Multi-scale intelligent prediction method and system for regional power grid load

Abstract

The application belongs to the technical field of smart grid, and specifically provides a regional power grid load multi-scale intelligent prediction method and system. The method mainly integrates multi-scale data such as activity plan, historical load, real-time passenger flow, regional signaling and traffic hub passenger flow, constructs a dynamic feedback load prediction method, deduces the future load change trend based on real-time passenger flow, and then automatically dynamically corrects the baseline prediction when the predicted load exceeds the safety threshold, and finally generates a precise load prediction result coordinated with the real-time passenger flow dynamics. The application effectively solves the problem that the power resources arranged in advance according to the activity plan time in the prior art are difficult to match in time with the load peak generated by the dynamic change of passenger flow, realizes more accurate prediction of the load peak period of the large activity area, and improves the timeliness and rationality of power resource allocation.

Description

Technical Field

[0001] This application belongs to the field of smart grid technology, and specifically relates to a multi-scale intelligent prediction method and system for regional power grid load. Background Technology

[0002] In regional power grid load forecasting, the current main basis for forming conventional load forecasts is historical electricity consumption data. When it is necessary to forecast the impact of large-scale events, power grid dispatchers may set a key period of attention before and after the planned start and end times of the event, based on the event's schedule, and make advance arrangements for power supply during that period by mainly referring to the electricity consumption records of similar events in the past.

[0003] However, when large-scale events are actually held, the specific times when people arrive at and leave the event site are directly affected by real-time traffic conditions, adjustments to the event schedule, and on-site dispersal arrangements. This means that the actual peak times for crowd flow and corresponding peak electricity consumption may deviate from the event plan time based on by dispatchers. Therefore, the power resources pre-allocated according to the plan time may not match the actual peak load time, resulting in insufficient power supply during the actual peak time or the failure to fully utilize the resources reserved during the planned time, thus affecting the rationality of power grid dispatch. Summary of the Invention

[0004] This application provides a multi-scale intelligent prediction method and system for regional power grid load, which effectively solves the problem in the prior art that the power resources pre-arranged according to the event plan time are difficult to match with the actual load peak caused by the dynamic changes in the flow of people. It realizes more accurate prediction of the load peak period in large event areas, enabling the power grid dispatch to make decisions based on the load curve that is more synchronized with the actual flow of people, thereby improving the timeliness and rationality of power resource allocation.

[0005] To achieve the above objectives, this application adopts the following technical solution:

[0006] Firstly, this application provides a multi-scale intelligent forecasting method for regional power grid load, comprising: acquiring raw data at the event scale, including the event planning time and historical typical power grid load start time of the target area; acquiring raw data at the short-term scale, including the future baseline load forecast curve of the target area, to obtain a first forecast curve; acquiring raw data at the real-time scale, including real-time active outbound passenger flow, real-time regional signaling density data, and real-time transportation hub passenger flow; comparing the event planning time with the historical typical load start time to generate a risk time window; fusing and trajectory extrapolating the real-time active outbound passenger flow, real-time regional signaling density data, and real-time transportation hub passenger flow to generate a time-series distribution of passenger arrival intensity in the future period; when the current time step is within the risk time window, superimposing the time-series distribution of passenger arrival intensity and the power grid load value of the first forecast curve in the risk time window to generate a coupled load curve; when the peak value of the coupled load curve continuously exceeds a preset safety threshold, correcting the load value and change trend of the first forecast curve in the risk time window according to the time-series distribution of passenger arrival intensity to generate a second forecast curve; and fusing the second forecast curve with the time-series distribution of passenger arrival intensity to obtain a coordinated load forecasting result.

[0007] Further, the process of obtaining the future baseline load forecast curve for the target area to obtain the first forecast curve includes: obtaining historical load data of the power grid in the target area within a preset historical period; calculating the load forecast value for a specified future period based on the historical load data of the power grid using a time series forecasting algorithm; and generating the first forecast curve based on all load forecast values.

[0008] Furthermore, the planned event time is compared with the historical typical load start time to generate a risk time window, including: calculating the time difference between the planned event time and the historical typical grid load start time; and dividing the time interval around the planned event time. As a risk window Represents the planned time of the event. This represents the time difference.

[0009] Furthermore, by fusing and trajectory extrapolating real-time outbound passenger flow, real-time regional signaling density data, and real-time transportation hub passenger flow, a temporal distribution of passenger arrival intensity for future periods is generated. This includes: gridding the real-time regional signaling density data and calculating its spatial gradient to obtain passenger flow direction field data; performing differential calculations on the real-time transportation hub passenger flow at adjacent time steps to obtain the instantaneous rate of change of passenger flow; generating an initial spatial distribution of passenger flow based on the real-time regional signaling density data, the instantaneous rate of change of passenger flow, and real-time outbound passenger flow; and calculating the predicted spatial distribution of passenger flow based on the initial spatial distribution of passenger flow and the passenger flow direction field data. , and Representing the first and At each time step in the grid cell The predicted number of people, Representing the Predicted pedestrian flow at each time step Represents the time step. and The field data representing the direction of pedestrian movement are respectively in and Components in direction, Based on the predicted spatial distribution of pedestrian flow, calculate the estimated total number of people in the target area at each time step, and generate the time-series distribution of pedestrian arrival intensity in future time periods. , Representing the The estimated total number of people in the target area at each time step, where the target area is the region centered on the target power grid load area. .

[0010] Furthermore, based on the real-time regional signaling density data, the instantaneous rate of change in passenger flow, and the real-time active outbound passenger flow, an initial spatial distribution of passenger flow is generated, including: calculating the normalized personnel density value of each spatial grid unit based on the real-time regional signaling density data; and calculating the initial base value of the personnel distribution for each spatial grid unit based on the personnel density value, the instantaneous rate of change in passenger flow, and the real-time active outbound passenger flow. ,and , Represents in grid cells The initial population distribution base value, Represents real-time passenger flow exiting the station. Represents grid cell The normalized population density value, This represents the sum of the normalized personnel density values ​​for all grid cells. Represents the instantaneous rate of change in passenger flow. This represents a preset baseline constant for passenger flow; the initial passenger flow spatial distribution is obtained by normalizing the baseline value of the initial passenger flow distribution. , Represents the initial spatial distribution of pedestrian flow. This represents the sum of the initial population distribution base values ​​for all grid cells.

[0011] Furthermore, the time-series distribution of passenger arrival intensity and the grid load value of the first prediction curve within the risk time window are superimposed to generate a coupled load curve. This includes: converting the estimated number of arrivals at each time step into an equivalent load increment based on the time-series distribution of passenger arrival intensity; extracting the baseline load value for each time step within the risk time window from the first prediction curve; calculating the correction factor for each time step within the risk time window based on the equivalent load increment, the baseline load value, and the instantaneous rate of change of passenger flow; and adjusting the equivalent load increment based on the correction factor. , This represents the adjusted equivalent load increment. This represents the equivalent load increment before adjustment. The equivalent load increment is algebraically added to the baseline load value at the same time step to obtain the coupled load curve.

[0012] Furthermore, based on the equivalent load increment, baseline load value, and instantaneous passenger flow change rate, correction factors are calculated for each time step within the risk time window. This includes: calculating the average baseline load value within the risk event window to obtain the average load value; calculating the ratio of the equivalent load increment to the average load value within the risk time window to obtain the passenger flow impact intensity ratio; and calculating intermediate factors based on the passenger flow impact intensity ratio, instantaneous passenger flow change rate, and average load value. , , Represents the proportion of the intensity of the impact of pedestrian flow. Represents the average load value. This represents the preset load threshold. The representative is used to prevent the denominator from being 0. The smallest positive number, Represents the instantaneous rate of change in passenger flow; for intermediate factors Amplitude limiting is performed to obtain the correction factor.

[0013] Furthermore, the load values ​​and trends of the first forecast curve within the risk time window are corrected to generate a second forecast curve. This includes: extracting the baseline load values ​​for each time step within the risk time window from the first forecast curve; obtaining the estimated number of arrivals for each time step from the time-series distribution of pedestrian arrival intensity; and calculating the load correction amount for each time step. , The load correction amount represents the time step t. This represents the estimated number of arrivals at time step t. The preset per capita energy consumption coefficient is represented; the baseline load value at each time step is added to the corresponding load correction amount to obtain the corrected load value, and a second prediction curve is generated based on all the corrected load values.

[0014] Furthermore, the second prediction curve is fused with the time-series distribution of passenger arrival intensity to obtain the coordinated load prediction result, including: smoothing the second prediction curve to obtain the trend load curve; calculating the fluctuation component of the estimated arrival number at each time step within the risk time window. , Representing the Each time step fluctuation component The average number of people expected to arrive is represented by the number of people arriving at all time steps within the risk time window. The trend load curve and the fluctuation component are added together at the same time step to obtain the final coordinated load forecast result.

[0015] Secondly, this application provides a regional power grid load multi-scale intelligent forecasting system, comprising:

[0016] Data acquisition module: used to acquire raw data at the event scale, including the event planning time and the historical typical power grid load start time of the target area; acquire raw data at the short-term scale, including the future baseline load forecast curve of the target area to obtain the first forecast curve; acquire raw data at the real-time scale, including real-time active outbound passenger flow, real-time regional signaling density data and real-time transportation hub passenger flow.

[0017] Risk assessment and passenger flow prediction module: It is used to compare the planned time of the event with the historical typical load start time to generate a risk time window; it integrates and extrapolates the trajectory based on real-time activity exit passenger flow, real-time regional signaling density data and real-time transportation hub passenger flow to generate the time series distribution of passenger flow arrival intensity in the future period.

[0018] Load coupling module: When the current time step is within the risk time window, it is used to superimpose the temporal distribution of the arrival intensity of the population and the grid load value of the first prediction curve in the risk time window to generate a coupled load curve.

[0019] Load correction module: When the peak value of the coupled load curve continuously exceeds the preset safety threshold, it corrects the load value and trend of the first prediction curve within the risk time window based on the temporal distribution of the arrival intensity of people, and generates a second prediction curve.

[0020] The results fusion module is used to fuse the second prediction curve with the time-series distribution of pedestrian arrival intensity to obtain the collaborative load prediction results.

[0021] Thirdly, a readable storage medium includes: computer program instructions stored in the readable storage medium, wherein the computer program instructions are read and executed by a processor to perform the steps of a regional power grid load multi-scale intelligent forecasting method.

[0022] The beneficial effects of this application are:

[0023] This application acquires data at three scales: event-based, short-term, and real-time. It determines risk windows based on event plans and historical information, and dynamically extrapolates future passenger flow intensity by integrating real-time passenger flow, signaling, and traffic data. This generates and dynamically corrects load forecast curves based on safety thresholds, effectively solving the problem in existing technologies where power resources pre-arranged according to event plans are difficult to match with actual load peaks caused by dynamic changes in passenger flow. This enables more accurate prediction of peak load periods in large event areas, allowing power grid dispatch to make decisions based on load curves that are more synchronized with actual passenger flow, thereby improving the timeliness and rationality of power resource allocation.

[0024] Other features and advantages of this application will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures pointed out in the description and the accompanying drawings. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of this application 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 some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0026] Figure 1 A flowchart illustrating the multi-scale intelligent forecasting method for regional power grid load in this application is shown.

[0027] Figure 2 This application illustrates a flowchart of the process for generating the temporal distribution of pedestrian arrival intensity over a future time period.

[0028] Figure 3 A schematic diagram of the process for generating the initial spatial distribution of pedestrian flow in this application is shown;

[0029] Figure 4 A schematic diagram of the process for generating the coupling load curve in this application is shown. Detailed Implementation

[0030] To address the problems raised in the background technology, this application constructs a dynamic feedback load forecasting method by integrating multi-scale data such as event plans, historical load, real-time passenger flow, regional signaling, and passenger flow at transportation hubs. Based on real-time passenger flow dynamics, it extrapolates future load change trends and automatically and dynamically corrects the baseline forecast when the predicted load exceeds a safe threshold. Ultimately, it generates accurate load forecasting results that are coordinated with real-time passenger flow dynamics, effectively addressing load peaks caused by the dynamic gathering and dispersal of crowds during large-scale events, and improving the timeliness and accuracy of power grid dispatching decisions.

[0031] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0032] In some embodiments, such as Figure 1 As shown, this application provides a multi-scale intelligent forecasting method for regional power grid load, including:

[0033] S1. Obtain raw data at the event scale, including the event planning time and the historical typical power grid load start time of the target area; obtain raw data at the short-term scale, including the future baseline load forecast curve of the target area to obtain the first forecast curve; obtain raw data at the real-time scale, including real-time active outbound passenger flow, real-time regional signaling density data and real-time transportation hub passenger flow.

[0034] The target area represents the specific geographical range where the power grid load needs to be predicted, such as the power supply area surrounding a large event venue. The event planning time represents the planned time of the large event. The historical typical power grid load start time represents the typical time point, derived from historical data analysis, when similar events cause significant changes in the power grid load of the target area.

[0035] The future baseline load forecast curve represents the conventional load forecast of the power system based on historical loads, and is used as the first forecast curve.

[0036] Real-time event exit passenger flow represents the sequence of people leaving the event venue in real time. Real-time regional signaling density data represents gridded data reflecting the relative population density of various locations within the target area, obtained through mobile communication network signaling. Real-time transportation hub passenger flow represents the sequence of people entering and leaving key transportation nodes within the target area in real time, such as subway stations and bus terminals.

[0037] S2. Compare the planned event time with the historical typical load start time to generate a risk time window. The risk time window is used to identify key monitoring periods where the power grid load may face high risks due to the impact of future activities and passenger flow. Based on the fusion and trajectory extrapolation of real-time activity outbound passenger flow, real-time regional signaling density data and real-time transportation hub passenger flow, predict how the passenger flow will arrive at the target area in the future period and generate the time series distribution of passenger flow arrival intensity in the future period. Each data point in the time series distribution of passenger flow arrival intensity represents the estimated total number of people arriving at the target area in a specific time step in the future.

[0038] S3. When the current time step is within the risk time window, the temporal distribution of the arrival intensity of the people and the power grid load value of the first prediction curve in the risk time window are superimposed to generate a coupled load curve.

[0039] Specifically, the values ​​of the temporal distribution of the arrival intensity of people within the risk window are converted into the corresponding equivalent power load increments through a preset conversion relationship. The load increment sequence is then algebraically superimposed point by point with the baseline load values ​​of the first prediction curve at the corresponding time points within the risk time window to generate a coupled load curve. The coupled load curve reflects the predicted load after superimposing the preliminary estimated impact of the people flow load.

[0040] S4. When the peak value of the coupled load curve continues to exceed the preset safety threshold, the load value and trend of the first prediction curve within the risk time window are corrected according to the temporal distribution of the arrival intensity of the flow of people, and a second prediction curve is generated.

[0041] The safety threshold is a preset warning load value, which reflects the critical level at which the load enters a higher risk range. For example, it can be the load value at the 90th percentile after sorting the historical daily peak load data of the region from smallest to largest.

[0042] When the peak value of the coupled load curve exceeds the preset safety threshold for multiple time steps, it indicates that the first prediction curve is insufficient to reflect the real risk. It is necessary to dynamically adjust and correct the load value and trend of the first prediction curve within the risk time window based on the temporal distribution of the arrival intensity of people, so as to generate a second prediction curve with higher accuracy.

[0043] S5. The second prediction curve is fused with the time-series distribution of passenger arrival intensity to obtain the coordinated load prediction result, which provides support for power grid dispatching decisions.

[0044] In some embodiments, obtaining the future baseline load forecast curve for the target area to obtain a first forecast curve includes:

[0045] S11. Obtain historical power grid load data for the target area within a preset historical period.

[0046] Specifically, historical power grid load data with a fixed time resolution is extracted from the target area within a preset historical period. This historical load data records the historical electricity consumption behavior of the target area. The historical period is, for example, the past 7 days or 30 days, and the time resolution is, for example, 15 minutes or 1 hour.

[0047] S12. Based on historical load data of the power grid, calculate the load forecast value for a specified future period using a time series forecast algorithm, and generate a first forecast curve based on all load forecast values.

[0048] Time series forecasting algorithms can be based on machine learning-based long short-term memory network models. For example, historical load time windows can be prepared as input to an LSTM model, and the model outputs load forecast values ​​for one or more future time points. The trained model parameters are used for forward calculation to obtain load forecast values ​​for all time steps within a specified future period. All load forecast values ​​are connected in chronological order to form a continuous first forecast curve. The first forecast curve represents the baseline load expectation for the target area in the future without considering current large-scale event disturbances.

[0049] In some embodiments, the event schedule time is compared with the historical typical load start time to generate a risk time window, including:

[0050] Sa21. Calculate the time difference between the planned event time and the historical typical power grid load start time.

[0051] Retrieve relevant data from historical records of multiple similar large-scale events that occurred in the target area over the past few years. For each historical event, analyze the power grid load curve on the day of the event and define an objective criterion to determine the actual load start time. For example, the moment when the daily power grid load curve first reaches 50% of the daily peak is determined as the actual load start time of this event. Calculate the arithmetic mean of the actual load start times of all historical events and define this mean as the typical historical power grid load start time.

[0052] Sa22. The risk time window aims to cover periods of potentially abnormal load fluctuations, determined by combining the event schedule time with historical offset patterns. For example, it can be defined as a time interval centered on the event schedule time. As a risk window Represents the planned time of the event. This represents the time difference.

[0053] In some embodiments, such as Figure 2 As shown, by fusing and extrapolating real-time outbound passenger flow, real-time regional signaling density data, and real-time transportation hub passenger flow, a time-series distribution of passenger arrival intensity in future periods is generated, including:

[0054] Sb21. The real-time regional signaling density data is gridded and the spatial gradient is calculated to obtain the field data of pedestrian movement direction; the passenger flow of the real-time transportation hub is differentially calculated between adjacent time steps to obtain the instantaneous change rate of passenger flow.

[0055] Specifically, the geographical area of ​​the target region is divided into multiple equally spaced square grids on a horizontal plane, with each square region serving as a grid unit, and indexed by row. and column indexes Unique identifier; real-time regional signaling density data represents each grid cell statistically obtained under this grid system. The signaling strength or number of users within the range is denoted as .

[0056] Based on the east-west direction The axis is in the north-south direction. The axis is calculated using the central difference method for each mesh cell. exist Rate of change of density in the axial direction and Rate of change of density in the axial direction Obtain the field data of pedestrian movement direction. The direction field data of human movement represents the direction and trend intensity of the macroscopic movement of people at various spatial locations.

[0057] Time series of passenger flow at real-time transportation hubs Perform differential calculations to determine the difference in passenger flow between adjacent time steps, and obtain the instantaneous rate of change in passenger flow. , The instantaneous rate of change of passenger flow reflects the instantaneous rate and direction of change of passenger flow and is used to measure the changes in the busyness of transportation hubs.

[0058] Sb22. Based on real-time regional signaling density data, instantaneous passenger flow change rate, and real-time active outbound passenger flow, generate an initial spatial distribution of passenger flow and estimate the spatial distribution of each grid cell within the target area of ​​the simulation. The initial population distribution is used to generate the initial spatial distribution of the population flow.

[0059] Sb23. Based on the initial spatial distribution and direction field data of pedestrian flow, the numerical discretization scheme of the advection equation is used for iterative solution to simulate the movement and diffusion of the crowd over time, and to calculate and predict the spatial distribution of pedestrian flow. For the ... Each time step, grid cell Number of people on Estimation can be performed using an explicit finite difference scheme: , and Representing the first and At each time step in the grid cell The predicted number of people, Representing the Predicted pedestrian flow at each time step Represents the time step. and The field data representing the direction of pedestrian movement are respectively in and Components in direction, .

[0060] Sb24. Define a region centered on the target power grid load area. Based on the predicted spatial distribution of pedestrian flow, the estimated total number of people within the target area at each time step is calculated, and a time-series distribution of pedestrian arrival intensity for future periods is generated. , , Representing the The estimated total number of people in the target area at each time step.

[0061] In some embodiments, such as Figure 3 As shown, based on real-time regional signaling density data, instantaneous passenger flow change rate, and real-time active exit passenger flow, an initial spatial distribution of passenger flow is generated, including:

[0062] Sb221. Calculate the normalized personnel density value for each spatial grid cell based on real-time regional signaling density data.

[0063] Calculate all grid cells The sum of values Calculate the normalized population density value for each grid cell. , The population density value represents the relative proportion of existing population density at each grid point at the current moment. This represents the sum of the personnel density values ​​for all grid cells.

[0064] Sb222. Based on the personnel density value, the instantaneous rate of change of passenger flow, and the real-time active outbound passenger flow, calculate the initial base value of the personnel distribution for each spatial grid cell. , Represents in grid cells The initial population distribution base value, Represents real-time passenger flow exiting the station. This represents a preset passenger flow baseline constant, used to measure the instantaneous rate of change in passenger flow. Whether it is significant, for example, the median of passenger flow data for the target transportation hub during the same period in history can be used as the passenger flow benchmark constant. To ensure Non-negative, for Perform truncation: .

[0065] Sb223. ​​Normalize the initial population distribution baseline values ​​to obtain the initial spatial distribution of population flow. , Representing the initial spatial distribution of pedestrian flow, that is, at the start of the simulation, the grid cells The estimated number of people on the site This represents the sum of the initial population distribution base values ​​for all grid cells.

[0066] In some embodiments, such as Figure 4 As shown, the temporal distribution of pedestrian arrival intensity and the grid load value of the first prediction curve within the risk time window are superimposed to generate a coupled load curve, including:

[0067] S31. Based on the temporal distribution of pedestrian arrival intensity The estimated arrivals at each time step are converted into equivalent load increments. , , The pre-defined per capita energy consumption coefficient can be determined by statistically analyzing the per capita energy consumption within the target range, such as... kW / person; extract baseline load values ​​for each time step within the risk time window from the first forecast curve. .

[0068] S32. Calculate the correction factor for each time step within the risk time window based on the equivalent load increment, baseline load value, and instantaneous passenger flow change rate.

[0069] S33. Adjust the equivalent load increment according to the correction factor. , This represents the adjusted equivalent load increment. This represents the equivalent load increment before adjustment. This represents the correction factor.

[0070] S34. The equivalent load increment and the baseline load value are algebraically added at the same time step to obtain the coupled load curve.

[0071] In some embodiments, the correction factor for each time step within the risk time window is calculated based on the equivalent load increment, baseline load value, and instantaneous passenger flow change rate, including:

[0072] S321. Calculate baseline load values The average load value is obtained by averaging the values ​​within the risk event window. .

[0073] S322. For each time step within the risk time window Calculate the equivalent load increment relative to average load value The proportion of pedestrian impact intensity is obtained from the ratio. .

[0074] S323. Based on the proportion of pedestrian impact intensity Instantaneous change rate of passenger flow and average load value Calculate intermediate factors , , This represents a preset load threshold used to adjust the influence of the load baseline. For example, it can be set to the sum of the historical average load during the same period on inactive days and twice the standard deviation. The representative is used to prevent the denominator from being 0. The smallest positive number, such as can be .

[0075] For intermediate factors Amplitude limiting is then performed to obtain the final correction factor. , , and These represent the preset lower and upper limits of the correction factor, respectively, and can be taken as follows: , .

[0076] In some embodiments, the load values ​​and trends of the first forecast curve within the risk time window are corrected to generate a second forecast curve, including:

[0077] S41. Extract the baseline load values ​​for each time step within the risk time window from the first forecast curve; obtain the estimated number of arrivals for each time step from the time-series distribution of pedestrian arrival intensity, and calculate the load correction for each time step. , This represents the load correction amount at time step t.

[0078] S42. Add the baseline load value at each time step to the corresponding load correction amount to obtain the corrected load value. , The result is obtained by calculating all time steps within the risk time window. Connect them in chronological order to form the second prediction curve.

[0079] In some embodiments, the second prediction curve is fused with the time-series distribution of pedestrian arrival intensity to obtain a coordinated load prediction result, including:

[0080] S51. Smooth the second forecast curve, such as by using a moving average method, to obtain the trend load curve. ; Calculate the fluctuation component of the estimated arrivals at each time step within the risk time window. , , Representing the Each time-step fluctuation component reflects the degree of deviation of the pedestrian flow intensity at that time step from the average level within the window. This represents the average number of estimated arrivals across all time steps within the risk time window.

[0081] S52. Trend Load Curve With fluctuation components The results are added together at the same time step to obtain the final coordinated load forecast. , All Connecting them in chronological order yields the final coordinated load forecast result.

[0082] In some embodiments, this application provides a regional power grid load multi-scale intelligent forecasting system, comprising:

[0083] Data acquisition module: used to acquire raw data at the event scale, including the event planning time and the historical typical power grid load start time of the target area; acquire raw data at the short-term scale, including the future baseline load forecast curve of the target area to obtain the first forecast curve; acquire raw data at the real-time scale, including real-time active outbound passenger flow, real-time regional signaling density data and real-time transportation hub passenger flow.

[0084] Risk assessment and passenger flow prediction module: It is used to compare the planned time of the event with the historical typical load start time to generate a risk time window; it integrates and extrapolates the trajectory based on real-time activity exit passenger flow, real-time regional signaling density data and real-time transportation hub passenger flow to generate the time series distribution of passenger flow arrival intensity in the future period.

[0085] Load coupling module: When the current time step is within the risk time window, it is used to superimpose the temporal distribution of the arrival intensity of the population and the grid load value of the first prediction curve in the risk time window to generate a coupled load curve.

[0086] Load correction module: When the peak value of the coupled load curve continuously exceeds the preset safety threshold, it corrects the load value and trend of the first prediction curve within the risk time window based on the temporal distribution of the arrival intensity of people, and generates a second prediction curve.

[0087] The results fusion module is used to fuse the second prediction curve with the time-series distribution of pedestrian arrival intensity to obtain the collaborative load prediction results.

[0088] In some embodiments, this application provides a readable storage medium, including: computer program instructions stored in the readable storage medium, wherein the computer program instructions are read and executed by a processor to perform the steps of a regional power grid load multi-scale intelligent forecasting method.

[0089] It should be noted that, in this application, 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 a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0090] Any references to memory, storage, database, or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory.

[0091] Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A multi-scale intelligent forecasting method for regional power grid load, characterized in that, include: Acquire raw data at the event scale, including event planning times and historical typical grid load start times for the target area; Obtain raw data at the short-term scale, including the future baseline load forecast curve for the target area, to obtain the first forecast curve; obtain raw data at the real-time scale, including real-time outbound passenger flow, real-time regional signaling density data, and real-time transportation hub passenger flow. The event's planned time is compared with the historical typical load start time to generate a risk time window; the real-time activity outbound passenger flow, real-time regional signaling density data, and real-time transportation hub passenger flow are fused and trajectory extrapolated to generate the temporal distribution of passenger flow arrival intensity in the future period. When the current time step is within the risk time window, the temporal distribution of the arrival intensity of the people and the power grid load value of the first prediction curve in the risk time window are superimposed to generate a coupled load curve; When the peak value of the coupled load curve continues to exceed the preset safety threshold, the load value and trend of the first prediction curve within the risk time window are corrected according to the temporal distribution of the arrival intensity of the crowd, and a second prediction curve is generated. The second prediction curve is fused with the time series distribution of pedestrian arrival intensity to obtain the coordinated load prediction result.

2. The method according to claim 1, characterized in that, Obtain the future baseline load forecast curve for the target area to obtain the first forecast curve, including: Obtain historical power grid load data for the target area within a preset historical period; Based on the historical load data of the power grid, the load forecast value for a specified future period is calculated using a time series forecast algorithm, and a first forecast curve is generated based on all load forecast values.

3. The method according to claim 1, characterized in that, The planned event time is compared with the historical typical load start time to generate a risk time window, including: Calculate the time difference between the planned time of the event and the start time of a typical historical power grid load; Divide the time intervals around the planned event time. As a risk window Represents the planned time of the event. This represents the time difference.

4. The method according to claim 1, characterized in that, Based on the fusion and trajectory extrapolation of real-time outbound passenger flow, real-time regional signaling density data, and real-time transportation hub passenger flow, a time-series distribution of passenger arrival intensity for future periods is generated, including: The real-time regional signaling density data is gridded and the spatial gradient is calculated to obtain the pedestrian flow direction field data; the passenger flow of the real-time transportation hub is differentially calculated between adjacent time steps to obtain the instantaneous change rate of passenger flow. Based on the real-time regional signaling density data, the instantaneous change rate of passenger flow, and the real-time active outbound passenger flow, an initial spatial distribution of passenger flow is generated; Based on the initial pedestrian spatial distribution and pedestrian movement direction field data, the predicted pedestrian spatial distribution is calculated. , and Representing the first and At each time step in the grid cell The predicted number of people, Representing the Predicted pedestrian flow at each time step Represents the time step. and The field data representing the direction of pedestrian movement are respectively in and Components in direction, ; Based on the predicted spatial distribution of pedestrian flow, the estimated total number of people in the target area at each time step is calculated, and the time series distribution of pedestrian arrival intensity in future time periods is generated. , Representing the The estimated total number of people in the target area at each time step, where the target area is the region centered on the target power grid load area. .

5. The method according to claim 4, characterized in that, Based on the real-time regional signaling density data, the instantaneous change rate of passenger flow, and the real-time active outbound passenger flow, an initial spatial distribution of passenger flow is generated, including: Based on the real-time regional signaling density data, calculate the normalized personnel density value of each spatial grid cell; Based on the aforementioned personnel density value, instantaneous passenger flow change rate, and real-time outbound passenger flow, calculate the initial population distribution baseline value for each spatial grid cell. ,and , Represents in grid cells The initial population distribution base value, Represents real-time passenger flow exiting the station. Represents grid cell The normalized population density value, This represents the sum of the normalized personnel density values ​​for all grid cells. Represents the instantaneous rate of change in passenger flow. Represents the preset passenger flow baseline constant; The initial population distribution baseline value is normalized to obtain the initial spatial distribution of population flow. , Represents the initial spatial distribution of pedestrian flow. This represents the sum of the initial population distribution base values ​​for all grid cells.

6. The method according to claim 1, characterized in that, The coupled load curve is generated by superimposing the temporal distribution of the arrival intensity of the population and the grid load value of the first prediction curve within the risk time window, including: Based on the temporal distribution of the arrival intensity of the people, the estimated number of arrivals at each time step is converted into an equivalent load increment; the baseline load value at each time step within the risk time window is extracted from the first prediction curve. Based on the equivalent load increment, baseline load value, and instantaneous passenger flow change rate, calculate the correction factor for each time step within the risk time window; The equivalent load increment is adjusted according to the correction factor. , This represents the adjusted equivalent load increment. This represents the equivalent load increment before adjustment. Represents the correction factor; The equivalent load increment and the baseline load value are algebraically added at the same time step to obtain the coupled load curve.

7. The method according to claim 6, characterized in that, Based on the equivalent load increment, baseline load value, and instantaneous passenger flow change rate, the correction factor for each time step within the risk time window is calculated, including: Calculate the average value of the baseline load value within the risk event window to obtain the average load value; The proportion of the equivalent load increment relative to the average load value within the risk time window is calculated to obtain the proportion of pedestrian impact intensity. Based on the aforementioned proportion of pedestrian impact intensity, instantaneous change rate of passenger flow, and average load value, the intermediate factor is calculated. , , Represents the proportion of the intensity of the impact of pedestrian flow. Represents the average load value. This represents the preset load threshold. The representative is used to prevent the denominator from being 0. The smallest positive number, Represents the instantaneous rate of change in passenger flow; For the intermediate factor Amplitude limiting is performed to obtain the correction factor.

8. The method according to claim 1, characterized in that, The load values ​​and trends of the first forecast curve within the risk time window are corrected to generate a second forecast curve, including: The baseline load values ​​for each time step within the risk time window are extracted from the first prediction curve; the estimated number of arrivals for each time step is obtained from the time-series distribution of pedestrian arrival intensity, and the load correction amount for each time step is calculated. , The load correction amount represents the time step t. This represents the estimated number of arrivals at time step t. This represents the preset per capita energy consumption coefficient; The baseline load value at each time step is added to the corresponding load correction amount to obtain the corrected load value. A second prediction curve is then generated based on all the corrected load values.

9. The method according to claim 1, characterized in that, The second prediction curve is fused with the time-series distribution of pedestrian arrival intensity to obtain the coordinated load prediction results, including: The second forecast curve is smoothed to obtain the trend load curve; the fluctuation component of the estimated arrivals at each time step within the risk time window is calculated. , Representing the Each time step fluctuation component This represents the average number of estimated arrivals across all time steps within the risk time window; The trend load curve and the fluctuation component are added together at the same time step to obtain the final coordinated load prediction result.

10. A regional power grid load multi-scale intelligent forecasting system, characterized in that, include: Data acquisition module: Used to acquire raw data at the event scale, including the event schedule time and historical typical power grid load start time for the target area; Obtain raw data at the short-term scale, including the future baseline load forecast curve for the target area, to obtain the first forecast curve; obtain raw data at the real-time scale, including real-time outbound passenger flow, real-time regional signaling density data, and real-time transportation hub passenger flow. Risk assessment and crowd forecasting module: used to compare the planned event time with the historical typical load start time to generate a risk time window; Based on the real-time activity exit passenger flow, real-time regional signaling density data and real-time transportation hub passenger flow, the time series distribution of passenger flow arrival intensity in future time periods is generated by fusion and trajectory extrapolation. Load coupling module: used to superimpose the temporal distribution of the arrival intensity of the people and the power grid load value of the first prediction curve in the risk time window when the current time step is in the risk time window, to generate a coupled load curve; Load correction module: When the peak value of the coupled load curve continuously exceeds the preset safety threshold, it corrects the load value and trend of the first prediction curve within the risk time window according to the temporal distribution of the arrival intensity of the crowd, and generates a second prediction curve. The result fusion module is used to fuse the second prediction curve with the time-series distribution of pedestrian arrival intensity to obtain the collaborative load prediction result.

Images