision-making method based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The specific methods for predicting changes in power grid load are as follows: Collect N hours of historical multi-source heterogeneous data, and dynamically adjust the feature weights according to the real-time weather type. During high temperature warnings, the temperature feature weight is increased to n1 times the original value; during continuous rainy weather, the humidity feature weight is increased to n2 times the original value; and during typhoon warnings, the wind speed feature weight is increased to n3 times the original value. The power grid coverage area is divided into M meteorological grids. The historical feature vectors are input into the gradient boosting tree prediction model. The weight of each grid is equal to the ratio of the load of that grid to the total load of the monitoring area. A recursive prediction strategy is adopted to output the load change value of each grid for the next N1 hours. Then, the prediction results are corrected by the residual through a long short-term memory network to form the final load prediction value ΔPd of the power grid monitoring area.

5. The power grid intelligent dispatching decision-making method based on multi-source heterogeneous data fusion according to claim 1, characterized in that, Equipment health factor, specifically includes: The system continuously monitors the average load rate of equipment within a preset time period. When the average load rate exceeds a set threshold, it is marked as a high-load state. The total load duration of the power grid equipment is determined, and the equipment health coefficient is determined by combining the total load duration and the high-load duration. The formula is then used to calculate the health coefficient. Let represent the x-th device, Zh(x) represent the total load duration of the grid equipment, and zh(x) represent the high load duration of the grid equipment.

6. The power grid intelligent dispatching decision-making method based on multi-source heterogeneous data fusion according to claim 1, characterized in that, Based on the standard risk model, generator output pre-adjustment, energy storage charging and discharging control, and interruptible load management are implemented in stages, and microgrid resources are coordinated through blockchain. The specific methods are as follows: At the generator control level, a pre-adjustment mechanism is activated when demand risk reaches medium to high risk. In load growth scenarios, output is increased by H1% of the predicted deficit, and in load decline scenarios, output is reduced by H2% of the predicted surplus. The adjustment process is constrained by the equipment risk status. Output increases are prohibited in high-risk equipment areas, and output adjustments in medium-risk equipment areas are limited to no more than h% of the unit's rated capacity. At the energy storage system control level, when the load forecast increases, the discharge mode is activated to supplement the power shortage, and when the load forecast decreases, the charging mode is switched to absorb the excess energy. The charging and discharging power adopts the maximum capacity operation mode in the low-risk area of ​​the equipment, and is derated to H3% of the rated power in the medium- and high-risk areas of the equipment. At the load management level, the interruptible load management procedure is activated only when the conditions of high demand risk, low equipment risk and forecast error of less than H4% for three consecutive time periods are met simultaneously. Commercial loads are interrupted first, while critical loads are kept powered.

7. A power grid intelligent dispatching decision system based on multi-source heterogeneous data fusion, applied to the power grid intelligent dispatching decision method based on multi-source heterogeneous data fusion as described in any one of claims 1-6, characterized in that, include: The data acquisition and load forecasting module collects multi-source heterogeneous data, constructs a meteorological feature vector set, predicts power grid load changes, and outputs load forecast values. The risk entropy generation module integrates load forecast values ​​and equipment health coefficients to construct a dual-risk collaborative analysis matrix, generate risk entropy values, and map them to standard risk patterns. The intelligent dispatching decision module performs generator output pre-adjustment, energy storage charging and discharging control, and interruptible load management in a tiered manner according to the standard risk model, and coordinates microgrid resources through blockchain.

Images