A method and system for generating a heat map of charging demand based on time series prediction

Abstract

The present application provides a method and system for generating a heat map of charging demand based on time series prediction, which relates to the technical field of heat map generation, and includes: establishing a mapping relationship between the network traffic fluctuation law and the time series characteristics of charging demand, and constructing a cross-domain correlation model by analyzing the periodic change law of data transmission volume; integrating piezoelectric vibration sensors to identify specific vibration modes caused by the plugging and unplugging actions of charging guns, and distinguishing the actual load operation state of charging piles from no-load false trigger events; deploying reactive power compensation devices on the distribution side to suppress the fluctuation of the power factor of the power grid and generate a steady-state power supply environment; performing spatio-temporal association on the output result of the cross-domain correlation model and the recognition result of the actual load operation state of the charging piles, and calculating the demand distribution density of the charging pile group; synchronously fusing the characteristic parameters of the specific vibration mode and the exclusion result of the no-load false trigger event to generate a multi-dimensional heat map. The technical solution provided by the present application improves the accuracy of charging demand prediction and heat map generation.

Description

Technical Field

[0001] This application relates to the technical field of heat map generation, and particularly to a method and system for generating a charging demand heat map based on time series prediction. Background Art

[0002] With the continuous increase in the penetration rate of electric vehicles in smart campuses, the charging demand shows characteristics of high variability and spatio-temporal heterogeneity, specifically manifested as significant concentration of charging demand periods, large fluctuations in equipment utilization rates, and obvious differences in peak and valley of power grid loads. To ensure the efficient operation of charging facilities and the stable operation of the power grid, it is necessary to combine multi-dimensional data such as user behavior patterns, campus traffic flow, and weather conditions to predict the spatio-temporal distribution of charging demand in order to optimize the charging resource scheduling; it is necessary to integrate heterogeneous data sources such as the operating status of charging piles, user reservation data, and power grid load parameters, break through the limitations of traditional single data models, and improve the prediction accuracy; it is necessary to balance the relationship between charging demand and the power grid's carrying capacity, avoid local overload or resource waste, and ensure that the prediction results match the power grid status.

[0003] Currently, a mainstream solution is a deep learning model based on spatio-temporal joint prediction, such as a prediction framework that combines long short-term memory network (LSTM) and convolutional neural network (CNN). This solution analyzes the time series characteristics of historical charging data and campus traffic flow, constructs a multi-dimensional input vector, uses LSTM to capture time dependence, CNN to extract spatial distribution features, and assigns prediction weights to different regions through an attention mechanism. For example, some studies generate demand distribution weights by introducing the mapping rules between user charging behavior patterns and power grid load periods, and then correct the prediction results by combining power grid load data. Such methods have been pilot applied in some smart campuses and can partially achieve the spatio-temporal distribution prediction of charging demand.

[0004] Although the above solutions have improved in prediction accuracy, their core problem lies in the lack of a closed-loop collaborative mechanism for physics, power grid, and user behavior. Existing models usually regard user behavior data, power grid parameters, and device status as independent inputs and do not establish a coupling relationship, resulting in prediction weights relying on static experience allocation and being difficult to adapt to scenario changes; the ability to identify the operating status of the device physical layer (such as charging pile failure rate, misreporting of plug and unplug events) is not integrated, resulting in prediction results containing noise and affecting the scheduling accuracy; the prediction model is not linked with power grid regulating devices and cannot correct the impact of power grid fluctuations on demand distribution, resulting in a deviation between the prediction results and the actual carrying capacity of the power grid and possibly triggering local overload risks. Summary of the Invention

[0005] This application provides a method and system for generating a charging demand heat map based on time series prediction to solve the problem of insufficient accuracy in charging demand prediction and heat map generation in the prior art.

[0006] In a first aspect, the present application provides a method for generating a heat map of charging demand based on time series prediction, including:

[0007] Establish a mapping relationship between the network traffic fluctuation law and the time series characteristics of charging demand, and analyze the periodic change law of the data transmission volume of communication nodes in the area where the charging pile is located based on the mapping relationship, so as to construct a cross-domain correlation model reflecting the coupling relationship between charging demand and network behavior;

[0008] Integrate a piezoelectric vibration sensor inside the support structure of the charging pile. Based on the amplitude difference of different frequency components in the mechanical vibration signal of the charging pile, identify the specific vibration mode caused by the plugging and unplugging actions of the charging gun, and distinguish the actual load operation state of the charging pile from the no-load false triggering event;

[0009] Deploy a reactive power compensation device on the distribution side. By adjusting the switching state of the compensation capacitor, suppress the power factor fluctuation of the power grid caused by the access of the charging pile cluster, and generate a steady-state power supply environment matching the charging demand prediction model;

[0010] Perform spatio-temporal association on the output result of the cross-domain correlation model and the recognition result of the actual load operation state of the charging pile, and combine the power parameters of the steady-state power supply environment to calculate the demand distribution density of the charging pile group in the target area;

[0011] According to the spatial gradient change characteristics of the demand distribution density, synchronously fuse the characteristic parameters of the specific vibration mode and the exclusion result of the no-load false triggering event to generate a multi-dimensional heat map including the prediction of charging pile availability and load bearing capacity, and the chromaticity mapping rule of the multi-dimensional heat map is negatively correlated with the power factor compensation state of the power grid.

[0012] Optionally, the step of performing spatio-temporal association on the output result of the cross-domain correlation model and the recognition result of the actual load operation state of the charging pile, and combining the power parameters of the steady-state power supply environment to calculate the demand distribution density of the charging pile group in the target area includes:

[0013] Divide the recognition result of the actual load operation state of the charging pile into a discrete event sequence according to a preset time slice, and the load state within each time slice in the discrete event sequence is associated with the peak interval of the network traffic data corresponding to the time slice;

[0014] According to the coupling weight of network traffic and charging demand output by the cross-domain correlation model, weight the load state within each time slice in the discrete event sequence to obtain a weighted result;

[0015] Establish a constraint relationship between the power parameters of the steady-state power supply environment and the maximum instantaneous load capacity of the charging pile, and perform capacity boundary correction on the weighted result based on the constraint relationship to generate the upper limit of the available power capacity of the charging pile node;

[0016] Overlay the upper limit of the available power capacity and the peak interval of the network traffic data in space and time, and calculate the demand distribution density of the charging pile group in the target area through density field calculation. The spatial resolution of the density field is jointly determined by the physical spacing of the charging pile nodes and the response delay time of the reactive power compensation device.

[0017] Optionally, according to the spatial gradient change characteristics of the demand distribution density, synchronously fuse the characteristic parameters of the specific vibration mode and the exclusion result of the no-load mis-trigger event to generate a multi-dimensional heat map including the prediction of charging pile availability and load-carrying capacity, including:

[0018] Discretize the spatial gradient change characteristics of the demand distribution density into multiple local density change rates according to the physical layout of the charging pile group. The direction vector of the local density change rate is associated with the vibration propagation direction of the charging gun plugging and unplugging action;

[0019] Extract the vibration frequency distribution interval in the characteristic parameters of the specific vibration mode that matches the charging gun plugging and unplugging action, and generate a vibration characteristic availability probability representing the availability state of the charging pile based on the energy proportion of each frequency band within the vibration frequency distribution interval;

[0020] According to the exclusion result of the no-load mis-trigger event, count the no-load mis-trigger frequency of the charging pile nodes within a preset time period, and combine the power parameters of the steady-state power supply environment to generate a no-load interference suppression factor reflecting the correction coefficient of the charging pile load-carrying capacity;

[0021] Perform spatial interpolation on the vibration characteristic availability probability and the no-load interference suppression factor according to the physical location of the charging piles, and overlay the vector distribution of the local density change rates to generate a multi-dimensional heat map including the prediction of charging pile availability and load-carrying capacity.

[0022] Optionally, establish a mapping relationship between the network traffic fluctuation law and the time-series characteristics of the charging demand, and construct a cross-domain correlation model reflecting the coupling relationship between the charging demand and the network behavior by analyzing the periodic change law of the data transmission volume of the communication nodes in the area where the charging piles are located, including:

[0023] Divide the network traffic monitoring units according to the physical coverage range of the communication nodes in the area where the charging piles are located, perform multi-scale periodic decomposition on the data transmission volume of the communication nodes based on the communication protocol type, and extract the dominant periodic components of the hourly, daily, and weekly data transmission volumes in each monitoring unit;

[0024] According to the temporal distribution characteristics of the charging gun plugging and unplugging events in the historical operation data of the charging piles, define the high-demand period, stable-demand period, and low-demand period of the charging demand, and establish the mapping rule between the charging demand period label and the amplitude-frequency characteristics of the dominant periodic component;

[0025] For each network traffic monitoring unit, match the phase fluctuation interval of the dominant periodic component with the duration window of the corresponding charging demand period label, and generate the dynamic correlation weight between the network traffic fluctuation rule and the charging demand temporal characteristic by introducing the amplitude-frequency characteristic mapping rule in the matching process;

[0026] Based on the correlation weight, construct the coupling relationship function between the charging demand and the network behavior. The temporal offset of the dominant periodic component of the network traffic in the coupling relationship function is used to correct the cross-domain correlation error, and output the structured cross-domain correlation model.

[0027] Optionally, the spatio-temporal superposition of the available power capacity upper limit and the peak interval of the network traffic data, and the demand distribution density of the charging pile group in the target area is calculated through density field, including:

[0028] Divide the spatial grid unit of the target area according to the physical distance between the charging pile nodes, discretely sample the available power capacity upper limit according to the spatial grid unit, and generate the local power capacity distribution;

[0029] Divide the peak interval of the network traffic data into multiple continuous time windows according to the time dimension, extract the spatial distribution characteristics of the network traffic data in each time window, and generate the network traffic spatial weight corresponding to the spatial grid unit;

[0030] Overlay the local power capacity distribution and the network traffic spatial weight according to the spatial grid unit to generate a density field reflecting the coupling relationship between the power capacity of the charging pile group and the network traffic;

[0031] Perform multi-scale convolution operations on the local power capacity distribution and the network traffic spatial weight according to the spatial grid unit, calculate the density increment of the spatial grid unit in the density field, and output the demand distribution density of the charging pile group in the target area.

[0032] Optionally, the spatial interpolation of the vibration feature availability probability and the no-load interference suppression factor according to the physical location of the charging pile, and the superposition of the vector distribution of the local density change rate to generate a multi-dimensional heat map including the charging pile availability prediction and the load-bearing capacity, including:

[0033] Construct a vibration feature interpolation field based on the directional characteristics of the physical location of the charging pile, and the interpolation weight of the vibration feature interpolation field is adjusted by the energy attenuation of the vibration propagation path in the vibration feature availability probability;

[0034] Construct an no-load interference correction field based on the spatial proximity relationship of the physical location of the charging pile, and the correction weight of the no-load interference correction field is controlled by the response time of the power grid reactive power compensation device in the no-load interference suppression factor;

[0035] Visualize the vibration feature interpolation field and the no-load interference correction field according to the physical location of the charging pile to generate an initial heat distribution;

[0036] Spatially superimpose the vector distribution of the local density change rate and the initial heat distribution to generate a multi-dimensional heat map including the prediction of the charging pile availability and the load-bearing capacity.

[0037] Optionally, a piezoelectric vibration sensor is integrated inside the support structure of the charging pile. Based on the amplitude difference of different frequency components in the mechanical vibration signal of the charging pile, identify the specific vibration mode triggered by the plugging and unplugging action of the charging gun to distinguish the actual load operation state of the charging pile from the no-load false trigger event, including:

[0038] Arrange piezoelectric vibration sensors in the stress concentration area of the charging pile support structure to collect multi-band vibration components in the mechanical vibration signal of the charging pile;

[0039] Extract the characteristic frequency band that matches the plugging and unplugging action of the charging gun from the multi-band vibration components, and generate the vibration energy distribution of the charging pile based on the amplitude change law of the characteristic frequency band;

[0040] According to the comparison result between the peak interval of the vibration energy distribution of the charging pile and the preset vibration energy threshold, identify the specific vibration mode triggered by the plugging and unplugging action of the charging gun;

[0041] Based on the spectral distribution characteristics and energy fluctuation trend of the specific vibration mode, distinguish the actual load operation state of the charging pile from the no-load false trigger event.

[0042] In a second aspect, the present application provides a charging demand heat map generation system based on time series prediction, including:

[0043] An analysis module for establishing a mapping relationship between the network traffic fluctuation law and the time series characteristics of the charging demand, and constructing a cross-domain correlation model reflecting the coupling relationship between the charging demand and the network behavior by analyzing the periodic change law of the data transmission volume of the communication nodes in the area where the charging pile is located;

[0044] An identification module, which is used to integrate piezoelectric vibration sensors inside the charging pile support structure. Based on the amplitude differences of different frequency components in the mechanical vibration signals of the charging pile, it identifies the specific vibration modes caused by the plugging and unplugging actions of the charging gun, and distinguishes the actual load operation state of the charging pile from the no-load false triggering events;

[0045] A suppression module, which is used to deploy reactive power compensation devices on the distribution side. By adjusting the switching states of the compensation capacitors, it suppresses the power factor fluctuations of the power grid caused by the access of the charging pile cluster and generates a steady-state power supply environment that matches the charging demand prediction model;

[0046] A calculation module, which is used to perform spatio-temporal correlation on the output result of the cross-domain correlation model and the identification result of the actual load operation state of the charging pile. Combining the power parameters of the steady-state power supply environment, it calculates the demand distribution density of the charging pile group in the target area;

[0047] A generation module, which is used to synchronously fuse the characteristic parameters of the specific vibration mode and the exclusion result of the no-load false triggering event according to the spatial gradient change characteristics of the demand distribution density, and generate a multi-dimensional heat map including the prediction of the charging pile availability and the load-bearing capacity. The chromaticity mapping rule of the multi-dimensional heat map is negatively correlated with the power factor compensation state of the power grid.

[0048] In a third aspect, an embodiment of the present application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for generating a charging demand heat map based on time series prediction as described in the first aspect above.

[0049] In a fourth aspect, an embodiment of the present application provides a computer storage medium, storing a computer program, and when the computer program is executed by a computer, it implements a method for generating a charging demand heat map based on time series prediction as described in the first aspect.

[0050] In the embodiments of the present application, a mapping relationship between the network traffic fluctuation law and the charging demand time series characteristics is established. By analyzing the periodic change law of the data transmission volume of communication nodes in the area where the charging pile is located, a cross-domain correlation model reflecting the coupling relationship between the charging demand and the network behavior is constructed; a piezoelectric vibration sensor is integrated inside the support structure of the charging pile. Based on the amplitude difference of different frequency components in the mechanical vibration signal of the charging pile, a specific vibration mode triggered by the plugging and unplugging action of the charging gun is identified to distinguish the actual load operation state of the charging pile from the no-load false triggering event; a reactive power compensation device is deployed on the distribution side. By adjusting the switching state of the compensation capacitor, the power factor fluctuation of the power grid caused by the access of the charging pile cluster is suppressed, and a steady power supply environment matching the charging demand prediction model is generated; the output result of the cross-domain correlation model is spatio-temporally correlated with the identification result of the actual load operation state of the charging pile, and combined with the power parameters of the steady power supply environment, the demand distribution density of the charging pile group in the target area is calculated; according to the spatial gradient change characteristics of the demand distribution density, the characteristic parameters of the specific vibration mode and the exclusion result of the no-load false triggering event are synchronously fused to generate a multi-dimensional heat map including the prediction of the charging pile availability and the load-bearing capacity, and the chromaticity mapping rule of the multi-dimensional heat map is negatively correlated with the power factor compensation state of the power grid.

[0051] The technical solution of the present application has the following beneficial effects:

[0052] By analyzing the periodic change law of the communication node data transmission volume, a mapping relationship between the network behavior and the charging demand is constructed to solve the benchmark modeling problem of multi-source data collaborative prediction and improve the accuracy of the charging demand time series prediction; based on the amplitude difference of the vibration signal frequency components, the actual load operation of the charging pile and the no-load false triggering event are accurately distinguished, the physical layer device state perception ability is enhanced, and the prediction noise interference is reduced; by adjusting the switching state of the compensation capacitor, the power factor fluctuation of the power grid is suppressed, and a stable electrical environment matching the grid response ability is provided for the demand prediction; the cross-domain model output is spatio-temporally correlated with the load state identification result, and combined with the grid power parameters, the demand distribution density is generated to realize the optimization of the accuracy and spatial resolution of the charging demand prediction; based on the spatial gradient characteristics of the demand distribution density, the vibration sensing data and the no-load exclusion result are fused to generate a visualization heat map including the availability prediction and the load-bearing capacity, and its chromaticity rule is linked with the grid compensation state, intuitively reflecting the grid bearing pressure.

[0053] Further, by dividing the actual load status of the charging pile into a discrete event sequence according to time slices and associating it with the peak interval of network traffic data, the load status is weighted based on the coupling weights output by the cross-domain model; the capacity boundary is corrected for the weighted result in combination with the grid power parameters to generate the upper limit of the available power capacity of the charging pile node; finally, through density field calculation, the available power capacity is spatially and temporally superimposed with the network traffic peak to output the demand distribution density, and its spatial resolution is jointly constrained by the physical distance between the charging piles and the grid response delay. Through multi-source data fusion (network traffic, load status, grid parameters) and weight allocation mechanism, the refined calculation of the charging demand distribution density is realized; combined with the collaborative constraint rules of the grid response delay and the physical distance of the equipment, it is ensured that the prediction result not only conforms to the grid regulation ability but also adapts to the actual layout of the charging piles, significantly improving the prediction accuracy and grid matching.

[0054] These aspects or other aspects of the present application will be more clearly understood in the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present application or in the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0056] Figure 1 The flowchart of a method for generating a heat map of charging demand based on time series prediction provided by the present application is shown;

[0057] Figure 2 The structural schematic diagram of a system for generating a heat map of charging demand based on time series prediction provided by the present application is shown;

[0058] Figure 3 The structural schematic diagram of a computing device provided by the present application is shown. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0059] In order to enable those skilled in the art of the present technology to better understand the solutions of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0060] In some of the processes described in the specification, claims, and the above-mentioned drawings of the present application, there are multiple operations that occur in a specific order. However, it should be clearly understood that these operations can be executed not in the order in which they appear herein or in parallel. The operation numbers, such as 101, 102, etc., are only used to distinguish different operations, and the numbers themselves do not represent any execution order. Additionally, these processes can include more or fewer operations, and these operations can be executed sequentially or in parallel. It should be noted that the descriptions such as "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent a sequential order, and do not limit that "first" and "second" are of different types.

[0061] The technical solution of the present application is applicable to the scenario of collaborative prediction of charging demand and network load in a smart park, and realizes charging demand prediction through multi-source data collaboration and cross-system linkage. First, establish a mapping model between the periodic fluctuations of network traffic and the time-series characteristics of charging demand to quantify the driving effect of network behavior on demand; combine piezoelectric vibration sensors to identify the specific vibration patterns (main frequency energy concentration, harmonic attenuation characteristics) of the charging gun plugging and unplugging actions to filter out no-load false trigger interference; synchronously deploy reactive power compensation devices to adjust the switching of capacitors to suppress the fluctuation of the power factor of the power grid and generate steady-state power supply parameters; perform spatio-temporal fusion of the network traffic weight, the load status recognition result, and the power grid capacity constraint to calculate the demand distribution density; finally, generate a multi-dimensional heat map based on the density gradient, device availability probability, and power grid status, and its chromaticity is negatively correlated with the power factor compensation value, intuitively reflecting the regional load pressure and the health of the power grid.

[0062] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0063] Figure 1 The following is a flowchart of a method for generating a charging demand heat map based on time-series prediction provided for the embodiments of the present application, as Figure 1 shown, the method includes:

[0064] 101. Establish a mapping relationship between the network traffic fluctuation law and the time-series characteristics of charging demand, and analyze the periodic change law of the data transmission volume of communication nodes in the area where the charging pile is located based on the mapping relationship, so as to construct a cross-domain correlation model reflecting the coupling relationship between charging demand and network behavior;

[0065] In this step, the cross-domain correlation model is a mathematical model that reflects the coupling relationship between network behavior and charging demand, established by analyzing the correlation between the network traffic data (such as peak data transmission volume, proportion of protocol types) of communication nodes in the area where the charging pile is located and the time-series characteristics of charging demand (such as frequency of charging gun plugging and unplugging events, power consumption curve), and is used to predict the trend of demand fluctuations.

[0066] In the embodiment of the present application, based on the established mapping relationship between the network traffic fluctuation law and the time-series characteristics of charging demand, multi-scale periodic decomposition is performed on the network traffic data of the communication nodes in the area where the charging pile is located to identify the periodic fluctuation laws at the hourly, daily, and weekly levels. For example, the occurrence time period and duration of traffic peaks are extracted through spectral analysis. Then, in combination with the time-series distribution of charging gun plugging and unplugging events in the historical operation data of the charging pile, high-frequency demand periods (such as morning and evening rush hours), stable demand periods, and low-demand periods are divided, and period labels are established. The network traffic periodic components are matched with the charging demand periods through a time alignment algorithm, and the similarity weights at different time scales are calculated. For the time-series offset problem between network traffic fluctuations and charging demand changes, based on the phase correlation characteristics of the mapping relationship, phase synchronization technology is used to compensate the time difference between the two. Finally, the period component matching weights and the time-series compensation parameters are integrated to generate a cross-domain correlation model.

[0067] Suppose in a smart park, the network traffic of the charging piles within the coverage area of the communication nodes reaches a peak (about 120 Gbps) every day from 18:00 to 20:00, corresponding to a sharp increase in charging demand during the off-work period (the frequency of charging gun plugging and unplugging increases to 5 times per minute). Based on the mapping relationship between network traffic and the time-series characteristics of charging demand, it is found through periodic decomposition that the daily periodic component of the network traffic remains at a high level from 18:00 to 19:30, while the charging demand reaches a peak from 18:30 to 20:00. The time alignment algorithm calculates the similarity weight between the two as 0.88. Based on the phase offset law of the mapping relationship, phase synchronization compensation detects that the network traffic reflects the change in charging demand 30 minutes in advance, and the output of the generated cross-domain correlation model shows that the prediction error drops from 8% to 2.5%.

[0068] 102. Integrate a piezoelectric vibration sensor inside the support structure of the charging pile. Based on the amplitude difference of different frequency components in the mechanical vibration signal of the charging pile, identify the specific vibration mode triggered by the charging gun plugging and unplugging actions, and distinguish the actual load operation state of the charging pile from the no-load false trigger event;

[0069] In this step, the specific vibration mode is the amplitude difference characteristic of different frequency components in the vibration signal triggered by the charging gun plugging and unplugging actions, such as the main frequency concentration interval, harmonic attenuation slope. The no-load false trigger event is an event in which the charging pile is misidentified as a load state due to non-real charging operations (such as human mis-touch, environmental interference).

[0070] In the embodiments of the present application, piezoelectric vibration sensors are deployed in the stress concentration areas of the charging pile support structure to collect mechanical vibration signals. The amplitude spectrum in the frequency band of 20 - 150 Hz is extracted through frequency domain conversion, and the main frequency range (such as 45 - 60 Hz) and harmonic characteristics corresponding to the charging gun plugging and unplugging actions are identified. Based on a preset energy attenuation threshold (such as the frequency point when the fundamental frequency energy decays to 50%), real charging events (smooth energy decay) and no-load false trigger events (rapid energy decay) are distinguished. Combining with the resonance characteristics of the charging pile mechanical structure (such as the fixed frequency response of the metal bracket), the judgment threshold is adjusted. For example, when the bracket screws are loose, the main frequency may shift by 5 Hz, and the frequency matching range needs to be corrected accordingly. Finally, the classification results of the actual load operation state of the charging pile and no-load false trigger events are output.

[0071] For example, continuing the above example, at 18:30, the vibration signal of a certain charging pile support structure is detected. Frequency domain analysis shows that the main frequency is 50 Hz (the characteristic frequency band of the charging gun plugging and unplugging action), and the harmonic energy decays to 35% of the fundamental frequency at 80 Hz, meeting the judgment conditions for real charging events (the threshold is set to less than 40% decay for no-load). In another event, the main vibration frequency is 88 Hz (close to the fan vibration frequency band), and the harmonic energy decays to 12% of the fundamental frequency at 100 Hz, triggering the exclusion of no-load false judgment. Combining with the mechanical damping characteristics of the loose bracket screws, the main frequency matching range is adjusted to 48 - 52 Hz. Finally, the recognition accuracy of the load state reaches 98.5%, and the false judgment rate drops to 1.3%.

[0072] 103. Deploy reactive power compensation devices on the distribution side. By adjusting the switching state of the compensation capacitors, the power factor fluctuation of the power grid caused by the access of the charging pile cluster is suppressed, and a steady-state power supply environment matching the charging demand prediction model is generated;

[0073] In this step, the reactive power compensation device is a device deployed on the distribution side to stabilize the power factor of the power grid by adjusting the switching state of the compensation capacitors. The steady-state power supply environment is a stable power supply state generated after the power factor fluctuation of the power grid is suppressed and matching the charging demand prediction model.

[0074] In the embodiments of the present application, the power factor, voltage volatility, and reactive power demand of the power grid are monitored, and the downward trend of the power factor caused by the access of the charging pile cluster is predicted (such as it may drop to 0.85 within the next 5 seconds). The reactive power to be compensated is calculated according to the instantaneous load capacity of the charging pile (such as 200 kVar), and the switching combination of the capacitor bank is optimally selected (such as switching 2 groups of 100 kVar capacitors). The stability of the power grid after compensation is verified through closed-loop control to ensure that the power factor is restored to the target range (such as 0.95 - 0.98), and the voltage volatility is lower than 3%. Finally, the steady-state power supply environment parameters are output, including the power factor compensation value, voltage stable range, and capacitor switching state.

[0075] For example, continuing with the previous example, the power distribution system of the smart park detected that the power factor dropped to 0.84 at 18:45 (the load rate of the charging pile cluster was 85%), and it was predicted that 220 kVar of reactive power needed to be compensated. The reactive power compensation device switched 2 groups of 100 kVar and 1 group of 30 kVar capacitors. After compensation, the power factor recovered to 0.96, and the voltage fluctuation rate dropped from 5% to 1.8%. Steady-state power supply environment parameters (power factor > 0.95, voltage fluctuation rate < 2%) were generated, which matched the predicted requirements of the cross-domain correlation model.

[0076] 104. Perform spatio-temporal association on the output result of the cross-domain correlation model and the recognition result of the actual load operation state of the charging piles, and calculate the demand distribution density of the charging pile group in the target area in combination with the power parameters of the steady-state power supply environment;

[0077] In this step, the demand distribution density is a quantization index of the charging demand density of the charging pile group changing with time and space in the target area. For example, it is the number of charging requests per minute per unit area.

[0078] In the embodiment of the present application, the network traffic weight output by the cross-domain correlation model (such as the peak period weight of 0.9) is mapped to the spatial grid where the charging pile is located according to a time window (such as 5 minutes) to form a demand trend driven by network traffic; in combination with the recognition result of the actual load state of the charging pile (such as 7 out of 10 charging piles in a certain grid are in the loaded state), the initial demand density is calculated (such as 1.4 vehicles / minute); the power parameters of the steady-state power supply environment are introduced (such as the maximum instantaneous load capacity of the power grid is 90% of the rated value), and the upper limit of the demand density is corrected through capacity constraints (such as adjusted from 1.5 vehicles / minute to 1.35 vehicles / minute). Finally, the spatial resolution is adjusted according to the physical distance between the charging piles (such as 10 meters) and the grid response delay (such as 2 seconds), and the demand distribution density is output.

[0079] For example, continuing with the previous example, the network traffic weight (0.88) output by the cross-domain correlation model is associated with the charging pile load state (70% load rate) according to a 10-meter spatial grid, and in combination with the grid capacity constraint coefficient (0.9), the initial demand density is calculated to be 1.6 vehicles / minute. After capacity correction, the output demand distribution density is 1.44 vehicles / minute, the spatial resolution is adjusted to a 10-meter grid, and a density peak of 1.9 vehicles / minute (triggering an alarm) is detected in area A, and the density in area B is 0.8 vehicles / minute (idle recommendation).

[0080] 105. According to the spatial gradient change characteristics of the demand distribution density, synchronously fuse the characteristic parameters of the specific vibration mode and the exclusion result of the no-load mis-trigger event to generate a multi-dimensional heat map including the prediction of the charging pile availability and the load-carrying capacity. The chromaticity mapping rule of the multi-dimensional heat map is negatively correlated with the power factor compensation state of the power grid.

[0081] In this step, the multi-dimensional heat map is a visualization chart including the prediction of the charging pile availability (equipment status), the load-carrying capacity (power grid constraint), and the demand density (spatiotemporal distribution). The chromaticity mapping rule is the negative correlation between the color depth in the heat map and the power factor compensation state of the power grid. For example, dark red indicates a low power factor (<0.9), and light green indicates a high power factor (>0.95).

[0082] In the embodiment of the present application, map the spatial gradient change characteristics of the demand distribution density to the base color of the heat map (such as the red gradient indicates the density level), overlay the availability probability of the charging pile (such as 30% transparency indicates a higher equipment failure risk) and the exclusion result of the no-load mis-trigger (such as the gray area indicates a low confidence level). Adjust the chromaticity offset according to the power factor compensation state of the power grid (such as the chromaticity is purple-shifted when the power factor is 0.88), and smooth the color change of adjacent areas through the spatial interpolation algorithm. Finally, generate an interactive multi-dimensional heat map, and the operation and maintenance personnel can click to query specific parameters (such as the load rate of a certain charging pile, the switching state of the compensation capacitor).

[0083] For example, continuing the above example, in the finally generated multi-dimensional heat map, area A shows dark red (demand density of 1.9 vehicles / minute), 40% transparency (availability probability of vibration characteristics of 60%), and purple-shifted chromaticity (power factor of 0.87), prompting the operation and maintenance personnel to schedule preferentially; area B shows light green (demand density of 0.8 vehicles / minute), 85% transparency (availability probability of vibration characteristics of 95%), and positive green chromaticity (power factor of 0.96), recommending users to go for charging; area C is marked in gray (similarity of the no-load mis-trigger exclusion result <0.6), shielding the interference of low-confidence data.

[0084] Steps 101-105 achieve the global optimization of charging demand prediction and power grid operation through cross-domain data collaborative modeling (associating network traffic and charging demand), physical and power grid closed-loop perception (linking vibration characteristic recognition and reactive power compensation), and multi-dimensional visualization drive (heat map rendering). Break through the limitation of a single data source, fuse network behavior, equipment status, and power grid parameters, significantly reduce prediction deviation and misjudgment interference; suppress the power fluctuation caused by the access of the charging pile cluster, ensure that the prediction result matches the power grid carrying capacity; synchronously reflect the charging demand distribution, equipment health, and power grid pressure status through the visual heat map, providing actionable insights for resource scheduling.

[0085] To further improve the spatio-temporal accuracy of charging demand prediction and ensure grid stability, this application addresses the time-varying correlation problem between network behavior and charging demand based on the event sequence division and weighting mechanism of time slices; introduces a grid capacity constraint correction model to break through the physical boundary limitations of traditional prediction; generates a high-resolution demand distribution through spatio-temporal superposition and density field calculation, and finally outputs actionable decision support. In some embodiments, spatio-temporally correlating the output result of the cross-domain correlation model with the recognition result of the actual load operation state of the charging piles, and combining the power parameters of the steady-state power supply environment to calculate the demand distribution density of the charging pile group in the target area includes:

[0086] 201. Divide the recognition result of the actual load operation state of the charging piles into discrete event sequences according to a preset time slice, and the load state within each time slice in the discrete event sequence forms an association with the peak interval of the network traffic data corresponding to the time slice;

[0087] In step 201, the discrete event sequence is a discretized state sequence obtained by dividing the recognition result of the actual load operation state of the charging piles according to a fixed time slice (such as 5 minutes), and the charging piles are marked as loaded or unloaded states within each time slice. The peak interval of the network traffic data is the peak range of the data transmission volume reached by the communication node within a specific time slice.

[0088] In the embodiments of this application, the recognition result of the actual load state of the charging piles is divided into time slices (such as a 5-minute window), and the proportion of the load state within each time slice is statistically calculated through a sliding window (such as 70% of the charging piles are loaded) to form a discrete event sequence. Synchronously extract the network traffic data corresponding to the time slice, and identify its peak interval (such as the traffic reaches 115 Gbps from 18:00 to 18:05). The periodic component of the network traffic data is decomposed by wavelet transform, and the correlation strength between the proportion of the load state and the peak interval of the traffic is calculated through the Pearson correlation coefficient (such as the correlation coefficient is 0.85) to form a load and traffic association mapping table in the time slice dimension.

[0089] 202. Weight the load state within each time slice in the discrete event sequence according to the coupling weight of the network traffic and the charging demand output by the cross-domain correlation model to obtain a weighted result;

[0090] In step 202, weighting means adjusting the weight of the load state in the discrete event sequence according to the coupling weight of the network traffic and the charging demand output by the cross-domain correlation model.

[0091] In the embodiments of the present application, based on the coupling weights output by the cross-domain correlation model (such as the mapping weights between the daily cycle components of network traffic and the charging demand), the time warping (DTW) algorithm is used to align the temporal offset between the load event sequence and the network traffic peak. The weight contribution degree of different time slices is quantified by the entropy weight method (such as the weight ratio in the peak period is increased to 1.2 times), the load state ratio is weighted and corrected (such as the original 70% load rate is adjusted to 78% after weighting), and a moving average filter is introduced to eliminate the short-term fluctuation interference, and the weighted result is obtained.

[0092] 203. Establish the constraint relationship between the power parameters of the steady-state power supply environment and the maximum instantaneous load capacity of the charging pile, and correct the capacity boundary of the weighted result based on the constraint relationship to generate the upper limit of the available power capacity of the charging pile node;

[0093] In step 203, the capacity boundary correction is to perform an upper limit constraint on the weighted load state according to the power parameters of the steady-state power supply environment (such as the maximum instantaneous load capacity of the power grid), and generate the upper limit of the available power capacity of the charging pile node.

[0094] In the embodiments of the present application, the power parameters of the steady-state power supply environment are collected (such as the power factor of 0.95 and the voltage fluctuation rate <2%), and the maximum load capacity that the power grid can bear in the next 5 seconds is predicted through Kalman filtering (such as 90% of the rated capacity). A linear constraint relationship between the total power demand of the charging pile group and the power grid capacity is established, and the Lagrangian relaxation algorithm is used to correct the boundary of the weighted load state (such as constraining the 78% load rate to 75%). Finally, the upper limit of the available power capacity of each charging pile node is output (such as the maximum power of a single pile is limited from 100 kW to 90 kW).

[0095] 204. Spatially and temporally superimpose the upper limit of the available power capacity and the peak interval of the network traffic data, and calculate the demand distribution density of the charging pile group in the target area through the density field. The spatial resolution of the density field is jointly determined by the physical distance between the charging pile nodes and the response delay time of the reactive power compensation device.

[0096] In step 204, the density field is a charging demand distribution model generated by spatially and temporally superimposing the upper limit of the available power capacity and the peak interval of the network traffic data, and its spatial resolution is jointly determined by the physical distance between the charging piles and the grid response delay.

[0097] In the embodiments of the present application, the upper limit of available power capacity is mapped into a power capacity distribution layer according to a spatial grid (such as 10 meters × 10 meters), and the peak interval of network traffic is mapped into a traffic weight layer according to time slices. The spatio-temporal Kriging interpolation algorithm is used to fuse the two layers of data to generate an initial density field. The spatial resolution base value is defined according to the physical distance between charging piles (such as the adjacent pile distance is 15 meters), and the time resolution compensation coefficient (such as 0.8) is calculated in combination with the response delay of the reactive power compensation device (such as 2 seconds). The smoothness of the initial density field is optimized through a convolutional neural network (CNN), and finally the demand distribution density is output.

[0098] The following is a specific example:

[0099] Suppose a smart park faces a sharp increase in charging demand during the evening peak period (18:00 - 20:00). The communication node monitors that the peak network traffic reaches 120 Gbps. Through the division of discrete event sequences, it is identified that 70% of the charging piles are in the loaded state within a certain 5-minute window. The peak interval of network traffic is 115 - 118 Gbps, and the correlation intensity is 0.88; the cross-domain correlation model outputs a coupling weight of 0.9 during the peak period, and the load rate is increased to 78% after weighted correction; the grid parameters show that the maximum load capacity that can be borne is 85% of the rated value, and the load rate is constrained to 75% through capacity boundary correction; finally, the demand distribution density is generated based on a 10-meter grid. Area A shows a demand density of 1.8 vehicles / minute (spatial resolution of 10 meters, time resolution of 2 seconds), triggering an overload warning.

[0100] Steps 201 - 204 achieve the deep coupling of charging demand prediction and grid operation through the multi-dimensional coordination of time sequence, space and power grid, fuse the network traffic weight and grid capacity constraints, and break through the neglect of physical boundaries by traditional models; through the linkage of capacity correction and reactive power compensation, ensure that the prediction results strictly match the grid bearing limit; the density field provides a high-resolution spatio-temporal distribution, supports minute-level response and resource scheduling; combined with the load status recognition and noise filtering mechanism, effectively resist sudden interference and data anomalies.

[0101] In order to further improve the physical perception accuracy and grid linkage ability of the charging demand heat map, a multi-dimensional visualization decision-making system is constructed by fusing equipment vibration characteristics, no-load interference suppression and density gradient field evolution. The spatial distribution analysis ability is enhanced based on the density gradient discretization mechanism of the physical layout of the charging piles; a visual heat map is generated by superimposing multi-dimensional data to realize the collaborative presentation of charging demand prediction and equipment health. In some embodiments, according to the spatial gradient change characteristics of the demand distribution density, the characteristic parameters of the specific vibration mode and the exclusion result of the no-load mis-trigger event are synchronously fused to generate a multi-dimensional heat map including the prediction of the availability of the charging pile and the load bearing capacity, including:

[0102] 301. Discretize the spatial gradient change characteristics of the demand distribution density into multiple local density change rates according to the physical layout of the charging pile group. The direction vector of the local density change rate is associated with the vibration propagation direction of the charging gun plugging and unplugging action.

[0103] In step 301, the local density change rate is a quantization index of the local area density change obtained by dividing the spatial gradient change of the demand distribution density according to the physical layout of the charging piles (such as the density increase or decrease amplitude per unit distance). The direction vector is a spatial vector representing the direction of the local density change, and its direction is consistent with the propagation direction of the vibration signal caused by the charging gun plugging and unplugging action.

[0104] In the embodiment of the present application, based on the physical layout of the charging pile group (such as grid distribution or star topology), a manifold learning algorithm is used to extract the spatial gradient characteristics of the demand distribution density and identify the boundary between the high-density area and the low-density area. The direction vector of the local density change rate is calculated by the direction field estimation technology. For example, using the energy attenuation model of the vibration signal propagation path (such as the attenuation rate of the main vibration frequency in the metal bracket is 2dB / m), the direction vector is aligned with the vibration propagation direction, and finally a distribution map of the discretized local density change rate is formed, where the vector length represents the change intensity and the direction reflects the density diffusion trend.

[0105] 302. Extract the vibration frequency distribution interval that matches the charging gun plugging and unplugging action from the characteristic parameters of the specific vibration mode, and generate a vibration characteristic availability probability representing the availability state of the charging pile based on the energy proportion of each frequency band within the vibration frequency distribution interval.

[0106] In step 302, the vibration frequency distribution interval is the characteristic frequency band range corresponding to the charging gun plugging and unplugging action in the vibration signal. The vibration characteristic availability probability is a quantization index of the charging pile availability calculated based on the energy proportion of each frequency band within the characteristic frequency band.

[0107] In the embodiment of the present application, independent component analysis (ICA) is performed on the original signal collected by the piezoelectric vibration sensor to separate the characteristic frequency band of the charging gun plugging and unplugging action (such as the main frequency of 45Hz and its harmonics). The energy proportion of each frequency band is calculated through the power spectral density (such as the energy proportion of 45Hz is 65% and the harmonic proportion of 60Hz is 20%), and a vibration energy distribution histogram is constructed. A support vector machine (SVM) classifier is used to train the relationship between the energy distribution and the device failure rate (such as when the energy dispersion degree > 30%, the failure probability increases), and the vibration characteristic availability probability (such as a normalized value between 0 and 1) is output.

[0108] 303. According to the elimination result of the no-load mis-triggering event, count the no-load mis-triggering frequency of the charging pile node within a preset time period, and combine the power parameters of the steady-state power supply environment to generate a no-load interference suppression factor reflecting the correction coefficient of the load-bearing capacity of the charging pile;

[0109] In step 303, the no-load interference suppression factor is a correction coefficient of the load-bearing capacity generated according to the no-load mis-triggering frequency and the power grid power parameters.

[0110] In the embodiment of the present application, count the number of no-load mis-triggering times of the charging pile node within a preset time period (such as 1 hour), combine the power factor compensation state (such as 0.92) and the voltage volatility (such as 2.5%) of the steady-state power supply environment, and predict the mis-triggering risk in the future period through a Hidden Markov Model (HMM). Based on the risk level (such as high risk, medium risk, low risk), assign the suppression factor weights (such as 0.7, 0.85, 1.0) to generate a no-load interference suppression factor, which is used to correct the upper limit of the predicted load-bearing capacity.

[0111] 304. Perform spatial interpolation on the vibration feature availability probability and the no-load interference suppression factor according to the physical location of the charging pile, and superimpose the vector distribution of the local density change rate to generate a multi-dimensional heat map including the charging pile availability prediction and the load-bearing capacity.

[0112] In step 304, the multi-dimensional heat map is a visualization chart integrating the charging pile availability probability, the load-bearing capacity correction coefficient, and the density gradient direction vector, and its chromaticity is negatively correlated with the power factor compensation state of the power grid.

[0113] In the embodiment of the present application, map the vibration feature availability probability to the transparency channel (such as high transparency indicates low availability), map the no-load interference suppression factor to the color saturation (such as low saturation indicates the need for load reduction), and map the local density change rate vector to the direction arrow. Use a spatial interpolation algorithm (such as inverse distance weighting) to fill the parameter gaps in the unmonitored areas, and synthesize a high-resolution heat map image through a Generative Adversarial Network (GAN). Introduce an attention mechanism to enhance the visual salience of the overload area (such as red flashing reminder), and finally generate an interactive multi-dimensional heat map. The chromaticity mapping rule of the multi-dimensional heat map is negatively correlated with the power factor compensation state of the power grid, that is, the lower the power factor (such as 0.85), the more the chromaticity tends to dark red, and the higher the power factor (such as 0.98), the more the chromaticity tends to light green.

[0114] The following is a specific example:

[0115] Suppose during the peak charging period on weekends in a smart park, the charging pile group is arranged in a circular layout. It is detected that the direction vector of the demand density gradient in area C is consistent with the vibration propagation path, and the energy ratio of the main frequency of the vibration signal at 50 Hz is 75% (the availability probability of the vibration characteristics is 0.9); the no-load false triggering frequency in area D is 15 times per hour, and a suppression factor of 0.75 is generated by combining the power factor of the power grid at 0.88; the availability probability of the vibration characteristics and the suppression factor are superimposed on the density gradient vector by spatial interpolation. The heat map shows that area C is light green (power factor 0.95, negatively correlated with the chromaticity rule), with a transparency of 10% (high availability), and the arrow points to the northwest. Area D is dark red (power factor 0.88, negatively correlated with the chromaticity rule), with a transparency of 40% (low availability), and the arrow vaguely indicates a risk, and the arrow points to prompt the operation and maintenance personnel to prioritize the expansion of area C.

[0116] Steps 301-304 achieve multi-dimensional and refined presentation of the charging demand heat map through the collaborative analysis of physics, the power grid, and spatial gradients, integrate vibration characteristics and density gradient directions, and break through the traditional heat map's neglect of equipment status; correct the load prediction through the suppression factor to ensure the strict matching of the heat map and the power grid status; multi-dimensional visualization parameters (chromaticity, transparency, direction arrow) provide a basis for composite decision-making; the statistical analysis of no-load false triggering and the analysis of vibration energy work together to filter out noise data and improve the reliability in complex scenarios.

[0117] In order to improve the accuracy of the correlation modeling between network traffic fluctuations and the time-series characteristics of charging demand, a cross-domain causal correlation model is constructed through multi-scale periodic decomposition and phase matching mechanisms. Monitoring units are divided based on the type of communication protocol, and the dominant periodic components are extracted; a coupling function that can correct time-series offsets is constructed to achieve accurate prediction of network behavior and charging demand. In some embodiments, the mapping relationship between the network traffic fluctuation pattern and the time-series characteristics of charging demand is established by analyzing the periodic change pattern of the data transmission volume of communication nodes in the area where the charging piles are located, and a cross-domain correlation model reflecting the coupling relationship between charging demand and network behavior is constructed, including:

[0118] 401. Divide the network traffic monitoring units according to the physical coverage range of the communication nodes in the area where the charging piles are located, perform multi-scale periodic decomposition on the data transmission volume of the communication nodes based on the type of communication protocol, and extract the dominant periodic components of the hourly, daily, and weekly data transmission volumes in each monitoring unit;

[0119] In step 401, the network traffic monitoring unit is a network traffic data collection area divided according to the physical coverage range (such as a radius of 200 meters) of the communication nodes in the area where the charging piles are located. The dominant periodic component is the periodic fluctuation component with the highest energy ratio on the hourly, daily, and weekly time scales after multi-scale decomposition of the network traffic data.

[0120] In the embodiments of the present application, network traffic monitoring units are divided according to the physical coverage radius of communication nodes (for example, a 4G base station covers 500 meters, and a Wi-Fi hotspot covers 100 meters) to ensure a one-to-one mapping between charging piles and communication nodes within each unit. Wavelet packet decomposition is performed on the data transmission volume based on the communication protocol type (such as TCP, UDP, MQTT), and periodic components at different time scales (hours, days, weeks) within each monitoring unit are extracted. Components with an energy ratio exceeding the threshold are selected as the dominant periodic components through power spectral density analysis.

[0121] 402. According to the temporal distribution characteristics of the charging gun plugging and unplugging events in the historical operation data of the charging pile, define the high-frequency demand period, stable demand period, and low-demand period of the charging demand, and establish a mapping rule between the charging demand period label and the amplitude-frequency characteristics of the dominant periodic component;

[0122] In step 402, the charging demand period label is the high-frequency demand period, stable demand period, and low-demand period divided according to the temporal distribution of the charging gun plugging and unplugging events. The amplitude-frequency characteristic mapping rule is the correspondence between the amplitude and frequency distribution characteristics of the dominant periodic component and the charging demand period label.

[0123] In the embodiments of the present application, the temporal distribution of the charging gun plugging and unplugging events in the historical data of the charging pile is statistically analyzed, and the K-means clustering algorithm is used to divide the demand period label (such as the clustering centers are 7:00 - 9:00, 12:00 - 14:00, 18:00 - 20:00). The amplitude-frequency characteristics of the dominant periodic component are extracted (such as the amplitude peak and frequency band width of the daily component at 18:00 - 20:00), and a mapping rule between the period label and the amplitude-frequency characteristics of the dominant periodic component is established through canonical correlation analysis (CCA).

[0124] 403. For each network traffic monitoring unit, match the phase fluctuation interval of the dominant periodic component with the duration window of the corresponding charging demand period label, and introduce the amplitude-frequency characteristic mapping rule during the matching process to generate the dynamic correlation weight between the network traffic fluctuation law and the charging demand temporal characteristics;

[0125] In step 403, the phase fluctuation interval is the phase offset range of the dominant periodic component on the time axis. The correlation weight is a quantitative index of the matching degree between the network traffic periodic component and the charging demand period.

[0126] In the embodiments of the present application, a sliding window analysis is performed on the phase fluctuation of the dominant period component (e.g., the window size is 1 hour) to detect the overlap between its peak time and the duration window of the demand period label. Based on the amplitude threshold and frequency band width range defined in the amplitude-frequency characteristic mapping rule, the phase matching tolerance is dynamically adjusted: if the amplitude exceeds the threshold and the frequency width meets the range (e.g., the amplitude in the high-frequency period > 100 Gbps, the frequency width is 10 - 20 Hz), a phase deviation of ±15 minutes is allowed, and a high-weight base value (e.g., 1.0) is assigned; if the frequency width exceeds the range (e.g., the frequency width in the low-frequency period > 5 Hz), the phase tolerance is reduced to ±5 minutes, and the weight base value is reduced based on the amplitude attenuation ratio (e.g., when the amplitude is 80 Gbps, the base value is 0.8). Finally, the dynamic correlation weight is generated by combining the phase overlap (e.g., a 90% overlap assigns a coefficient of 1.2, a 60% overlap assigns 0.8).

[0127] 404. Based on the correlation weight, construct a coupling relationship function between the charging demand and the network behavior. The time series offset of the dominant period component of the network traffic in the coupling relationship function is used to correct the cross-domain correlation error, and a structured cross-domain correlation model is output.

[0128] In step 404, the coupling relationship function is a mathematical expression describing the correlation between the time series characteristics of the dominant period component of the network traffic and the charging demand. Its input is the time series offset of the period component, and the output is the cross-domain correlation error correction value.

[0129] In the embodiments of the present application, a multiple regression function with the dynamic correlation weight as the coefficient is constructed. The time series offset of the dominant period component of the network traffic (e.g., the peak of the daily component lags by 20 minutes) is used as the independent variable, and the charging demand prediction error is used as the dependent variable. The function parameters are optimized by the gradient descent algorithm, and L1 regularization is introduced to suppress overfitting. The weight distribution is dynamically adjusted based on the amplitude-frequency characteristic mapping rule (e.g., a wider frequency signal allows a larger time series offset compensation), and finally a structured cross-domain model is output.

[0130] The following is a specific example:

[0131] Suppose that during holidays, the charging demand pattern in a smart park changes suddenly, and the communication node coverage area is divided into 3 monitoring units; Unit A detects that the amplitude peak of the daily dominant cycle component is 130 Gbps, the bandwidth is 18 Hz (meeting the high-frequency period mapping rule), the phase fluctuation interval is from 18:10 to 18:50, which overlaps with the demand period label from 18:00 to 19:00 with an overlap degree of 90%, and the weight base value is 1.0×1.2 = 1.2; The dominant component amplitude of Unit B is 60 Gbps, the bandwidth is 8 Hz (meeting the low valley period rule), the phase fluctuation interval is from 23:00 to 23:30, which overlaps with the demand period label from 22:00 to 24:00 with an overlap degree of 70%, and the weight base value is 0.6×0.9 = 0.54; The multi-regression function detects that the time series of Unit A is offset by +10 minutes, and that of Unit B is offset by -15 minutes. After adjusting the weights based on the bandwidth rule, a structured cross-domain model is output, and the prediction error is reduced from 12% to 3%.

[0132] Steps 401-404 achieve accurate modeling of the correlation between network traffic and charging demand through a technical link of multi-scale period analysis, amplitude-frequency phase coordination, and dynamic error correction. The cross-domain model integrates communication protocol characteristics and the evolution law of demand periods, significantly reducing the prediction deviation caused by time series offset; The dynamic weight mechanism adapts to the power grid load fluctuation, ensuring that the prediction result matches the real-time power supply capacity; The structured model supports fast switching between multiple scenarios (holidays, weekdays), reducing the dependence on manual parameter adjustment; The amplitude-frequency characteristic mapping rule and the phase fluctuation interval provide a basis for physical layer causal correlation, improving the reliability of operation and maintenance.

[0133] To improve the spatio-temporal resolution and accuracy of the charging demand distribution density, through spatial grid modeling and multi-scale data fusion, the collaborative calculation of power capacity and network traffic is realized, including grid division based on the physical distance between charging piles to generate local power distribution, constructing a density field in combination with the spatio-temporal characteristics of network traffic, and optimizing the density increment calculation through multi-scale convolution, and finally outputting a high-precision demand distribution density. In some embodiments, the spatio-temporal superposition of the available power capacity upper limit and the peak interval of network traffic data, and calculating the demand distribution density of the charging pile group in the target area through the density field, includes:

[0134] 501. Divide the spatial grid units of the target area according to the physical distance between the charging pile nodes, discretely sample the available power capacity upper limit according to the spatial grid units, and generate a local power capacity distribution;

[0135] In step 501, the local power capacity distribution is a power distribution map obtained by discretely sampling the available power capacity upper limit of the charging pile nodes according to the spatial grid units, reflecting the power supply capacity of each grid unit.

[0136] In the embodiments of the present application, the target area is divided into equally spaced spatial grid cells (such as 10 m × 10 m) according to the physical spacing of the charging pile nodes (such as the adjacent pile distance of 20 m). After discretely sampling the upper limit of the available power capacity of each charging pile (such as the maximum power of a single pile being 100 kW) through the inverse distance weighted interpolation algorithm, it is mapped to the corresponding grid cell. The interpolation result is corrected in combination with the grid load rate (such as 80%) (such as the power capacity of the overloaded grid cell being reduced to 80 kW) to generate the local power capacity distribution.

[0137] 502. Divide the peak interval of the network traffic data into multiple consecutive time windows in the time dimension, extract the spatial distribution characteristics of the network traffic data within each time window, and generate the network traffic spatial weight corresponding to the spatial grid cell.

[0138] In step 502, the network traffic spatial weight is the distribution weight of the peak interval of the network traffic data divided by the time window on the spatial grid cell, reflecting the spatio-temporal driving intensity of network behavior on the charging demand.

[0139] In the embodiments of the present application, the peak interval of the network traffic data (such as the traffic reaching 150 Gbps from 18:00 to 18:30) is sliced by time windows (such as 5 minutes), and the traffic spatial distribution characteristics of each communication node within each window (such as the proportion of TCP protocol traffic, the distribution of packet sizes) are extracted. After dimensionality reduction through principal component analysis (PCA), the weight coefficient corresponding to the spatial grid cell is generated (such as the weight of the central grid during peak hours being 0.9 and the weight of the edge grid being 0.6), forming the network traffic spatial weight.

[0140] 503. Superimpose the local power capacity distribution and the network traffic spatial weight according to the spatial grid cell to generate a density field reflecting the coupling relationship between the power capacity of the charging pile group and the network traffic.

[0141] In step 503, the density field is a charging demand density space model generated by superimposing the local power capacity distribution and the network traffic spatial weight, characterizing the coupling effect of the power capacity and the network traffic.

[0142] In the embodiments of the present application, the local power capacity distribution and the network traffic spatial weight matrix are superimposed point by point according to the grid cell, and the spatial weighted fusion algorithm (such as the entropy weight method) is used to balance the contribution ratio of the power capacity and the traffic weight (such as the power capacity accounting for 60% and the traffic weight accounting for 40%). Spatial autocorrelation analysis (Moran's I index) is introduced to detect the hot spots (such as high-density aggregation areas) of the density field, generating the density field.

[0143] 504. Perform multi-scale convolution operations on the local power capacity distribution and the network traffic spatial weights according to the spatial grid cells, calculate the density increment of the spatial grid cells in the density field, and output the demand distribution density of the charging pile group in the target area.

[0144] In step 504, the density increment is the change in the demand density of the spatial grid cells calculated through multi-scale convolution operations in the density field, reflecting the short-term demand fluctuation trend.

[0145] In the embodiments of the present application, multi-scale convolution kernels (such as 3×3, 5×5, 7×7) are designed to perform sliding window scanning on the density field to extract density gradient features in different spatial ranges (such as local mutations, regional trends). The density increment in the future time window (such as 5 minutes) is predicted through a long short-term memory network (LSTM), and the prediction error is corrected by combining Kalman filtering to output the density increment. Finally, the density increment is superimposed on the current density field to generate the updated demand distribution density.

[0146] The following is a specific example:

[0147] Suppose the charging pile group in a smart park is star-shaped, with a spacing of 18 meters. In step 501, the target area is divided into 12-meter × 12-meter spatial grids. The charging piles in the central area are dense (spacing 10 meters), generating a local power capacity upper limit of 150 kW (load rate 65%), and the sparse area at the edge is 100 kW; in step 502, it is monitored that the network traffic peak surges to 180 Gbps (live data transmission) from 15:00 to 15:15. The spatial weight of the central grid traffic reaches 0.98, and that of the edge grid is 0.6; in step 503, after superimposing the power capacity and the traffic weight, the generated density field shows that the demand density of the central grid is 2.2 vehicles / minute, triggering a yellow warning; in step 504, through the scanning of the multi-scale convolution kernel, the density increment of the central grid is detected (+0.5 vehicles / minute within 5 minutes). The LSTM predicts that the density will exceed 3.0 vehicles / minute in the next 10 minutes, and automatically activates the power allocation of the adjacent idle grids, temporarily increasing the power capacity of the edge grid to 120 kW to divert the pressure.

[0148] Steps 501 - 504 realize the refined calculation of the charging demand density through the technical path of grid-based power distribution, traffic weight fusion, and multi-scale density evolution. The combination of grid division and multi-scale convolution accurately captures local demand mutations; the increment prediction and correction mechanism support minute-level density updates; the balance between power capacity and network traffic weight ensures the physical executability of the prediction results; and different park topologies are adapted through the grid-based model to reduce the deployment cost.

[0149] To enhance the collaborative expression ability of the heat map for device status and grid response, a multi-dimensional visualization engine is constructed through vibration propagation path energy attenuation modeling and reactive power compensation response control. A vibration interpolation field is generated based on the directional characteristics of charging piles; a multi-dimensional heat map is generated by superimposing the density gradient vector, realizing three-dimensional visualization of device availability, grid status, and demand density. In some embodiments, spatially interpolating the vibration feature availability probability and the no-load interference suppression factor according to the physical location of the charging piles, and superimposing the vector distribution of the local density change rate to generate a multi-dimensional heat map including charging pile availability prediction and load-carrying capacity, includes:

[0150] 601. Construct a vibration feature interpolation field based on the directional characteristics of the physical location of the charging piles, and the interpolation weight of the vibration feature interpolation field is adjusted by the energy attenuation of the vibration propagation path in the vibration feature availability probability;

[0151] In step 601, the vibration feature interpolation field is a spatial interpolation model constructed based on the directional characteristics of the physical location of the charging piles, and its interpolation weight is adjusted by the energy attenuation rate of the vibration signal propagation path.

[0152] In the embodiments of the present application, according to the directional characteristics of the physical layout of the charging piles (such as the attenuation difference of the vibration propagation path of the north-south charging gun), the Kriging interpolation algorithm is used to construct a spatial interpolation field. The energy attenuation curve is fitted by the measured data of the vibration sensor (such as the attenuation coefficient of the metal bracket is 1.5 dB / m, and that of the concrete bracket is 2.2 dB / m), and the interpolation weights of adjacent charging pile nodes are adjusted to finally generate a vibration feature interpolation field reflecting the spatial distribution of device availability.

[0153] 602. Construct a no-load interference correction field based on the spatial proximity relationship of the physical location of the charging piles, and the correction weight of the no-load interference correction field is controlled by the response time of the grid reactive power compensation device in the no-load interference suppression factor;

[0154] In step 602, the no-load interference correction field is a correction model constructed based on the spatial proximity relationship of the charging piles, and its weight is controlled by the response time of the grid reactive power compensation device in the no-load interference suppression factor.

[0155] In the embodiments of the present application, the no-load mis-triggering frequency of the charging pile nodes (such as a certain node mis-triggers 8 times per hour) is statistically analyzed, combined with the response time of the grid reactive power compensation device (such as a delay of 3 seconds), and the spatial autoregressive model (SAR) is used to calculate the correction weight. The correction effect under different response times is predicted through Monte Carlo simulation (such as the weight is 0.85 when the delay is 2 seconds and 0.7 when the delay is 5 seconds), and a no-load interference correction field matrix is generated.

[0156] 603. Visualize the vibration feature interpolation field and the no-load interference correction field according to the physical location of the charging pile to generate an initial thermal distribution;

[0157] In step 603, the initial thermal distribution is an initial distribution map generated by superimposing the vibration feature interpolation field and the no-load interference correction field according to the physical location of the charging pile, reflecting the preliminary results of equipment availability and load correction.

[0158] In the embodiment of the present application, a point-by-point product operation is performed on the vibration feature interpolation field (availability probability 0 - 1) and the no-load interference correction field (correction coefficient 0 - 1). The voxelization rendering technology is used to map the result to a thermal base color (for example, red indicates high availability and low interference). The spatial missing values are filled by bilinear interpolation to generate a smoothly transitioning initial thermal distribution.

[0159] 604. Perform a spatial superposition of the vector distribution of the local density change rate and the initial thermal distribution to generate a multi-dimensional thermal map including the prediction of the charging pile availability and the load-bearing capacity.

[0160] In step 604, the multi-dimensional thermal map is a visualization chart generated by superimposing the vector distribution (such as the arrow direction) of the local density change rate and the initial thermal distribution, and its chromaticity is negatively correlated with the power factor compensation state of the power grid.

[0161] In the embodiment of the present application, the vector direction of the local density change rate is extracted (such as the density diffusion in the southeast direction), and the manifold learning algorithm (such as t-SNE) is used to project the vector field onto a two-dimensional plane. The vector arrows are superimposed on the initial thermal distribution through the transparency blending algorithm, and the chromaticity channel is rendered according to the power factor (for example, 0.88 is mapped to dark red), and finally a multi-dimensional thermal map including the prediction of the charging pile availability and the load-bearing capacity is generated.

[0162] The following is a specific example:

[0163] Suppose the charging piles in a smart park are radially arranged. The charging piles on the north side face the main road (the vibration propagation path attenuation is 1.2 dB / m), and the south side faces the parking lot (the attenuation is 2.0 dB / m). In step 601, a vibration feature interpolation field is constructed, with the interpolation weight of 0.92 on the north side and 0.75 on the south side; in step 602, it is detected that the no-load false trigger frequency on the south side is 12 times per hour, and the grid response delay is 4 seconds, obtaining a correction weight of 0.68 and constructing a no-load interference correction field; through step 603, the generated initial thermal distribution shows light green (availability 0.9) on the north side and orange-yellow (availability 0.7) on the south side; in step 604, the vector direction of the local density change rate (pointing to the central area on the north side) is superimposed, and the output multi-dimensional thermal map prompts the operation and maintenance personnel to send mobile charging vehicles to the north side to avoid overload in the center.

[0164] To improve the accuracy and anti-interference ability of charging pile load status recognition, a physical layer perception closed-loop is constructed through multi-band analysis and energy evolution modeling of vibration signals. Effective feature extraction is enhanced based on vibration signal acquisition in stress concentration areas. By combining spectral characteristics and energy trends, load and no-load interference are distinguished to achieve highly robust status recognition. In some embodiments, a piezoelectric vibration sensor is integrated inside the charging pile support structure. Based on the amplitude differences of different frequency components in the mechanical vibration signal of the charging pile, a specific vibration mode triggered by the plugging and unplugging actions of the charging gun is identified to distinguish the actual load operation state of the charging pile from no-load false trigger events, including:

[0165] 701. Arrange piezoelectric vibration sensors in the stress concentration areas of the charging pile support structure to collect multi-band vibration components in the mechanical vibration signal of the charging pile;

[0166] In step 701, the stress concentration area: the area in the charging pile support structure where significant vibration is likely to occur due to mechanical stress concentration. The multi-band vibration components are the vibration components in multiple frequency intervals obtained after the frequency domain decomposition of the mechanical vibration signal collected by the piezoelectric vibration sensor.

[0167] In the embodiments of the present application, piezoelectric vibration sensors are deployed at the bolt connection points and bracket welding points of the charging pile support structure, and the original vibration signal is collected through a high sampling rate (such as 10 kHz). The wavelet packet decomposition technique is used to decompose the signal into multiple frequency bands (such as sub-bands at 10 Hz intervals), and the time-domain amplitude sequence of each frequency band is extracted. Key frequency bands are selected through energy ratio screening (such as the frequency bands with an energy ratio > 5% are retained) to form multi-band vibration components.

[0168] 702. Extract the characteristic frequency bands in the multi-band vibration components that match the plugging and unplugging actions of the charging gun, and generate the vibration energy distribution of the charging pile based on the amplitude change law of the characteristic frequency bands;

[0169] In step 702, the vibration energy distribution is the evolution curve of the square integral value of the amplitude of each sub-band within the characteristic frequency band over time.

[0170] In the embodiments of the present application, independent component analysis (ICA) is performed on the multi-band vibration components to separate the characteristic frequency bands of the plugging and unplugging actions of the charging gun (such as the main frequency of 45 Hz and its second harmonic of 90 Hz). The short-time energy of each sub-band within the characteristic frequency band is calculated (such as the sum of the squared amplitudes in a 50 ms window), and the vibration energy distribution of the charging pile is generated. The energy curve is smoothed through moving average filtering to eliminate instantaneous impact noise (such as the sudden vibration caused by a vehicle passing by).

[0171] 703. According to the comparison result between the peak interval of the vibration energy distribution of the charging pile and a preset vibration energy threshold, identify the specific vibration mode triggered by the plugging and unplugging actions of the charging gun;

[0172] In step 703, the preset vibration energy threshold is the energy critical value set according to historical data statistics. The specific vibration mode is a vibration signal mode that meets the energy threshold condition and has specific spectral characteristics.

[0173] In the embodiment of the present application, the peak energy distribution of the characteristic frequency band of historical charging events is statistically analyzed (for example, the fundamental frequency energy peak interval is 70 - 100 dB), and a threshold is set (for example, when the fundamental frequency energy > 75 dB, it is an effective event). The support vector machine (SVM) is used to classify the energy distribution curve (such as unimodal, bimodal, multimodal), and the unimodal concentrated energy curve that conforms to the charging gun plugging and unplugging action is screened. The spectral flatness index (such as < 0.3 is effective) is combined to verify the mode consistency, and the specific vibration mode is identified.

[0174] 704. Based on the spectral distribution characteristics and energy fluctuation trend of the specific vibration mode, distinguish the actual load operation state of the charging pile from the no-load false trigger event.

[0175] In step 704, the spectral distribution characteristic is the energy concentration degree and harmonic attenuation law of the vibration signal in the frequency domain. The energy fluctuation trend is the stability or mutation characteristic of the vibration signal energy evolving over time.

[0176] In the embodiment of the present application, principal component analysis (PCA) is performed on the spectrum of the specific vibration mode, and key features such as the proportion of fundamental frequency energy and harmonic attenuation slope are extracted. The energy fluctuation trend is modeled by the hidden Markov model (HMM) (for example, the energy of a real charging event rises and falls slowly, and the energy of a no-load event has a rapid spike), and a joint discrimination rule is set to distinguish the actual load operation state of the charging pile from the no-load false trigger event: when the proportion of fundamental frequency energy > 65% and the energy fluctuation trend is stable, it is determined as load operation; when the fundamental frequency energy is dispersed (proportion < 50%) or the fluctuation is mutated, it is determined as a no-load false trigger.

[0177] The following is a specific example:

[0178] A charging pile is temporarily added in a smart park, and vibration signals are detected at the bolts of its metal bracket. Through wavelet packet decomposition in step 701, the frequency bands of 45 Hz, 90 Hz, and 135 Hz are extracted to form a multi-band vibration component data set; in step 702, ICA is used to separate the fundamental frequency energy of 45 Hz with a proportion of 70%, and the vibration energy distribution of the charging pile is calculated after calculating the short-time energy of each sub-band; in step 703, SVM classifies it as a unimodal concentrated type, and the fundamental frequency energy peak is 82 dB (exceeding the threshold of 75 dB), forming a specific vibration mode; through step 704, PCA shows that the proportion of fundamental frequency energy is 72%, and HMM determines that the energy trend is stable, confirming it as a real load. For another vibration event, the main frequency is 60 Hz, the energy is dispersed (proportion 48%), and HMM detects an energy mutation and marks it as a no-load false trigger.

[0179] Figure 2The figure shows a schematic structural diagram of a charging demand heat map generation system based on time series prediction provided by an embodiment of the present application. As Figure 2 shown, the system includes:

[0180] An analysis module 21, configured to establish a mapping relationship between the network traffic fluctuation law and the time series characteristics of the charging demand, and construct a cross-domain correlation model reflecting the coupling relationship between the charging demand and the network behavior by analyzing the periodic change law of the data transmission volume of the communication nodes in the area where the charging piles are located;

[0181] An identification module 22, configured to integrate piezoelectric vibration sensors inside the support structure of the charging pile, and based on the amplitude difference of different frequency components in the mechanical vibration signal of the charging pile, identify the specific vibration mode caused by the plugging and unplugging actions of the charging gun, and distinguish the actual load operation state of the charging pile from the no-load false trigger event;

[0182] A suppression module 23, configured to deploy a reactive power compensation device on the distribution side, and by adjusting the switching state of the compensation capacitor, suppress the power factor fluctuation of the power grid caused by the access of the charging pile cluster, and generate a steady-state power supply environment matching the charging demand prediction model;

[0183] A calculation module 24, configured to perform spatio-temporal association on the output result of the cross-domain correlation model and the identification result of the actual load operation state of the charging pile, and combine the power parameters of the steady-state power supply environment to calculate the demand distribution density of the charging pile group in the target area;

[0184] A generation module 25, configured to generate a multi-dimensional heat map including the prediction of the charging pile availability and the load-bearing capacity according to the spatial gradient change characteristics of the demand distribution density, and synchronously fuse the characteristic parameters of the specific vibration mode and the exclusion result of the no-load false trigger event. The chromaticity mapping rule of the multi-dimensional heat map is negatively correlated with the power factor compensation state of the power grid.

[0185] Figure 2 The described charging demand heat map generation system based on time series prediction can execute Figure 1 the charging demand heat map generation method described in the embodiment shown, and its implementation principle and technical effects will not be elaborated again. For the charging demand heat map generation system in the above embodiment, the specific manners in which each module and unit perform operations have been described in detail in the embodiment related to the method, and will not be elaborated here.

[0186] In a possible design, Figure 2 the charging demand heat map generation system in the embodiment shown can be implemented as a computing device. As Figure 3 shown, the computing device may include a storage component 31 and a processing component 32;

[0187] The storage component 31 stores one or more computer instructions, and the one or more computer instructions are for the processing component 32 to call and execute.

[0188] The processing component 32 is used for the above Figure 1 A method for generating a heat map of charging demand based on time series prediction according to the above embodiment.

[0189] Among them, the processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above method.

[0190] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0191] Of course, the computing device may necessarily further include other components, such as an input / output interface, a display component, a communication component, etc.

[0192] The input / output interface provides an interface between the processing component and the peripheral interface module, and the peripheral interface module may be an output device, an input device, etc.

[0193] The communication component is configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0194] Among them, the computing device may be a physical device or an elastic computing host provided by a cloud computing platform, etc. At this time, the computing device may refer to a cloud server, and the above processing component, storage component, etc. may be basic server resources leased or purchased from the cloud computing platform.

[0195] The embodiment of the present application also provides a computer storage medium storing a computer program, and when the computer program is executed by a computer, it can implement the above Figure 1 A method for generating a heat map of charging demand based on time series prediction according to the above embodiment.

[0196] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0197] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0198] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0199] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. A method for generating a heat map of charging demand based on time series prediction, characterized in that Including: Establish a mapping relationship between the network traffic fluctuation law and the time-series characteristics of charging demand, and analyze the periodic change law of the data transmission volume of communication nodes in the area where the charging pile is located based on the mapping relationship, so as to construct a cross-domain correlation model reflecting the coupling relationship between charging demand and network behavior; Integrate a piezoelectric vibration sensor inside the support structure of the charging pile, and identify the specific vibration mode caused by the plugging and unplugging action of the charging gun based on the amplitude difference of different frequency components in the mechanical vibration signal of the charging pile, so as to distinguish the actual load operation state of the charging pile from the no-load mis-trigger event; Deploy a reactive power compensation device on the distribution side, and suppress the power factor fluctuation of the power grid caused by the access of the charging pile cluster by adjusting the switching state of the compensation capacitor, and generate a steady-state power supply environment matching the charging demand prediction model; Perform spatio-temporal association on the output result of the cross-domain correlation model and the recognition result of the actual load operation state of the charging pile, and combine the power parameters of the steady-state power supply environment to calculate the demand distribution density of the charging pile group in the target area; According to the spatial gradient change characteristics of the demand distribution density, synchronously fuse the characteristic parameters of the specific vibration mode and the exclusion result of the no-load mis-trigger event, and generate a multi-dimensional heat map including the prediction of charging pile availability and load-bearing capacity; 2. The method according to claim 1, characterized in that, The performing spatio-temporal association on the output result of the cross-domain correlation model and the recognition result of the actual load operation state of the charging pile, and combining the power parameters of the steady-state power supply environment to calculate the demand distribution density of the charging pile group in the target area includes: Divide the recognition result of the actual load operation state of the charging pile into a discrete event sequence according to a preset time slice, and the load state within each time slice in the discrete event sequence is associated with the network traffic data peak interval corresponding to the time slice; Weight the load state within each time slice in the discrete event sequence according to the coupling weight of network traffic and charging demand output by the cross-domain correlation model to obtain a weighted result; Establish a constraint relationship between the power parameters of the steady-state power supply environment and the maximum instantaneous load capacity of the charging pile, and correct the capacity boundary of the weighted result based on the constraint relationship to generate the upper limit of the available power capacity of the charging pile node; Perform spatio-temporal superposition on the upper limit of the available power capacity and the network traffic data peak interval, and calculate and output the demand distribution density of the charging pile group in the target area through density field calculation, and the spatial resolution of the density field is jointly determined by the physical distance between the charging pile nodes and the response delay time of the reactive power compensation device; 3. The method according to claim 1, wherein The according to the spatial gradient change characteristics of the demand distribution density, synchronously fusing the characteristic parameters of the specific vibration mode and the exclusion result of the no-load mis-trigger event, and generating a multi-dimensional heat map including the prediction of charging pile availability and load-bearing capacity includes: Discretize the spatial gradient change characteristics of the demand distribution density into multiple local density change rates according to the physical layout of the charging pile group, and the direction vector of the local density change rate is associated with the vibration propagation direction of the charging gun plugging and unplugging action; Extract the vibration frequency distribution interval that matches the plugging and unplugging action of the charging gun among the characteristic parameters of the specific vibration mode, and generate a vibration characteristic availability probability characterizing the availability state of the charging pile based on the energy proportion of each frequency band within the vibration frequency distribution interval; According to the exclusion result of the no-load false triggering event, count the no-load false triggering frequency of the charging pile node within a preset time period, and generate a no-load interference suppression factor reflecting the load-bearing capacity correction factor of the charging pile in combination with the power parameters of the steady-state power supply environment; Perform spatial interpolation on the vibration characteristic availability probability and the no-load interference suppression factor according to the physical location of the charging pile, and superimpose the vector distribution of the local density change rate to generate a multi-dimensional heat map including the charging pile availability prediction and the load-bearing capacity; 4. The method according to claim 1, characterized in that The establishment of the mapping relationship between the network traffic fluctuation law and the time-series characteristics of the charging demand, and the analysis of the periodic change law of the data transmission volume of the communication nodes in the area where the charging pile is located based on the mapping relationship to construct a cross-domain correlation model reflecting the coupling relationship between the charging demand and the network behavior, including: Divide the network traffic monitoring unit according to the physical coverage range of the communication nodes in the area where the charging pile is located, perform multi-scale periodic decomposition on the data transmission volume of the communication nodes based on the communication protocol type, and extract the dominant periodic components of the hourly, daily, and weekly data transmission volumes within each monitoring unit; According to the time-series distribution characteristics of the charging gun plugging and unplugging events in the historical operation data of the charging pile, define the high-frequency demand period, stable demand period, and low-demand period of the charging demand, and establish the mapping rule between the charging demand period label and the amplitude-frequency characteristics of the dominant periodic component; For each network traffic monitoring unit, match the phase fluctuation interval of the dominant periodic component with the duration window of the corresponding charging demand period label, and generate the dynamic correlation weight between the network traffic fluctuation law and the time-series characteristics of the charging demand by introducing the amplitude-frequency characteristic mapping rule during the matching process; Based on the correlation weight, construct the coupling relationship function between the charging demand and the network behavior. The time-series offset of the dominant periodic component of the network traffic in the coupling relationship function is used to correct the cross-domain correlation error, and a structured cross-domain correlation model is output.

5. The method according to claim 2, wherein The spatio-temporal superposition of the available power capacity upper limit and the network traffic data peak interval, and the calculation of the demand distribution density of the charging pile group in the target area through density field, including: Divide the spatial grid unit of the target area according to the physical distance between the charging pile nodes, discretely sample the available power capacity upper limit according to the spatial grid unit, and generate the local power capacity distribution; Divide the network traffic data peak interval into multiple continuous time windows in the time dimension, extract the spatial distribution characteristics of the network traffic data within each time window, and generate the network traffic spatial weight corresponding to the spatial grid unit; Superimpose the local power capacity distribution and the network traffic spatial weight according to the spatial grid unit to generate a density field reflecting the coupling relationship between the power capacity of the charging pile group and the network traffic; Perform multi-scale convolution operations on the local power capacity distribution and the network traffic spatial weights according to the spatial grid cells, calculate the density increment of the spatial grid cells in the density field, and output the demand distribution density of the charging pile group in the target area.

6. The method according to claim 3, wherein The spatial interpolation of the vibration feature availability probability and the no-load interference suppression factor according to the physical positions of the charging piles, and the superposition of the vector distribution of the local density change rate to generate a multi-dimensional heat map including the charging pile availability prediction and the load-bearing capacity, includes: Construct a vibration feature interpolation field based on the directional features of the physical positions of the charging piles, and the interpolation weights of the vibration feature interpolation field are adjusted by the energy attenuation of the vibration propagation path in the vibration feature availability probability; Construct a no-load interference correction field based on the spatial proximity relationship of the physical positions of the charging piles, and the correction weights of the no-load interference correction field are controlled by the response time of the grid reactive power compensation device in the no-load interference suppression factor; Perform visualization processing on the vibration feature interpolation field and the no-load interference correction field according to the physical positions of the charging piles to generate an initial heat distribution; Perform spatial superposition on the vector distribution of the local density change rate and the initial heat distribution to generate a multi-dimensional heat map including the charging pile availability prediction and the load-bearing capacity.

7. The method according to claim 1, characterized in that, Integrate piezoelectric vibration sensors inside the charging pile support structure, and based on the amplitude differences of different frequency components in the mechanical vibration signal of the charging pile, identify the specific vibration mode caused by the charging gun plugging and unplugging actions to distinguish the actual load operation state of the charging pile from the no-load mis-trigger event, includes: Arrange piezoelectric vibration sensors in the stress concentration area of the charging pile support structure to collect multi-band vibration components in the mechanical vibration signal of the charging pile; Extract the characteristic frequency band that matches the charging gun plugging and unplugging action from the multi-band vibration components, and generate the charging pile vibration energy distribution based on the amplitude change law of the characteristic frequency band; Identify the specific vibration mode caused by the charging gun plugging and unplugging action according to the comparison result between the peak interval of the charging pile vibration energy distribution and the preset vibration energy threshold; Distinguish the actual load operation state of the charging pile from the no-load mis-trigger event based on the spectral distribution characteristics and energy fluctuation trend of the specific vibration mode.

8. A charging demand heat map generation system based on time series prediction, characterized in that, Includes: An analysis module for establishing a mapping relationship between the network traffic fluctuation law and the charging demand time series characteristics, and constructing a cross-domain correlation model reflecting the coupling relationship between the charging demand and the network behavior by analyzing the periodic change law of the data transmission volume of the communication nodes in the area where the charging piles are located; An identification module for integrating piezoelectric vibration sensors inside the charging pile support structure, and based on the amplitude differences of different frequency components in the mechanical vibration signal of the charging pile, identifying the specific vibration mode caused by the charging gun plugging and unplugging actions to distinguish the actual load operation state of the charging pile from the no-load mis-trigger event; A suppression module for deploying reactive power compensation devices on the distribution side, and by adjusting the switching state of the compensation capacitors, suppressing the power factor fluctuation of the power grid caused by the access of the charging pile cluster and generating a steady-state power supply environment matching the charging demand prediction model; A calculation module, configured to perform spatio-temporal association on the output result of the cross-domain correlation model and the recognition result of the actual load operation state of the charging pile, and calculate the demand distribution density of the charging pile group in the target area in combination with the power parameters of the steady-state power supply environment; A generation module, configured to synchronously fuse the characteristic parameters of the specific vibration mode and the exclusion result of the no-load mis-trigger event according to the spatial gradient change characteristics of the demand distribution density, and generate a multi-dimensional heat map including the prediction of the charging pile availability and the load-bearing capacity, wherein the chromaticity mapping rule of the multi-dimensional heat map is negatively correlated with the power factor compensation state of the power grid.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for generating a charging demand heat map based on time series prediction according to any one of claims 1 to 7.

10. A computer storage medium, characterized in that, A computer program is stored, and when the computer program is executed by a computer, it implements a method for generating a charging demand heat map based on time series prediction according to any one of claims 1 to 7.

Images