New energy charging pile dynamic scheduling method, system, equipment and medium

Abstract

The invention relates to a new energy charging pile dynamic scheduling method, system and device and a medium, and the method comprises the steps: collecting the real-time load data of a power grid, user charging request information, battery health state data and historical charging data, and carrying out the classification and aggregation of charging behavior characteristics, and generating behavior grouping labels; demand prediction and power safety boundary calculation are carried out based on a preset power prediction model in combination with the battery health state, and a power prediction result is formed; detecting power grid load conflicts and optimizing power distribution to generate a preliminary scheduling scheme; performing multi-target power optimization adjustment on the conflict points to generate a charging power scheme, and issuing and executing the charging power scheme; and finally, dynamically updating prediction model parameters according to charging pile feedback data. According to the method, the problems of neglect of battery health change, insufficient charging behavior distinguishing, response lag, weak collaborative optimization capability and the like in the prior art are solved, and collaborative optimization of power grid stability, equipment safety and charging efficiency is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of new energy charging control, in particular to a new energy charging pile dynamic scheduling method, system, device and medium. BACKGROUND

[0002] With the popularization of electric vehicles and the development of smart grids, the dynamic scheduling technology of new energy charging piles has become a core link of urban energy management. This kind of technology mainly focuses on how to coordinate the relationship between charging demand and grid carrying capacity in complex grid environment, while taking into account the safety of battery equipment and user experience. Its core lies in dynamically adjusting the charging pile power distribution strategy through real-time data collection and analysis, in order to realize the multiple goals of stable operation of power grid, prolongation of equipment life and improvement of charging efficiency. The current mainstream method usually distributes power based on grid load prediction and simple priority rules, trying to find a balance point between power supply capacity and user demand.

[0003] However, the existing technology still has significant limitations in practical application. First, the traditional method fails to fully consider the dynamic changes of battery health status on the limitation of charging power, resulting in accelerated aging and even thermal runaway risk of battery pack in long-term use. Secondly, in the scene of frequent grid load fluctuation, most scheduling schemes lack fine differentiation of multiple types of charging behavior, making it difficult to effectively balance the charging needs of emergency service vehicles and ordinary users. More importantly, the existing system generally has a response lag problem, when the actual load deviates from the predicted value, it cannot quickly adjust the power distribution strategy, which is easy to cause local power grid overload or resource idling. In addition, the lack of coordination optimization capability of charging pile cluster makes it difficult to balance between grid stability and charging efficiency in the overall scheduling scheme, especially in peak periods, often resulting in a sharp increase in charging waiting time or grid fluctuation exceeding the standard. SUMMARY

[0004] Based on this, the purpose of the present application is to provide a new energy charging pile dynamic scheduling method, system, device and medium which can real-time coordinate the grid carrying capacity, battery safety constraints and user demand priority, and has a dynamic feedback adjustment mechanism.

[0005] The purpose of the present application is achieved by the following scheme:

[0006] In a first aspect, the present application provides a new energy charging pile dynamic scheduling method, comprising the following steps:

[0007] S1: processing the collected real-time grid load data, user charging request information, battery health status data and historical charging data, classifying and aggregating the charging behavior characteristics according to the charging period characteristics, power preference characteristics and vehicle type characteristics, and generating behavior grouping labels;

[0008] S2: processing the behavior grouping label based on a preset power prediction model, time series prediction of charging demand combined with battery health state data and historical charging data and calculation of power safety boundary to generate a power prediction result containing grouping demand prediction value and charging pile available power range;

[0009] S3: processing the power prediction result, detecting grid load conflict and optimizing power distribution of the charging pile to generate a preliminary scheduling scheme containing conflict point identification;

[0010] S4: processing the preliminary scheduling scheme, multi-objective power optimization adjustment for the charging pile associated with the conflict point identification to generate a charging power scheme meeting the grid stability index and sending the charging power scheme to the charging pile controller, the charging power scheme being used to indicate the power output value;

[0011] S5: processing the actual load data fed back by the charging pile controller, calculating load deviation and judging whether the load deviation exceeds a threshold value to update the parameter weight of the power prediction model.

[0012] In one of the embodiments, the new energy charging pile dynamic scheduling method provided by the application specifically includes the following steps of S1:

[0013] S11: performing noise filtering processing on the collected real-time grid load data, using a sliding window filtering algorithm to eliminate transient interference signals in the grid load fluctuation to generate a standardized grid load sequence;

[0014] S12: performing feature analysis processing on the collected user charging request information, extracting charging time period distribution features, power demand level features and vehicle type code features to generate a user behavior feature vector;

[0015] S13: performing correlation analysis processing on the collected battery health state data and historical charging data, performing charging behavior similarity clustering combined with the grid load sequence and the user behavior feature vector, calculating the Euclidean distance between the feature vectors and setting a similarity threshold value, merging adjacent core areas with a distance less than the threshold value to form a charging behavior cluster to generate a behavior grouping label, the behavior grouping label being used to identify a vehicle group with similar charging behavior.

[0016] In one of the embodiments, the new energy charging pile dynamic scheduling method provided by the application specifically includes the following steps of S2:

[0017] S21: performing time series analysis processing on the behavior grouping label and the historical charging data, calculating the charging demand change trend of each behavior grouping based on a preset power prediction model to generate a grouping demand prediction value;

[0018] S22: Perform health constraint mapping processing on the battery health state data, calculate the maximum allowed charging power according to the battery health degree attenuation curve and the temperature change relationship, and generate a dynamic power safety boundary;

[0019] S23: Perform collaborative optimization processing on the grouped demand prediction value and the dynamic power safety boundary, adjust the excess prediction value based on the preset power grid capacity threshold and the transformer load rate, and generate a power prediction result containing the grouped demand prediction value and the charging pile available power range.

[0020] In one embodiment, the new energy charging pile dynamic scheduling method provided by the application specifically includes the following steps:

[0021] S31: Perform conflict detection processing on the grouped demand prediction value in the power prediction result, analyze whether the power grid load peak value exceeds the preset safety threshold, whether the battery temperature change rate exceeds the preset temperature rise limit, and whether the high-priority vehicle concentration exceeds the set proportion threshold, mark the charging pile that meets any condition as a conflict node, and generate a conflict point identification set;

[0022] S32: Perform power allocation optimization processing on the charging piles associated with the conflict point identification set, assign a dynamic weight coefficient based on the vehicle service type data in the user charging request information, and construct an optimization objective function that minimizes the weighted waiting time under the constraint of the charging pile available power range in the power prediction result;

[0023] S33: Perform linear programming solution processing on the optimization objective function, iteratively calculate the optimal solution that meets the power grid load stability condition using the simplex method, generate a preliminary scheduling scheme containing the conflict point identification, and the preliminary scheduling scheme is used to indicate the initial power allocation value of each charging pile.

[0024] In one embodiment, the new energy charging pile dynamic scheduling method provided by the application specifically includes the following steps:

[0025] S41: Perform parameter initialization on the conflict point identification in the preliminary scheduling scheme, construct a particle swarm position vector based on the power setting value of the associated charging pile, and generate an optimization parameter matrix containing position and velocity parameters;

[0026] S42: Perform fitness evaluation on the optimization parameter matrix, calculate the weighted comprehensive index of the power grid load variance and the battery temperature rise rate, and generate an evolution direction vector;

[0027] S43: Perform iterative optimization processing on the evolution direction vector, update the particle position and velocity parameters through a loop, compare the current fitness with the historical optimal value, continuously adjust the particle swarm state until the power grid stability index is met, generate a charging power scheme and send it to the charging pile controller.

[0028] In one of the embodiments, the application provides a new energy charging pile dynamic scheduling method, and a calculation formula of a weighted comprehensive index of the method is as follows:

[0029]

[0030] Wherein, F is the weighted comprehensive index, is the grid load variance is the battery temperature rise rate, ΔT is the temperature change, Δt is the time change, and α and β are weight coefficients.

[0031] In one of the embodiments, the S5 of the new energy charging pile dynamic scheduling method provided by the application specifically includes the following steps:

[0032] S51: deviation calculation and processing are performed on real-time load data fed back by a charging pile controller, an actual load deviation rate index is constructed based on a grouping demand prediction value of a power prediction result, and dynamic deviation data is generated;

[0033] S52: the dynamic deviation data is continuously analyzed and processed, an abnormal state in which a deviation rate exceeds a preset deviation threshold and lasts for a set time window is detected, and a model update triggering instruction is generated;

[0034] S53: parameter adjustment processing is performed on the model update triggering instruction, a prediction error of the power prediction model is calculated based on a gradient descent algorithm, and a weight parameter of the power prediction model is adjusted.

[0035] In a second aspect, the application provides a new energy charging pile dynamic scheduling system, which is configured with the following modules:

[0036] A charging behavior grouping module is configured to process collected grid real-time load data, user charging request information, battery health state data and historical charging data, classify and aggregate charging behavior characteristics according to charging time period characteristics, power preference characteristics and vehicle type characteristics, and generate behavior grouping labels;

[0037] A power demand prediction module is configured to process the behavior grouping labels based on a preset power prediction model, combine the battery health state data and the historical charging data to perform time series prediction on charging demand and calculate a power safety boundary, and generate a power prediction result containing a grouping demand prediction value and a charging pile available power range;

[0038] A load conflict scheduling module is configured to process the power prediction result, detect grid load conflicts and optimize power distribution of charging piles, and generate a preliminary scheduling scheme containing conflict point identification;

[0039] The power optimization adjustment module is configured to process the preliminary scheduling scheme, perform multi-objective power optimization adjustment on the charging piles associated with the conflict points, generate a charging power scheme meeting the power grid stability index, and send the charging power scheme to the charging pile controller, where the charging power scheme is configured to indicate a power output value.

[0040] The model parameter updating module is configured to process the actual load data fed back by the charging pile controller, calculate a load deviation, determine whether the load deviation exceeds a threshold, and update a parameter weight of the power prediction model.

[0041] In a third aspect, the present application provides a computer device, which comprises a memory and a processor, the memory stores a computer program, and the processor implements any of the new energy charging pile dynamic scheduling methods described above when executing the computer program.

[0042] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement any of the new energy charging pile dynamic scheduling methods described above.

[0043] In summary, the new energy charging pile dynamic scheduling method provided by the present application can realize fine differentiation of user groups based on dynamic classification and aggregation of charging behavior characteristics, and overcome the defects of traditional methods that do not consider the differences in charging modes. Through the time series prediction mechanism of battery health state data, a dynamic safety boundary constraint can be established in the power prediction stage to ensure that the charging process does not accelerate battery aging or induce thermal runaway risk. The directed detection and optimization processing of the power grid load conflict can realize early identification of overload risk and resource conflict, and combined with the weight allocation strategy based on the vehicle service type, the differentiated needs of emergency service vehicles and ordinary users can be considered. In the multi-objective power optimization adjustment stage, an adaptive weight mechanism is used to simultaneously optimize power grid fluctuation suppression and battery temperature rise control to achieve dynamic balance of power supply quality and equipment safety. The model parameter updating mechanism based on actual load deviation can build a prediction-execution-feedback closed-loop control system to effectively solve the response lag problem of traditional schemes. Finally, through the instruction closed loop of the charging pile controller, the power distribution can be dynamically adjusted at a second level to achieve the triple goals of minimizing power grid fluctuations, maximizing battery life, and optimizing user waiting time, thereby providing a systematic solution for energy dynamic scheduling in high-density charging scenarios.

[0044] For better understanding and implementation, the present application will be described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 A flowchart of a new energy charging pile dynamic scheduling method provided by an embodiment of the present application is shown in the figure.

[0046] Figure 2A flowchart for generating a charging power scheme is provided for an embodiment of the present application.

[0047] Figure 3 A structural diagram of a new energy charging pile dynamic scheduling system is provided for another embodiment of the present application. DETAILED DESCRIPTION

[0048] In order to facilitate the understanding of the present application, the present application will be described more fully below with reference to the accompanying drawings. The preferred embodiments of the present application are shown in the drawings. However, the present application can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive.

[0049] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs. The terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0050] In one embodiment, as shown in Figure 1 A new energy charging pile dynamic scheduling method is provided, and the present embodiment takes the method applied to a terminal as an example. It can be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is realized through the interaction of the terminal and the server. In the present embodiment, the method includes the following steps:

[0051] S1: processing the collected real-time load data of the power grid, user charging request information, battery health state data and historical charging data, classifying and aggregating the charging behavior characteristics according to the charging period characteristics, power preference characteristics and vehicle type characteristics, and generating behavior grouping labels.

[0052] Specifically, the system collects real-time power grid load data through smart meters, substation monitoring systems and built-in sensors of charging piles deployed on the grid side, including current load power, voltage fluctuation, frequency deviation and transformer load rate and other parameters; user charging request information is reported by the charging pile human-computer interaction interface or mobile application, including charging start time, expected completion time, requested charging capacity, user identity and other information; battery health state (SOH) data is uploaded through the vehicle-mounted battery management system (BMS) via communication protocols (such as GB / T 27930, ISO 15118), covering battery internal resistance, maximum available capacity, cycle count, temperature distribution and health score and other key indicators; historical charging data is retrieved from the background database, including past charging period distribution, average charging power, single charging capacity, interruption frequency and other statistical information.

[0053] Exemplarily, the system cleans, normalizes and performs feature engineering processing on the above-mentioned multi-source data, extracts charging period features through time series analysis, divides each day into multiple time windows, and calculates the charging frequency and duration proportion of each user in each period; power preference features are quantified by statistical parameters such as average output power, maximum power usage frequency and power change slope during the user's historical charging process; vehicle type features are classified according to user registration information or BMS-reported vehicle model data, such as private cars, online car hailing, taxis, buses and logistics vehicles, and combined with their typical battery capacity, daily driving distance and charging habits to establish type labels.

[0054] Preferably, the system classifies and aggregates charging behavior features based on charging period features, power preference features and vehicle type features. Charging period features are divided according to power grid load characteristics, power preference features are divided according to charging pile output power, and vehicle type features are divided according to use scenarios and priorities. The system can use an improved K-means clustering algorithm to realize feature aggregation, taking the three types of features as clustering dimensions, determining the number of clusters by the elbow rule, introducing a weighting coefficient in the clustering process to ensure that the behavior features of high-priority vehicles are accurately distinguished, and generating a unique behavior grouping label after clustering, which contains period, power preference, vehicle type and priority information.

[0055] S2: Based on the preset power prediction model, the behavior grouping label is processed, the charging demand is time series predicted combined with the battery health state data and the historical charging data, and the power safety boundary is calculated, generating a power prediction result containing grouping demand prediction value and charging pile available power range.

[0056] Specifically, the system constructs a composite prediction model that integrates multi-source information based on time series prediction theory and battery electrochemical characteristic modeling principles, to achieve accurate prediction of future charging demand and safe power boundary determination. The technical core of the model is to introduce behavioral grouping labels as prior knowledge into the prediction process, improving the model's ability to capture group behavior trends, while combining battery physical constraints to ensure the engineering feasibility of the prediction results. The preset power prediction model is a deep learning-based sequence-to-sequence (Seq2Seq) architecture, consisting of an encoder and a decoder with an embedded attention mechanism in between.

[0057] Preferably, the encoder of the model can use a long short-term memory network (LSTM) or a gated recurrent unit (GRU) structure to extract the time-dependent relationships in historical charging data. The input sequence includes the cumulative charging power, the number of charging users, and the average charging duration of each behavioral group at the past N time steps. External covariates such as real-time load of the power grid, weather conditions, holiday indicators, etc. are also encoded as auxiliary feature vectors, which are input into the encoder along with the main sequence. Similarly, the decoder can use an LSTM / GRU structure to receive the context vector output by the encoder and combine the typical charging pattern prior knowledge represented by the behavioral grouping label to generate the group-level charging demand prediction values for the next M time steps. The attention mechanism allows the model to dynamically focus on the most relevant historical segments in the input sequence during the decoding process. For example, when predicting the demand of the "ride-hailing daytime power supplement" group, the model can focus on the historical charging peak data during the lunch break on weekdays.

[0058] Illustratively, the system inputs the input matrix into the trained model to predict the total charging demand of each behavioral group in the future scheduling period, i.e., the group demand prediction value. The system calculates the power safety boundary, which includes the power grid constraint boundary, the battery safety boundary, and the charging pile hardware boundary. The power grid constraint boundary takes the real-time available load of the distribution network as the upper limit and zero as the lower limit, and the real-time available load of the distribution network is determined by the difference between the distribution transformer rated capacity and the non-charging load power. The battery safety boundary is dynamically calculated based on the battery health state data, and the upper limit is adjusted according to the battery health state, the remaining power, and the single battery temperature. The charging pile hardware boundary takes the rated power of the charging pile as the upper limit and the power value that avoids idle loss as the lower limit. The available power range of the charging pile is the intersection of the three. The system stores the power prediction results in the form of a structured data set, which includes the group ID, the prediction period, the group demand prediction value, the charging pile ID, the available power upper limit, the available power lower limit, and the prediction confidence.

[0059] S3: processing the power prediction result, detecting the grid load conflict and optimizing the power distribution of the charging pile, and generating a preliminary scheduling scheme containing the conflict point identification.

[0060] Specifically, in the conflict detection link, the system aggregates the demand prediction value of all behavior groups in each time slice to obtain the total demand power, compares the total demand power with the upper limit of the available power of the grid, and if the total demand power exceeds the upper limit of the available power of the grid, it is determined that there is a grid load conflict, and the system records the conflict time slice, the behavior groups and charging pile numbers involved, and the excess power related information; if the demand prediction value of a charging pile exceeds its available power range, it is determined that there is a single pile power conflict, and the system separately marks the charging pile.

[0061] After completing the conflict detection, the system optimizes the power distribution of the charging pile based on the preset optimization target and constraint conditions, the optimization target is to maximize the demand satisfaction rate of high priority groups, and the constraint conditions include that the total distribution power does not exceed the upper limit of the available power of the grid, the single pile distribution power is within its available power range, and the demand satisfaction rate of different priority groups meets the preset requirements. Preferably, the system can use a greedy algorithm to implement power distribution, preferentially satisfying the demand of the highest priority group, and then sequentially distributing the demands of other groups in order of priority. If the demand of a group exceeds the remaining available power, the system proportionally reduces the distribution power of low power preference charging piles in the group until the total distribution power meets the constraint conditions. After optimization, the system generates a preliminary scheduling scheme, which contains time slices, charging pile numbers, distribution powers, conflict point identifications, priorities, etc. The conflict point identification needs to clearly identify the conflict type, conflict time slice, and excess power related information.

[0062] S4: processing the preliminary scheduling scheme, performing multi-objective power optimization adjustment on the charging piles associated with the conflict point identification, generating a charging power scheme that meets the grid stability index and sending it to the charging pile controller, and the charging power scheme is used to indicate the power output value.

[0063] Specifically, the core objectives of multi-objective optimization include grid stability, battery safety and user experience. The grid stability objective involves indicators such as post-dispatch grid load fluctuation, voltage deviation, frequency deviation, etc. The battery safety objective involves indicators such as charging current, charging voltage, battery temperature, etc. The user experience objective involves indicators such as the deviation between actual charging time and user expected time, charging completion rate, etc. Each objective is included in the optimization consideration according to a pre-set weight. Preferably, the system can implement multi-objective optimization using the NSGA-II algorithm, encode the charging pile distribution power involved in the conflict point as a chromosome, then initialize the population, and the generated initial solution must satisfy the single-pile power range constraint. The system constructs a fitness function based on the grid stability indicator, battery safety indicator, and user experience indicator, and calculates the fitness value of each solution through the function. Then, the roulette wheel selection method is used to select the parent generation, and the single-point crossover and Gaussian mutation operations are used to generate the child generation, combined with non-dominated sorting and congestion calculation to retain the optimal solution. After iterating until the pre-set termination condition is met, the system selects a solution that takes into account the three objectives from the Pareto optimal solution set as the final adjustment scheme.

[0064] After the adjustment scheme is determined, the system verifies the grid stability indicators of the optimized distribution power, simulates the post-dispatch grid operating state through simulation means, and confirms that the relevant indicators meet the requirements. After passing the verification, the system generates a charging power scheme, which contains information such as charging pile number, time slice, power output value, and safety verification result. The system uses a pre-set communication protocol to send the charging power scheme to the charging pile controller, ensuring reliable message transmission. The charging pile controller receives the scheme and analyzes it, adjusts the output power through PWM signal control of the power component, and realizes precise power control.

[0065] S5: Process the actual load data fed back by the charging pile controller, calculate the load deviation and determine whether it exceeds the threshold, and update the parameter weight of the power prediction model.

[0066] Specifically, the charging pile controller collects the output power in real time through the built-in current sensing device and voltage sensing device, and aggregates the average power by time slice as the actual load data of that time slice to ensure that the data is representative. The system uses a feedback mechanism that combines regular feedback and abnormal triggering. Under normal circumstances, the charging pile controller uploads the actual load data to the dispatch server through the wireless communication component at a period synchronized with the time slice. If the actual power exceeds the pre-set deviation range of the distributed power, immediate feedback is triggered to ensure that the system responds quickly to abnormalities.

[0067] After receiving the actual load data, the system calculates the load deviation and judges whether it exceeds the threshold value. Specifically, the system calculates the absolute deviation and the relative deviation for each charging pile in each time slice. The absolute deviation is the absolute value of the actual load and the allocated power, and the relative deviation is the percentage of the absolute deviation and the allocated power. At the same time, the system calculates the average relative deviation of all charging piles in the same time slice, which is used as the overall deviation index of the time slice. The system sets the deviation threshold value based on the grid fault tolerance capability and the charging demand accuracy requirement. If the relative deviation does not exceed the threshold value, it is determined that the deviation is normal. If the relative deviation exceeds the threshold value, it is determined that the deviation is out of limit, triggering the parameter update of the power prediction model.

[0068] Preferably, the system can use a preset optimizer to update the LSTM module parameters in the power prediction model, taking the mean square error of the load deviation as the loss function, and adjusting the input layer weight, hidden layer weight and output layer weight through back propagation. For the ARIMA module, the system re-estimates the model parameters based on the time slice data of the deviation out of limit, and determines the parameter value using the maximum likelihood estimation method. The system keeps the update period and the feedback period synchronized. If the deviation is normal for consecutive multiple periods, the update period is extended to avoid overfitting of the model. The updated model parameters are stored in the cloud model parameter library, supporting version control. The latest parameters are automatically called by the system during the next prediction, ensuring that the model prediction accuracy is dynamically improved with actual operation data.

[0069] In summary, the new energy charging pile dynamic scheduling method provided by the present application can realize fine differentiation of user groups based on dynamic classification and aggregation of charging behavior characteristics, overcoming the defects of traditional methods that do not adequately consider the differences in charging modes. Through the time series prediction mechanism of integrating battery health state data, a dynamic safety boundary constraint can be established during the power prediction stage to ensure that the charging process does not accelerate battery aging or induce thermal runaway risk. The directed detection and optimization processing of grid load conflicts can achieve early identification of overload risk and resource conflicts, and combined with the weight allocation strategy based on vehicle service type, the differentiated needs of emergency service vehicles and ordinary users can be considered. In the multi-objective power optimization adjustment stage, an adaptive weight mechanism is used to simultaneously optimize grid fluctuation suppression and battery temperature rise control to achieve dynamic balance between power supply quality and equipment safety. The model parameter update mechanism based on actual load deviation can construct a closed-loop control system of prediction-execution-feedback, effectively solving the response lag problem of traditional solutions. Finally, through the instruction closed loop of the charging pile controller, the power allocation can be dynamically adjusted at the second level, achieving the triple goals of minimizing grid fluctuations, maximizing battery life and optimizing user waiting time, providing a systematic solution for energy dynamic scheduling in high-density charging scenarios.

[0070] In one embodiment, the new energy charging pile dynamic scheduling method provided by the present application comprises the following steps:

[0071] S11: Noise filtering processing is performed on the collected real-time power grid load data, and a sliding window filtering algorithm is used to eliminate transient interference signals in the power grid load fluctuation, to generate a standardized power grid load sequence.

[0072] Specifically, the system receives a real-time load data stream from a power grid monitoring device, which contains power, current and voltage measurements of multiple monitoring points in the power distribution network, and timestamps the obtained raw data to ensure that the data from different locations have a unified time reference. Preferably, the system can process the data stream using a sliding window filtering algorithm, and the length of the sliding window is determined according to the dynamic response characteristics of the power grid.

[0073] At each time step, the system calculates the weighted average value of the data in the window, and the weight distribution is set according to the time-varying characteristics of the signal. Transient interference signals are characterized by short-time amplitude mutations, with a duration shorter than the characteristic time scale of normal load changes. Through the smoothing effect of the sliding window, the system reduces the impact of transient interference on the data sequence. The system normalizes the filtered data to eliminate the dimensional differences between different monitoring points. The normalization process can use a linear transformation method to map the data from each point to a unified numerical range, generating a standardized power grid load sequence that reflects the macroscopic variation trend of the power grid load and removes high-frequency noise and short-term fluctuations.

[0074] S12: Feature analysis processing is performed on the collected user charging request information to extract charging period distribution features, power demand level features and vehicle type coding features, to generate a user behavior feature vector.

[0075] Specifically, the system performs feature analysis processing on the collected user charging request information, which is obtained from the charging pile human-computer interaction terminal or the user terminal application, and contains the basic information submitted by the user, such as the expected charging start time, the expected charging end time, the target SOC, the charging mode selection, the vehicle brand and model, etc. The system extracts the charging period distribution feature, determines the time period belonging of each charging request according to the user's expected charging start time and end time, and combines the time period categories divided according to the power grid load characteristics, to convert the time period belonging information into a quantifiable feature value, forming the charging period distribution feature, which reflects the distribution rule of the user's charging behavior in the time dimension.

[0076] Subsequently, the system extracts the power demand level feature, associates the power output range of the charging pile according to the user-selected charging mode and the difference between the target SOC and the current SOC, divides different power demand levels, each level corresponds to a power demand range, and the system converts the power demand level into a numerical feature to form the power demand level feature. The feature reflects the user's demand intensity for charging power, wherein the charging mode includes fast charging, ordinary charging, and slow charging. At the same time, the system extracts the vehicle type coding feature, encodes the vehicle type according to the user-provided vehicle brand and model, and combines the classification standard of the vehicle use scene to generate a unique vehicle type code. After converting the code into a feature value, the vehicle type coding feature is formed. The feature distinguishes the charging demand priority of different types of vehicles. After completing the extraction of the three types of features, the system combines the charging time period distribution feature, the power demand level feature, and the vehicle type coding feature in a predetermined order to construct a multi-dimensional data structure and generate a user behavior feature vector. The vector contains the core information of the user's charging behavior in the time, power, and vehicle type dimensions.

[0077] S13: Correlation analysis and processing of the collected battery health status data and historical charging data, charging behavior similarity clustering combined with the power grid load sequence and the user behavior feature vector, calculation of the Euclidean distance between the feature vectors and setting of a similarity threshold, merging of adjacent core areas with a distance less than the threshold to form a charging behavior cluster, generation of a behavior grouping label, and the behavior grouping label used to identify a vehicle group with similar charging behavior.

[0078] Specifically, the system simultaneously acquires battery health status data and historical charging data, the battery health status data including battery capacity, internal resistance, and cycle count, the historical charging data including past charging start and end times, charging amount, and power curve, and time-aligns and correlates the battery health status data with the historical charging data to establish the relationship between battery status and charging behavior. The system combines the user behavior feature vector, the battery health status, and the historical charging mode into a comprehensive feature vector, calculates the Euclidean distance between the comprehensive feature vectors, and the Euclidean distance reflects the similarity degree of different user charging behaviors.

[0079] Illustratively, the system sets a similarity threshold, which is determined according to the distribution density of the feature space and the clustering target. The system uses a density-based clustering method to divide the feature vectors with a distance less than the similarity threshold into the same cluster. The system identifies the core area in the feature space, which is composed of high-density feature vectors. The system merges adjacent core areas to form a charging behavior cluster, each charging behavior cluster corresponding to a typical charging behavior mode, and the system generates a unique behavior grouping label for each charging behavior cluster, which is associated with the feature center of the cluster.

[0080] In one of the embodiments, the application provides a new energy charging pile dynamic scheduling method S2, which specifically comprises the following steps:

[0081] S21: Time series analysis and processing are performed on the behavior grouping label and historical charging data, the charging demand change trend of each behavior grouping is calculated based on a preset power prediction model, and a grouping demand prediction value is generated.

[0082] Specifically, the system performs time dimension alignment processing on the behavior grouping label and historical charging data, classifies the historical charging data according to the behavior grouping label, ensures that each grouping corresponds to a unique set of historical charging data, and avoids confusion of data in different groupings. The system performs time series decomposition on the classified historical charging data, divides the data into continuous time series segments according to a preset time granularity, extracts the charging demand peak, valley, average demand and other time series characteristics in each segment, and labels the corresponding power grid load background, seasonal factors and other associated conditions of each time series segment, forming a structured time series feature set.

[0083] Preferably, the system can call a preset power prediction model, input the time series feature set corresponding to the behavior grouping label into the model, the model learns the demand change rule in the historical time series data, calculates the charging demand change trend of each behavior grouping in a future preset scheduling period, and associates the inherent characteristics of the behavior grouping, such as the charging frequency corresponding to the vehicle type and the demand intensity corresponding to the power preference, in the trend calculation process. The system integrates the demand estimation values at multiple time nodes output by the model according to the demand change trend, generates the total charging demand quantization result of each behavior grouping in the corresponding prediction period, i.e., the grouping demand prediction value, and associates the corresponding behavior grouping ID and prediction period information with the prediction value.

[0084] S22: Health constraint mapping processing is performed on the battery health state data, the maximum allowed charging power is calculated according to the battery health degree attenuation curve and the temperature change relationship, and a dynamic power safety boundary is generated.

[0085] Specifically, the system first analyzes the health degree parameter in the battery health state data, calls the preset battery health degree attenuation curve, which reflects the law of the change of the battery health degree with the use time and the cycle number, determines the allowed charging power basic constraint value under the current battery health degree according to the corresponding position of the health degree on the attenuation curve, that is, the higher the health degree, the closer the basic constraint value to the rated charging power of the battery, and the lower the health degree, the lower the basic constraint value, so as to slow down the further attenuation of the health degree. At the same time, the system associates the temperature parameter in the battery health state data, establishes the corresponding relationship between the temperature change and the power adjustment coefficient, when the current temperature of the battery is in the preset suitable interval, the power adjustment coefficient is 1, that is, the basic constraint value is not additionally limited; when the temperature exceeds the upper limit of the suitable interval, the adjustment coefficient is reduced according to the temperature exceeding range, and the allowed charging power is correspondingly reduced; when the temperature is lower than the lower limit of the suitable interval, the adjustment coefficient is also fine-tuned according to the temperature deviation, so as to avoid the damage to the battery caused by high-power charging at low temperature.

[0086] Preferably, the system multiplies the basic constraint value corresponding to the health degree and the adjustment coefficient corresponding to the temperature to calculate the maximum allowed charging power under the current state of the battery, and at the same time, combines the charging cycle number and the voltage balance parameter of the battery to correct the calculation result, that is, the more the cycle number and the worse the voltage balance, the greater the correction amplitude of the maximum allowed charging power. The system takes the corrected maximum allowed charging power as the core parameter, combines the power tolerance characteristics of the battery in different charging stages, generates a power safety boundary that dynamically changes with the charging process, and the boundary needs to be associated with the battery unique identifier and the calculation time stamp to ensure that the boundary corresponding to each battery has real-time effectiveness.

[0087] S23: The grouped demand prediction value and the dynamic power safety boundary are cooperatively optimized, the excess prediction value is adjusted based on the preset power grid capacity threshold and the transformer load rate, and a power prediction result containing the grouped demand prediction value and the available power range of the charging pile is generated.

[0088] Specifically, the system calls the grouped demand prediction value from the prediction data temporary library, the dynamic power safety boundary from the battery constraint database, and at the same time, acquires the preset power grid capacity threshold and the transformer load rate parameter, the power grid capacity threshold reflects the upper limit of the total power that can be distributed to the charging pile by the current power distribution network, and the transformer load rate reflects the current load intensity that can be tolerated by the distribution transformer equipment, and the system needs to balance the demand, safety and power grid carrying capacity through cooperative optimization.

[0089] Exemplarily, the system preliminarily compares the demand prediction value of each behavior grouping with the dynamic power safety boundary of the corresponding battery, screens out the groupings whose demand prediction values exceed the corresponding dynamic power safety boundary, and marks them as to-be-adjusted groupings; for the groupings whose demand prediction values do not exceed the dynamic power safety boundary, the system temporarily retains the prediction values thereof as initial reserved values. The system secondly aggregates the demand prediction values of all the groupings, calculates a total demand prediction value, and compares the total demand prediction value with the preset power grid capacity threshold and the maximum allowed load corresponding to the transformer load rate - if the total demand prediction value does not exceed the power grid capacity threshold and does not cause the transformer load rate to exceed the standard, the system directly takes the current prediction value and the dynamic power safety boundary as the basic data; if the total demand prediction value exceeds any index, it is determined that there is an excess prediction, and the excess part needs to be adjusted. The system adjusts the excess prediction value in accordance with the principle of "giving priority to low-risk batteries with high health attenuation and high-priority vehicle groupings", that is, first reducing the demand prediction value of the grouping corresponding to the battery with high health attenuation risk, and then adjusting the prediction values of other to-be-adjusted groupings in order according to the priority of the vehicle type. During the adjustment process, the system needs to ensure that the final prediction value of each grouping does not exceed the dynamic power safety boundary corresponding thereto, and that the total demand prediction value does not exceed the power grid capacity threshold and the upper limit of the transformer load rate.

[0090] After the system adjustment is completed, the system integrates the final demand prediction values of the groupings, determines the available power range of each charging pile in combination with the dynamic power safety boundary, and generates a power prediction result containing the behavior grouping ID, the grouping demand prediction value, the charging pile ID, and the charging pile available power range.

[0091] In one of the embodiments, the S3 of the new energy charging pile dynamic scheduling method provided by the application specifically includes the following steps:

[0092] S31: performing conflict detection processing on the grouping demand prediction values in the power prediction result, analyzing whether the power grid load peak value exceeds the preset safety threshold, whether the battery temperature change rate exceeds the preset temperature rise limit, and whether the concentration of high-priority vehicles exceeds the set proportion threshold, marking the charging piles that meet any condition as conflict nodes, and generating a conflict point identification set.

[0093] Specifically, the system receives the group demand prediction value in the power prediction result, and calculates the total load power of each time period in combination with the real-time monitoring data of the power grid. The system compares the total load power with the preset safety threshold point by point, and the safety threshold is determined according to the rated parameters and operation specifications of the power distribution equipment. When the predicted load at a certain time point exceeds the threshold, the system determines that there is a risk of power grid overload, and marks all charging piles connected to the same feeder as potential conflict nodes. The system also obtains the temperature change rate information in the battery health state data, which is calculated by the temperature difference between consecutive time points. The system compares the temperature change rate with the preset temperature rise limit value, which is set according to the thermal stability characteristics of the battery material. When the temperature change rate exceeds the limit value, the system considers that there is a risk of thermal runaway, and includes the charging pile where the corresponding vehicle is located in the conflict node set.

[0094] Further, the system analyzes the distribution density of high-priority vehicles, which are identified by the service type data in the user charging request, including public transportation, emergency support and other categories. The system calculates the proportion of the number of high-priority vehicles in a unit geographical area. When the proportion exceeds the set proportion threshold, the system determines that local resource competition may lead to service delay, and also marks the related charging piles as conflict nodes. All marked charging piles and their corresponding time intervals are collected to form a conflict point identification set. The conflict point identification includes device number, time range, trigger condition type and associated behavior grouping label. The identification set is used as the input basis for subsequent optimization processing, to locate the space-time area that needs to adjust the power allocation.

[0095] S32: Perform power allocation optimization processing on the conflict point identification set associated charging piles, assign dynamic weight coefficients based on the vehicle service type data in the user charging request information, and construct an optimization objective function of minimizing weighted waiting time under the constraint of available power range of the charging pile in the power prediction result.

[0096] Specifically, the system classifies and analyzes the vehicle service type data, and assigns dynamic weight coefficients to different types of vehicles according to the preset service priority rules. Under the same service type, the weight coefficient is adjusted according to the charging demand urgency, and the higher the urgency, the greater the weight coefficient. Among different service types, the basic weight coefficient is determined according to the preset priority order, and the basic weight coefficient of high-priority service type is higher than that of low-priority service type. The value of the dynamic weight coefficient needs to ensure that the weight difference of different types of vehicles can be quantitatively reflected in the subsequent optimization process.

[0097] Exemplarily, the system secondly determines the constraint conditions of the power allocation, including: the power allocation value of each charging pile needs to be within the available power range of the corresponding charging pile, i.e. not lower than the lower limit of the available power and not higher than the upper limit of the available power; the total power allocation value of all conflict node charging piles associated with the same power grid node in the same period does not exceed the preset capacity limit of the power grid node; the power allocation value corresponding to the high-priority service type vehicle needs to meet its minimum charging demand, i.e. not lower than the minimum charging power requirement of the vehicle of this type.

[0098] Based on the above-obtained dynamic weight coefficients and constraint conditions, the system constructs an optimization objective function, the core objective of which is to minimize the weighted waiting time, which is calculated by the sum of the product of the waiting time of each vehicle and the corresponding dynamic weight coefficient, wherein the waiting time is calculated by the charging pile power allocation value, the vehicle expected charging amount and the charging efficiency, the variable of the objective function is the power allocation value of each conflict node charging pile, and the function expression needs to integrate the constraint conditions to ensure that all operating limits are met during the solving process.

[0099] S33: Linear programming solving process is performed on the optimization objective function, simplex method is used to iteratively calculate the optimal solution that meets the power grid load stability condition, and a preliminary scheduling scheme containing conflict point identification is generated, which is used to indicate the initial power allocation value of each charging pile.

[0100] Specifically, the system converts the optimization objective function and the constraint conditions into the standard form of linear programming, arranges the objective function into a linear expression of maximization or minimization, converts the inequality constraint into an equality constraint, determines the non-negative constraint condition of the variable, and determines the coefficient matrix in the objective function, the coefficient matrix in the constraint condition and the constant term vector. Preferably, the system performs iterative calculation by simplex method, and the calculation steps are as follows:

[0101] First step, determine the initial basic feasible solution, select the slack variable as the initial basic variable, and calculate the objective function value corresponding to the initial basic feasible solution;

[0102] Second step, calculate the test number, determine whether the current basic feasible solution is the optimal solution by the test number, if the test number meets the optimality condition, the current solution is the optimal solution, if not, go to the next step;

[0103] Third step, determine the in-base variable and the out-base variable, select the variable with the optimal test number as the in-base variable, and determine the out-base variable by the minimum ratio method to realize the iterative update of the basic variable;

[0104] Fourthly, repeat the calculation of the test number and the base variable updating process until a solution that satisfies the optimality condition is obtained. During the iterative solution process, the system synchronously verifies whether the solution satisfies the power grid load stability condition, i.e., whether the fluctuation amplitude of the power grid load corresponding to the total power distribution value obtained by the solution is within the preset stable range. If the solution satisfies the power grid load stability condition, the solution is determined as the final optimal solution. If the solution does not satisfy the power grid load stability condition, the power grid capacity parameter in the constraint condition is adjusted, and the iterative solution is re-executed until an optimal solution that meets the requirements is obtained.

[0105] The system determines the power distribution values of the charging piles of each conflict node corresponding to the optimal solution as initial power distribution values, integrates the initial power distribution values, conflict point identifiers, charging pile IDs, execution time periods, constraint condition satisfaction conditions and other information, and generates a preliminary scheduling scheme in a preset format. In the scheme, the initial power distribution values of each conflict node charging pile and the corresponding service vehicle type need to be clearly defined.

[0106] In one embodiment, as shown in FIG. 4, the new energy charging pile dynamic scheduling method provided by the present application includes the following steps: Figure 2

[0107] S41: Parameter initialization is performed on the conflict point identifiers in the preliminary scheduling scheme, a particle swarm position vector is constructed based on the power setting values of the associated charging piles, and an optimization parameter matrix containing position and velocity parameters is generated.

[0108] Specifically, the system receives the conflict point identifiers contained in the preliminary scheduling scheme, extracts the power distribution values of the charging piles associated with the conflict points in each time unit, arranges these power distribution values in the order of charging pile number and time sequence, and constructs an initial position vector of the particle swarm optimization algorithm. The vector represents a potential power configuration state in a multi-dimensional solution space. The system synchronously generates an initial velocity vector for each particle. Each component of the velocity vector represents the change direction and amplitude of the corresponding power variable in the solution space. The value range of the velocity is determined by the dynamic ability of the charging pile power adjustment.

[0109] The system combines the position vector and the velocity vector into an optimization parameter matrix. The number of rows of the matrix corresponds to the size of the particle swarm, and the number of columns corresponds to the total number of optimization variables, i.e., the product of the number of charging piles participating in the adjustment and the time step. The optimization parameter matrix also includes additional areas for storing individual historical optimal positions and group historical optimal positions. These areas are used to compare the current solution with the historical optimal solution during the iteration process. The system determines the set of charging piles participating in the optimization according to the power grid topology and load distribution, and includes their power variables in the optimization dimension to ensure that the optimization process covers all nodes that may affect the stability of the power grid.

[0110] ​S42: Perform fitness evaluation on the optimization parameter matrix, calculate the weighted comprehensive index of grid load variance and battery temperature rise rate, and generate the evolution direction vector.

[0111] Specifically, the system performs fitness evaluation on the power distribution scheme represented by each particle in the optimization parameter matrix, and the evaluation process is based on the comprehensive index of grid operation state and battery safety state. The system first calculates the total load of the grid at each time step according to the power output value corresponding to the particle position, and calculates the variance of the load sequence in the time dimension to measure the degree of load fluctuation, the variance calculation formula is as follows:

[0112]

[0113] Wherein, represents the grid load variance, P total (t) represents the total load power at time point tt, represents the average load power in the entire time sequence, and N represents the total number of time steps. The system also obtains the temperature data uploaded by the vehicle battery management system connected to the charging pile, calculates the temperature change rate per unit time as a quantitative indicator of battery thermal safety, and the temperature rise rate calculation formula is as follows:

[0114]

[0115] Wherein, ΔT represents the temperature change, Δt represents the time change, and T(t1) and T(t2) are the battery temperatures at two consecutive monitoring times. The system performs weighted summation on the grid load variance and the battery temperature rise rate to form a weighted comprehensive index, and the calculation formula is as follows:

[0116]

[0117] Wherein, F is the weighted comprehensive index, and α and β are weight coefficients, respectively representing the relative importance of grid stability and battery thermal safety in the comprehensive evaluation, and the weight coefficients are determined by system configuration parameters. Wherein, the lower the fitness value, the better the power distribution scheme performs in maintaining grid stability and controlling battery temperature rise, and the system determines the evolution direction of the particle swarm according to the fitness value, and the evolution direction vector is guided by the individual historical optimal solution and the group historical optimal solution, and is used to drive the subsequent speed and position update.

[0118] S43: Perform iterative optimization on the evolution direction vector, update the particle position and speed parameters through loop, compare the current fitness with the historical optimal value, and continuously adjust the particle swarm state until the grid stability index is met, generate the charging power scheme and send it to the charging pile controller.

[0119] Specifically, the system iteratively updates the velocity and position parameters of the particle swarm according to the evolutionary direction vector, and the velocity updating process follows the standard form of the particle swarm algorithm, and the calculation formula is as follows:

[0120] v ij (k+1) = v ij (k) + c1r1(p ij -x ij (k)) + c2r2(p gj -x ij (k)

[0121] Wherein, v ij (k) represents the current speed of the i-th particle in the j-th dimension, x ij (k) represents the current position of the i-th particle in the j-th dimension, p ij represents the historical optimal position of the i-th particle, p gj represents the historical optimal position of the group in the j-th dimension, c1 and c2 are learning factors, and r1 and r2 are random numbers in the interval [0, 1]. The position updating formula is as follows:

[0122] x ij (k+1) = x ij (k) = v ij (k+1)

[0123] The system recalculates the fitness value of the updated position after each iteration, and compares it with the historical optimal value of the individual and the group. If it is better than the historical value, replace the corresponding optimal solution. The system continues to perform the iteration process until the preset convergence condition is met, and the convergence condition includes that the rate of change of the fitness value is lower than the threshold, the maximum number of iterations is reached, or the power grid stability index enters the allowed range. When the termination condition is met, the system outputs the power distribution scheme corresponding to the current group optimal position as the charging power scheme, which includes the accurate output power instruction of each charging pile in each control period. The system sends the charging power scheme to the controller of the corresponding charging pile through the communication interface, and the controller adjusts the output state of the power module according to the instruction to execute the scheduling plan.

[0124] In one embodiment, the new energy charging pile dynamic scheduling method provided by the application comprises the following steps:

[0125] S51: Perform deviation calculation and processing on the real-time load data fed back by the charging pile controller, construct the actual load deviation rate index based on the grouped demand prediction value of the power prediction result, and generate dynamic deviation data.

[0126] Specifically, the system receives real-time load data from the charging pile controller, which contains the actual output power values of each charging pile in consecutive time units. The system groups these actual output powers by behavior and collects them, and aligns them with the grouped demand prediction values in the generated power prediction results, forming a comparison sequence of predicted and actual values in the corresponding time period. The system calculates the deviation between the actual load and the prediction value in each time unit, and calculates the relative deviation rate based on the prediction value. The calculation formula of the deviation rate is as follows:

[0127]

[0128] Where δ(t) represents the actual load deviation rate at time point t, P actual (t) represents the actual load power at that time point, P pred (t) represents the grouped demand prediction value at the corresponding time point. The system normalizes the deviation rate to eliminate the magnitude difference caused by the size of the group, and generates a dynamic deviation data sequence that records the output error change of the prediction model in different time periods.

[0129] S52: Continuously analyze the dynamic deviation data, detect the abnormal state where the deviation rate exceeds the preset deviation threshold and lasts for a set time window, and generate a model update trigger instruction.

[0130] Specifically, the system continuously analyzes the dynamic deviation data sequence to detect whether there is an abnormal state that deviates from the normal range. The system sets a deviation threshold to determine whether the deviation at a single time point exceeds the acceptable range. The threshold is determined by the system configuration parameters and reflects the tolerance limit of the prediction accuracy. Preferably, the system can use a sliding time window to scan the continuous deviation rate data. The length of the time window is determined by the system running period and the data sampling frequency, and is used to judge the persistence of the deviation. When the length of time that the deviation rate continuously exceeds the preset deviation threshold reaches the set time window, the system determines that there is a significant model deviation, and generates a model update trigger instruction. The instruction contains the trigger time, the behavior group label involved, the deviation duration and the corresponding input feature vector, which are used to locate the part of the model that needs to be updated.

[0131] S53: Parameter adjustment processing is performed on the model update trigger instruction, and the prediction error of the power prediction model is calculated based on the gradient descent algorithm to adjust the weight parameters of the power prediction model.

[0132] Specifically, the system receives the model update trigger instruction, extracts the input feature data and actual load value in the corresponding time period, and constructs a new training sample set, which is used to correct the parameters of the power prediction model. Preferably, the system calls the preset power prediction model, inputs the input feature data into the model, obtains the new prediction output, and calculates the prediction error with the actual load value. The error function adopts the form of mean square error:

[0133]

[0134] wherein E represents the prediction error, P model (t) represents the output value of the model at time point tt, and M represents the sample number. The system calculates the partial derivative of the error with respect to the model weight parameter based on the gradient descent algorithm, and adjusts the weight according to the following update rule:

[0135]

[0136] wherein represents the i-th layer j-th weight parameter at the k-th iteration, and η is the learning rate, which is determined by the system configuration parameter. After the system completes the weight parameter update, the new parameter is written into the model storage area to replace the original parameter, realizing the online learning and adaptive adjustment of the model. After the adjustment is completed, the system stores the updated weight parameter set in the model parameter library and marks it as the "effective version". The subsequent power prediction process automatically calls the version parameter to realize the dynamic optimization of the power prediction model precision, and ensure that the model can adapt to the change trend of the actual running data.

[0137] It should be understood that although each step in the flowchart involved in each embodiment as described above is displayed in sequence according to the arrow, these steps are not necessarily executed in sequence according to the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in each embodiment as described above can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or steps or stages in other steps.

[0138] Based on the same inventive concept, the application also provides a new energy charging pile dynamic scheduling system for implementing the new energy charging pile dynamic scheduling method described above. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme described in the above method, so the specific limitations in one or more new energy charging pile dynamic scheduling system embodiments provided below can refer to the limitations of the new energy charging pile dynamic scheduling method described above, which will not be repeated here.

[0139] Preferably, as Figure 3 indicated, the application provides a new energy charging pile dynamic scheduling system 600, which is configured with the following modules:

[0140] The charging behavior grouping module 610 is used for processing the collected power grid real-time load data, user charging request information, battery health state data and historical charging data, classifying and aggregating the charging behavior characteristics according to the charging period characteristics, power preference characteristics and vehicle type characteristics, and generating behavior grouping labels;

[0141] The power demand prediction module 620 is used for processing the behavior grouping labels based on a preset power prediction model, combining the battery health state data and the historical charging data to predict the time series of the charging demand and calculate the power safety boundary, and generating a power prediction result containing the grouping demand prediction value and the charging pile available power range;

[0142] The load conflict scheduling module 630 is used for processing the power prediction result, detecting the power grid load conflict and optimizing the power distribution of the charging pile, and generating a preliminary scheduling scheme containing conflict point identification;

[0143] The power optimization adjustment module 640 is used for processing the preliminary scheduling scheme, performing multi-objective power optimization adjustment on the charging pile associated with the conflict point identification, generating a charging power scheme meeting the power grid stability index and sending it to the charging pile controller, and the charging power scheme is used to indicate the power output value;

[0144] The model parameter updating module 650 is used for processing the actual load data fed back by the charging pile controller, calculating the load deviation and judging whether it exceeds the threshold, and updating the parameter weight of the power prediction model.

[0145] Preferably, the charging behavior grouping module 610 provided by the application is configured with the following units:

[0146] The power grid load filtering unit is used for noise filtering processing of the collected power grid real-time load data, using a sliding window filtering algorithm to eliminate transient interference signals in the power grid load fluctuation, and generating a standardized power grid load sequence;

[0147] The charging request analysis unit is configured to perform feature analysis processing on the collected user charging request information, extract charging time period distribution features, power demand level features and vehicle type code features, and generate a user behavior feature vector;

[0148] The charging behavior clustering unit is configured to perform correlation analysis processing on the collected battery health state data and historical charging data, perform charging behavior similarity clustering in combination with a power grid load sequence and a user behavior feature vector, calculate Euclidean distances between feature vectors and set a similarity threshold, merge adjacent core areas with distances less than the threshold to form a charging behavior cluster, and generate a behavior grouping label, which is used to identify a vehicle group with similar charging behaviors.

[0149] Preferably, the power demand prediction module 620 provided by the application is configured with the following units:

[0150] The grouping demand prediction unit is configured to perform time series analysis processing on the behavior grouping label and the historical charging data, calculate charging demand trends of each behavior group based on a preset power prediction model, and generate a grouping demand prediction value;

[0151] The dynamic power boundary unit is configured to perform health constraint mapping processing on the battery health state data, calculate a maximum allowed charging power according to a battery health degree decay curve and a temperature change relationship, and generate a dynamic power safety boundary;

[0152] The power prediction optimization unit is configured to perform collaborative optimization processing on the grouping demand prediction value and the dynamic power safety boundary, adjust an excess prediction value based on a preset power grid capacity threshold and a transformer load rate, and generate a power prediction result containing the grouping demand prediction value and a charging pile available power range.

[0153] Preferably, the load conflict scheduling module 630 provided by the application is configured with the following units:

[0154] The conflict point detection unit is configured to perform conflict detection processing on the grouping demand prediction value in the power prediction result, analyze whether a power grid load peak value exceeds a preset safety threshold, whether a battery temperature change rate exceeds a preset temperature rise limit value, and whether a high-priority vehicle concentration exceeds a set proportion threshold, mark a charging pile that meets any condition as a conflict node, and generate a conflict point identification set;

[0155] The power allocation optimization unit is configured to perform power allocation optimization processing on the charging piles associated with the conflict point identification set, assign a dynamic weight coefficient based on vehicle service type data in the user charging request information, and construct an optimization objective function of minimizing weighted waiting time under the constraint of the charging pile available power range in the power prediction result.

[0156] The scheduling scheme solving unit is configured to perform linear programming solving processing on the optimization objective function, iteratively calculate an optimal solution satisfying the power grid load stability condition by using a simplex method, and generate a preliminary scheduling scheme containing conflict point identifiers, the preliminary scheduling scheme being used to indicate initial power allocation values of each charging pile.

[0157] Preferably, the power optimization adjustment module 640 provided by the application is configured with the following units:

[0158] The conflict parameter initialization unit is configured to initialize parameters of the conflict point identifiers in the preliminary scheduling scheme, construct a particle swarm position vector based on power set values of associated charging piles, and generate an optimization parameter matrix containing position and velocity parameters;

[0159] The fitness evaluation unit is configured to evaluate the fitness of the optimization parameter matrix, calculate a weighted comprehensive index of the power grid load variance and the battery temperature rise rate, and generate an evolution direction vector.

[0160] The iterative optimization scheme unit is configured to perform iterative optimization processing on the evolution direction vector, update the particle position and velocity parameters by looping, compare the current fitness with the historical optimal value, continuously adjust the particle swarm state until the power grid stability index is satisfied, generate a charging power scheme, and send the charging power scheme to the charging pile controller.

[0161] Preferably, the model parameter updating module 650 provided by the application is configured with the following units:

[0162] The load deviation calculation unit is configured to perform deviation calculation processing on real-time load data fed back by the charging pile controller, construct an actual load deviation rate index based on grouped demand prediction values of the power prediction result, and generate dynamic deviation data.

[0163] The deviation anomaly detection unit is configured to perform continuous analysis processing on the dynamic deviation data, detect an abnormal state in which the deviation rate exceeds a preset deviation threshold and lasts for a set time window, and generate a model update triggering instruction.

[0164] The model weight adjustment unit is configured to perform parameter adjustment processing on the model update triggering instruction, calculate prediction errors of the power prediction model based on a gradient descent algorithm, and adjust weight parameters of the power prediction model.

[0165] In one embodiment, the application further provides a computer device, which comprises a memory and a processor, the memory stores a computer program, and the processor implements the new energy charging pile dynamic scheduling method described above when executing the computer program.

[0166] In one embodiment, the application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the new energy charging pile dynamic scheduling method described above.

[0167] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the present specification and the features of the different embodiments or examples without contradiction.

[0168] For the device embodiment, since it basically corresponds to the method embodiment, the relevant part is described in the part of the method embodiment. The above described device embodiment is only schematic, wherein the components illustrated as separate components can or can not be physically separate, and the components illustrated as a unit can or can not be physical units, i.e. can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purpose of the present disclosure according to the actual needs. Those skilled in the art can understand and implement it without creative labor.

[0169] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of various changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A new energy charging pile dynamic scheduling method, characterized in that: The following steps are involved: S1: Processes the collected real-time grid load data, user charging request information, battery health status data, and historical charging data, classifies and aggregates charging behavior characteristics based on charging period characteristics, power preference characteristics, and vehicle type characteristics, and generates behavior group labels; S2: Processing the behavior group labels based on a preset power prediction model, combining the battery health status data and the historical charging data to perform a time series prediction of charging demand and calculate the power safety margin, generating a power prediction result containing the group demand prediction value and the available power range of the charging pile; S3: Processing the power prediction results, detecting grid load conflicts and optimizing the power distribution of charging piles, and generating a preliminary scheduling plan including conflict point identifiers; S4: Processing the preliminary scheduling plan, performing multi-objective power optimization and adjustment on the charging piles associated with the conflict point identifiers, generating a charging power plan that meets the grid stability index, and sending the plan to the charging pile controller. The charging power plan is used to indicate the power output value; S5: Processing the actual load data fed back by the charging pile controller, calculating the load deviation and determining whether it exceeds a threshold, and updating the parameter weights of the power prediction model.

2. The method according to claim 1, characterized in that Said S1 comprises: S11: performing noise filtering on the collected real-time grid load data, using a sliding window filtering algorithm to eliminate transient interference signals in grid load fluctuations, and generating a standardized grid load sequence; S12: Perform feature analysis on the collected user charging request information to extract charging period distribution features, power demand level features, and vehicle type coding features to generate a user behavior feature vector; S13: Perform correlation analysis on the collected battery health status data and historical charging data, perform charging behavior similarity clustering based on the grid load sequence and the user behavior feature vector, calculate the Euclidean distance between feature vectors and set a similarity threshold, merge adjacent core areas with a distance less than the threshold to form a charging behavior cluster, and generate a behavior grouping label, which is used to identify groups of vehicles with similar charging behaviors.

3. The method according to claim 1, characterized in that The S2 includes: S21: performing time series analysis on the behavior group labels and the historical charging data, calculating the charging demand change trend of each behavior group based on a preset power prediction model, and generating a group demand prediction value; S22: Perform health constraint mapping on the battery health status data, calculate the maximum allowable charging power based on the battery health decay curve and temperature change relationship, and generate a dynamic power safety margin; S23: Coordinated optimization processing is performed on the group demand forecast value and the dynamic power safety boundary, and the excess forecast value is adjusted based on a preset grid capacity threshold and transformer load rate to generate a power forecast result containing the group demand forecast value and the available power range of the charging pile.

4. The method according to claim 1, wherein The S3 includes: S31: Perform conflict detection on the group demand forecast values ​​in the power forecast results to analyze whether the grid load peak exceeds a preset safety threshold, whether the battery temperature change rate exceeds a preset temperature rise limit, and whether the concentration of high-priority vehicles exceeds a set ratio threshold. Charging piles that meet any of the conditions are marked as conflict nodes, and a conflict point identification set is generated. S32: Performing power allocation optimization processing on the charging piles associated with the conflict point identifier set, allocating dynamic weight coefficients based on the vehicle service type data in the user charging request information, and constructing an optimization objective function for minimizing weighted waiting time under the constraints of the available power range of the charging piles in the power prediction result; S33: Performing linear programming solution processing on the optimization objective function, using the simplex method to iteratively calculate the optimal solution that meets the grid load stability condition, and generating a preliminary scheduling plan including conflict point identification, wherein the preliminary scheduling plan is used to indicate the initial power allocation value of each charging pile.

5. The method according to claim 1, wherein The S4 includes: S41: Initializing parameters of the conflict point identifiers in the preliminary scheduling plan, constructing a particle swarm position vector based on the power setting values ​​of the associated charging piles, and generating an optimization parameter matrix including position and speed parameters; S42: performing fitness evaluation on the optimization parameter matrix, calculating a weighted comprehensive index of grid load variance and battery temperature rise rate, and generating an evolution direction vector; S43: Iteratively optimize the evolution direction vector, cyclically update the particle position and velocity parameters, compare the current fitness with the historical optimal value, continuously adjust the particle swarm state until the grid stability index is met, generate a charging power plan and send it to the charging pile controller.

6. The method according to claim 5, characterized in that The calculation formula of the weighted comprehensive index is: Among them, F is the weighted comprehensive index, is the grid load variance, is the battery temperature rise rate, ΔT is the temperature change, Δt is the time change, and α and β are weight coefficients.

7. The method according to any one of claims 1 to 6, characterized in that The S5 includes: S51: performing deviation calculation on the real-time load data fed back by the charging pile controller, constructing an actual load deviation rate indicator based on the group demand prediction value of the power prediction result, and generating dynamic deviation data; S52: Continuously analyzing and processing the dynamic deviation data to detect an abnormal state in which the deviation rate exceeds a preset deviation threshold and persists for a set time window, and generating a model update trigger instruction; S53: performing parameter adjustment processing on the model update trigger instruction, calculating the prediction error of the power prediction model based on the gradient descent algorithm, and adjusting the weight parameters of the power prediction model.

8. A new energy charging pile dynamic scheduling system, characterized in that: The system comprises: The charging behavior grouping module is used to process the collected real-time grid load data, user charging request information, battery health status data, and historical charging data. It classifies and aggregates charging behavior characteristics based on charging period characteristics, power preference characteristics, and vehicle type characteristics to generate behavior grouping labels; A power demand prediction module is configured to process the behavior group labels based on a preset power prediction model, perform time series prediction of charging demand based on the battery health status data and the historical charging data, calculate the power safety margin, and generate a power prediction result containing the group demand prediction value and the available power range of the charging pile; A load conflict scheduling module is used to process the power prediction results, detect grid load conflicts, optimize the power distribution of charging piles, and generate a preliminary scheduling plan including conflict point identification; a power optimization and regulation module, configured to process the preliminary scheduling plan, perform multi-objective power optimization and regulation on the charging piles associated with the conflict point identifiers, generate a charging power plan that meets the grid stability index, and send the plan to the charging pile controller, wherein the charging power plan indicates the power output value; The model parameter updating module is used to process the actual load data fed back by the charging pile controller, calculate the load deviation and determine whether it exceeds the threshold, and update the parameter weights of the power prediction model.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images