A Method and System for Analyzing the Charging Cost of a Liquid-Cooled Ultra-Fast Charging Terminal

Abstract

An embodiment of the present invention discloses a method and system for analyzing the charging cost of a liquid-cooled ultra-fast charging terminal, belonging to the technical field of data analysis. Through multi-dimensional data collection and in-depth behavior analysis, this method can accurately depict the charging behavior characteristics of users, effectively improve the accuracy of charging demand prediction and resource scheduling, and can formulate a differentiated service strategy to enhance the personalized adaptation ability of charging services by converting user behavior characteristics into quantifiable parameters. In addition, the generation and execution mechanism of the cross-terminal dynamic scheduling strategy can optimize the allocation efficiency of power resources in the space-time dimension and alleviate the local overload problem during peak hours, while the coordinated adjustment function of real-time power supply parameters and cost calculation logic can balance the grid operation stability while ensuring the charging efficiency. In summary, this method can significantly improve the resource utilization efficiency of charging facilities.

Description

Technical Field

[0001] The embodiments of the present invention belong to the technical field of data analysis, and particularly relate to a method and system for analyzing the charging cost of a liquid-cooled ultra-fast charging terminal. Background Art

[0002] A liquid-cooled ultra-fast charging terminal is an advanced charging device of great significance in the field of electric vehicle charging. It uses liquid-cooling technology to solve the heat dissipation problem during the charging process. During ultra-fast charging, due to the high charging power, the traditional air-cooling method often fails to meet the heat dissipation requirements. However, the liquid-cooling technology can efficiently carry away heat through the circulation of the coolant inside the charging device, thus ensuring the stability and safety of the charging device during high-power charging.

[0003] The liquid-cooled ultra-fast charging terminal can achieve a charging power far exceeding that of ordinary charging devices, greatly shortening the charging time of electric vehicles, which plays a key role in improving the usability of electric vehicles. For example, some liquid-cooled ultra-fast charging terminals can supplement a large amount of electric energy for electric vehicles in a short time, meeting the urgent charging needs of users and making scenarios such as long-distance travel of electric vehicles more feasible.

[0004] However, with the wide application of liquid-cooled ultra-fast charging terminals, some problems have emerged. Specifically, existing technical solutions often struggle to maintain the stable operation of the power grid while ensuring charging efficiency, and thus it is difficult to guarantee the resource utilization efficiency of charging facilities. Summary of the Invention

[0005] The embodiments of the present invention provide a method and system for analyzing the charging cost of a liquid-cooled ultra-fast charging terminal, which can solve or partially solve the technical problems involved in the above background art.

[0006] The embodiments of the present invention provide a method for analyzing the charging cost of a liquid-cooled ultra-fast charging terminal, which is applied to the charging middle platform system of the liquid-cooled ultra-fast charging terminal. The method includes: collecting multiple groups of operation status data sets of the liquid-cooled ultra-fast charging terminal, where the operation status data sets include records of the chronological change of charging power, the distribution of user charging cycles, and the corresponding segmented cost settlement results; performing multi-dimensional charging behavior analysis on the operation status data sets to generate a user behavior feature set including charging time period preference patterns, power adjustment rules, and cost response intensities; constructing charging preference tags for the in-vehicle charging end based on the user behavior feature set, where the charging preference tags include time period priority weights, power stability thresholds, and cost tolerance intervals; generating a dynamic scheduling strategy across liquid-cooled ultra-fast charging terminals according to the charging preference tags, where the dynamic scheduling strategy includes power reallocation paths and cost compensation rules; and optimizing the real-time power supply parameters and cost calculation logic of the target liquid-cooled ultra-fast charging terminal according to the dynamic scheduling strategy to achieve the coordinated improvement of charging efficiency and resource utilization rate.

[0007] An embodiment of the present invention provides a charging middle platform system for a liquid-cooled ultra-fast charging terminal, including at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the above method.

[0008] An embodiment of the present invention provides a readable storage medium, on which a program or instructions are stored, and when the program or instructions are executed by a processor, the steps of the above method are implemented.

[0009] Through multi-dimensional data collection and in-depth behavior analysis, the embodiment of the present invention can accurately depict the charging behavior characteristics of users, effectively improve the accuracy of charging demand prediction and resource scheduling, and can formulate a differentiated service strategy to enhance the personalized adaptation ability of charging services by converting user behavior characteristics into quantifiable parameters. In addition, the generation and execution mechanism of the cross-terminal dynamic scheduling strategy can optimize the allocation efficiency of power resources in the space-time dimension and alleviate the local overload problem during peak hours, while the coordinated adjustment function of real-time power supply parameters and cost calculation logic can balance the grid operation stability while ensuring the charging efficiency. In summary, the embodiment of the present invention can significantly improve the resource utilization efficiency of charging facilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 It is a flowchart of a method for analyzing the cost of charging a liquid-cooled ultra-fast charging terminal provided by an embodiment of the present invention.

[0011] Figure 2 It is a schematic structural diagram of a charging middle platform system for a liquid-cooled ultra-fast charging terminal provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0012] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without creative efforts shall fall within the protection scope of the embodiments of the present invention.

[0013] In the embodiments of the present invention, terms such as "first" and "second" are used to distinguish similar objects, rather than to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the embodiments of the present invention can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are usually of the same category, and the number of objects is not limited. For example, the first object can be one or multiple. In addition, in the embodiments of the present invention, "and / or" means at least one of the connected objects, and the character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0014] Figure 1 Disclosed is a method for analyzing the charging cost of a liquid-cooled ultra-fast charging terminal, which is applied to the charging middle platform system of the liquid-cooled ultra-fast charging terminal. The method includes the following steps 110-step 150.

[0015] Step 110: Collect multiple groups of operating state data sets of the liquid-cooled ultra-fast charging terminal. The operating state data set includes records of the chronological change of the charging power, the distribution of the user charging cycle, and the corresponding segmented cost settlement results.

[0016] In the embodiments of the present invention, the operating state data set refers to a multi-dimensional data set of the charging process recorded in real time through the sensor network of the liquid-cooled ultra-fast charging terminal, including records of the chronological change of the charging power, the distribution of the user charging cycle, and the segmented cost settlement results. Among them, the record of the chronological change of the charging power stores the instantaneous power values in the form of a time series (such as the fluctuating data of 150kW to 200kW collected per second), reflecting the dynamic change of the power load during the charging process; the distribution of the user charging cycle records the start time, duration, and charging amount of each user's charging (such as the charging behavior of the user from 19:00 to 21:00 every day), which is used to analyze the charging habits; the segmented cost settlement result is associated with the electricity price policy and the actual payment data of the user (such as the segmented record of the user paying 1.2 yuan / kWh and 0.8 yuan / kWh under the peak-valley electricity price), and the segmented cost settlement result can be used for cost sensitivity analysis.

[0017] Specifically, the charging middle platform system of the liquid-cooled ultra-fast charging terminal captures the chronological change record of the charging power in real time through the multi-channel sensor network deployed at the terminal, including the instantaneous power value, the fluctuation amplitude, and the duration in seconds. For example, a certain terminal records the process of the power value gradually decreasing from the initial peak of 200kW to 150kW during a continuous 30-minute charging cycle, forming a time series curve containing 600 data points.

[0018] In addition, the user charging cycle distribution data is collected through the terminal operation interface, covering the start timestamp, end timestamp, and charging amount of each charging. For example, user A starts charging from 19:00 to 21:00 every day for a continuous week, and the charging amount is stable in the range of 80kWh to 85kWh.

[0019] Furthermore, the segmented fee settlement result is generated by the charging module, recording the stepped changes in electricity prices during different time periods and the actual payment amounts of users. For example, under the policy of a peak-time electricity price of 1.5 yuan per kilowatt-hour and a valley-time electricity price of 0.8 yuan per kilowatt-hour, user B has a segmented charging record due to crossing the peak-valley time period during a certain charging. The above data is synchronized to the central data storage node through an encrypted transmission protocol, and a timestamp alignment algorithm is used to eliminate the time deviation of multi-terminal data, ensuring the temporal consistency of the dataset. For abnormal data (such as invalid records with power suddenly dropping to zero), the charging middle platform system of the liquid-cooled ultra-fast charging terminal performs cleaning through a sliding window mean filtering algorithm. For example, interpolation repair is performed on abnormal points where the power value deviates from the mean by more than 30% within 5 consecutive seconds.

[0020] Step 120: Perform multi-dimensional charging behavior analysis on the operating state dataset to generate a user behavior feature set including charging time period preference patterns, power adjustment rules, and cost response intensities.

[0021] In the embodiment of the present invention, multi-dimensional charging behavior analysis is a process of extracting user behavior features based on the operating state dataset through an algorithm model. The charging time period preference pattern refers to the charging frequency and duration distribution of users in different time periods (such as the charging proportion of a certain user group exceeding 40% during the evening peak period), and the high-frequency time period interval is identified through a density clustering algorithm; the power adjustment rule describes the behavioral characteristics of users actively or passively adjusting the charging power (such as the average response time when the power drops from 200 kW to 150 kW), and the typical power curve is matched through a dynamic time warping algorithm; the cost response intensity quantifies the sensitivity of users to cost changes (such as the probability of power reduction increasing by 25% when the electricity price increases by 0.1 yuan), and is calculated by combining the behavior trigger rate and the delay time.

[0022] Specifically, the charging middle platform system of the liquid-cooled ultra-fast charging terminal performs multi-dimensional analysis on the preprocessed dataset. First, a density-based DBSCAN clustering algorithm is used to analyze the time period preference of the charging start timestamp. For example, after dividing 100,000 charging records by hour granularity, it is identified that the proportion of charging records in the time period from 19:00 to 21:00 reaches 45%, which is significantly higher than other time periods, and is defined as the high-frequency charging time period group.

[0023] The extraction of the power adjustment rule relies on the morphological analysis of the power time series curve. The charging middle platform system of the liquid-cooled ultra-fast charging terminal matches the typical power change pattern through a dynamic time warping algorithm (DTW). For example, it is detected that a certain type of user is used to manually adjusting the power from 200 kW to 150 kW within 10 minutes after charging starts, and the recurrence rate of this behavior in historical data reaches 72%.

[0024] Further, the calculation of the cost response intensity combines segmented cost data with user behavior adjustment records. For example, statistics show that when the real-time electricity price exceeds 1.2 yuan per kilowatt-hour, 65% of users will trigger a power reduction operation, and the average response delay is 3.2 seconds. The liquid-cooled ultra-fast charging terminal charging middleware system generates a user behavior feature set by constructing a three-dimensional feature space (time period distribution density, power adjustment frequency, cost response speed), and each feature dimension is normalized to eliminate the influence of dimensions.

[0025] Step 130: Construct a charging preference label for the in-vehicle charging terminal based on the user behavior feature set, where the charging preference label includes a time period priority weight, a power stability threshold, and a cost tolerance interval.

[0026] In the embodiment of the present invention, the charging preference label is an aggregated expression of user behavior features, including a time period priority weight, a power stability threshold, and a cost tolerance interval. The time period priority weight is calculated by weighting the time period coverage frequency and the power density (for example, a certain time period is given a weight of 0.7 due to high-frequency use and high power demand); the power stability threshold defines the acceptable power fluctuation range for users (for example, fluctuations within ±5% are regarded as stable states), and is determined based on historical adjustment behavior statistics; the cost tolerance interval delimits the acceptance boundary of users for cost changes (for example, the tolerable interval is 1.0 yuan to 1.2 yuan), and is generated through the correlation analysis of cost distribution and behavior feedback.

[0027] For example, based on the user behavior feature set, the liquid-cooled ultra-fast charging terminal charging middleware system can construct a charging preference label through a weighted fusion algorithm.

[0028] (1) The calculation of the time period priority weight comprehensively considers the time period coverage frequency and the average power density within the time period. For example, for a certain user, the charging frequency ratio in the time period from 19:00 to 21:00 is 38%, and the average power density reaches 180 kW. After linear weighting, the weight of this time period is set to 0.62.

[0029] (2) The power stability threshold is determined by statistically analyzing the acceptable power fluctuation range for users. For example, it is found through analysis that the adjustment probability of a certain user group increases to 85% when the power fluctuation exceeds ±8%, so the threshold is set to ±7.5%.

[0030] (3) The cost tolerance interval is generated using a sliding window dynamic calibration algorithm. For example, after dividing the cost data into three sub-intervals of 0.6 - 0.8 yuan, 0.8 - 1.0 yuan, and 1.0 - 1.2 yuan, the proportions of users terminating charging in each interval are 5%, 12%, and 28% respectively. Based on this, 1.1 yuan is defined as the high tolerance boundary. The label parameters are dynamically updated through an incremental learning mechanism. For example, when it is detected that a certain user still maintains high-power charging at an electricity price of 1.15 yuan for three consecutive times, the upper limit of the cost tolerance interval of the system is adjusted from 1.2 yuan to 1.25 yuan.

[0031] Step 140: Generate a dynamic scheduling strategy for the cross-liquid-cooled ultra-fast charging terminal according to the charging preference label, where the dynamic scheduling strategy includes a power redistribution path and a cost compensation rule.

[0032] In the embodiment of the present invention, the dynamic scheduling strategy is a resource allocation and cost regulation rule generated by the charging middle platform system of the liquid-cooled ultra-fast charging terminal according to the charging preference label. The power redistribution path refers to the topological structure for optimizing power distribution among multiple terminals (such as diverting 20% of the power of an overloaded terminal to an adjacent idle terminal), which is dynamically generated based on the load prediction model and grid constraint conditions; the cost compensation rule is a compensation mechanism designed to balance the cost difference caused by power adjustment (such as compensating users whose costs exceed the limit due to diversion by 20% of the deviation value), which is triggered in combination with the cost tolerance interval and user priority grading.

[0033] In this step, the charging middle platform system of the liquid-cooled ultra-fast charging terminal generates a cross-terminal dynamic scheduling strategy according to the charging preference label. The formulation of the power redistribution path relies on the load prediction model. For example, it is predicted that the total regional demand power will reach 15 MW during the period from 19:00 to 21:00 the next day, while the grid supply capacity is 12 MW, and the system automatically generates a path plan to divert 3 MW of demand to adjacent terminals. The cost compensation rule adopts a hierarchical response mechanism. For example, for users whose costs increase due to power diversion, when the deviation value is within the tolerance interval (less than 0.5 yuan), a fixed ratio compensation (20% of the deviation value) is adopted, and when it exceeds the interval, a dynamic compensation algorithm (30% of the deviation value + basic compensation) is activated. The multi-objective optimization algorithm is introduced in the strategy generation process, and indicators such as the terminal load balance degree (target value > 0.85), user satisfaction (predicted value > 4.2 / 5), and grid stability (volatility < 5%) are weighed at the same time, and the optimal strategy is selected through Pareto front analysis.

[0034] Step 150: Optimize the real-time power supply parameters and cost calculation logic of the target liquid-cooled ultra-fast charging terminal according to the dynamic scheduling strategy to achieve the coordinated improvement of charging efficiency and resource utilization rate.

[0035] In the embodiment of the present invention, the real-time power supply parameters are a set of operating parameters adopted by the liquid-cooled ultra-fast charging terminal when executing the dynamic scheduling strategy, including power upper limit, priority queue, response delay, etc. For example, during peak hours, a power upper limit of 220 kW is set for high-priority users, and low-priority users are limited to 180 kW. The cost calculation logic refers to the algorithm process for generating user costs according to real-time electricity prices and compensation rules (such as superimposing basic electricity prices, diversion compensation, and peak-valley discounts), which needs to meet the requirements of calculation real-time and accuracy (such as millisecond-level response and ±0.5% error control).

[0036] Charging efficiency and resource utilization rate are the core indicators for measuring system performance. Charging efficiency is calculated by the ratio of the amount of electric energy transmitted per unit time to the theoretical maximum value (e.g., the actual transmission efficiency is increased from 85% to 92%); the resource utilization rate comprehensively evaluates the collaborative utilization level of power facilities, terminal equipment, and grid load (e.g., the regional grid load rate is optimized from 75% to 88%). The two are synergistically improved through the continuous optimization of the dynamic scheduling strategy. For example, while ensuring the charging speed of users, the peak load of the grid is reduced.

[0037] In the actual application process, the dynamic scheduling strategy can be sent to the liquid-cooled ultra-fast charging terminal for execution through control instructions. The power supply parameter optimization adopts a fuzzy control algorithm. For example, when it is detected that the real-time load of a certain terminal reaches 90%, the initial power of newly connected users is automatically limited from 200kW to 180kW, and the difference of 20kW is dynamically allocated to high-priority users. The cost calculation logic integrates a real-time electricity price prediction module. For example, when it is predicted that the photovoltaic power generation will increase suddenly the next day in combination with weather forecast data, a power price reduction strategy is generated 6 hours in advance, and the valley-time electricity price is adjusted from 0.8 yuan to 0.75 yuan. The system performs a global status scan every 5 minutes. When it is detected that the deviation between the actual resource utilization rate and the predicted value exceeds 8%, the policy regeneration process is triggered. The execution effect evaluation adopts the A / B test mechanism. For example, 10% of the terminals are designated as the control group to maintain the original policy. After the experimental group adopts the new policy, the resource utilization rate is increased by 12% and the user complaint rate is decreased by 18%. After verifying the effectiveness of the policy, it is pushed in full volume.

[0038] Based on the above content, the technical solutions described in steps 110 - 150 are demonstrated through an application scenario example below.

[0039] In the liquid-cooled ultra-fast charging demonstration area of a city with a high penetration rate of new energy vehicles, at 19:00 on a certain weekday evening, a charging peak arrives, and the full-link intelligent regulation of the charging middle platform system of the liquid-cooled ultra-fast charging terminal is started. At this time, the real-time load rates of 12 liquid-cooled ultra-fast charging terminals in the area reach 92%. The system first collects the operation status data set from the terminal sensor array: the power time series record shows that the power value of terminal 3 fluctuates and drops from 200kW to 165kW during the period from 18:55 to 19:10, corresponding to a charging volume of 28kWh; the user charging cycle data shows that user C has charged 5 times in the 19:00 - 20:30 period in the past 7 days, with an average single charging volume of 72kWh; the segmented cost record shows that the cost of this user's last charge increased by 19.6 yuan because it crossed the 19:00 peak-time electricity price node (from 0.8 yuan to 1.2 yuan per kilowatt-hour).

[0040] When the system performs multi-dimensional analysis on massive data, through the improved OPTICS clustering algorithm, it identifies that there are three typical charging time period groups in this area: the morning commuting group (07:30 - 09:00, accounting for 18%), the lunchtime charging group (12:00 - 13:30, accounting for 23%), and the evening peak group (19:00 - 21:30, accounting for 59%). For the power fluctuation curve of user C, dynamic time warping matching finds that there is a regular behavior of actively reducing the power to 80% of the rated value after charging for 10 minutes, and it is triggered and marked as a "high-cost sensitive user". The cost response intensity calculation module detects that when the electricity price exceeds 1.1 yuan for this user, the power reduction probability reaches 83%, and the average response delay is 2.8 seconds, which is significantly higher than the regional average of 65% and 4.5 seconds.

[0041] Based on the above characteristics, the system constructs the charging preference labels for user C: in the time period priority weight, the evening peak time period is assigned a weight of 0.68 (40% of the time period frequency ratio and 60% of the power density contribution in the calculation formula); the power stability threshold is set at ±6% (because the probability of charging interruption surges when the power fluctuation exceeds 7% in its historical data); the cost tolerance interval is dynamically calibrated to 0.9 - 1.15 yuan (originally 0.8 - 1.2 yuan, and the upper limit is extended because it continued to charge at 1.18 yuan twice recently). At the same time, the regional heat map shows that there are 4 terminals within 500 meters around terminal 3 in an idle state, and the load rate is only 35%.

[0042] The dynamic scheduling strategy engine immediately generates an optimization plan: among the current 200kW power distribution of terminal 3, 35kW is shunted to terminal 5 (idle rate 65%) through the topological network, and a cost critical warning of 1.15 yuan is set for user C. The cost compensation rule module calculates that the charging time is extended by 8 minutes due to the shunt, and 25% of the deviation value of 0.7 yuan (predicted cost difference) is automatically compensated as points into the user's account. The real-time power supply parameter regulation unit synchronously updates the maximum output power of terminal 3 to 165kW and reserves 10kW of elastic capacity for high-priority users.

[0043] After implementing this strategy, the system monitors that the overall load balance degree of the area has increased from 0.62 to 0.79. The actual charging cost of user C has decreased by 11.3% compared with the prediction, and the charging completion time has only been extended by 6 minutes (originally predicted 8 minutes). During the low valley period at 1:00 am the next day, the system automatically starts the reverse optimization of the compensation strategy: the shunted power of terminal 5 is called back to terminal 3, and the remaining 12kWh of charging demand is completed using the low valley electricity price (0.6 yuan) to achieve the optimal allocation of cross-time resources. After 30 days of operation verification, the peak utilization rate of charging piles in this area has decreased from 88% to 76%, the average user satisfaction score has increased from 4.1 to 4.7, and the peak-valley difference rate of the power grid has decreased by 14.6%, fully demonstrating the technical value from data perception to strategy closed-loop.

[0044] In an optional embodiment, the multi-dimensional charging behavior analysis of the operating state data set to generate a user behavior feature set including charging period preference patterns, power adjustment rules, and cost response intensities includes:

[0045] Screen a selected subset of charging records that meet the preset integrity conditions from the operating state data set, where the selected subset of charging records includes the power fluctuation trajectory and corresponding cost segmentation data within a complete charging cycle;

[0046] Conduct a charging period distribution analysis on the selected subset of charging records to identify high-frequency charging period groups and corresponding power peak densities;

[0047] Extract the trigger condition parameters of power sudden drop events in the power fluctuation trajectory to generate a dynamic event sequence including the sudden drop time point and the sudden drop amplitude;

[0048] Perform spatio-temporal correlation analysis on the high-frequency charging period group and the dynamic event sequence to determine the response relationship of the cost segmentation data to the power adjustment behavior;

[0049] Generate the period preference pattern, power adjustment rule, and cost response intensity in the user behavior feature set based on the response relationship and the power peak density.

[0050] In this embodiment, the technical solution of the liquid-cooled ultra-fast charging terminal charging middleware system for multi-dimensional charging behavior analysis of the operating state data set specifically involves the steps of constructing a user behavior feature set through data screening, feature extraction, and correlation analysis. The liquid-cooled ultra-fast charging terminal charging middleware system first screens the subset of charging records according to the preset integrity conditions, specifically including charging cycle integrity verification and data continuity verification. The charging cycle integrity verification requires that a single charging record must include the start timestamp, end timestamp, and complete power fluctuation trajectory, excluding fragmented records caused by equipment failures or communication interruptions; the data continuity verification uses a timestamp continuity detection algorithm to ensure that the interval between adjacent data points in the power fluctuation trajectory does not exceed a set threshold (such as 1 second). The selected subset of charging records after screening includes charging cycle data with clear start and end boundaries and complete fluctuation characteristics. For example, a charging process lasting 45 minutes includes 2700 continuous power sampling points and 24 corresponding segmented billing records.

[0051] When performing a charging period distribution analysis on a selected subset of charging records, the charging middle - platform system of the liquid - cooled ultra - fast charging terminal divides the charging start time into time slots according to a preset time granularity (such as 15 minutes), and uses an improved density - based clustering algorithm to identify high - frequency charging period groups. The specific process includes: calculating the distribution density of the number of charging records in each time slot, setting a density threshold (such as exceeding 2 times the standard deviation of the mean) to screen out candidate time slots; performing a time - window expansion process on adjacent candidate time slots to generate continuous high - frequency charging period intervals (such as 19:00 - 21:30). The power peak density calculation module then counts the occurrence frequency and the proportion of the duration of the power peak in the power fluctuation trajectory within each high - frequency period interval. For example, within a certain period interval, it is detected that the power peak appears 3.2 times per minute on average, and the peak maintenance duration accounts for 68% of the total duration.

[0052] In the power sudden - drop event detection stage, the charging middle - platform system of the liquid - cooled ultra - fast charging terminal performs morphological analysis on the power fluctuation trajectory, and defines a power sudden - drop event as a continuous power value drop exceeding a preset threshold (such as 15%) and the duration being shorter than a set window (such as 30 seconds). The trigger condition parameter extraction module records the start time point, end time point, and the absolute value of the power difference of each sudden - drop event. For example, it is detected that during a certain charging process, at 19:12:35, the power suddenly drops from 180 kW to 145 kW, and the sudden - drop amplitude is 19.4%. The dynamic event sequence generation unit sorts the sudden - drop events according to the time stamp and constructs an event chain structure with spatio - temporal attributes. For example, a structured sequence containing event numbers, sudden - drop amplitudes, and the time interval between adjacent events is generated.

[0053] In the spatio - temporal correlation analysis stage, the charging middle - platform system of the liquid - cooled ultra - fast charging terminal performs multi - dimensional mapping of the high - frequency charging period group and the dynamic event sequence: establishing a first analysis dimension with the high - frequency period interval as the time baseline, and a second analysis dimension with the sudden - drop time point as the event base point. The segmented cost data is mapped to the corresponding dimension through the time - stamp alignment algorithm to generate a cost - change heat map. For example, when it is detected that the electricity price rises from 0.8 yuan to 1.2 yuan in the 19:00 - 20:00 period, it is associated that the occurrence frequency of sudden - drop events within this period increases by 42%. The response relationship calculation engine analyzes the correlation between the sudden - drop amplitude and the cost fluctuation value. For example, when the cost increases by 0.1 yuan, the average sudden - drop amplitude increases by 5.3%, and the response delay shortens by 1.8 seconds. Based on this, a cost response intensity index is constructed (such as defining the response coefficient as the change rate of the power adjustment amplitude caused by a unit change in cost).

[0054] The generation process of the user behavior feature set integrates the above multi-dimensional analysis results: the time period preference pattern is calculated by weighted calculation of the high-frequency charging time period interval and its power peak density. For example, the time period weight coefficient of a certain user group in the time period from 19:00 to 21:00 is 0.72. The power adjustment rule is feature-encoded through the statistical distribution of the frequency and amplitude of sudden drop events (such as defining the adjustment intensity as the ratio of the mean value of the sudden drop amplitude to the standard deviation). The cost response intensity generates a three-dimensional feature vector by integrating the cost-related response coefficient and the delay time decay factor. The feature set is standardized to eliminate the dimension difference and is reduced in dimension through principal component analysis to form an interpretable feature space. For example, the original 12-dimensional data is compressed to 3 main component dimensions, retaining more than 85% of the information volume.

[0055] In this way, through multi-dimensional data correlation analysis, deep feature extraction of user charging behavior is realized, improving the accuracy of charging demand prediction and resource scheduling. The spatio-temporal correlation model based on the dynamic event sequence enhances the quantitative evaluation ability of the impact of cost policies. The standardized feature set generation mechanism provides a unified decision-making input benchmark for cross-terminal collaborative scheduling, effectively supporting the collaborative optimization of charging efficiency and resource utilization rate.

[0056] In a preferred technical solution, the analysis of the charging time period distribution of the selected subset of charging records to identify the high-frequency charging time period group and the corresponding power peak density includes: clustering the charging start times in the selected subset of charging records according to a preset time granularity to form multiple time period distribution clusters;

[0057] Calculating the time period coverage frequency of each time period distribution cluster, where the time period coverage frequency is determined by the ratio of the number of charging records in the cluster to the total number of records;

[0058] Screening the time period distribution clusters with the time period coverage frequency exceeding the preset frequency threshold as candidate heat time period groups;

[0059] Expanding the charging start times in the candidate heat time period groups by a time window to generate continuous active charging time period intervals;

[0060] Statistically analyzing the peak occurrence frequency and the duration ratio of the power fluctuation trajectory in the active charging time period interval to generate a corresponding power peak density matrix.

[0061] Furthermore, the extraction of the trigger condition parameters of the power sudden drop events in the power fluctuation trajectory to generate a dynamic event sequence including the sudden drop time point and the sudden drop amplitude includes:

[0062] Detecting the section where the continuous power values in the power fluctuation trajectory drop by more than the preset sudden drop threshold as a candidate sudden drop event;

[0063] Recording the start time point, end time point and absolute value of the power difference of each candidate sudden drop event;

[0064] Calculate the ratio of the absolute value of the power difference to the starting power value as the sudden drop amplitude;

[0065] Sort the candidate sudden drop events in a time series according to the starting time point to generate a dynamic event sequence;

[0066] Analyze the interval duration distribution of adjacent events in the dynamic event sequence to determine the trigger frequency and associated characteristics of the sudden drop events.

[0067] In the above technical solution, the charging middle platform system of the liquid-cooled ultra-fast charging terminal analyzes the charging time period distribution of the selected charging record subset and extracts the technical solution of the power sudden drop event parameters, and realizes the deep feature mining of user behavior through data clustering and dynamic event modeling. When the system performs density clustering on the charging start time according to the preset time granularity, it uses a density estimation algorithm based on the Gaussian kernel function to divide the time period distribution clusters. The preset time granularity is set as a configurable parameter, and the typical value is 15 minutes to ensure that the time period division has sufficient time resolution. For example, after dividing the charging records of 30 consecutive days into 2880 time period slots at 15-minute intervals, the system calculates the probability density distribution of the number of charging records in each time period slot and identifies the time period clusters with a density exceeding the preset frequency threshold. The calculation formula for the time period coverage frequency is the ratio of the number of charging records in the cluster to the total number of records. For example, if a certain time period cluster contains 1200 records and the total number of records is 10000, the time period coverage frequency is 12%. The screening of the candidate heat time period group is dynamically adjusted according to the preset frequency threshold. When detecting changes in the regional charging load, the system automatically increases the threshold from the initial value of 8% to 10% to focus on the core time period.

[0068] When expanding the time window of the candidate heat time period group, the system uses a sliding window merging algorithm to eliminate the discrete cluster gaps. For example, a certain candidate heat time period group contains two adjacent time period clusters of 19:15 - 19:30 and 19:30 - 19:45. After merging, a continuous active charging time period interval of 19:15 - 19:45 is generated. The calculation module of the power peak density then counts the peak attributes of the power fluctuation trajectory within this interval: the peak occurrence frequency is based on the number of power extreme values detected within a one-minute statistical period (such as 3.2 peaks are detected per minute in a certain time period), and the duration ratio is obtained by calculating the ratio of the peak maintenance duration to the total duration (such as the peak maintenance duration reaches 68%). The construction of the power peak density matrix uses a two-dimensional coordinate system, with the horizontal axis being the time dimension (such as sliced by 5 minutes) and the vertical axis being the power level (such as stratified by 50kW intervals). The matrix element value is the peak density index of the corresponding space-time unit (such as an index of 0.85 for a certain unit indicates the density degree of the power level at that time).

[0069] The detection of power dip events is based on the first derivative analysis of the power fluctuation trajectory. The system defines the dip threshold as the percentage of the continuous power drop amplitude exceeding the rated value (e.g., 15%) and the duration being less than the set window (e.g., 30 seconds). The identification process of candidate dip events uses a sliding window difference algorithm: calculate the power change rate of adjacent data points with a step size of 1 second, mark the starting time point of the event when the cumulative drop amplitude exceeds the threshold, and record the ending time point until the power stabilizes or the drop terminates. For example, if it is detected that the power continuously drops from 180 kW to 153 kW (a 15% drop) within 5 seconds during a certain charging process, it is marked as a valid dip event. The extraction of trigger condition parameters includes recording the starting time point, ending time point, and the absolute value of the power difference (e.g., the difference of 27 kW in this example). The dip amplitude is calculated as the ratio of the power difference to the starting power value (27 / 180 = 15%).

[0070] The generation of the dynamic event sequence is achieved through a timestamp sorting algorithm, arranging the candidate dip events in ascending order of the starting time point to form an ordered event chain. For example, during a charging cycle, three dip events are detected with timestamps {19:12:35}, {19:18:20}, {19:25:44} respectively, then an event sequence including the timing position and the dip amplitude is generated. The statistical analysis of the trigger frequency is based on the distribution characteristics of the event interval duration. The system uses the kernel density estimation method to calculate the probability density of the interval time between adjacent events. For example, it is detected that the probability of the interval time being in the range of 3 - 5 minutes reaches 65%. The correlation feature analysis module identifies the multi - event correlation pattern through event co - occurrence detection. For example, it is found that the probability of a certain type of user triggering a power reduction again within 10 minutes after a dip event increases by 40%.

[0071] Applying the above - mentioned embodiments, accurately identify high - frequency charging periods through density clustering and time - window expansion, providing the ability to locate spatio - temporal hotspots for resource scheduling; reveal the triggering rules of power dip behaviors based on dynamic event sequence modeling, enhancing the prediction accuracy of the system for users' cost - sensitive behaviors; the construction of the power peak density matrix can realize the quantitative characterization of the charging load distribution, thus supporting the collaborative optimization of power grid stability analysis and terminal power allocation strategies.

[0072] In another preferred technical solution, the spatio - temporal correlation analysis of the high - frequency charging period group and the dynamic event sequence to determine the response relationship of the cost - segmented data to the power adjustment behavior includes:

[0073] Establish a first analysis layer with the active charging period interval as the time dimension;

[0074] Establish a second analysis layer with the dip time points in the dynamic event sequence as the event dimension;

[0075] Map the segmented cost data to the first analysis layer according to the timestamp to generate a heat distribution map of cost changes;

[0076] Match and correlate the sudden drop amplitude with the cost fluctuation value in the segmented cost data to generate a cost response intensity index;

[0077] Construct a spatio-temporal correlation response relationship matrix based on the heat distribution map of cost changes and the cost response intensity index.

[0078] In this technical solution, the charging middle platform system of the liquid-cooled ultra-fast charging terminal constructs a cost response relationship model through the spatio-temporal correlation analysis of high-frequency charging time period groups and dynamic event sequences, forming a multi-dimensional decision support framework. The system first establishes a first analysis layer with the active charging time period interval as the time dimension. This analysis layer divides the continuous time axis into analysis units matching the high-frequency charging time periods. For example, for the identified active time period interval of 19:00 - 21:00, the system divides it into 24 time slices with 5 minutes as the basic unit, and each slice embeds charging power, cost data, and user behavior labels. The division granularity of the time dimension is dynamically adjusted according to the period activity. In high-density periods (such as 19:30 - 20:30), a finer 2-minute slice is used, while in low-density periods (such as 21:00 - 21:30), a 10-minute slice is used to optimize the allocation of computing resources. Each time slice integrates multi-dimensional data features, including the average power value, cost change gradient, and user group distribution density, forming the basic structure of the spatio-temporal data cube.

[0079] The second analysis layer constructs discrete analysis nodes with the sudden drop time points in the dynamic event sequence as the event dimension. The system performs spatio-temporal encoding on the sudden drop events, maps the sudden drop time point of each event to the corresponding time slice, and associates event attribute parameters (such as a 15% sudden drop amplitude and a 5-minute interval time). The analysis nodes in the event dimension express the correlation between events through a directed graph structure. For example, it is defined that if the interval time between event A (12% sudden drop amplitude) and event B (18% sudden drop amplitude) is less than a preset threshold (such as 8 minutes), a causal relationship edge is established and a weight coefficient (such as 0.65) is assigned. The node attributes in the event graph structure include the sudden drop amplitude, the triggered user type, and the associated cost segment identifier, forming an event-driven analysis network.

[0080] The segmented cost data is mapped to the time slice unit of the first analysis layer through the timestamp alignment algorithm to generate a cost change heat distribution map. This heat map uses color depth coding technology to express the intensity of cost fluctuations. For example, when it is detected at the 19:30 time slice that the electricity price jumps from 0.8 yuan to 1.2 yuan, the corresponding unit is displayed as a dark red mark, while the stable cost interval is displayed as light blue. The generation process of the heat distribution map introduces a sliding window smoothing algorithm to eliminate the interference of instantaneous fluctuations. For example, the cost gradient values of three adjacent time slices are weighted and averaged (weight coefficients 0.5 - 0.3 - 0.2) to ensure that the visual expression of the cost trend conforms to the actual policy adjustment rhythm.

[0081] The matching and correlation between the sudden drop amplitude and the cost fluctuation value are realized by a two-way mapping mechanism. In the forward mapping process, the cost fluctuation value at the time point when the sudden drop event occurs is extracted (such as the cost increase of 0.15 yuan corresponding to the event time point 19:12:35), and the linear proportional relationship between the sudden drop amplitude and the cost change is calculated (such as an average power reduction of 6.2% is triggered for every 0.1 yuan increase in cost). The reverse mapping analyzes the user behavior feedback delay after the cost adjustment (such as the average response time for the user to restore the original power is 4.3 seconds after the cost drops by 0.1 yuan). The system constructs a cost-power response surface model to quantify the incentive effect of different cost intervals on the power adjustment behavior. For example, the cost response sensitivity is increased to 1.8 times the normal level in the 1.0 - 1.2 yuan interval.

[0082] The construction of the spatio-temporal correlation response relationship matrix integrates the heat distribution map and the response intensity index to form a three-dimensional data structure: the first dimension is the time slice sequence, the second dimension is the event type code, and the third dimension is the response intensity level. The filling rule of the matrix elements is determined by the product of the event density and the response intensity within the time slice. For example, if 3 sudden drop events are detected in a certain time slice during the period from 19:15 to 19:20 and the average response intensity is 0.75, then the corresponding matrix cell value is 2.25. This matrix supports a dynamic update mechanism. When a new charging record or cost policy change is detected, the matrix weight coefficient is adjusted through an incremental learning algorithm (such as a new event updates a certain cell value from 2.25 to 2.38) to ensure that the model continuously adapts to the evolution of the behavior pattern.

[0083] It can be seen that by accurately depicting the dynamic influence law of the cost policy on user behavior through the spatio-temporal two-dimensional correlation model, it provides data support for the formulation of differentiated charging strategies; the constructed multi-dimensional response relationship matrix realizes the quantitative expression of complex behavior patterns, enhancing the system's prediction and response capabilities to sudden load fluctuations; the two-way mapping mechanism reveals the interaction mechanism between cost and power adjustment, improving the economy of resource scheduling strategies and user acceptance, and ultimately achieving the coordinated optimization of charging service intelligence and grid stability.

[0084] In an exemplary embodiment, constructing a charging preference label for an in-vehicle charging terminal based on the user behavior feature set includes:

[0085] Extracting a charging time period preference pattern from the user behavior feature set, where the charging time period preference pattern is determined by the time period weights and power peak density distribution of the active charging time period interval;

[0086] Analyzing the user adjustment behavior records of the cost segmented data in different cost intervals to generate a cost tolerance interval;

[0087] Determining a power stability threshold according to the trigger frequency and cost response intensity index of the sudden drop event in the dynamic event sequence;

[0088] Fusing the time period weights, power stability threshold, and cost tolerance interval in multiple dimensions to generate a charging preference label for the in-vehicle charging terminal;

[0089] Dynamically adjusting the parameter weights in the charging preference label through a real-time charging data stream update mechanism, where the parameter weights are synchronously updated based on the load fluctuations of the liquid-cooled ultra-fast charging terminal and the grid cost strategy.

[0090] Specifically, analyzing the user adjustment behavior records of the cost segmented data in different cost intervals to generate a cost tolerance interval includes:

[0091] Dividing the cost segmented data into multiple consecutive sub-cost intervals according to the cost amount;

[0092] Counting the number of charging behavior adjustments and the corresponding adjustment types in each sub-cost interval, where the adjustment types include active power reduction, charging pause, and terminal switching operations;

[0093] Calculating the adjustment trigger rate of each sub-cost interval, where the adjustment trigger rate is determined by the ratio of the number of adjustments to the number of charging times;

[0094] Generating a cost tolerance interval according to the adjustment trigger rate and the sub-cost interval span, where the cost tolerance interval includes a first tolerance boundary and a second tolerance boundary;

[0095] Dynamically calibrating the cost tolerance interval through a sliding window algorithm to eliminate the interference of abnormal data on the interval range.

[0096] In the embodiment of the present invention, the steps of constructing the charging preference tags for the on-vehicle charging end in the liquid-cooled ultra-fast charging terminal charging middle platform system achieve the accurate characterization of user charging preferences through a multi-dimensional behavior feature fusion and dynamic parameter adjustment mechanism. Specifically, when the liquid-cooled ultra-fast charging terminal charging middle platform system extracts the charging time period preference pattern from the user behavior feature set, it performs weighted calculation by comprehensively considering the time period weight and the power peak density distribution in the active charging time period interval. The determination of the time period weight is based on the ratio of the time period coverage frequency to the contribution degree of the power peak. For example, a certain time period is given a weight value of 0.72 due to a coverage frequency ratio of 38% and a power peak density index of 0.85. The power peak density distribution is obtained by statistically analyzing the extreme value distribution characteristics of the power fluctuation trajectory in the active time period interval. For example, during the time period from 19:00 to 20:00, a power peak fluctuation of 3.2 times per minute is detected, and its density level is marked as the high-density level. The quantitative expression of the time period preference pattern adopts a three-dimensional feature vector form, including the time period start time offset, the duration coefficient, and the power density attenuation factor, to ensure the distinguishability and comparability of different time period patterns.

[0097] Furthermore, the generation of the cost tolerance interval relies on the hierarchical analysis of the cost segmented data. The system divides the cost segmented data into continuous sub-cost intervals at a preset amount interval (such as 0.2 yuan), for example, three intervals of 0.6 - 0.8 yuan, 0.8 - 1.0 yuan, and 1.0 - 1.2 yuan. The user adjustment behavior records within each sub-cost interval are extracted through the event log analysis module, specifically including power active down adjustment events (such as from 200kW to 180kW), charging pause events (such as a pause duration exceeding 5 minutes), and terminal switching operations (such as switching to an adjacent low-rate terminal). The calculation of the adjustment trigger rate uses the sliding window statistical method. For example, in the 0.8 - 1.0 yuan interval, 36 adjustment events are detected among 120 charging behaviors, so the adjustment trigger rate is 30%. The trigger rate threshold is dynamically set according to historical data. When the adjustment trigger rate of a certain sub-interval exceeds 1.5 times the standard deviation of the mean, it is marked as a high-sensitivity interval.

[0098] It is understandable that the determination of the first tolerance boundary and the second tolerance boundary adopts a double-threshold decision-making mechanism. The first tolerance boundary is defined as the cost critical point where the adjustment trigger rate first exceeds the preset threshold. For example, if the trigger rate jumps from 15% to 32% in the range of 0.8 - 1.0 yuan, then 0.8 yuan is set as the first boundary; the second tolerance boundary is the critical point where the trigger rate continuously exceeds the threshold and shows a monotonically increasing trend. For example, if the trigger rate linearly increases from 40% to 55% in the range of 1.0 - 1.2 yuan, 1.1 yuan is taken as the second boundary. The sliding window dynamic calibration algorithm weights historical data through a time decay factor. For example, the exponential weighted moving average method is adopted, giving 0.7 weight to recent data and 0.3 weight to historical data, to eliminate the interference of abnormal fluctuations (such as special rate policies during holidays) on the interval range. When the detected boundary point deviation exceeds the preset tolerance (such as ±0.05 yuan) during the calibration process, boundary re-drawing is triggered. For example, when it is detected that the user's adjustment behavior surges at the cost point of 1.05 yuan for three consecutive billing cycles, the system adjusts the second boundary from 1.1 yuan to 1.05 yuan.

[0099] In addition, the determination of the power stability threshold is achieved by integrating the sudden drop event trigger frequency and the cost response intensity index of the dynamic event sequence. The calculation of the trigger frequency is based on the number of sudden drop events occurring within a unit time window. For example, 0.8 sudden drop events are detected per minute during peak hours, and its frequency level is marked as the high-frequency level. The cost response intensity index is determined by the correlation coefficient between the sudden drop amplitude and the cost fluctuation value. For example, when the cost increases by 0.1 yuan, the average sudden drop amplitude increases by 7.2%, and the correlation coefficient reaches 0.78. The system uses fuzzy logic rules to map the trigger frequency and the response intensity to the power stability threshold. For example, the combined trigger threshold for high frequency and high response is tightened to ±5%, and the combined trigger threshold for low frequency and low response is relaxed to ±10%. The dynamic adjustment mechanism of the threshold is linked to the power grid stability index. When it is detected that the regional power grid load rate exceeds 85%, the threshold is automatically shrunk by 20% to suppress power fluctuations.

[0100] Specifically, the multi-dimensional fusion process of the charging preference label adopts a feature weighted splicing algorithm. After the time period weight, the power stability threshold, and the cost tolerance interval are respectively normalized to the [0, 1] interval values, a dynamic weight coefficient is assigned according to the real-time load status (such as the time period weight coefficient is increased to 0.6 during peak hours, and the power stability coefficient is decreased to 0.3). The label generation engine encodes the three types of parameters into a structured data object. For example, the time period weight of 0.72, the power stability threshold of ±7%, and the cost tolerance interval of 0.8 - 1.1 yuan are combined to form a label instance. The real-time charging data stream update mechanism adjusts the parameter weights through an incremental learning model. For example, when it is detected that a certain user frequently switches terminals at the cost point of 1.05 yuan, the system increases the update frequency of its cost tolerance interval to once per minute, and at the same time decreases the time period weight coefficient by 0.1 to reflect the change in the behavior pattern.

[0101] With such a design, accurate modeling of charging preference labels is achieved through multi-dimensional parameter fusion and dynamic calibration mechanisms, improving the matching degree between charging strategies and user needs; the tolerance interval generation method based on cost behavior correlation analysis enhances the system's adaptability to market policy changes; the dynamic adjustment function of the power stability threshold effectively balances the freedom of user behavior and the grid stability requirements, providing a core decision-making basis for intelligent charging scheduling, and ultimately achieving the dual goals of personalized charging services and optimized power resources.

[0102] In an alternative embodiment, the generating a dynamic scheduling strategy across liquid-cooled ultra-fast charging terminals according to the charging preference label includes:

[0103] Analyze the time period weight, power stability threshold, and cost tolerance interval in the charging preference label;

[0104] Obtain the real-time operating load data of multiple liquid-cooled ultra-fast charging terminals and the grid power supply constraint conditions;

[0105] Based on the time period weight, predict the charging demand distribution and power demand peak within the target time window;

[0106] Generate a power shunt path among terminals according to the power demand peak and the real-time operating load data;

[0107] Combine the cost tolerance interval to adjust the cost compensation coefficient in the power shunt path, and generate a dynamic scheduling strategy across terminals.

[0108] In the following steps, the predicting the charging demand distribution and power demand peak within the target time window based on the time period weight includes:

[0109] Extract the active charging time period interval and time period coverage frequency in the time period weight;

[0110] Construct a prediction input sequence with the time period coverage frequency as the time axis and the power peak density as the intensity axis;

[0111] Analyze the periodic characteristics in the input sequence through a time series pattern recognition algorithm, and generate a charging demand prediction curve within the target time window;

[0112] Identify the local extreme points in the charging demand prediction curve as the power demand peaks;

[0113] Determine the charging demand distribution according to the time stamps corresponding to the local extreme points and the terminal location information.

[0114] In addition, after generating the dynamic scheduling strategy across terminals, the method further includes:

[0115] Monitor the execution effect data of the dynamic scheduling strategy in real time, where the execution effect data includes the resource utilization rate improvement rate and the user satisfaction index;

[0116] Compare the difference value between the execution effect data and the preset optimization target to generate a strategy adjustment parameter;

[0117] Feed back the strategy adjustment parameter to the update mechanism of the charging preference label to form a closed-loop optimization link;

[0118] Gradually converge to the optimal scheduling strategy through multiple rounds of iterative optimization to achieve the adaptive matching of charging resource allocation and user needs.

[0119] In this embodiment, the technical solution of the liquid-cooled ultra-fast charging terminal charging middle platform system to generate a cross-terminal dynamic scheduling strategy realizes the efficient allocation of charging resources through multi-dimensional parameter integration and a closed-loop optimization mechanism. When the system analyzes the charging preference label, it extracts the active charging time period interval and the time period coverage frequency from the time period weight. For example, it is identified that the time period weight coefficient from 19:00 to 21:00 is 0.75, and the time period coverage frequency reaches 45%. The power stability threshold is determined comprehensively according to the triggering frequency of sudden drop events in the dynamic event sequence (such as 0.6 times per minute) and the cost response intensity index (such as a power reduction of 8.2% triggered by a cost increase of 0.1 yuan). For example, the threshold is set to ±7% to balance the user behavior freedom and the power grid stability requirements. The cost tolerance interval determines the boundary values (such as setting the first tolerance boundary of 1.05 yuan and the second boundary of 1.15 yuan) by analyzing the adjustment trigger rate distribution of sub-cost intervals (such as a trigger rate of 38% in the interval of 1.0 - 1.2 yuan), and dynamically calibrates the interval range in combination with the sliding window algorithm to eliminate the interference of abnormal data (such as the boundary offset caused by the mutation of charging behavior during holidays).

[0120] When obtaining real-time operating load data, the system collects the instantaneous power values, temperature status, and queue lengths of each node through the terminal sensor network. For example, it is monitored that the load rate of Terminal 3 reaches 95%, while the load rate of the adjacent Terminal 5 is only 58%. The power grid power supply constraint condition analysis module synchronously obtains the real-time power supply capacity of the regional power grid (such as the maximum output power of 12 MW), the line transmission loss rate (such as 3.2%), and the stability index (such as the voltage fluctuation threshold of ±5%), forming a multi-dimensional constraint matrix. The demand prediction model based on time period weights constructs an input sequence with the time period coverage frequency as the time axis and the power peak density as the intensity axis. For example, the data of the 19:00 - 21:00 time period in the past 30 days is sliced by 15 minutes to generate an input sequence containing 96 data points. The time series pattern recognition algorithm uses a long short-term memory network (LSTM) to mine the periodic features in the sequence (such as the power demand increasing by 23% during the daily peak time period), generating a charging demand prediction curve for the target time window (such as 19:00 - 21:00 the next day). The curve local extreme point detection module identifies the power demand peak through derivative analysis (such as predicting a peak demand of 15.2 MW at 19:45), and combines the terminal geographical location information (such as Terminal 3 being located in the center of the commercial area) to generate a spatial distribution heat map (such as the demand in the commercial area accounting for 62%).

[0121] The generation of the power shunt path is based on a graph theory optimization algorithm, modeling the terminal nodes as weighted vertices and the power grid lines as edges. The system calculates the optimal shunt path according to the power demand peak (such as 15.2 MW) and the real-time load data (such as the available power margin of Terminal 3 being 200 kW). For example, 180 kW of the load of Terminal 3 is allocated to Terminal 5 through a low-loss path (transmission loss rate of 1.8%) in the topological network. The adjustment of the cost compensation coefficient is dynamically set according to the cost tolerance interval. For example, when the shunt causes the user cost to increase by 0.8 yuan (exceeding the compensation trigger point of the second tolerance boundary of 1.15 yuan), a hierarchical compensation mechanism is activated (basic compensation rate of 20% + dynamic adjustment rate of 5%). The output of the dynamic scheduling strategy includes a power distribution instruction set (such as adjusting the power upper limit of Terminal A to 200 kW and increasing Terminal B to 180 kW) and a cost compensation parameter table (such as the compensation coefficient of user group C being 0.25), and is sent to each terminal execution unit through an encrypted control channel for execution.

[0122] The execution effect monitoring module collects in real time the resource utilization improvement rate (e.g., the regional average load rate is optimized from 85% to 78%) and the user satisfaction index (e.g., the satisfaction score rises from 4.1 to 4.6). The difference value calculation engine compares the actual data with the preset optimization goals (such as the load balancing degree goal of 0.8 and the satisfaction threshold of 4.5), and generates policy adjustment parameters (such as the time period weight coefficient is lowered by 0.1 and the compensation rate is increased by 3%). When the parameters are fed back to the charging preference label update mechanism, an incremental learning algorithm is used to adjust the label parameters. For example, when it is detected that the boundary of the cost tolerance interval deviates by more than the tolerance (such as 0.05 yuan), the interval recalibration process is triggered. During the multi-round iterative optimization process, the system saves the historical policy set (such as versions V1.2 to V1.5) through the policy version management module, and filters the optimal policy based on the convergence condition (such as the difference value of three consecutive rounds of optimization is less than 2%). The finally formed adaptive matching mechanism can dynamically adjust the power allocation granularity (such as from the 50kW level to the 10kW level) and the compensation rules to achieve high-precision adaptation of resource allocation and user needs.

[0123] Thus, through multi-dimensional data fusion and closed-loop feedback mechanism, the precise generation and continuous optimization of dynamic scheduling strategies are realized, improving the global allocation efficiency of charging resources; the prediction model based on spatio-temporal feature analysis enhances the adaptability to complex demand fluctuations, ensuring the stability of power grid operation; the synergistic effect of the hierarchical compensation mechanism and the dynamic parameter adjustment function balances the economic interests of users and the operating costs of the system, promoting the coordinated development of charging service intelligence and power grid sustainability.

[0124] Under an optional technical idea, optimizing the real-time power supply parameters and cost calculation logic of the liquid-cooled supercharging terminal according to the dynamic scheduling strategy optimization target includes:

[0125] Determine the threshold allocable power value of the target terminal according to the power shunt path in the dynamic scheduling strategy;

[0126] Adjust the cost ladder calculation rule of the target terminal in combination with the cost tolerance interval in the charging preference label;

[0127] Real-time monitor the power usage rate of the target terminal and the user behavior feedback signal;

[0128] Dynamically optimize the priority allocation mechanism in the power supply parameters to ensure the power supply priority of users with active time period weights;

[0129] Eliminate the cost difference caused by power shunt through the cost balancing algorithm to maintain the stability of the user's cost expenditure within the cost tolerance interval.

[0130] Among them, eliminating the cost difference caused by power shunt through the cost balancing algorithm includes:

[0131] Identify user groups whose cost deviation caused by power splitting exceeds a preset threshold;

[0132] Extract the boundary values of the cost tolerance interval in the charging preference tags of the user group;

[0133] Calculate the degree of deviation between the cost deviation and the boundary of the cost tolerance interval;

[0134] Generate a dynamic compensation coefficient based on the degree of deviation, and the compensation coefficient has a positive correlation with the degree of deviation;

[0135] Embed the dynamic compensation coefficient into the cost calculation logic to achieve real-time correction of cost deviation.

[0136] Based on this design idea, the cold ultra-fast charging terminal charging middle platform system optimizes the technical solutions of real-time power supply parameters and cost calculation logic according to the dynamic scheduling strategy, and realizes the efficient utilization of charging resources and the precise control of user costs through the multi-dimensional parameter dynamic adaptation and real-time feedback mechanism. When the system determines the threshold allocable power value of the target terminal according to the power splitting path in the dynamic scheduling strategy, it adopts a joint optimization algorithm based on load balance degree and grid constraints. For example, when it is detected that the total load demand in the area where the target terminal is located is 15.2 MW while the grid power supply upper limit is 13.8 MW, the system calculates the optimal splitting ratio through the topological path loss model (such as a line loss rate of 1.8%), and distributes the over-limit 1.4 MW load to the available capacity of adjacent terminals according to the priority. The setting of the threshold allocable power value takes into account the maximum bearing capacity of the terminal equipment (such as the rated power of a single terminal is 250 kW) and the real-time temperature status (such as automatically reducing the power upper limit by 10% when the heat dissipation efficiency decreases), forming a dynamically adjusted safe operation boundary.

[0137] The adjustment depth of the cost ladder calculation rule deeply integrates the cost tolerance interval parameters in the charging preference tags. The system maps the boundary values of the cost tolerance interval (such as 0.8 - 1.1 yuan) to the ladder tariff calculation engine, and automatically triggers the tariff smoothing algorithm when the real-time electricity price approaches or exceeds the boundary. For example, when the cost approaches the 1.1 yuan upper limit, the system uses the piecewise linear interpolation method to adjust the tariff growth gradient from the conventional 0.05 yuan / level to 0.03 yuan / level, delaying the speed at which the user cost reaches the tolerance boundary. The dynamic adjustment module of the ladder rule synchronously monitors the user behavior feedback signal. For example, when it is detected that the power reduction frequency of a certain user group increases by 15% after the tariff adjustment, the tariff ladder interval is automatically shrunk (such as from 0.1 yuan to 0.08 yuan) to enhance the response sensitivity of the strategy.

[0138] The real-time monitoring module collects power usage rate metrics through a distributed sensor network deployed at the terminal, specifically including instantaneous power values, cumulative energy consumption, and power fluctuation variances. For example, when it is monitored that the power usage rate of a certain terminal continuously exceeds 90% during the period from 19:30 to 20:00, the system automatically generates a load warning signal. The parsing of user behavior feedback signals uses natural language processing and pattern recognition technologies to analyze the satisfaction scores (such as 1 - 5 stars) and text feedback (such as "The charging speed is too slow") submitted by users through the terminal interface, and quantifies them into behavior feedback indices (such as a satisfaction index of 4.2 corresponding to a behavior parameter of 0.85). The priority allocation mechanism for power supply parameters is dynamically adjusted according to time period weights and user levels. For example, for a VIP user group with a high time period weight (0.75), an additional 15% power flexibility capacity is allocated during peak hours to ensure that their charging efficiency is not affected by load fluctuations.

[0139] The core of the cost balancing algorithm lies in identifying and eliminating cost deviations caused by power shunting. The system calculates the absolute value of the deviation by comparing the original charging plan cost of the user with the actual cost after shunting (for example, for user A, the planned cost is 58 yuan, the actual cost is 63 yuan, and the deviation is 5 yuan). The preset threshold is dynamically adjusted according to the width of the cost tolerance interval (for example, when the interval width is 0.3 yuan, the threshold is set to 1.5 yuan). When it is detected that the deviation exceeds the threshold, a multi-level compensation process is initiated. For example, for a user group with a deviation of 5 yuan, the second boundary value of its cost tolerance interval is extracted (such as 1.1 yuan), and the deviation degree is calculated as (5 - 1.1) / 1.1 = 3.55 times, generating a dynamic compensation coefficient of 0.35 (the coefficient is positively correlated with the deviation degree but limited by a saturation function at the upper limit). When the compensation logic is embedded in the cost calculation engine, a hierarchical superposition mechanism is adopted. For example, the compensation amount is superimposed on the basic rate (such as 5 yuan deviation × 0.35 coefficient = 1.75 yuan compensation), and it is returned in real time in the form of user account points or electricity bill deductions.

[0140] The generation process of the dynamic compensation coefficient introduces a time decay factor and a behavior feedback weight. For example, for users whose charging costs exceed the standard three times in a row due to shunting, the system gradually increases the compensation coefficient (from 0.2 to 0.4), and at the same time adjusts the weight according to the user's historical satisfaction score (such as a user with a score of 4.5 has a weight set to 0.8, and a user with a score of 3.0 has a weight set to 0.5). The cost correction data is synchronized to the billing database in real time to ensure the accuracy and traceability of the bill data. The system performs a global cost balancing status scan every 5 minutes. When it is detected that the overall deviation rate of the region exceeds 2%, a compensation rule re-optimization process is triggered. For example, the reference value of the compensation coefficient is adjusted from 0.3 to 0.35.

[0141] Thus, through the synergistic effect of the dynamic threshold power allocation and the intelligent cost compensation mechanism, the problem of out-of-control user costs caused by power shunting is effectively solved, ensuring the fairness and economy of charging services; the priority allocation mechanism based on real-time monitoring data significantly improves the charging experience of high-demand users and enhances the system's service differentiation ability; the adaptive compensation strategy of the cost balancing algorithm precisely matches the user cost expenditure with the psychological expectation on the premise of maintaining the controllability of operating costs. The closed-loop optimization mechanism promotes the evolution of charging resource allocation from static preset to dynamic intelligence by continuously calibrating power supply parameters and cost rules, providing core technical support for building a flexible and scalable intelligent charging network, and ultimately achieving the dual goals of efficient utilization of power resources and improvement of user satisfaction.

[0142] In an extended embodiment, the multi-round iterative optimization gradually converges to the optimal scheduling strategy to achieve the adaptive matching of charging resource allocation and user demand, including:

[0143] Monitor the real-time execution effect data set generated during the execution of the current dynamic scheduling strategy, where the real-time execution effect data set includes the resource utilization rate fluctuation curve of each terminal and the user satisfaction feedback sequence;

[0144] Layer by layer, compare the real-time execution effect data set with the preset optimization goal to identify the deviation degree of resource utilization rate and the difference value of satisfaction;

[0145] Generate a multi-dimensional strategy adjustment parameter set based on the deviation degree of resource utilization rate and the difference value of satisfaction, where the parameter set includes the time period weight correction coefficient, the power shunting compensation factor, and the cost tolerance calibration amount;

[0146] Inject the multi-dimensional strategy adjustment parameter set into the update link of the charging preference label, trigger the dynamic calibration of the label parameters, and generate an optimized version of the charging preference label;

[0147] According to the optimized version of the charging preference label, re-parse the time period priority weight, the power stability threshold, and the cost tolerance interval, and reconstruct the input conditions of the dynamic scheduling strategy;

[0148] Based on the reconstructed input conditions, parametrically adjust the power reallocation path and the cost compensation rule to generate an updated scheduling strategy adapted to the current resource state and user demand;

[0149] Execute the updated scheduling strategy and start a new round of execution effect monitoring to form a closed-loop feedback optimization link;

[0150] Judge whether the current updated scheduling strategy reaches the optimal balance state of resource allocation and user demand through the preset convergence condition: if not, trigger the next round of iterative optimization; if so, terminate the optimization process and lock the optimal scheduling strategy.

[0151] Specifically, the liquid-cooled ultra-fast charging terminal charging middle platform system realizes a technical solution for the adaptive matching of charging resource allocation and user needs through multiple rounds of iterative optimization, and constructs a dynamic closed-loop intelligent scheduling system. The system collects and integrates the resource utilization rate fluctuation curves and user satisfaction feedback sequences of each terminal in real time to form a full-dimensional execution effect data set covering the equipment operation status and user behavior responses. The generation of the resource utilization rate fluctuation curve is based on the instantaneous power values, load rates, and temperature data uploaded by the terminal sensor network. The instantaneous fluctuations are smoothed through a sliding window mean algorithm (such as calculating the mean with a 5-minute window) to generate a smooth curve reflecting the load trend. The construction of the user satisfaction feedback sequence adopts multi-source data fusion technology, integrating terminal interface rating data (such as 1-5 star ratings), semantic analysis results of voice feedback (such as "satisfied with the charging speed" mapped to a satisfaction index of 0.8), and behavior pattern derivation parameters (such as the inverse correlation coefficient between the charging interruption rate and the power adjustment frequency) to form a normalized satisfaction index.

[0152] The difference comparison engine matches the real-time data set with the preset optimization goals layer by layer. The calculation of the resource utilization deviation degree adopts a standard deviation weighted model. For example, when the regional target load balance degree is 0.8 and the actual value is 0.65, the deviation degree is quantified as (0.8 - 0.65) / 0.8 × 100% = 18.75%. The satisfaction difference value is calculated by comparing the square of the difference between the actual satisfaction index and the target threshold (such as 4.2 / 5.0). For example, when the actual value is 3.9, the difference value is (4.2 - 3.9)² = 0.09. The generation module of the multi-dimensional strategy adjustment parameter set dynamically generates time period weight correction coefficients (such as for every 5% increase in the load deviation degree, the weight is reduced by 0.1), power shunt compensation factors (such as for every 0.1 increase in the satisfaction difference value, the compensation rate is increased by 2%), and cost tolerance calibration amounts (such as when the user behavior feedback index drops by 10%, the tolerance interval is expanded by 0.05 yuan) according to the combined weights of the deviation degree and the difference value.

[0153] The dynamic calibration process of the charging preference label adopts an incremental parameter injection mechanism. The system weights and fuses the time period weight correction coefficient with the original label parameters. For example, the original time period weight of 0.75 is updated to 0.675 after adding the correction coefficient of -0.1. The adjustment of the power stability threshold expands or contracts dynamically according to the compensation factor. For example, when it is detected that the power shunt causes a decrease in satisfaction, the threshold is relaxed from ±7% to ±10% to reduce the frequency of user behavior intervention. The application of the cost tolerance calibration amount is achieved by translating the interval boundary. For example, when the calibration amount is +0.05 yuan, the original interval of 0.8 - 1.1 yuan is expanded to 0.8 - 1.15 yuan. The label update link adopts a version control mechanism, retaining historical parameter snapshots (such as versions V2.1 to V2.3) to support rollback operations, ensuring that the system can quickly recover to a stable state in case of abnormal parameter adjustment.

[0154] When reconstructing the input conditions of the dynamic scheduling strategy, the system re-parses the optimized tag parameters: the calculation of the time period priority weight introduces a load deviation compensation factor. For example, the reciprocal term of the deviation (1 / deviation) is added to the weight calculation formula during peak hours, so that the time period weight in high-load areas gets an additional boost; the determination of the power stability threshold integrates real-time power grid fluctuation data. For example, when the detected voltage volatility in a region exceeds 2%, the threshold is automatically shrunk by 20% to enhance power grid stability; the boundary calibration of the cost tolerance interval combines the historical strategy execution effect. For example, for the interval segments with satisfaction differences still existing after three consecutive calibrations, the exponential decay algorithm is used to reduce the calibration step size to improve the convergence speed.

[0155] The optimization of the power reallocation path adopts an adaptive graph network model, and superimposes a real-time load heat map on the original topological structure. For example, when the load rate of a certain terminal jumps from 75% to 92%, the system adds virtual nodes to its associated path to expand the shunt path selection space. The parametric adjustment of the cost compensation rule introduces a satisfaction feedback weight. For example, a double compensation coefficient is given to high-satisfaction user groups (rating > 4.5), while a basic compensation rate is adopted for low-satisfaction groups (rating < 3.0). The output of the updated scheduling strategy includes a dynamic power distribution matrix and a hierarchical compensation rule table, which are synchronized to each terminal execution unit through an encrypted control channel, and a version compatibility verification mechanism (such as verifying the matching of protocol hash values) is enabled to ensure the secure deployment of the strategy.

[0156] The operation of the closed-loop feedback optimization link adopts an event-driven mode. When the new strategy has been executed for a preset period (such as 30 minutes) or a key indicator mutation is detected (such as a sudden 15% drop in the satisfaction index), a new round of optimization iteration is triggered. The convergence condition determination module comprehensively evaluates the effects of multiple rounds of optimization: the deviation of resource utilization rate needs to be reduced by less than 1% in three consecutive iterations, the satisfaction difference value needs to be stable within the ±5% interval of the target threshold, and the change rate of the cost tolerance calibration amount is lower than 0.5%. When the above conditions are met, the system locks the current strategy version as the optimal solution, starts the persistent storage process, solidifies the strategy parameters into the distributed database, and releases the computing resources occupied by the optimization process.

[0157] It can be seen that through the closed-loop feedback and multi-round iterative optimization mechanism, the continuous self-improvement of the charging resource allocation strategy is realized, significantly enhancing the system's adaptability to the dynamic environment; the intelligent generation and injection mechanism of multi-dimensional strategy adjustment parameters ensures the deep coupling of user behavior characteristics and grid operation status, enhancing the accuracy and robustness of the scheduling strategy; the adaptive convergence determination model effectively balances the optimization efficiency and result stability, avoiding falling into local optimal solutions; the version control and rollback functions improve the system's fault tolerance, ensuring continuous and reliable services. Finally, an intelligent charging management platform with autonomous evolution ability is constructed, achieving a high degree of unity between the global optimal allocation of power resources and the satisfaction of users' personalized needs, providing core technical support for the operation of large-scale charging facilities under the new power system.

[0158] In a further extensible embodiment, parameterizing the adjustment of the power redistribution path and cost compensation rules based on the reconstructed input conditions to generate an updated scheduling strategy adapted to the current resource status and user requirements includes:

[0159] Obtain the updated value of the time period priority weight, the adjustment amount of the power stability threshold, and the calibration amount of the cost tolerance in the reconstructed input conditions;

[0160] Analyze the matching relationship between the updated value of the time period priority weight and the current terminal load distribution to determine the priority allocation coefficient in the power redistribution path;

[0161] Dynamically expand or contract the power fluctuation tolerance range in the power redistribution path according to the adjustment amount of the power stability threshold to generate path constraint conditions adapted to the current grid stability status;

[0162] Combined with the calibration amount of the cost tolerance, perform gradient correction on the compensation coefficient in the cost compensation rule to generate compensation rule adjustment parameters matching the user's cost sensitivity;

[0163] Input the priority allocation coefficient, path constraint conditions, and compensation rule adjustment parameters into the strategy generation engine to generate a candidate scheduling strategy set including the updated power shunt ratio and cost compensation mechanism;

[0164] Based on the candidate scheduling strategy set, conduct a multi-objective trade-off evaluation of the terminal load balance degree, cost deviation tolerance degree, and predicted user satisfaction value, and select the candidate strategy with the highest comprehensive score as the updated scheduling strategy;

[0165] Verify the compatibility of the updated scheduling strategy with the historical execution effect data to ensure that the adjustment range of the strategy parameters is within the preset safety threshold;

[0166] Output the updated scheduling strategy that passes the verification and embed it into the real-time control system of the liquid-cooled ultra-fast charging terminal for execution.

[0167] In this embodiment, the technical solution of the liquid-cooled ultra-fast charging terminal charging middle platform system for parametric adjustment of the power reallocation path and cost compensation rules after reconstructing the input conditions realizes the dynamic adaptation of the scheduling strategy through multi-dimensional parameter collaborative optimization and multi-objective decision-making mechanism. When the system obtains the updated value of the time period priority weight after reconstruction, it uses a spatio-temporal matching algorithm to analyze its correlation with the current terminal load distribution. For example, when it is detected that the weight of a certain time period is updated from 0.75 to 0.68, the system determines the change in the load density of the corresponding area in this time period (such as from 85% to 72%) through the load heat map matching algorithm, and then calculates the priority distribution coefficient in the power reallocation path. The generation of the priority distribution coefficient adopts a weighted decay model, which combines the time period weight value (0.68), the load density difference (13%), and the historical strategy execution effect (such as the success rate of the past three optimizations in this time period is 92%) to generate a dynamic coefficient (such as 0.65), which is used to control the resource acquisition priority of different terminals during the power shunt process.

[0168] The application of the power stability threshold adjustment amount is realized through a dynamic constraint generator. The system adjusts the power fluctuation tolerance range according to the threshold change direction (expansion or contraction). For example, when the threshold expands from ±7% to ±10%, the maximum allowable fluctuation amplitude in the path constraint condition is synchronously increased, allowing the terminal to maintain operation within a larger power fluctuation range (such as allowing the instantaneous power to fluctuate from 180kW to 198kW without triggering adjustment); conversely, when the threshold shrinks to ±5%, the system automatically tightens the path selection criteria and only selects terminals with a load volatility lower than 5% as shunt nodes. The dynamic adjustment of the constraint conditions is deeply coupled with the power grid stability index. For example, when the frequency fluctuation of the regional power grid exceeds 0.5Hz, the threshold contraction mechanism is triggered to reduce the fluctuation tolerance range by 30% to give priority to ensuring the safety of the power grid.

[0169] The gradient correction process of the cost compensation rule combines the cost tolerance calibration amount and user behavior feedback data. The system maps the calibration amount (such as the interval boundary expands from 1.1 yuan to 1.15 yuan) to the adjustment gradient of the compensation coefficient. For example, an increasing curve of the compensation coefficient is set in the interval of 1.1 - 1.15 yuan (the compensation rate increases by 0.5% for each 0.01 yuan interval). The gradient correction engine synchronously analyzes the user's historical behavior data and assigns a higher compensation sensitivity to users with high-frequency adjustments (such as users with more than 15 monthly average power adjustment times) (such as the coefficient weighting factor is increased to 1.2). The adjustment parameters of the compensation rule are generated through a non-linear interpolation algorithm to ensure a smooth transition of the compensation rate at the boundary of the tolerance interval and avoid abnormal user perception caused by rate jumps.

[0170] After receiving the priority allocation coefficient, path constraint conditions, and compensation rule adjustment parameters, the policy generation engine uses a multi-objective optimization algorithm to generate a set of candidate scheduling policies. The determination of the power splitting ratio is based on a mixed-integer programming model, which maximizes the load balancing degree index under the premise of meeting the path constraint conditions. For example, for a regional network with 12 terminals, the engine may generate three candidate solutions: Solution A with a splitting ratio of 18% and a load balancing degree of 0.82, Solution B with a splitting ratio of 25% and a balancing degree of 0.78, and Solution C with a splitting ratio of 15% and a balancing degree of 0.85. The construction of the cost compensation mechanism adopts a hybrid architecture of a rule engine and a machine learning model. For example, for high-cost sensitive user groups (compensation coefficient > 0.3), a reinforcement learning model is used to predict the optimal compensation strategy, while for ordinary user groups, a rule-based static compensation table is adopted.

[0171] The multi-objective trade-off evaluation module jointly scores the terminal load balancing degree, cost deviation tolerance, and predicted user satisfaction of the candidate policy set. The load balancing degree is quantified by calculating the deviation degree of the regional terminal load variance from the ideal value (e.g., when the variance drops from 120 to 95, the balancing degree increases by 15%); the cost deviation tolerance is evaluated based on the expected cost overrun ratio after the policy is executed (e.g., the predicted proportion of overrun users drops from 8% to 3%); the predicted user satisfaction value is derived through a behavior feedback model, integrating the relevance between historical scoring data and policy parameters (e.g., when the compensation coefficient increases by 0.1, the predicted satisfaction increases by 0.2). The evaluation process uses the Pareto optimal frontier analysis method to screen out candidate policies that are non-inferior solutions in all three objective dimensions. For example, a comprehensive optimal policy with a load balancing degree of 0.83, a cost deviation tolerance of 4%, and a predicted satisfaction value of 4.3 is selected.

[0172] In the policy compatibility verification stage, the system analyzes the matching degree between the updated scheduling policy and the historical execution effect database. The verification indicators include the safety threshold of the parameter adjustment range (e.g., the single adjustment of the time period weight does not exceed ±0.15), the load mutation rate limit (e.g., the change rate of the power splitting ratio does not exceed 20% per hour), and the continuity of the compensation rule (e.g., the difference in the compensation coefficient between adjacent policy versions is less than 0.05). When it is detected that the power splitting ratio of a certain candidate policy increases by 28% compared to the historical maximum (exceeding the preset safety threshold of 25%), the system automatically triggers the policy correction process to ensure a smooth transition by reducing the splitting gradient (e.g., adjusting from 5% / minute to 3% / minute). After passing the verification, the updated scheduling policy is encrypted with a digital signature and sent to the terminal control system through a low-latency communication protocol. The embedded execution module uses a real-time priority preemption mechanism to ensure the timely response of policy instructions.

[0173] Designed in this way, through dynamic parameter injection and multi-objective optimization mechanism, the precise adaptation of the scheduling strategy is realized, significantly improving the flexibility and adaptability of charging resource allocation; the intelligent adjustment function of the power fluctuation tolerance range effectively balances the grid stability and the user behavior freedom, enhancing the robustness of the system operation; the combination of the gradient compensation model and the compatibility verification mechanism ensures the smooth and reliable process of cost regulation, avoiding the deterioration of user experience caused by policy mutation; the low-latency response characteristic of the embedded real-time control system guarantees the timeliness of policy execution, providing core technical support for the dynamic allocation of resources in complex charging scenarios. The overall solution constructs a complete technical closed-loop from policy generation to execution verification, promoting the continuous evolution of charging services towards intelligence and adaptability.

[0174] In summary, through multi-dimensional data collection and in-depth behavior analysis in the embodiments of the present invention, the accurate characterization of user charging behavior characteristics is realized, effectively improving the accuracy of charging demand prediction and resource scheduling. The constructed charging preference label system converts user behavior characteristics into quantifiable parameters, providing a core basis for formulating differentiated service strategies and enhancing the personalized adaptation ability of charging services. The generation and execution mechanism of the cross-terminal dynamic scheduling strategy optimize the allocation efficiency of electric power resources in the space-time dimension, alleviating the local overload problem during peak hours. The coordinated adjustment function of real-time power supply parameters and cost calculation logic balances the grid operation stability while ensuring charging efficiency, reducing the operation cost. The overall solution forms a closed-loop optimization link from data perception to policy execution, significantly improving the resource utilization efficiency of charging facilities.

[0175] Furthermore, Figure 2 FIG. is a schematic structural diagram of a liquid-cooled ultra-fast charging terminal charging middle platform system 200 provided by an embodiment of the present invention. As Figure 2 shown, the liquid-cooled ultra-fast charging terminal charging middle platform system 200 includes a processor 210. The processor 210 can call and run a computer program from a memory to implement the method in the embodiments of the present invention.

[0176] Optionally, as Figure 2 shown, the liquid-cooled ultra-fast charging terminal charging middle platform system 200 may further include a memory 230. Among them, the processor 210 can call and run a computer program from the memory 230 to implement the method in the embodiments of the present invention.

[0177] Among them, the memory 230 can be an independent device from the processor 210 or integrated in the processor 210.

[0178] Optionally, as Figure 2As shown, the charging middle platform system 200 of the liquid-cooled ultra-fast charging terminal may further include a transceiver 220. The processor 210 may control the transceiver 220 to interact with other devices. Specifically, it may send information or data to other devices, or receive information or data sent by other devices.

[0179] Optionally, the charging middle platform system 200 of the liquid-cooled ultra-fast charging terminal may implement the corresponding processes of the storage engine or components (such as processing modules) in the storage engine or the device deployed with the storage engine in each method of the embodiments of the present invention. For the sake of brevity, details are not described herein again.

[0180] It should be understood that the processor in the embodiments of the present invention may be an integrated circuit chip with signal processing capabilities.

[0181] It can be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. It should be noted that the memory of the systems and methods described herein is intended to include, but is not limited to, suitable types of memory.

[0182] On the above basis, a readable storage medium is provided. Programs or instructions are stored on the readable storage medium, and when the programs or instructions are executed by a processor, the steps of the above methods are implemented.

[0183] It should be noted that in this article, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or device including that element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present invention is not limited to performing functions in the order shown or discussed. It may also include performing functions in a substantially simultaneous manner or in a reverse order according to the functions involved. For example, the described methods may be performed in an order different from that described, and various steps may be added, omitted, or combined. Additionally, the features described with reference to certain examples may be combined in other examples.

[0184] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-described example methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the embodiments of the present invention, in essence, or the part that contributes to the prior art can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions for causing a terminal (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the embodiments of the present invention.

[0185] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the embodiments of the present invention are not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the embodiments of the present invention, those of ordinary skill in the art can also make many forms without departing from the purpose of the embodiments of the present invention and the scope protected by the embodiments of the present invention, and all of them belong to the protection scope of the embodiments of the present invention.

Claims

1. A method for analyzing the cost of charging a liquid-cooled supercharging terminal, characterized in that: The method comprises: Collecting multiple sets of operating status data sets of liquid-cooled supercharging terminals, wherein the operating status data sets include charging power time series change records, user charging cycle distribution, and corresponding segment fee settlement results; Perform multi-dimensional charging behavior analysis on the operating status data set to generate a user behavior feature set including charging time period preference mode, power adjustment law and cost response strength: select a selected charging record subset that meets the preset integrity condition from the operating status data set, the selected charging record subset includes the power fluctuation trajectory and the corresponding cost segmentation data within the complete charging cycle; perform charging time period distribution analysis on the selected charging record subset to identify high-frequency charging time period groups and corresponding power peak density; extract trigger condition parameters of power sudden drop events in the power fluctuation trajectory to generate a dynamic event sequence including sudden drop time point and sudden drop amplitude; perform spatiotemporal correlation analysis on the high-frequency charging time period group and the dynamic event sequence to determine the response relationship of the cost segmentation data to the power adjustment behavior; generate the time period preference mode, power adjustment law and cost response strength in the user behavior feature set based on the response relationship and power peak density; wherein the cost response strength index is obtained by matching and correlating the sudden drop amplitude with the cost fluctuation value in the cost segmentation data; A charging preference tag for the on-board charger is constructed based on the user behavior feature set, wherein the charging preference tag includes a time period priority weight, a power stability threshold, and a cost tolerance interval; the power stability threshold is used to characterize the power fluctuation range acceptable to the user, and is determined based on historical adjustment behavior statistics, specifically according to the trigger frequency of sudden drop events in the dynamic event sequence and the cost response intensity index; Generate a dynamic scheduling strategy across liquid-cooled supercharging terminals according to the charging preference tag, wherein the dynamic scheduling strategy includes a power redistribution path and a cost compensation rule; The real-time power supply parameters and cost calculation logic of the target liquid-cooled supercharging terminal are optimized according to the dynamic scheduling strategy to achieve a coordinated improvement in charging efficiency and resource utilization.

2. The method according to claim 1, characterized in that: The performing charging period distribution analysis on the selected charging record subset to identify high-frequency charging period groups and corresponding power peak densities includes: Density clustering of the charging start time in the selected charging record subset according to a preset time granularity to form a plurality of time period distribution clusters; Calculate the time period coverage frequency of each time period distribution cluster, where the time period coverage frequency is determined by the ratio of the number of charging records in the cluster to the total number of records; Filter the time period distribution clusters whose time period coverage frequency exceeds the preset frequency threshold as candidate thermal time period groups; Expand the time window of the charging start time in the candidate thermal time period group to generate a continuous active charging time period interval; Counting the peak occurrence frequency and duration ratio of the power fluctuation trajectory within the active charging period, and generating a corresponding power peak density matrix; The extracting the trigger condition parameters of the power sudden drop event in the power fluctuation trajectory to generate a dynamic event sequence including a sudden drop time point and a sudden drop amplitude includes: Detecting a section in the power fluctuation trajectory where the continuous power value drops by more than a preset sudden drop threshold as a candidate sudden drop event; Record the starting time point, ending time point and absolute value of power difference of each candidate sudden drop event; Calculate the ratio of the absolute value of the power difference to the initial power value as the sudden drop amplitude; Sorting the candidate sudden drop events in time series according to the starting time point to generate a dynamic event sequence; Analyze the interval duration distribution of adjacent events in the dynamic event sequence to determine the triggering frequency and correlation characteristics of sudden drop events.

3. The method according to claim 1, characterized in that The step of performing spatiotemporal correlation analysis on the high-frequency charging time period group and the dynamic event sequence to determine the response relationship between the cost segmentation data and the power adjustment behavior includes: Establish the first analysis layer with the active charging period interval as the time dimension; Establish the second analysis layer with the sudden drop time point in the dynamic event sequence as the event dimension; Mapping the cost segmentation data to the first analysis layer according to the timestamp to generate a cost change thermal distribution map; Matching and associating the sudden drop amplitude with the cost fluctuation value in the cost segmentation data to generate a cost response intensity index; A spatiotemporal correlation response relationship matrix is ​​constructed based on the cost change heat distribution map and the cost response intensity index.

4. The method according to claim 1, characterized in that: The step of constructing a charging preference label for the vehicle charging terminal based on the user behavior feature set includes: Extracting a charging period preference pattern from the user behavior feature set, wherein the charging period preference pattern is determined by the period weight and power peak density distribution of the active charging period interval; Analyze the user adjustment behavior records in different fee ranges of the fee segmentation data to generate a fee tolerance range; Determining a power stability threshold according to a trigger frequency of a sudden drop event in the dynamic event sequence and a cost response intensity index; The time period weight, power stability threshold and cost tolerance interval are integrated in multiple dimensions to generate a charging preference label for the on-board charging terminal; Dynamically adjust the parameter weights in the charging preference tag through a real-time charging data stream update mechanism, where the parameter weights are synchronously updated based on the load fluctuations of the liquid-cooled supercharging terminal and the grid cost strategy; The step of analyzing the user adjustment behavior records of the fee segmentation data in different fee intervals to generate a fee tolerance interval includes: Dividing the expense segmentation data into a plurality of continuous sub-expense intervals according to the expense amount; Count the number of charging behavior adjustments and the corresponding adjustment types within each sub-cost interval, where the adjustment types include active power reduction, charging suspension, and terminal switching operations; Calculating an adjustment trigger rate for each sub-cost interval, wherein the adjustment trigger rate is determined by a ratio of the number of adjustments to the number of charging times; generating a cost tolerance interval according to the adjustment trigger rate and the sub-cost interval span, wherein the cost tolerance interval includes a first tolerance boundary and a second tolerance boundary; The cost tolerance interval is dynamically calibrated through a sliding window algorithm to eliminate the interference of abnormal data on the interval range.

5. The method according to claim 1, characterized in that: The generating a dynamic scheduling strategy across liquid-cooled supercharging terminals according to the charging preference tag includes: Parsing the time period weight, power stability threshold and cost tolerance range in the charging preference tag; Obtain real-time operating load data and grid power supply constraints of multiple liquid-cooled supercharging terminals; Predicting charging demand distribution and power demand peak within a target time window based on the time period weights; generating a power distribution path between terminals according to the power demand peak and the real-time operating load data; The cost compensation coefficient in the power split path is adjusted in combination with the cost tolerance interval to generate a dynamic scheduling strategy across terminals.

6. The method according to claim 5, characterized in that The predicting of charging demand distribution and power demand peak within the target time window based on the time period weights includes: Extracting the active charging time interval and the time period coverage frequency in the time period weight; Construct a forecast input sequence with the period coverage frequency as the time axis and the power peak density as the intensity axis; Analyzing the periodic characteristics in the input sequence through a time series pattern recognition algorithm to generate a charging demand prediction curve within a target time window; Identifying a local extreme point in the charging demand prediction curve as a power demand peak; The charging demand distribution is determined according to the timestamp corresponding to the local extreme point and the terminal location information.

7. The method according to claim 1, characterized in that The real-time power supply parameters and cost calculation logic of the target liquid-cooled supercharging terminal are optimized according to the dynamic scheduling strategy, including: Determine a threshold allocatable power value of the target terminal according to the power split path in the dynamic scheduling strategy; Adjust the cost ladder calculation rule of the target terminal in combination with the cost tolerance interval in the charging preference tag; Real-time monitoring of the target terminal's power usage and user behavior feedback signals; Dynamically optimize the priority allocation mechanism in the power supply parameters to ensure power supply priority for users with active time period weights; Eliminate the cost differences caused by power diversion through the cost balance algorithm, and maintain the stability of user cost expenditure within the cost tolerance range; The cost balancing algorithm is used to eliminate the cost difference caused by power splitting, including: Identify user groups whose cost deviation due to power diversion exceeds a preset threshold; Extracting a cost tolerance interval boundary value in the charging preference tag of the user group; Calculate the degree of deviation of the cost deviation from the cost tolerance interval boundary; generating a dynamic compensation coefficient according to the degree of deviation, wherein the compensation coefficient is positively correlated with the degree of deviation; The dynamic compensation coefficient is embedded in the cost calculation logic to achieve real-time correction of cost deviation.

8. The method according to claim 5, characterized in that After the cross-terminal dynamic scheduling strategy is generated, the method further includes: Real-time monitoring of the execution effect data of the dynamic scheduling strategy, wherein the execution effect data includes resource utilization improvement rate and user satisfaction index; Compare the difference between the execution effect data and the preset optimization target to generate strategy adjustment parameters; Feeding back the strategy adjustment parameters to the updating mechanism of the charging preference tag to form a closed-loop optimization link; Through multiple rounds of iterative optimization, the optimal scheduling strategy is gradually converged to achieve adaptive matching of charging resource allocation and user needs.

9. A liquid-cooled supercharging terminal charging middle station system, characterized in that: The method comprises at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method according to any one of claims 1 to 8.

Images