Multi-port hybrid charging pile power pool reinforcement learning scheduling method and system

Abstract

The application discloses a multi-port hybrid charging pile power pool reinforcement learning scheduling method and system, and relates to the technical field of computer data processing. The multi-port hybrid charging pile power pool reinforcement learning scheduling method and system comprises the following steps: S1, periodically collecting port scheduling data and performing preprocessing on the port scheduling data; S2, constructing and training a vehicle target SOC prediction model, outputting a vehicle target SOC, evaluating the charging urgency of each connected vehicle, and dividing a scheduling priority level; S3, constructing a port power distribution model, outputting a power distribution value of each charging port, and delivering the power distribution value to a power control interface for execution; and S4, evaluating a power scheduling execution deviation, and entering a protection mode when an evaluation anomaly occurs. The application solves the problem that fixed proportional power distribution and alternating current and direct current port topology isolation result in the inability to flexibly schedule the total power of charging piles, and cause high-power resources to be occupied by high-battery vehicles and charging of low-battery vehicles to be limited.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer data processing, in particular to a multi-port hybrid charging pile power pool reinforcement learning scheduling method and system. BACKGROUND

[0002] With the rapid popularization of new energy vehicles, charging piles as the core facilities of electric vehicle energy supply, their operation efficiency and power distribution strategy directly affect the charging experience of users and the utilization rate of power grid resources. In the prior art, various methods have been proposed for the operation monitoring and power distribution of charging piles, but there are still problems of insufficient power utilization, rigid scheduling and lack of adaptive optimization in complex scenarios.

[0003] For example, the invention with publication number CN118885703B proposes a running state detection and analysis method suitable for charging pile high-power devices, belonging to the technical field of charging piles, including the following steps: step S1: obtaining detection data of charging piles in different scenarios based on charging pile detection equipment, establishing a data model library; step S2: classifying the detection data according to the charging gun power and type, preprocessing the classified detection data to handle missing values and noise data; step S3: embedding the preprocessed detection data into the data model, using the least squares method fitting in the high-order polynomial fitting algorithm to process the preprocessed detection data, and finally calculating the optimal fitting curve; step S4: deploying the optimal fitting curve, evaluating the fitting result, obtaining the current detection data of the charging pile, comparing and analyzing the fitting result of the optimal fitting curve with the current detection data, and judging the existing faults and possible faults of the charging pile.

[0004] For example, the invention with publication number CN119416651B provides a multi-gun head charging pile power distribution method, medium and system, belonging to the technical field of charging piles, including: first, obtaining charging pile historical operation data, establishing a mathematical model containing multiple error terms, such as charging power, time, battery capacity, etc. Then build a hybrid model, including a multi-branch neural network and a mathematical model, to adapt to charging piles of different sizes. The power distribution model is obtained by training with historical data. In the actual charging process, input real-time data and adjust power distribution, while collecting feedback data. Based on real-time data and prediction error, the lightweight Lora model is used to continuously fine-tune the fusion sub-network, and the main model is updated regularly.

[0005] However, the above technical scheme has the following shortcomings: first, the power distribution strategy is still biased towards static modeling, lacking dynamic perception of vehicle battery state, urgency demand and environmental conditions; second, the scheduling algorithm lacks flexibility in the multi-port hybrid scene, making it difficult to achieve optimal utilization of the power pool as a whole; third, there is a lack of feedback loop at the execution level, resulting in the model being unable to learn and adaptively adjust according to abnormal execution conditions.

[0006] Therefore, in view of the above problems, there is an urgent need for a multi-port hybrid charging pile power pool reinforcement learning scheduling method and system. SUMMARY

[0007] Technical problems solved

[0008] In view of the shortcomings of the prior art, the present application provides a multi-port hybrid charging pile power pool reinforcement learning scheduling method and system, which solves the problem of fixed proportion power distribution and AC / DC port topology isolation leading to the inability to flexibly schedule the total power of the charging pile, resulting in high power resource occupation by high power vehicles and charging limitation of low power vehicles.

[0009] Technical scheme

[0010] To achieve the above purpose, the present application realizes the following technical scheme: a multi-port hybrid charging pile power pool reinforcement learning scheduling method and system, comprising: S1, periodically collecting port scheduling data, and performing synchronous alignment, abnormality elimination, dimension elimination and normalization processing on the port scheduling data to obtain preprocessed port scheduling data; S2, based on the preprocessed port scheduling data, constructing and training a vehicle target SOC prediction model, inputting real-time port scheduling data into the prediction model, outputting vehicle target SOC, and evaluating the charging urgency of each connected vehicle to divide the scheduling priority level; S3, calling the port scheduling data and sorting results of the connected vehicles in the instant scheduling queue, constructing a port power distribution model, outputting the power distribution value of each charging port, and issuing it to the power control interface to execute real-time power scheduling; S4, after power distribution execution, evaluating the deviation in the power scheduling execution process, when the evaluation result is normal, continuing to execute charging, when the evaluation result is abnormal, entering protection mode, and retriggering the target SOC prediction and power distribution process, realizing closed-loop optimization of power scheduling.

[0011] Further, the port scheduling data is periodically collected, and the port scheduling data is subjected to synchronous alignment, abnormality elimination, dimension elimination and normalization processing to obtain preprocessed port scheduling data. The specific steps are as follows: taking a fixed-width sliding time window as a sampling period, periodically collecting the port scheduling data, the port scheduling data including vehicle battery SOC, vehicle battery voltage, vehicle battery temperature, grid voltage, charging current, charging pile output power and ambient temperature; through constructing a synchronous sampling scheduling mechanism based on multi-source data access priority, the asynchronous data in the port scheduling data is subjected to timing alignment; through adopting an exponential moving average filtering algorithm, the port scheduling data is subjected to noise suppression and signal smoothing; through a univariate outlier detection algorithm of Z-Score method, the port scheduling data is subjected to abnormal point elimination; through a minimum-maximum normalization algorithm based on quantile interval scaling, the port scheduling data is subjected to cross-dimension dimension elimination and data distribution unification.

[0012] Further, based on the preprocessed port scheduling data, a vehicle target SOC prediction model is constructed and trained, and the specific steps of inputting real-time port scheduling data into the prediction model to output the vehicle target SOC are as follows: extracting the preprocessed historical port scheduling data, adopting a supervised learning method, and constructing a vehicle target SOC prediction model based on a multi-layer feedforward neural network; taking the port scheduling data as input features and the vehicle target SOC when the vehicle actually completes the charging task as a supervision label to constitute a prediction training sample set; and taking the minimum mean square error as a loss function, the vehicle target SOC prediction model is trained; the real-time port scheduling data is input into the trained vehicle target SOC prediction model to output the vehicle target SOC.

[0013] Further, the specific steps of evaluating the charging urgency of each access vehicle are as follows: taking the square of the difference between the vehicle target SOC and the vehicle battery SOC, dividing by the vehicle battery temperature plus one to obtain an electric quantity difference temperature control factor; dividing the rated output current by the sum of the charging current and a minimum term, and taking the ratio to the power of an electric current suppression index to obtain a unit current utilization term; taking the ratio of the grid voltage to the vehicle battery voltage plus one and taking the natural logarithm to obtain a voltage matching degree term; multiplying the electric quantity difference temperature control factor, the unit current utilization term and the voltage matching degree term in turn to obtain a scheduling urgency evaluation value.

[0014] Further, the specific steps of dividing the scheduling priority level are as follows: real-time comparison of the scheduling urgency evaluation value D and the multi-level scheduling threshold D1 and D2, and the access vehicle is divided into three priority queues: when D≤D1, the vehicle is assigned to a scheduling waiting queue, and the periodic power allocation is suspended; when D1 D2, the vehicle is assigned to a normal scheduling queue and enters a normal power scheduling process; when D≥D2, the vehicle is assigned to an immediate scheduling queue and is sorted from large to small according to the scheduling urgency evaluation value.

[0015] Further, the specific steps of calling the port scheduling data and the sorting result of the access vehicle in the instant scheduling queue, constructing the port power allocation model, outputting the power allocation value of each charging port, and issuing to the power control interface to execute real-time power scheduling are as follows: calling the port scheduling data, the scheduling urgency evaluation value and the sorting result of the access vehicle in the instant scheduling queue, selecting the deep deterministic policy gradient algorithm as the training framework, and constructing the port power allocation model based on the policy network and the value network group; using the historical port scheduling data and the actual power allocation record to construct the allocation training sample set, constructing a composite reward function that fuses the scheduling deviation, the temperature control pressure and the voltage deviation in the training process, and obtaining the function output value; taking the function output value as the negative reward cumulative value, driving the port power allocation model to execute parameter optimization and policy convergence through the policy gradient update mechanism; inputting the port scheduling data and the sorting result of the access vehicle in the instant scheduling queue into the trained port power allocation model, and outputting the power allocation value corresponding to each charging port; and taking the power allocation value as the execution instruction, issuing to each corresponding charging port through the power control interface for execution.

[0016] Further, the specific steps of constructing the composite reward function that fuses the scheduling deviation, the temperature control pressure and the voltage deviation, and obtaining the function output value are as follows: multiplying the square of the last period power allocation value by the charging current as the energy load item; subtracting the charging current from the port rated current, adding a small item, and then multiplying by the scheduling urgency evaluation value to obtain the normalized scheduling reference item; dividing the energy load item by the normalized scheduling reference item to obtain the scheduling deviation penalty item; dividing the absolute value of the difference between the vehicle battery temperature and the ambient temperature by the sum of the vehicle battery temperature and the small item to obtain the temperature control pressure item; dividing the absolute value of the difference between the grid voltage and the vehicle battery voltage by the sum of the grid voltage, the vehicle battery voltage and the small item to obtain the voltage deviation item; adding the scheduling deviation penalty item, the temperature control pressure item and the voltage deviation item in turn to obtain the function output value.

[0017] Further, the specific steps of evaluating the deviation during power scheduling execution after power allocation are as follows: after power allocation, the real-time charging pile output power and the corresponding port scheduling data are extracted, the absolute value of the difference between the grid voltage and the vehicle battery voltage is multiplied by the charging current to obtain the voltage deviation current response term; the voltage deviation current response term is divided by the sum of the vehicle battery temperature and the minimum term to obtain the temperature control normalization term; the power allocation value is divided by the sum of the difference between the port rated current and the charging current and the minimum term to obtain the current load amplification term; the temperature control normalization term and the current load amplification term are multiplied to obtain the scheduling load response term; the absolute value of the difference between the current period vehicle battery SOC and the last period vehicle battery SOC is divided by the sum of the sampling time interval and the minimum term, and squared to obtain the SOC change rate term; the difference between the charging pile output power and the power allocation value is divided by the sum of the power allocation value and the minimum term, and squared to obtain the power execution deviation term; the scheduling load response term, the SOC change rate term and the power execution deviation term are added to obtain the scheduling abnormality evaluation value.

[0018] Further, when the evaluation result is normal, the charging is continued, when the evaluation result is abnormal, the protection mode is entered, and the target SOC prediction and power allocation process are retriggered to realize the specific steps of closed-loop optimization of power scheduling: the scheduling abnormality evaluation value is compared with the abnormal threshold in real time, when the scheduling abnormality evaluation value is less than or equal to the abnormal threshold, it is determined that the current scheduling behavior is normal, and the charging task is continued; when the scheduling abnormality evaluation value is greater than the abnormal threshold, it is determined that the current scheduling behavior is abnormal, the current charging port power output is suspended, the protection mode is entered, and the real-time port scheduling data is extracted to re-predict the vehicle target SOC and recalculate the power allocation value; in the protection mode, the battery voltage change rate and the vehicle battery temperature are monitored in real time, when the battery voltage change rate is greater than the voltage change threshold, the V2G limited power mechanism is started; when the vehicle battery temperature is less than the battery temperature threshold, the conventional fast charging mode is switched to the pulse heating mode, the battery internal resistance heat is used to quickly warm up, and when the vehicle battery temperature is greater than the battery temperature threshold, the conventional fast charging mode is switched back; at the same time, the port scheduling data and the sorting result of the current period are added to the distribution training sample set as new samples for offline iterative training of the subsequent port power allocation model, forming a self-learning closed-loop mechanism based on scheduling performance feedback.

[0019] The second aspect of the application provides a multi-port hybrid charging pile power pool reinforcement learning scheduling system, comprising: a port data acquisition and preprocessing module, a vehicle demand priority evaluation module, a charging power intelligent distribution module, and a scheduling exception closed-loop optimization module, wherein: the port data acquisition and preprocessing module is used to periodically acquire port scheduling data, and perform synchronous alignment, exception elimination, dimension elimination and normalization processing on the port scheduling data to obtain preprocessed port scheduling data; the vehicle demand priority evaluation module is used to construct and train a vehicle target SOC prediction model based on the preprocessed port scheduling data, input real-time port scheduling data into the prediction model, output vehicle target SOC, and evaluate the charging urgency of each connected vehicle to divide the scheduling priority level; the charging power intelligent distribution module is used to call the port scheduling data and sorting results of the connected vehicles in the instant scheduling queue, construct a port power distribution model, output the power distribution value of each charging port, and issue it to the power control interface to perform real-time power scheduling; the scheduling exception closed-loop optimization module is used to evaluate the deviation during power scheduling execution after power distribution execution, continue charging when the evaluation result is normal, and enter protection mode when the evaluation result is abnormal, and retrigger the target SOC prediction and power distribution process to realize closed-loop optimization of power scheduling.

[0020] Advantages

[0021] The application has the following advantages:

[0022] (1) The multi-port hybrid charging pile power pool reinforcement learning scheduling method and system constructs a scheduling urgency evaluation value, takes multi-dimensional data such as vehicle target SOC, vehicle battery SOC, vehicle battery temperature, current, and voltage as input, calculates the charging urgency of each vehicle in real time, and divides the scheduling priority level accordingly, so that the power distribution is no longer dependent on a fixed proportion or static rules, but is dynamically adjusted according to the vehicle state, environmental conditions, and power grid situation, realizing differentiated distribution and effectively avoiding the problem of long-term occupation of high-power resources by high-quantity vehicles.

[0023] (2) The multi-port hybrid charging pile power pool reinforcement learning scheduling method and system constructs a port power distribution model of power pooling, unifies alternating current ports and direct current ports into power pool management, and performs global optimization combined with reinforcement learning strategy, so that each port as a dynamic unit in the power pool can realize power cross-port scheduling under the premise of meeting the constraint conditions, thereby breaking through the limitation caused by topology isolation, making the total power resource flow reasonably between ports, and significantly improving the overall power utilization rate of the system.

[0024] (3) The multi-port hybrid charging pile power pool reinforcement learning scheduling method and system, through the emergency degree sorting queue mechanism, makes the low SOC or long time access vehicle automatically enter the instant scheduling queue, preferentially obtains the power distribution, and combines the real-time optimization ability of the power distribution model, preferentially guarantees the vehicle in urgent need of power supplement when the total power is limited, thereby minimizing the user waiting time, relieving the charging limitation of the low power vehicle, and significantly improving the charging efficiency and user experience.

[0025] (4) The multi-port hybrid charging pile power pool reinforcement learning scheduling method and system, by constructing the scheduling abnormality evaluation value, fuses the voltage deviation current response term, the SOC change rate term, the power execution deviation term and other indicators into a multi-dimensional detection system, and quantifies the deviation in the scheduling execution process in real time; when the abnormality evaluation value exceeds the threshold value, the system automatically triggers the protection mechanism, executes the power limiting or suspends the output, and re-performs the power distribution calculation, thereby effectively preventing the risks of overcurrent, overheating and voltage mismatch, and guaranteeing the operation safety of the charging pile and the vehicle battery. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 The multi-port hybrid charging pile power pool reinforcement learning scheduling method flow chart;

[0027] Figure 2 The multi-port hybrid charging pile power pool reinforcement learning scheduling system structure diagram;

[0028] Figure 3 The charging queue determination schematic diagram based on the scheduling emergency degree evaluation value;

[0029] Figure 4 The fault response and temperature control recovery flow chart in the protection mode. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0031] Please refer to Figures 1-4The embodiment of the application provides a technical scheme: a multi-port hybrid charging pile power pool reinforcement learning scheduling method and system, comprising: S1, periodically collecting port scheduling data, and performing synchronous alignment, abnormality elimination, dimension elimination and normalization processing on the port scheduling data to obtain preprocessed port scheduling data; S2, based on the preprocessed port scheduling data, a vehicle target SOC prediction model is constructed and trained, real-time port scheduling data is input into the prediction model, the vehicle target SOC is output, the charging urgency of each access vehicle is evaluated, and a scheduling priority level is divided; S3, the port scheduling data and the sorting result of the access vehicle in the instant scheduling queue are called, a port power allocation model is constructed, the power allocation value of each charging port is output, and is sent to a power control interface to execute real-time power scheduling; S4, after power allocation execution, the deviation in the power scheduling execution process is evaluated, when the evaluation result is normal, charging is continued, when the evaluation result is abnormal, a protection mode is entered, and the target SOC prediction and power allocation process are triggered again, and closed-loop optimization of power scheduling is realized.

[0032] Specifically, the specific steps of periodically collecting port scheduling data and performing synchronous alignment, abnormality elimination, dimension elimination and normalization processing on the port scheduling data to obtain preprocessed port scheduling data are as follows: a fixed-width sliding time window is set as a sampling period, the port scheduling data is periodically collected according to a set sampling frequency and a time stamp, the port scheduling data includes vehicle battery SOC, vehicle battery voltage, vehicle battery temperature, grid voltage, charging current, charging pile output power and environmental temperature, wherein each item of port scheduling data is bound to a unique time stamp and retains original accuracy; by constructing a synchronous sampling scheduling mechanism based on multi-source data access priority, interpolation prediction and time sequence alignment are performed on asynchronous data according to the refresh rate and time lag characteristics of each port scheduling data, so that frame-level synchronization of each port scheduling data in the sampling period is realized; by using an exponential moving average filtering algorithm, historical trends and current values are fused, noise suppression and signal smoothing processing are performed on the port scheduling data sequence, and the stability and continuity of the original port scheduling data are improved; further, by using a univariate outlier detection algorithm of the Z-Score method, extreme data points deviating from the standard deviation by more than a set threshold in the port scheduling data are marked and eliminated, so as to reduce the influence of abnormal values on the subsequent modeling performance; finally, by using a quantile interval scaling combined with a min-max normalization algorithm, the port scheduling data in different dimensions are normalized and mapped and the scales are unified, so as to realize the dimensionless expression of each type of port scheduling data in the feature space, and provide high-quality standardized input for subsequent prediction models and decision modules.

[0033] In this embodiment, by constructing a synchronous sampling scheduling mechanism based on multi-source data access priority, and combining an exponential moving average filtering algorithm, a Z-Score univariate outlier detection algorithm, and a quantile interval scaling and min-max normalization algorithm, high-precision preprocessing of port scheduling data is realized, the time sequence consistency and data distribution uniformity of asynchronous scheduling data are significantly improved, the input data quality and modeling robustness of the subsequent vehicle target SOC prediction model and power allocation model are enhanced, thereby improving the real-time and reliability of the scheduling decision process.

[0034] Specifically, based on the preprocessed port scheduling data, a vehicle target SOC prediction model is constructed and trained, and the specific steps of inputting real-time port scheduling data into the prediction model and outputting the vehicle target SOC are as follows: extracting historical port scheduling data after synchronization alignment, anomaly elimination, dimension elimination and normalization processing, using a supervised learning method, constructing a vehicle target SOC prediction model based on a multi-layer feedforward neural network structure, the feedforward neural network structure consists of an input layer, multiple hidden layers and an output layer; wherein the input layer is used to receive the port scheduling data as the input features of the model; the hidden layer uses a ReLU activation function to realize nonlinear mapping, and the number of nodes in each hidden layer is dynamically adjusted according to the training error to improve the fitting ability of the model to the feature combination relationship; the output layer uses a linear activation function to output the target SOC prediction value of the corresponding access vehicle; the parameters in the port scheduling data are used as input features, and the vehicle target SOC collected when the vehicle actually completes the charging task is used as a supervision label to form a prediction training sample set with complete structure and sufficient sample distribution; and the least mean square error is used as a loss function, combined with a back propagation algorithm and a gradient descent optimization strategy, to train the vehicle target SOC prediction model for parameter iteration and error minimization. After the model training is completed, the real-time collected and preprocessed port scheduling data is input into the vehicle target SOC prediction model to output the vehicle target SOC prediction result of the corresponding access vehicle at the current time.

[0035] In this embodiment, by introducing preprocessed port scheduling data as input features, combining a multi-layer feedforward neural network structure to realize nonlinear feature fitting, using the least mean square error as a loss function, and cooperating with a back propagation algorithm and a gradient descent optimization strategy, the accuracy and stability of the vehicle target SOC prediction result are effectively improved; further, high-accuracy recognition of the charging demand of the access vehicle is realized, providing high-reliability data support for subsequent scheduling urgency evaluation and power allocation strategies, and enhancing the real-time perception ability and adaptive scheduling performance of the system in a multi-port hybrid charging scene.

[0036] Specifically, the specific steps of evaluating the charging urgency of each access vehicle are as follows: first, square the difference between the predicted vehicle target SOC and the real-time collected vehicle battery SOC to represent the deviation between the current vehicle desired charging level and the actual state of charge; then divide the squared difference value by the sum of the vehicle battery temperature and the constant one to introduce the adjustment effect of the vehicle battery thermal state on the charging demand, and obtain the charge difference temperature control factor. Secondly, divide the port rated output current by the sum of the charging current and the minimum term, which is a minimum positive real number not equal to zero, to avoid numerical anomalies caused by the denominator approaching zero; and take the power function value of the ratio with the current suppression index as the exponent to form the unit current utilization term, which quantifies the regulation ability of the current current use efficiency. The current suppression index is obtained by analyzing the deviation distribution of the charging current and the rated output current in the historical port scheduling data using a statistical fitting algorithm based on quantile interval, and the value range is (0-1]. Thirdly, take the natural logarithm of the ratio of the real-time grid voltage and the vehicle battery voltage plus 1 to obtain the voltage matching degree term, which represents the influence of the charging system voltage adaptability on the scheduling behavior. Finally, multiply the charge difference temperature control factor, the unit current utilization term and the voltage matching degree term in the order of sequence to fuse the common influence of multi-dimensional factors, and output the scheduling urgency evaluation value which comprehensively represents the charging urgency of the access vehicle.

[0037] The specific calculation formula of the scheduling urgency evaluation value is:

[0038] ;

[0039] In the formula, represents the scheduling urgency evaluation value, represents the vehicle battery SOC, represents the vehicle target SOC, represents the vehicle battery temperature, represents the charging current, represents the rated output current, represents the vehicle battery voltage, represents the grid voltage, represents the minimum term, represents the current suppression index.

[0040] In the embodiment, Table 1 is a dispatch urgency evaluation value data table, which lists the key characteristic variables and corresponding dispatch urgency evaluation values of five access vehicles in the current dispatch cycle. The key variables used include: vehicle battery SOC, vehicle target SOC, vehicle battery temperature, charging current, rated output current, vehicle battery voltage, and grid voltage. The current suppression index is uniformly set to 0.1, and the specific data is as follows: in vehicle 1, the vehicle battery SOC is 50, the vehicle target SOC is 80, the vehicle battery temperature is 28, the charging current is 40, the rated output current is 120, the vehicle battery voltage is 360, and the grid voltage is 400, and the corresponding dispatch urgency evaluation value is 25.88; in vehicle 2, the vehicle battery SOC is 45, the vehicle target SOC is 90, the vehicle battery temperature is 32, the charging current is 70, the rated output current is 150, the vehicle battery voltage is 380, and the grid voltage is 400, and the corresponding dispatch urgency evaluation value is 47.62; in vehicle 3, the vehicle battery SOC is 35, the vehicle target SOC is 85, the vehicle battery temperature is 25, the charging current is 30, the rated output current is 100, the vehicle battery voltage is 350, and the grid voltage is 400, and the corresponding evaluation value is 82.63; in vehicle 4, the vehicle battery SOC is 50, the vehicle target SOC is 95, the vehicle battery temperature is 35, the charging current is 80, the rated output current is 150, the vehicle battery voltage is 380, and the grid voltage is 400, and the corresponding dispatch urgency evaluation value is 43.07; in vehicle 5, the vehicle battery SOC is 40, the vehicle target SOC is 85, the vehicle battery temperature is 30, the charging current is 50, the rated output current is 120, the vehicle battery voltage is 365, and the grid voltage is 400, and the corresponding dispatch urgency evaluation value is 52.75.

[0041] Table 1 Dispatch urgency evaluation value data table

[0042]

[0043] As Figure 3 shown, the distribution of dispatch urgency evaluation values of five access vehicles in the same sampling cycle is shown. The horizontal axis represents the access vehicle number, and the vertical axis represents the dispatch urgency evaluation value D. In the figure, different colored broken line points identify the dispatch queue to which the vehicle belongs: green represents the dispatch waiting queue, blue represents the regular dispatch queue, and red represents the immediate dispatch queue. As can be seen from the figure, the dispatch urgency evaluation value of vehicle 1 is 25.88, which is lower than the first-level dispatch threshold, and belongs to the dispatch waiting queue; the dispatch urgency evaluation values of vehicle 2, vehicle 4, and vehicle 5 are 47.62, 43.07, and 52.75 respectively, which are between the first-level threshold and the second-level threshold, and belong to the regular dispatch queue; the dispatch urgency evaluation value of vehicle 3 is 82.63, which is higher than the second-level dispatch threshold, and is determined as the immediate dispatch queue, which should be given priority in power resource allocation.Figure 3 The effect of classifying the charging demand of the multiple vehicles by scheduling the emergency evaluation value is directly reflected.

[0044] In the embodiment, by constructing a scheduling emergency evaluation model fusing the target SOC of the vehicle, the SOC of the vehicle battery, the temperature of the vehicle battery, the charging current, the rated output current of the port, the current suppression index, the grid voltage and the voltage of the vehicle battery, the quantitative evaluation of the current charging demand state of the vehicle is realized. In the evaluation process, three types of characteristic variables, including the power difference temperature control factor, the unit current utilization item and the voltage matching degree item, are introduced to ensure the comprehensive coverage of the model in the energy state, thermal state and voltage adaptability dimensions. The method improves the discrimination accuracy of the actual charging demand of the vehicle, enhances the rationality of the scheduling priority division, provides more refined and dynamic scheduling basis for subsequent port power allocation, and significantly improves the overall operation efficiency and adaptability of resource allocation of the system.

[0045] Specifically, the specific steps of dividing the scheduling priority level are as follows: real-time comparison of the scheduling emergency evaluation value D and the multi-level scheduling threshold D1 and D2, and the access vehicle is divided into three different scheduling priority level queues according to the charging urgency level: when the scheduling emergency evaluation value D is less than or equal to the first-level scheduling threshold D1, it is determined that the current charging demand of the corresponding access vehicle does not have urgency, the vehicle is marked as a low priority level and is assigned to a scheduling waiting queue, and the power allocation instruction of the current sampling period is suspended; when the scheduling emergency evaluation value D is greater than the first-level scheduling threshold D1 and less than the second-level scheduling threshold D2, it is determined that the corresponding access vehicle has medium charging urgency, and it is divided into a regular scheduling queue and enters a normal power scheduling process to accept regular power allocation; wherein, the normal power scheduling process refers to queuing and buffering the port scheduling data of the access vehicle in the regular scheduling queue, and scheduling the vehicles one by one by using sequential polling, inputting the port scheduling data of the vehicle into the trained port power allocation model, calculating the power allocation value combined with the current system load state, and executing the charging task in batches after the immediate scheduling queue is scheduled, so as to ensure that the medium priority vehicle completes energy supply without interfering with high priority scheduling. When the scheduling emergency evaluation value D is greater than or equal to the second-level scheduling threshold D2, it is determined that the corresponding access vehicle has significant charging urgency, and is included in the immediate scheduling queue, and all access vehicles in the immediate scheduling queue are sorted in real time in the order from large to small according to the scheduling emergency evaluation value, so as to prioritize the charging resource allocation demand of high urgency vehicles and improve the responsiveness and accuracy of scheduling.

[0046] In this embodiment, by means of real-time comparison mechanism based on dispatching urgency evaluation value D and multi-level dispatching threshold D1 and D2, the access vehicles are divided into dispatching waiting queue, regular dispatching queue and immediate dispatching queue, ensuring that different charging urgency vehicles obtain differentiated resource dispatching strategy; wherein, low urgency vehicles are suspended for dispatching to avoid resource waste, medium urgency vehicles obtain orderly energy supply through sequential polling and system load coordination, and high urgency vehicles realize priority guarantee of charging resources through immediate response of reinforcement learning model, which significantly improves the accuracy, responsiveness of power dispatching and dynamic adaptability of overall system load.

[0047] Specifically, the port scheduling data and the sorting results of the access vehicles in the immediate dispatching queue are called to construct a port power allocation model, output the power allocation values of each charging port, and be issued to the power control interface to execute real-time power scheduling. The specific steps are as follows: the port scheduling data, dispatching urgency evaluation value and sorting results corresponding to the access vehicles in the immediate dispatching queue are called, the deep deterministic policy gradient algorithm is selected as the training framework, and the port power allocation model composed of policy network and value network is constructed, the policy network is used to generate power allocation strategy, and the value network is used to estimate the expected return of the current strategy; the historical port scheduling data and the corresponding actual power allocation records after synchronization alignment, abnormality elimination, dimension elimination and normalization processing are used to construct the allocation training sample set; during the training process, a composite reward function integrating scheduling bias, temperature control pressure and voltage bias is constructed, the composite reward function quantifies the influence of power allocation strategy on system operation state, generates function output value as negative reward cumulative value, and drives the port power allocation model to execute continuous optimization of port power allocation model parameters and gradual convergence of final strategy through policy gradient update mechanism; the port scheduling data and the sorting results of the access vehicles in the immediate dispatching queue in the current period are input into the trained port power allocation model, the corresponding power allocation values of each charging port are generated through forward propagation, the dynamic optimal allocation of resources is realized under the premise of meeting the safety, power grid capacity and vehicle charging capacity constraints; and the power allocation values are used as power control instructions, which are issued to each charging port through the corresponding power control interface of the charging controller to drive the charging pile to execute real-time power scheduling and complete the landing execution of the dispatching target in the current period.

[0048] In this embodiment, by calling the port scheduling data and sorting results of the access vehicles in the instant scheduling queue, a port power allocation model based on the deep deterministic policy gradient algorithm is constructed, and a composite reward function integrating the scheduling bias, temperature control pressure and voltage bias is introduced to realize efficient training and adaptive optimization of the port power allocation strategy. This method not only improves the response accuracy of the scheduling urgency evaluation value, but also dynamically outputs the power allocation value corresponding to each charging port according to the system load state, which can effectively guarantee the priority allocation of charging resources for high-priority vehicles, enhance the real-time and fine level of scheduling decision, and significantly improve the resource allocation efficiency and charging service quality of the multi-port hybrid charging pile under complex working conditions.

[0049] Specifically, the specific steps of obtaining the function output value of the composite reward function integrating the scheduling bias, temperature control pressure and voltage bias are as follows: first, square the power allocation value allocated to the charging port in the last period and multiply it by the real-time charging current to construct an energy load term reflecting the energy output load of the current power allocation, which is used to measure the system power consumption intensity; second, subtract the real-time charging current from the port rated current and add a minimum term to construct a current residual capacity factor, and multiply the current residual capacity factor by the scheduling urgency evaluation value at the current time to generate a normalized scheduling reference term under a unified standard to represent the reasonable power proportion of each access vehicle under the current system load condition; third, divide the energy load term by the normalized scheduling reference term to form a scheduling bias penalty term for measuring the allocation accuracy deviation degree; then, extract the absolute value of the difference between the vehicle battery temperature and the ambient temperature, and divide the absolute value by the sum of the vehicle battery temperature and a minimum term to obtain a temperature control pressure term for measuring the system thermal stress level; further, calculate the absolute value of the difference between the grid voltage and the vehicle battery voltage, and divide it by the sum of the grid voltage, the vehicle battery voltage and a minimum term to form a voltage bias term reflecting the voltage adaptation state; finally, add the scheduling bias penalty term, the temperature control pressure term and the voltage bias term to obtain the function output value as the comprehensive feedback index of the composite reward function in the port power allocation model, which is used to drive the gradient update and policy iteration in the strategy optimization process.

[0050] wherein the specific calculation formula of the function output value is:

[0051] ;

[0052] In the formula, represents the function output value, represents the last period power allocation value, represents the scheduling urgency evaluation value, represents the charging current, represents the port rated current, represents the vehicle battery temperature, represents an ambient temperature, represents a grid voltage, represents a vehicle battery voltage, represents a minimum term.

[0053] In the embodiment, by constructing a composite reward function that fuses a scheduling deviation penalty term, a temperature control pressure term and a voltage deviation term, the last period power allocation value, the charging current, the port rated current, the scheduling urgency evaluation value, the vehicle battery temperature, the ambient temperature, the grid voltage and the vehicle battery voltage are fully integrated to comprehensively quantify the power scheduling accuracy, the thermal safety margin and the voltage adaptation degree, so as to accurately guide the gradient optimization direction and the strategy convergence path in the port power allocation model training process, and effectively improve the adaptability, robustness and execution efficiency of the power scheduling strategy in the multi-port high-load operation scenario.

[0054] Specifically, the specific steps of evaluating the deviation in the power scheduling execution process after the power allocation is executed are as follows: after the power allocation is executed, the real-time charging pile output power, the grid voltage, the vehicle battery voltage, the vehicle battery temperature, the charging current, the port rated current, the vehicle battery SOC and the power allocation value are extracted; first, the absolute value of the difference between the grid voltage and the vehicle battery voltage is taken as the voltage offset, and multiplied by the charging current in the current sampling period to obtain a voltage deviation current response term reflecting the actual influence of voltage deviation on charging behavior; second, the voltage deviation current response term is divided by the sum of the vehicle battery temperature and the minimum term to obtain a temperature control normalization term affected by temperature regulation; further, the power allocation value is divided by the difference between the port rated current and the current charging current and then added to the minimum term to obtain a current load amplification term measuring the current load level of the port; then, the temperature control normalization term and the current load amplification term are multiplied to obtain a scheduling load response term for representing the interaction between scheduling load pressure and thermal risk; subsequently, the absolute value of the difference between the vehicle battery SOC in the current sampling period and the vehicle battery SOC in the last period is divided by the sum of the sampling time interval and the minimum term, and the ratio is squared to obtain an SOC change rate term characterizing the change speed of the charge state; then, the difference between the actual output power of the charging pile and the corresponding power allocation value is divided by the sum of the power allocation value and the minimum term, and the ratio is squared to obtain a power execution deviation term for reflecting the execution deviation degree of the power scheduling; finally, the scheduling load response term, the SOC change rate term and the power execution deviation term are sequentially added to form a scheduling abnormality evaluation value for evaluating the accuracy of power scheduling execution, the system thermal stability and the efficiency of power transmission, which is used to support the effectiveness judgment and protection mechanism triggering strategy of subsequent scheduling behavior of the system.

[0055] wherein the specific calculation formula of the scheduling abnormality evaluation value is:

[0056] ;

[0057] wherein, denotes a dispatch abnormality evaluation value, denotes a charging current, denotes a grid voltage, denotes a vehicle battery voltage, denotes a vehicle battery temperature, denotes a power allocation value, denotes a port rated current, denotes a vehicle battery SOC, denotes a previous cycle vehicle battery SOC, denotes a sampling time interval, denotes a minimum term, denotes a charging pile output power.

[0058] In the embodiment, by constructing the dispatch abnormality evaluation value, the dispatch load response term, the SOC change rate term and the power execution deviation term are fused, the quantitative evaluation of the power allocation execution process in the dimensions of load pressure, power dynamic change and power execution accuracy is realized, the potential abnormal working conditions in the dispatch process can be accurately identified, the robustness and reliability of the power dispatch behavior are improved, the scientific basis and judgment basis are provided for the subsequent protection mode triggering and dispatch closed-loop optimization, so as to guarantee the thermal safety of system operation, the stability of power transmission and the execution consistency of charging process.

[0059] Specifically, when the evaluation result is normal, the charging is continued to be executed, when the evaluation result is abnormal, the protection mode is entered, and the target SOC prediction and power allocation process are retriggered, and the specific steps of closed-loop optimization of power dispatch are as follows: the dispatch abnormality evaluation value is compared with the abnormal threshold in real time, when the dispatch abnormality evaluation value is less than or equal to the abnormal threshold, it is determined that the current dispatch behavior is in a normal running state, the controller maintains the power output of the charging port, and the current charging task is continued to be executed; when the dispatch abnormality evaluation value is greater than the abnormal threshold, it is identified that the current dispatch behavior has abnormal risk, the power output instruction of the current charging port is immediately suspended, the control logic is switched to the protection mode running state, the port dispatch data collected at the current time is extracted, the vehicle target SOC prediction model is re-executed, the updated vehicle target SOC result is obtained, and the power allocation value of the corresponding port is re-calculated based on the real-time dispatch queue and the sorting result, so as to correct the dispatch deviation; in the protection mode, if the charging task is completed, the power output of the charging port is resumed, and the charging task is continued to be executed; if the charging task is not completed, the power output of the charging port is maintained, and the charging task is continued to be executed. Figure 4As shown, the system monitors two safety indicators of the vehicle battery: the rate of change of battery terminal voltage and the vehicle battery temperature, through high-speed sampling. When the rate of change of battery voltage exceeds the voltage change threshold, the V2G power limiting mechanism is activated. The upper limit of the discharge command is dynamically lowered through the power control interface to limit the output power, thereby effectively suppressing the risk of lithium dendrite formation and preventing abnormal battery overcurrent. When the vehicle battery temperature is lower than the battery temperature threshold, the charging control strategy is switched from the conventional fast charging mode to the pulse heating mode. An intermittent current excitation strategy is used to rapidly increase the cell temperature by utilizing the in-situ heating mechanism of the battery internal resistance until the vehicle battery temperature exceeds the battery temperature threshold. Then, the system automatically switches back to the conventional fast charging process and restores the standard charging path. At the same time, the port scheduling data processed in this cycle and its priority in the scheduling ranking are added as structured new samples to the allocation training sample set of the port power allocation model to participate in subsequent offline reinforcement learning iterative training tasks. This continuously optimizes the strategy parameters of the port power allocation model, ultimately forming a power scheduling adaptive optimization mechanism that integrates scheduling performance feedback, autonomous learning, and closed-loop control.

[0060] In this implementation scheme, a judgment mechanism based on the comparison of scheduling anomaly evaluation values ​​and anomaly thresholds is constructed to achieve high-precision identification of power scheduling behavior status. This ensures that the charging port maintains stable power output when scheduling behavior is normal and switches to protection mode immediately when scheduling behavior is abnormal, effectively preventing overload risks and strategy failures. In protection mode, high-frequency monitoring indicators of battery voltage change rate and vehicle battery temperature are introduced to further refine the safety response strategy. The battery operation safety and low-temperature adaptability are improved through V2G power limiting mechanism and pulse heating mode. At the same time, new samples are constructed and the training sample set is updated using the current period port scheduling data and sorting results, so that the port power allocation model can be continuously optimized in offline iteration. Finally, a power scheduling adaptive optimization mechanism with scheduling anomaly identification capability, safety response capability, and data-driven update capability is formed, comprehensively improving the intelligence, robustness, and adaptability of the system of this invention.

[0061] like Figure 2As shown, the second aspect of the application provides a multi-port hybrid charging pile power pool reinforcement learning scheduling system, which comprises a port data acquisition and preprocessing module, a vehicle demand priority evaluation module, a charging power intelligent allocation module and a scheduling exception closed-loop optimization module, wherein: the port data acquisition and preprocessing module is used to periodically acquire port scheduling data, and perform synchronous alignment, exception elimination, dimension elimination and normalization processing on the port scheduling data to obtain preprocessed port scheduling data; the vehicle demand priority evaluation module is used to construct and train a vehicle target SOC prediction model based on the preprocessed port scheduling data, input real-time port scheduling data into the prediction model, output vehicle target SOC, evaluate the charging urgency of each connected vehicle, and divide the scheduling priority level; the charging power intelligent allocation module is used to call the port scheduling data and sorting results of the connected vehicles in the instant scheduling queue, construct a port power allocation model, output the power allocation value of each charging port, and issue it to the power control interface to execute real-time power scheduling; the scheduling exception closed-loop optimization module is used to evaluate the deviation during power scheduling execution after power allocation execution, continue charging when the evaluation result is normal, and enter protection mode when the evaluation result is abnormal, and retrigger the target SOC prediction and power allocation process to realize closed-loop optimization of power scheduling.

[0062] As shown, Figure 4 As shown, the second aspect of the application provides a multi-port hybrid charging pile power pool reinforcement learning scheduling system, which comprises a port data acquisition and preprocessing module, a vehicle demand priority evaluation module, a charging power intelligent allocation module and a scheduling exception closed-loop optimization module, wherein: the port data acquisition and preprocessing module is used to periodically acquire port scheduling data, and perform synchronous alignment, exception elimination, dimension elimination and normalization processing on the port scheduling data to obtain preprocessed port scheduling data; the vehicle demand priority evaluation module is used to construct and train a vehicle target SOC prediction model based on the preprocessed port scheduling data, input real-time port scheduling data into the prediction model, output vehicle target SOC, evaluate the charging urgency of each connected vehicle, and divide the scheduling priority level; the charging power intelligent allocation module is used to call the port scheduling data and sorting results of the connected vehicles in the instant scheduling queue, construct a port power allocation model, output the power allocation value of each charging port, and issue it to the power control interface to execute real-time power scheduling; the scheduling exception closed-loop optimization module is used to evaluate the deviation during power scheduling execution after power allocation execution, continue charging when the evaluation result is normal, and enter protection mode when the evaluation result is abnormal, and retrigger the target SOC prediction and power allocation process to realize closed-loop optimization of power scheduling.

[0063] In the embodiment, by constructing a multi-port mixed charging pile power pool reinforcement learning scheduling system comprising a port data acquisition preprocessing module, a vehicle demand priority evaluation module, a charging power intelligent allocation module and a scheduling exception closed-loop optimization module, an end-to-end intelligent charging scheduling process is formed. Based on the periodically collected port scheduling data, the system completes data synchronization alignment, exception elimination, dimension elimination and normalization processing, ensuring the quality and consistency of the input data; by constructing a vehicle target SOC prediction model and combining a scheduling urgency evaluation mechanism, the charging demand of the accessed vehicle is dynamically determined and the priority level is divided, improving the accuracy of resource allocation; relying on the policy gradient algorithm to construct a port power allocation model, the intelligent regulation and control of the power value of each charging port is realized, improving the flexibility and response speed of the system scheduling; through the scheduling exception evaluation mechanism and the protection mode switching strategy, the stability and safety of the power scheduling process are ensured, and combined with the data-driven model self-learning mechanism, the continuous optimization and closed-loop update of the scheduling strategy parameters are realized, comprehensively improving the intelligent scheduling ability and system adaptability of the invention under complex working conditions.

[0064] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another, without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0065] The preferred embodiments of the application disclosed above are only used to help explain the application. The preferred embodiments do not describe all the details of the application, nor limit the application to the specific embodiments described. Obviously, many modifications and variations can be made to the application based on the content of the specification. The specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the application, so that those skilled in the art can well understand and utilize the application. The application is limited only by the claims and their full scope and equivalents.

Claims

1. A reinforcement learning scheduling method for the power pool of a multi-port hybrid charging pile, characterized in that, Includes the following steps: S1 periodically collects port scheduling data and performs synchronization alignment, anomaly removal, dimension elimination and normalization processing on the port scheduling data to obtain preprocessed port scheduling data; S2. Based on the preprocessed port scheduling data, construct and train the vehicle target SOC prediction model, input the real-time port scheduling data into the prediction model, output the vehicle target SOC, evaluate the charging urgency of each access vehicle, and classify the scheduling priority level. S3 calls the port scheduling data and sorting results of the connected vehicles in the real-time scheduling queue, constructs the port power allocation model, outputs the power allocation value of each charging port, and sends it to the power control interface to execute real-time power scheduling. S4. After power allocation is executed, the deviation during the power scheduling process is evaluated. If the evaluation result is normal, charging continues. If the evaluation result is abnormal, the protection mode is entered and the target SOC prediction and power allocation process is re-triggered to achieve closed-loop optimization of power scheduling. The specific steps for achieving closed-loop optimization of power scheduling—continuing charging when the evaluation result is normal, entering protection mode when the evaluation result is abnormal, and re-triggering the target SOC prediction and power allocation process—are as follows: The scheduling anomaly assessment value is compared with the anomaly threshold in real time. When the scheduling anomaly assessment value is less than or equal to the anomaly threshold, the current scheduling behavior is determined to be normal and the charging task continues. When the scheduling anomaly assessment value is greater than the anomaly threshold, the current scheduling behavior is determined to be abnormal, the power output of the current charging port is suspended, the system enters protection mode, and real-time port scheduling data is extracted to recalculate the vehicle target SOC prediction and power allocation value. In protection mode, the battery voltage change rate and vehicle battery temperature are monitored in real time. When the battery voltage change rate is greater than the voltage change threshold, the V2G power limiting mechanism is activated. When the vehicle battery temperature is less than the battery temperature threshold, the regular fast charging mode is switched to pulse heating mode, which uses the battery internal resistance to generate heat to quickly raise the temperature until the vehicle battery temperature is greater than the battery temperature threshold, at which point the mode is switched back to regular fast charging mode. Meanwhile, the port scheduling data and sorting results of the current period are added as new samples to the allocation training sample set for offline iterative training of the subsequent port power allocation model, forming a self-learning closed-loop mechanism based on scheduling performance feedback.

2. The reinforcement learning scheduling method for the power pool of a multi-port hybrid charging pile according to claim 1, characterized in that: The specific steps for periodically collecting port scheduling data and performing synchronization alignment, anomaly removal, dimension elimination, and normalization processing on the port scheduling data to obtain preprocessed port scheduling data are as follows: A fixed-width sliding time window is used as a sampling period to periodically collect port scheduling data, which includes vehicle battery SOC, vehicle battery voltage, vehicle battery temperature, grid voltage, charging current, charging pile output power, and ambient temperature. By constructing a synchronous sampling scheduling mechanism based on the access priority of multi-source data, the asynchronous data in the port scheduling data is aligned in time series; by adopting the exponential sliding average filtering algorithm, the noise of the port scheduling data is suppressed and the signal is smoothed; by using the univariate outlier detection algorithm of the Z-Score method, the abnormal points in the port scheduling data are removed; by combining the quantile interval scaling with the min-max normalization algorithm, the cross-dimensional dimension elimination and data distribution unification of the port scheduling data are performed.

3. The reinforcement learning scheduling method for the power pool of a multi-port hybrid charging pile according to claim 1, characterized in that: Based on the preprocessed port scheduling data, the vehicle target SOC prediction model is constructed and trained. The specific steps of inputting the real-time port scheduling data into the prediction model and outputting the vehicle target SOC are as follows: Extract the historical port scheduling data after preprocessing, and adopt the supervised learning method to construct a vehicle target SOC prediction model based on a multi-layer feedforward neural network; use the port scheduling data as the input feature, and use the vehicle target SOC when the vehicle actually completes the charging task as the supervised label to form a prediction training sample set; And use the least mean square error as the loss function to train the vehicle target SOC prediction model; input the real-time port scheduling data into the trained vehicle target SOC prediction model to output the vehicle target SOC.

4. The reinforcement learning scheduling method for the power pool of a multi-port hybrid charging pile according to claim 1, characterized in that: The specific steps of evaluating the charging urgency of each access vehicle are as follows: Square the difference between the vehicle target SOC and the vehicle battery SOC, and then divide it by the vehicle battery temperature plus one to obtain the power difference temperature control factor; divide the rated output current by the sum of the charging current and the minimum term, and take the power function value with the current suppression index as the exponent of the ratio to obtain the unit current utilization term; take the natural logarithm of the sum of the ratio of the grid voltage to the vehicle battery voltage plus one to obtain the voltage matching degree term; Multiply the power difference temperature control factor, the unit current utilization term and the voltage matching degree term in sequence to obtain the scheduling urgency evaluation value.

5. The reinforcement learning scheduling method for the power pool of a multi-port hybrid charging pile according to claim 1, characterized in that: The specific steps of dividing the scheduling priority levels are as follows: Compare the scheduling urgency evaluation value D with the multi-level scheduling thresholds D1 and D2 in real time, and divide the access vehicles into three priority queues: When D ≤ D1, the vehicle is assigned to the scheduling waiting queue, and the power distribution for the current cycle is postponed; When D1 < D < D2, the vehicle is assigned to the regular scheduling queue and enters the normal power scheduling process; When D ≥ D2, the vehicle is assigned to the immediate scheduling queue and sorted from large to small according to the scheduling urgency evaluation value.

6. The reinforcement learning scheduling method for the power pool of a multi-port hybrid charging pile according to claim 1, characterized in that: The specific steps of calling the port scheduling data and sorting results of the access vehicles in the immediate scheduling queue, constructing a port power distribution model, outputting the power distribution values of each charging port, and sending them to the power control interface to perform real-time power scheduling are as follows: Call the port scheduling data, scheduling urgency evaluation value and sorting results corresponding to the access vehicles in the immediate scheduling queue, select the deep deterministic policy gradient algorithm as the training framework, and construct a port power distribution model based on the policy network and value network group; Construct an allocation training sample set by using the historical port scheduling data and actual power distribution records. During the training process, construct a composite reward function that combines scheduling deviation, temperature control pressure and voltage deviation to obtain the function output value; Using the function output value as the cumulative negative reward value, the port power allocation model is driven to perform parameter optimization and policy convergence through the policy gradient update mechanism. The port scheduling data and sorting results of the vehicles connected in the real-time scheduling queue are input into the trained port power allocation model, and the power allocation value corresponding to each charging port is output. The power allocation value is then used as an execution instruction and sent to each corresponding charging port for execution through the power control interface.

7. The reinforcement learning scheduling method for the power pool of a multi-port hybrid charging pile according to claim 6, characterized in that: The specific steps for constructing a composite reward function that integrates scheduling deviation, temperature control pressure, and voltage deviation, and obtaining the function output value are as follows: The energy load term is obtained by squaring the power allocation value of the previous cycle and multiplying it by the charging current. The normalized scheduling benchmark term is obtained by subtracting the charging current from the rated current at the port, adding a minimum term, and then multiplying it by the scheduling urgency assessment value. The scheduling deviation penalty term is obtained by dividing the energy load term by the normalized scheduling benchmark term. The temperature control pressure term is obtained by dividing the absolute value of the difference between the vehicle battery temperature and the ambient temperature by the sum of the vehicle battery temperature and the minimum term. The voltage deviation term is obtained by dividing the absolute value of the difference between the grid voltage and the vehicle battery voltage by the sum of the grid voltage, the vehicle battery voltage, and the minimum term. The scheduling deviation penalty term, the temperature control pressure term, and the voltage deviation term are added together in sequence to obtain the function output value.

8. The reinforcement learning scheduling method for the power pool of a multi-port hybrid charging pile according to claim 1, characterized in that: After the power allocation is executed, the specific steps for evaluating the deviation during the power scheduling execution process are as follows: After power allocation is executed, the real-time charging pile output power and corresponding port scheduling data are extracted. The absolute value of the difference between the grid voltage and the vehicle battery voltage is multiplied by the charging current to obtain the voltage deviation current response term. The voltage deviation current response term is divided by the sum of the vehicle battery temperature and the minimum term to obtain the temperature control normalization term. The power allocation value is divided by the sum of the difference between the port rated current and the charging current and the minimum term to obtain the current load amplification term. The temperature control normalization term is multiplied by the current load amplification term to obtain the scheduling load response term. The absolute value of the difference between the vehicle battery SOC of the current cycle and the vehicle battery SOC of the previous cycle is divided by the sum of the sampling time interval and the minimum term, and the square is taken to obtain the SOC change rate term. The difference between the charging pile output power and the power allocation value is divided by the sum of the power allocation value and the minimum term, and the square is taken to obtain the power execution deviation term. The scheduling load response term, the SOC change rate term, and the power execution deviation term are added together to obtain the scheduling anomaly evaluation value.

9. A multi-port hybrid charging pile power pool reinforcement learning scheduling system, characterized in that: include: The system includes a port data acquisition and preprocessing module, a vehicle demand priority assessment module, a charging power intelligent allocation module, and a scheduling anomaly closed-loop optimization module, among which: The port data acquisition and preprocessing module is used to periodically acquire port scheduling data and perform synchronization alignment, anomaly removal, dimension elimination and normalization processing on the port scheduling data to obtain preprocessed port scheduling data. The vehicle demand priority assessment module is used to construct and train a vehicle target SOC prediction model based on preprocessed port scheduling data, input real-time port scheduling data into the prediction model, output the vehicle target SOC, assess the charging urgency of each connected vehicle, and classify scheduling priority levels. The intelligent charging power allocation module is used to call the port scheduling data and sorting results of the connected vehicles in the real-time scheduling queue, construct the port power allocation model, output the power allocation value of each charging port, and send it to the power control interface to execute real-time power scheduling. The scheduling anomaly closed-loop optimization module is used to evaluate the deviation in the power scheduling process after power allocation is executed. When the evaluation result is normal, charging continues. When the evaluation result is abnormal, it enters protection mode and re-triggers the target SOC prediction and power allocation process to achieve closed-loop optimization of power scheduling. The specific steps for achieving closed-loop optimization of power scheduling—continuing charging when the evaluation result is normal, entering protection mode when the evaluation result is abnormal, and re-triggering the target SOC prediction and power allocation process—are as follows: The scheduling anomaly assessment value is compared with the anomaly threshold in real time. When the scheduling anomaly assessment value is less than or equal to the anomaly threshold, the current scheduling behavior is determined to be normal and the charging task continues. When the scheduling anomaly assessment value is greater than the anomaly threshold, the current scheduling behavior is determined to be abnormal, the power output of the current charging port is suspended, the system enters protection mode, and real-time port scheduling data is extracted to recalculate the vehicle target SOC prediction and power allocation value. In protection mode, the battery voltage change rate and vehicle battery temperature are monitored in real time. When the battery voltage change rate is greater than the voltage change threshold, the V2G power limiting mechanism is activated. When the vehicle battery temperature is less than the battery temperature threshold, the regular fast charging mode is switched to pulse heating mode, which uses the battery internal resistance to generate heat to quickly raise the temperature until the vehicle battery temperature is greater than the battery temperature threshold, at which point the mode is switched back to regular fast charging mode. Meanwhile, the port scheduling data and sorting results of the current period are added as new samples to the allocation training sample set for offline iterative training of the subsequent port power allocation model, forming a self-learning closed-loop mechanism based on scheduling performance feedback.

Images