Electric vehicle charging load space-time prediction method based on MDP and coupling effect, storage medium and electronic equipment

Abstract

The invention provides an electric vehicle charging load space-time prediction method based on an MDP and a coupling effect, a storage medium and electronic equipment. The method comprises the steps of collecting initial data of an electric vehicle in a road network; classifying the electric vehicles and constructing corresponding travel models, and generating a dynamic travel path of each electric vehicle by using a Markov decision process; the residual charge state of the electric vehicle is calculated in real time through an energy consumption model considering multi-factor coupling; continuously monitoring the residual charge state of the electric vehicle, if the residual charge state is lower than a preset charging trigger threshold value, triggering a charging behavior, and guiding the electric vehicle to go to a charging station near a current node or a next node; the starting time, the charging duration, the charging power and the geographic position of all the triggered charging events are counted, and a charging load space-time distribution matrix is generated. According to the method, the charging load space-time distribution characteristics of different vehicle types, functional areas and travel days can be accurately presented.

Description

A spatiotemporal prediction method, storage medium, and electronic device for electric vehicle charging load based on MDP and coupling effect. Technical Field

[0001] This invention belongs to the field of electric vehicle and power grid interaction technology, and proposes a spatiotemporal prediction method for electric vehicle charging load based on Markov decision process (MDP) and coupling effect, as well as a storage medium and electronic device. Background Technology

[0002] Guided by the strategic goal of "carbon peaking and carbon neutrality," electric vehicles (EVs), as the core carrier of clean energy applications, have experienced rapid industrial growth. However, large-scale EV charging loads are characterized by prominent spatiotemporal randomness and complex coupling of influencing factors. Their disorderly access has significantly impacted the safety, stability, and power quality of the power grid. Therefore, accurately capturing the spatiotemporal distribution patterns of EV charging loads has become a crucial foundation for ensuring optimized grid operation and improving the capacity for renewable energy absorption. Current domestic and international research has not yet effectively solved these practical problems, and the following technical bottlenecks remain:

[0003] Limitations of travel behavior modeling: Existing studies mostly rely on fixed travel chains or static road network information to build models, which neither fully considers the subjectivity and randomness of drivers' travel decisions nor dynamically integrates key information such as real-time traffic flow and road condition changes. This leads to significant deviations between EV spatiotemporal transfer path predictions and actual travel scenarios. For example, some studies use shortest path algorithms to plan trips, ignoring the dynamic adjustments to routes caused by road congestion, making it difficult to reflect real travel patterns.

[0004] Insufficient consideration of coupling factors: Electric vehicle charging load is influenced by multiple factors, including vehicle performance, travel environment, and date type. However, existing models often analyze single factors in isolation, failing to establish multi-factor coupled energy consumption models and load prediction mechanisms. For example, regarding the impact of temperature on energy consumption, some studies only consider this factor in isolation, neglecting the cumulative effect of vehicle speed changes due to road congestion on energy consumption per unit mileage. This results in significant errors in energy consumption calculations, consequently affecting the accuracy of charging demand assessment.

[0005] Insufficient load forecasting accuracy: Existing methods do not adequately characterize the differences in travel characteristics among different types of electric vehicles and fail to differentiate charging demand patterns across different functional areas. This results in forecasts that cannot reflect the amplitude and distribution differences in charging load across the three dimensions of "vehicle type-region-time period." Furthermore, traditional models do not dynamically update road network information and vehicle distribution, making it difficult to adapt to real-time changes in traffic and charging scenarios.

[0006] To address the aforementioned issues, there is an urgent need to develop a method for predicting the spatiotemporal distribution of electric vehicle charging load that takes into account multi-factor coupling, dynamic travel route planning, and differentiated travel characteristics, in order to improve prediction accuracy and practical application value. This has become a pressing problem to be solved. Summary of the Invention

[0007] In view of this, the present invention provides a spatiotemporal prediction method, storage medium and electronic device for electric vehicle charging load based on Markov decision process (MDP) and coupling effect, so as to solve the problems existing in the prior art.

[0008] This invention provides a spatiotemporal prediction method for electric vehicle charging load based on MDP and coupling effects, comprising:

[0009] Collect initial data on electric vehicles in the road network;

[0010] Electric vehicles are classified and corresponding travel models are constructed. Based on the travel models and the initial data, the departure time and dwell time of each electric vehicle at different nodes in the urban road network are calculated. The dynamic travel path of each electric vehicle is generated using a Markov decision process.

[0011] Based on the dynamic travel route, the remaining state of charge of electric vehicles is calculated in real time through an energy consumption model that takes into account multiple factors, including traffic flow, road conditions, date type and ambient temperature.

[0012] The remaining charge status of each electric vehicle is continuously monitored and compared with a preset charging trigger threshold. If the predicted remaining charge status is lower than the preset charging trigger threshold, the charging behavior is triggered, guiding the electric vehicle to the charging station near the current or next node.

[0013] The start time, charging duration, charging power, and geographical location of all triggered charging events are statistically analyzed, and a spatiotemporal distribution matrix of charging load is generated based on the relevant data of the charging events.

[0014] Preferably, the initial data of the electric vehicle includes the vehicle model, initial state of charge, origin, and battery parameters.

[0015] Further optimization involves calculating the departure time and dwell time of electric vehicles at different nodes in the urban road network, including the following steps:

[0016] The urban road network is abstracted as a topological structure consisting of nodes and paths connecting the nodes. The nodes include random network nodes and fixed network nodes. The random network nodes include residential areas, work areas and other business areas. The fixed network nodes include bus starting stations and bus ending stations. All nodes are equipped with charging facilities.

[0017] Electric vehicles are classified according to their travel characteristics and operating modes, and corresponding travel models are constructed.

[0018] Based on the initial data of electric vehicles and their corresponding travel models, the departure time and dwell time of the electric vehicles at different nodes in the urban road network are calculated.

[0019] Further preferred, the electric vehicles are divided into three categories: private cars, taxis, and buses.

[0020] Further optimization involves fitting the travel time and dwell time of electric vehicles at each node using a Gaussian mixture model.

[0021] Further optimization involves the following steps for calculating parameters μ and σ in the Gaussian mixture model:

[0022] Acquire vehicle trajectory data, vehicle status data, and charging record data, and classify them according to vehicle type and date type. Parse the continuous time series into stay events and travel events, extract the arrival time, departure time, stay duration, and travel duration of each electric vehicle travel event, and construct a stay time sample set and a travel time sample set.

[0023] Perform K-means clustering on the samples and initialize with cluster centers. Initialize with the variance of each cluster sample ;

[0024] Calculate the non-standardized posterior for each sample and each component, and normalize it to obtain the responsibility degree;

[0025] The number of valid samples is calculated based on the aforementioned responsibility level, and the model parameters are updated using the number of valid samples. and .

[0026] Further optimization involves using Markov decision processes to generate dynamic travel routes for electric vehicles, as follows:

[0027] 1) Define the state function ,in, Let t be the current node and t be the current time. The current state of charge of the electric vehicle. For environmental parameters;

[0028] 2) Define the action function : Used to indicate selecting the next edge / adjacent node starting from the current node;

[0029] 3) Calculate the transition probability of the electric vehicle from the current node to all adjacent nodes using the path with the shortest remaining travel time as the reward function;

[0030] 4) Using the Monte Carlo sampling method, based on the calculated transition probability distribution, randomly select the next road segment for the electric vehicle, and repeat steps 3) and 4) until the destination is reached or charging is triggered, thereby generating the dynamic travel path of the electric vehicle.

[0031] Further optimization involves the following steps for real-time calculation of the remaining state of charge of an electric vehicle:

[0032] Calculate the current vehicle speed based on the current road congestion information;

[0033] The energy consumption of the electric vehicle is calculated based on its current speed and using an energy consumption model that takes into account the city road class and date type.

[0034] The energy consumption of the electric vehicle is nonlinearly corrected based on ambient temperature.

[0035] The present invention also provides a storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described method for spatiotemporal prediction of electric vehicle charging load based on MDP and coupling effect.

[0036] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for spatiotemporal prediction of electric vehicle charging load based on MDP and coupling effect.

[0037] The beneficial effects of the spatiotemporal prediction method for electric vehicle charging load based on MDP and coupling effect provided by this invention are as follows:

[0038] 1. By integrating real-time traffic congestion, date type, and temperature through MDP theory, road network information and route planning are dynamically updated. This can accurately reflect the decision-making behavior of car owners in different travel scenarios, such as avoiding congestion and temporary charging. Compared with traditional shortest path methods or static travel chain models, it helps to reduce the prediction error of single travel chain duration, and the number of node transfers is more in line with actual travel patterns.

[0039] 2. Constructing an energy consumption model that couples multiple factors such as traffic flow, road conditions, date type, and temperature, and distinguishing the superimposed effects of different road grades and temperatures on energy consumption, can improve the accuracy of energy consumption calculation and provide a precise energy consumption basis for judging charging demand.

[0040] 3. Modeling the differentiated travel characteristics of private cars, taxis, and buses, and distinguishing the functional differences of residential areas, work areas, and leisure areas, it can accurately predict the charging load distribution of different date types (weekdays / non-weekdays) and different areas.

[0041] 4. The prediction results can be directly used for charging station site selection, orderly charging and discharging scheduling and power grid planning, providing data support for the coordinated operation of EVs and the power grid, efficiently helping to "shave peaks and fill valleys" and empowering the achievement of carbon emission reduction targets. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below.

[0043] Figure 1 is a flowchart of the spatiotemporal prediction method for electric vehicle charging load based on MDP and coupling effect provided by the present invention. Detailed Implementation

[0044] The present invention will be further described below with reference to specific embodiments.

[0045] As shown in Figure 1, this invention provides a method for predicting the spatiotemporal distribution of electric vehicle charging load based on MDP and coupling effects, including the following steps:

[0046] S1: Collect initial data of electric vehicles in the road network, wherein the initial data includes vehicle type, initial state of charge (SOC), origin, battery parameters, etc.

[0047] S2: Classify electric vehicles and construct corresponding travel models. Based on the travel models and the initial data, calculate the departure time and dwell time of each electric vehicle at different nodes in the urban road network. Use Markov decision process to generate dynamic travel paths for each electric vehicle.

[0048] The calculation of the departure time and dwell time of electric vehicles at different nodes in the urban road network includes the following steps:

[0049] S21: The urban road network is abstracted into a topological structure consisting of nodes and paths connecting the nodes. The nodes include random network nodes and fixed network nodes. The random network nodes include residential areas (H), work areas (W), and other business areas (O). The other business areas include shopping, dining, social and other activity areas. The fixed network nodes include bus starting stations (S) and bus ending stations (E). All nodes are equipped with charging facilities. The spatial displacement process between nodes corresponds to the actual driving trajectory of electric vehicles.

[0050] S22: Classify electric vehicles according to their travel characteristics and operating modes, and construct corresponding travel models;

[0051] Electric vehicles are mainly divided into three categories: private cars, taxis, and buses;

[0052] For private car travel models, based on Markov decision process theory, a multi-objective travel chain is constructed, represented by "home (H) - workplace (W) - other affairs (O) - home (H)";

[0053] For the taxi travel model, due to the high randomness of its travel destination, a two-dimensional normal distribution sampling algorithm is used to randomly generate the destination of each trip. At the same time, models for average passenger carrying time and passenger waiting time (whose waiting time also follows a normal distribution) are established, and Monte Carlo methods are used for random sampling to simulate its continuous operation process;

[0054] For the bus travel model, since the bus routes and timetables are fixed, the arrival time and possible charging needs at each fixed stop are calculated based on the preset fixed routes (such as SE) and departure intervals.

[0055] Among them, the travel time of electric vehicles and the duration of stay at each node The model is fitted using a Gaussian mixture model, where the probability density function of the Gaussian mixture model is:

[0056] ;

[0057] In the formula: μ and σ are the mean and variance at time t, respectively.

[0058] Using historical travel trajectory data of electric vehicles (city-level electric vehicle travel trajectory data of more than 3 months, including departure time, stay time and destination type) as samples, the EM algorithm is used to calculate the mean μ and variance σ of the model parameters in an iterative manner. Different μ and σ can reflect the travel habits of different car owner groups.

[0059] The steps for calculating the model parameters μ and σ using the EM algorithm in an iterative manner are as follows:

[0060] 1) Constructing a sample set: Acquire vehicle trajectory data, vehicle status data, and charging record data, and classify them according to vehicle type and date type. Parse the continuous time series into stay events and travel events, and extract the arrival time of each electric vehicle's travel event. Departure time Duration of stay Travel duration Construct a sample set of stay time and a sample set of travel time, specifically as follows: N represents the number of historical electric vehicle travel times;

[0061] 2) Initialize parameters: Perform K-means clustering on the samples and initialize parameters using the cluster centers. Initialize with the variance of each cluster sample ;

[0062] 3) Calculate posterior responsibility: Calculate the posterior responsibility for each sample. With each component The degree of responsibility is obtained by non-standardized posterior and normalization;

[0063] The formula for calculating the non-standardized posterior is as follows:

[0064] ;

[0065] The normalization formula is as follows:

[0066] ;

[0067] 4) Update parameters using the degree of responsibility: Calculate the number of valid samples based on the degree of responsibility and update the model parameters. and We obtained the probability distribution of travel time and stay time for different vehicle types and date types;

[0068] The formula for calculating the number of valid samples is as follows:

[0069] ;

[0070] The formula for updating the parameters is as follows:

[0071] ;

[0072] ;

[0073] ;

[0074] S23: Based on the initial data of the electric vehicle and its corresponding travel model, calculate the departure time and dwell time of the electric vehicle at different nodes in the urban road network;

[0075] One method for generating dynamic travel routes for electric vehicles using Markov decision processes is as follows:

[0076] 1) Define the state function ,in, Let t be the current node and t be the current time. The current state of charge of the electric vehicle. These are environmental parameters, including current speed / road condition information, etc.

[0077] 2) Define the action function : Used to indicate selecting the next edge / adjacent node starting from the current node;

[0078] 3) The reward function is the path with the shortest remaining travel time. Make action decisions by calculating the transition probabilities of the electric vehicle from the current node to all neighboring nodes. To simulate uncertain real-world traffic environments, in which,

[0079] ;

[0080] ;

[0081] In the formula: Indicates the state Below, U represents the expected remaining time to reach the destination from the current location, where U is the node's distance from the current location. The number of all feasible paths to the destination; M is the number of nodes the next step passes through in all feasible paths. Number of possible solutions; K is the number of nodes that the node transitions through in the current solution; , Let be the lengths of the k-th path segment under U and M, respectively; , These represent the EV speeds on the k-th path segment under conditions U and M, respectively.

[0082] 4) Using the Monte Carlo sampling method, based on the calculated transition probability distribution, randomly select the next road segment for the electric vehicle, and repeat steps 3) and 4) until the destination is reached or charging is triggered, generating a dynamic travel path that closely matches the actual traffic conditions;

[0083] S3: Based on the dynamic travel path, the remaining state of charge (SOC) of the electric vehicle is calculated in real time through an energy consumption model that takes into account multiple factors, including traffic flow, road conditions, date type and ambient temperature.

[0084] The specific steps are as follows:

[0085] S31: Calculate the current vehicle speed based on current road congestion information. Specifically: using traffic simulation and historical data, assuming that electric vehicles have the same speed within a certain threshold, calculate the number of electric vehicles traveling per unit distance. Determine the travel speed v of electric vehicles on road j at time t. Update the number of electric vehicles on each road segment at the beginning of each scheduling cycle. At the end of each node transfer, the travel speed v of electric vehicles is adjusted, i.e., v is dynamically updated.

[0086] ;

[0087] In the formula: Let be the initial number of EVs in road j at time t, and k be the correction parameter (to adapt to congestion simulation caused by insufficient EV penetration). The number of EVs for road j within the scheduling period. This is the current road length. According to... Determine the driving speed v of electric vehicles; reduce vehicle speed in congested areas and reduce vehicle speed to close to the road's design speed in uncongested areas.

[0088] S32: Based on the current speed v of the electric vehicle, the energy consumption of the electric vehicle is calculated using an energy consumption model. This model considers the influence of urban road classification (expressway, arterial road, secondary arterial road, local road) and date type (weekday, non-weekday), reflecting a quantitative relationship between energy consumption per unit mileage and v. The formula is:

[0089] ;

[0090] In the formula: The baseline value of energy consumption per unit mileage is given for different date characteristics d (d1 is a working day, d2 is a non-working day), road class r (r1 is an expressway, r2 is a main road, r3 is a secondary road, r4 is a local road), and driving speed v. The energy consumption coefficient is related to road class (reflecting the impact of physical attributes of different road classes such as expressways and arterial roads on energy consumption). Associate energy consumption coefficients with date types (adapting to the differentiated effects of weekday peak congestion and non-weekday traffic dispersion on energy consumption). Based on energy consumption offset;

[0091] S33: The energy consumption of the electric vehicle is nonlinearly corrected based on the ambient temperature, wherein the correction model is as follows:

[0092] ;

[0093] In the formula: y represents the fitting parameters; y is a natural number; c is the current ambient temperature.

[0094] Since the effect of ambient temperature on electric vehicle battery energy consumption is non-linear, the energy consumption of electric vehicles at actual temperatures is corrected by using 20℃ as the standard temperature (where electric vehicles have the highest energy efficiency and most stable performance). Combined with road grade energy consumption, the energy consumption per unit mileage of road r at temperature c is obtained:

[0095] ;

[0096] In the formula: Standard road energy consumption based on 20°C; The energy consumption per unit mile for electric vehicles (EVs) under temperature c and road grade r. Energy consumption per unit distance for electric vehicles (EVs) at 20°C;

[0097] S4: Continuously monitor the remaining state of charge of each electric vehicle and compare the remaining state of charge with the preset charging trigger threshold. If the predicted remaining state of charge is lower than the preset charging trigger threshold, the charging behavior is triggered to guide the electric vehicle to the charging station near the current or next node.

[0098] The preset charging trigger threshold is 20%.

[0099] S5: Calculate the start time, charging duration, charging power, and geographical location of all triggered charging events, and generate a spatiotemporal distribution matrix of charging load based on the relevant data of the charging events.

[0100] The spatiotemporal distribution matrix of charging load can intuitively display the total power of electric vehicle charging load in various areas of the city at different times of the day, providing direct and reliable data support for power grid dispatching and charging station planning.

[0101] The beneficial effects of the spatiotemporal prediction method for electric vehicle charging load based on MDP and coupling effect provided by this invention are as follows:

[0102] 1. By integrating real-time traffic congestion, date type, and temperature through MDP theory, road network information and route planning are dynamically updated. This can accurately reflect the decision-making behavior of car owners in different travel scenarios, such as avoiding congestion and temporary charging. Compared with traditional shortest path methods or static travel chain models, it helps to reduce the prediction error of single travel chain duration, and the number of node transfers is more in line with actual travel patterns.

[0103] 2. Constructing an energy consumption model that couples multiple factors such as traffic flow, road conditions, date type, and temperature, and distinguishing the superimposed effects of different road grades and temperatures on energy consumption, can improve the accuracy of energy consumption calculation and provide a precise energy consumption basis for judging charging demand.

[0104] 3. Modeling the differentiated travel characteristics of private cars, taxis, and buses, and distinguishing the functional differences of residential areas, work areas, and leisure areas, it can accurately predict the charging load distribution of different date types (weekdays / non-weekdays) and different areas.

[0105] 4. The prediction results can be directly used for charging station site selection, orderly charging and discharging scheduling and power grid planning, providing data support for the coordinated operation of EVs and the power grid, efficiently helping to "shave peaks and fill valleys" and empowering the achievement of carbon emission reduction targets.

[0106] The present invention also provides a storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described method for spatiotemporal prediction of electric vehicle charging load based on MDP and coupling effect.

[0107] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for spatiotemporal prediction of electric vehicle charging load based on MDP and coupling effect.

[0108] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0109] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.

[0110] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0111] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0112] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0113] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A spatiotemporal prediction method for electric vehicle charging load based on MDP and coupling effect, characterized in that, include: Initial data of electric vehicles in the road network is collected; electric vehicles are classified and corresponding travel models are constructed. Based on the travel models and the initial data, the departure time and dwell time of each electric vehicle at different nodes in the urban road network are calculated, and a dynamic travel path for each electric vehicle is generated using a Markov decision process. Based on the dynamic travel path, the remaining state of charge of electric vehicles is calculated in real time using an energy consumption model that considers multiple factors, including traffic flow, road conditions, date type, and ambient temperature. The remaining state of charge of each electric vehicle is continuously monitored and compared with a preset charging trigger threshold. If the predicted remaining state of charge is lower than the preset charging trigger threshold, charging behavior is triggered, guiding the electric vehicle to a charging station near the current or next node. The start time, charging duration, charging power, and geographical location of all triggered charging events are statistically analyzed, and a spatiotemporal distribution matrix of charging load is generated based on the relevant data of the charging events.

2. The spatiotemporal prediction method for electric vehicle charging load based on MDP and coupling effect according to claim 1, characterized in that, The initial data for the electric vehicle includes the vehicle model, initial state of charge, origin, and battery parameters.

3. The spatiotemporal prediction method for electric vehicle charging load based on MDP and coupling effect according to claim 1, characterized in that, Calculating the departure time and dwell time of electric vehicles at different nodes in an urban road network includes the following steps: The urban road network is abstracted into a topology consisting of nodes and paths connecting them. The nodes include random network nodes and fixed network nodes. Random network nodes include residential areas, work areas, and other business areas. Fixed network nodes include bus origin and destination stations. All nodes are pre-equipped with charging facilities. Electric vehicles are classified according to their travel characteristics and operating modes, and corresponding travel models are constructed. Based on the initial data of the electric vehicles and their corresponding travel models, the departure time and dwell time of the electric vehicles at different nodes in the urban road network are calculated.

4. The spatiotemporal prediction method for electric vehicle charging load based on MDP and coupling effect according to claim 1, characterized in that, The electric vehicles are divided into three categories: private cars, taxis, and buses.

5. The spatiotemporal prediction method for electric vehicle charging load based on MDP and coupling effect according to claim 1, characterized in that, The travel time and dwell time of electric vehicles at each node were fitted using a Gaussian mixture model.

6. The spatiotemporal prediction method for electric vehicle charging load based on MDP and coupling effect according to claim 1, characterized in that, The calculation steps for parameters μ and σ in the Gaussian mixture model are as follows: Obtain vehicle trajectory data, vehicle status data, and charging record data, and classify them according to vehicle type and date type. Parse the continuous time series into stay events and travel events, extract the arrival time, departure time, stay duration, and travel duration for each electric vehicle's travel event, and construct stay time sample sets and travel time sample sets. Perform K-means clustering on the samples, initializing the clusters with cluster centers. Initialize with the variance of each cluster sample ； The unstandardized posterior for each sample and each component is calculated and normalized to obtain the responsibility score; the number of valid samples is calculated based on the responsibility score, and the model parameters are updated using the number of valid samples. and 。 7. The spatiotemporal prediction method for electric vehicle charging load based on MDP and coupling effect according to claim 1, characterized in that, The method for generating dynamic travel routes for electric vehicles using Markov decision processes is as follows: 1) Define the state function. ,in, Let t be the current node and t be the current time. The current state of charge of the electric vehicle. 1) Define environmental parameters; 2) Define action functions. : Used to indicate the selection of the next edge / adjacent node from the current node; 3) Calculate the transition probability of the electric vehicle from the current node to all adjacent nodes using the path with the shortest remaining travel time as the reward function; 4) Using the Monte Carlo sampling method, randomly select the next road segment for the electric vehicle according to the calculated transition probability distribution, and repeat steps 3) and 4) until the destination is reached or charging is triggered, thereby generating the dynamic travel path of the electric vehicle.

8. The spatiotemporal prediction method for electric vehicle charging load based on MDP and coupling effect according to claim 1, characterized in that, Real-time calculation of the remaining state of charge of an electric vehicle includes the following steps: calculating the current vehicle speed based on the current road congestion information; calculating the energy consumption of the electric vehicle based on the current vehicle speed and using an energy consumption model, wherein the energy consumption model considers the urban road level and date type; and performing nonlinear correction on the energy consumption of the electric vehicle based on the ambient temperature.

9. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the spatiotemporal prediction method for electric vehicle charging load based on MDP and coupling effect as described in any one of claims 1-8.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the spatiotemporal prediction method for electric vehicle charging load based on MDP and coupling effect as described in any one of claims 1-8.

Images