Traffic-power network space-time coupling optimization method and system based on user preference empowerment

Abstract

The invention relates to the technical field of electric power-traffic network cooperative operation, and discloses a traffic-electric power network space-time coupling optimization method and system based on user preference empowerment, and the method comprises the steps: collecting and preprocessing the behavior preference data of electric vehicle user charging and traveling in a predetermined region; the method comprises the following steps: modeling and quantifying electric vehicle user preferences based on an information enhanced generative adversarial network deep learning architecture, and generating preference weight vectors with spatial-temporal characteristics; constructing a traffic network model by using the weighted directed graph, establishing a power distribution network model, and constructing a traffic-power coupling network model based on the space-time mapping relationship; and constructing a multi-target optimization model, and solving the multi-target optimization model by using an optimization algorithm based on the user preference weight of the electric vehicle to realize dynamic optimization of charging scheduling and path planning. According to the invention, efficient coupling optimization of traffic and a power network is realized, and the overall performance and user satisfaction of the system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power-transportation network coordinated operation, and in particular to a method and system for optimizing the spatiotemporal coupling of transportation-power networks based on user preference empowerment. Background Art

[0002] Globally, with growing environmental awareness and technological advancements, the market share of electric vehicles has increased annually. The widespread adoption of electric vehicles not only helps reduce greenhouse gas emissions but also reduces dependence on fossil fuels. However, the widespread use of electric vehicles also brings new challenges, particularly in the management of transportation and power networks. The charging needs of electric vehicles impose new requirements on load distribution in power networks, while the layout and use of charging stations directly impact the dynamics of traffic flow.

[0003] With the increasing popularity of electric vehicles (EVs), the coupling between transportation and power networks has become increasingly prominent. The demand for EV charging puts pressure on power networks. Furthermore, EV charging behavior, the layout and use of charging stations, and other factors influence the dynamics of traffic flow. Therefore, studying the spatiotemporal coupling optimization of transportation and power networks is of great significance. Traditionally, traffic flow and power load management are typically performed independently, lacking consideration of the complex interactions between the two. Traffic management systems primarily focus on optimizing vehicle flow, while power systems focus on load balancing and ensuring a stable power supply. However, EVs are not only vehicles but also mobile power load units. Therefore, effectively coordinating traffic flow and power load has become a pressing issue.

[0004] At present, the research on the coupling of transportation and power networks mainly focuses on the following aspects:

[0005] 1. Independent Optimization: Traditional research approaches typically separate traffic flow optimization and power load management, without considering their mutual impact. While this approach may achieve certain optimization effects in each area, it lacks overall system coordination and efficiency.

[0006] 2. Simple coupling models: Some studies have attempted to simply couple traffic and power networks, but most are static models that fail to dynamically reflect user behavior and network changes. These models often assume that traffic flows and power loads are fixed and lack the ability to respond to real-time data.

[0007] For example, patent publication number CN117649082A discloses a method for constructing a traffic-electricity coupling network model, using model reliability indicators calculated from real-time operating data of the coupling network and user-acceptable scheduling properties calculated from facial image information to determine the specific path for the scheduling solution. This method, in terms of user characteristics, calculates user scheduling properties solely based on collected facial image information, without considering the user's historical charging habits and driving style. Therefore, it is difficult to accurately capture user characteristics based solely on facial information.

[0008] Patent publication number CN117372057A proposes a differentiated regulation method. By acquiring network parameters, operating parameters, and scenario data from the power-transportation coupling system, it decouples price signals based on the regulation targets and sets independent price regulation signals for fuel vehicle routing guidance, electric vehicle routing guidance, and electric vehicle charging guidance (i.e., differentiated regulation). This method also fails to consider users' subjective driving and charging preferences, making it difficult to accurately guide users.

[0009] Therefore, existing optimization methods lack consideration of user behavior and fail to fully account for the complex behavioral interactions of users in transportation and electricity use, resulting in low system efficiency. In addition, most coupling models are static and cannot dynamically reflect user behavior and network changes, making it difficult to cope with complex situations in actual applications. The personalized needs and preferences of users are not fully quantified and weighted, resulting in deficiencies in the model in providing personalized services and adaptability. These shortcomings limit the coordinated optimization of transportation and power networks and the improvement of user experience. In response to these problems, the present invention proposes a transportation-power network spatiotemporal coupling optimization method and system based on user preference empowerment. Summary of the Invention

[0010] In response to the problems in the related art, the present invention proposes a transportation-power network spatiotemporal coupling optimization method and system based on user preference empowerment to overcome the above-mentioned technical problems existing in the existing related art.

[0011] This invention aims to address the challenges of existing urban transport-power optimization technologies involving electric vehicles, which face difficulties in accurately quantifying user preferences and lack effective modeling of the coupling between transport and power networks. Existing methods often ignore the heterogeneity of user behavior, resulting in insufficient personalization and adaptability. By quantifying and weighting user preferences, this invention optimizes the spatiotemporal coupling of transport and power networks, improving the efficiency of dynamic traffic flows and charging station queues, meeting personalized user needs while reducing the burden on transport and power systems.

[0012] To this end, the specific technical solutions adopted in the present invention are as follows:

[0013] According to one aspect of the present invention, a method for optimizing the spatiotemporal coupling of a transportation-power network based on user preference weighting is provided, comprising the following steps:

[0014] S1. Collecting charging and travel behavior preference data of electric vehicle users in a predetermined area and preprocessing the collected behavior preference data;

[0015] S2, based on the information-enhanced generative adversarial network deep learning architecture, models and quantifies electric vehicle user preferences and generates preference weight vectors with spatiotemporal characteristics;

[0016] S3. Use weighted directed graphs to build a transportation network model and a distribution network model. Based on the time-space mapping relationship, a transportation-power coupling network model is constructed.

[0017] S4. Construct a multi-objective optimization model. Based on the preference weights of electric vehicle users, use the optimization algorithm to solve the multi-objective optimization model to achieve dynamic optimization of charging scheduling and path planning.

[0018] Furthermore, the charging and travel behavior preference data includes charging behavior data, driving behavior data, traffic network data and power load data;

[0019] Among them, charging behavior data includes charging time, charging frequency, charging location, charging duration and charging amount; driving behavior data includes daily mileage, driving time and driving speed; traffic network data includes real-time road conditions and congestion status; power load data includes charging facility load and electricity price fluctuations.

[0020] Furthermore, the information-enhanced generative adversarial network deep learning architecture models and quantifies electric vehicle user preferences, and generates a preference weight vector with spatiotemporal characteristics, including the following steps:

[0021] S21. Take the multi-source heterogeneous time series data stream of electric vehicle users' charging behavior data, driving behavior data, traffic network data, and power load data as input data, and use the improved Informer encoder to extract key time series features from the input data;

[0022] S22. Construct a generative adversarial network with information enhancement function, wherein the generator is used to receive feature representation and random noise as input, generate user preference features through multi-layer nonlinear transformation, and the discriminator is used to distinguish the generated preference features from real user preference samples;

[0023] S23. Use the adaptive mapping layer to convert the preference features output by the generator into a standardized preference vector, and dynamically adjust the preference vector based on the real-time behavior data of the electric vehicle.

[0024] Furthermore, each time step in the multi-source heterogeneous time series data stream contains a charging behavior feature vector, a driving behavior feature vector, a traffic network state vector and a power load feature vector.

[0025] Furthermore, the improved Informer encoder is an encoder optimized by the attention mechanism and multi-head self-attention layer, which is used to efficiently capture the long-range spatiotemporal dependencies of charging mode, load changes, and location distribution.

[0026] Furthermore, the method of constructing a transportation network model using a weighted directed graph and establishing a distribution network model, and constructing a transportation-power coupling network model based on a time-space mapping relationship includes the following steps:

[0027] S31. Construct a transportation network model using a weighted directed graph, wherein nodes in the transportation network model represent transportation hubs, edges represent road connections, and weights include attributes of distance and travel time;

[0028] S32. Establish a distribution network model, where nodes in the distribution network model correspond to charging facilities, edges correspond to transmission lines, and attributes include capacity constraints and load distribution;

[0029] S33. Based on the transportation network model and the distribution network model, a transportation-power coupling network model is constructed in combination with the time-space mapping relationship to achieve the coordinated evolution of the transportation and power network states.

[0030] Furthermore, the expression of the traffic-power coupling network model is:

[0031] G={G PN , G TN , G PTN}

[0032] The expression of the distribution network model is:

[0033]

[0034] The expression of the traffic network model is:

[0035]

[0036] The expression of the coupled layer model is:

[0037]

[0038] In the formula, G, G PN , G TN , G PTN They represent the coupled network model, distribution network model, transportation network model and coupled layer model respectively, N PN 、E PN 、F PNThey represent the node set, line set and flow set of the distribution network respectively, i P represents the i-th node in the distribution network, n represents the total number of nodes in the distribution network, represents the section between nodes i and j in the distribution network, represents the line flow connecting node i and node j in the distribution network, N TN 、E TN 、F TN 、T TN They represent the node set, road section set, road section flow set and road section travel time set of the traffic network respectively, i T represents the i-th node in the transportation network, m represents the total number of nodes in the transportation network, represents the road segment between node i and node j in the traffic network, Indicates road section traffic flow, Indicates road section The travel time, E PTN represents the set of virtual connection lines between the distribution network and the transportation network, represents the segment between node i and node j in the coupled network, F PTN Represents the charging station operating status set, Represents the lines in the coupled network Virtual trends and traffic conditions.

[0039] Furthermore, the multi-objective optimization model is constructed, and based on the electric vehicle user preference weights, an optimization algorithm is used to solve the multi-objective optimization model to achieve dynamic optimization of charging scheduling and path planning, which includes the following steps:

[0040] S41. Establish a multi-objective optimization framework oriented towards maximizing system effectiveness and improving user satisfaction, and obtain a multi-objective optimization model;

[0041] S42, determining an adaptive optimization algorithm based on decision variables of preference weights, charging costs, and time consumption of electric vehicle users;

[0042] S43. Use an adaptive optimization algorithm to solve a multi-objective optimization model to achieve comprehensive optimization of charging path costs, traffic congestion levels, and grid operation costs.

[0043] Furthermore, the objective function of the multi-objective optimization model is expressed as:

[0044] min(αC user +βC congestion +γC grid )

[0045]

[0046] Where C user represents the charging path cost of electric vehicle users, C congestion Indicates the degree of traffic congestion, C grid represents the operating cost of the power grid, α, β, and γ represent the weight coefficients of the charging path cost of electric vehicle users, the weight coefficients of the traffic congestion degree, and the weight coefficients of the power grid operating cost, respectively. M represents the total number of electric vehicles participating in the optimization, (i, j) represents the edge connected by the two endpoints i and j in the traffic-power coupling network, and R n represents the charging path of the nth electric vehicle user, c ij represents the power transmission cost on edge (i, j). ij represents the traffic flow on edge (i, j), p k represents the electricity price of the kth charging station, q n，k represents the charging amount of the nth electric vehicle user at the kth charging station, E t represents the set of all edges in the transportation network, u ij represents the traffic capacity on edge (i, j), E e represents the set of all edges in the power network, P ij represents the transmission power on edge (i, j).

[0047] According to another aspect of the present invention, a transportation-power network spatiotemporal coupling optimization system based on user preference empowerment is provided, which includes a behavior preference data acquisition module, a preference weight vector generation module, a transportation power network modeling module, and a multi-objective optimization solution module;

[0048] The behavior preference data collection module is used to collect charging and travel behavior preference data of electric vehicle users in a predetermined area and pre-process the collected behavior preference data;

[0049] The preference weight vector generation module is used to model and quantify electric vehicle user preferences based on an information-enhanced generative adversarial network deep learning architecture, and generate a preference weight vector with spatiotemporal characteristics;

[0050] The transportation and power network modeling module is used to construct a transportation network model using a weighted directed graph, establish a distribution network model, and construct a transportation-power coupling network model based on a time-space mapping relationship;

[0051] The multi-objective optimization solution module is used to construct a multi-objective optimization model, and based on the electric vehicle user preference weights, uses an optimization algorithm to solve the multi-objective optimization model to achieve dynamic optimization of charging scheduling and path planning.

[0052] The beneficial effects of the present invention are:

[0053] 1) The present invention uses a generative adversarial network (Informer-GAN) deep learning architecture based on information enhancement to model and quantify user preferences, improve personalization and adaptability, and meet the needs of different users. At the same time, a dynamic coupling model of transportation and power networks is constructed to fully consider the interactive impact of the two and improve the overall system efficiency. Through dynamic optimization algorithms, electric vehicle charging paths are intelligently planned to achieve real-time optimization of traffic flows and charging station queues to adapt to changes in actual applications. Through these innovations, the present invention overcomes the shortcomings of the existing technology, achieves efficient coupling optimization of transportation and power networks, and improves the overall performance of the system and user satisfaction. Among similar methods, traditional algorithms based on collaborative filtering algorithms and matrix decomposition methods are difficult to handle the sparsity and temporal nature of data; deep learning methods based on fully connected neural networks and recurrent neural networks have low training efficiency, are prone to overfitting, and have poor interpretability; and traditional GANs and conditional GANs (CGANs) often show instability in data generation and are difficult to handle complex scenarios. The core advantage of Informer-GAN over traditional methods lies in: through the information enhancement mechanism and improved adversarial training architecture, it achieves deep fusion of multi-source data and dynamic feature extraction, which not only solves the training instability problem of traditional GAN, but also can more accurately capture and predict the dynamic preference changes of users in charging scenarios.

[0054] 2) From the perspective of coordinated optimization of transportation and power networks, the present invention quantifies user preferences and assigns weights through Informer-GAN, which can more effectively manage traffic flows and charging station queues. It can predict charging demand in different areas and time periods in advance, and guide some users to charging stations with lower loads in advance for popular charging stations. This not only alleviates traffic congestion around popular charging stations and reduces waiting time for vehicles in queues, but also avoids load peaks at charging stations and improves the utilization efficiency of charging facilities. At the same time, the optimized charging route planning reduces the invalid mileage of vehicles, reduces traffic energy consumption, realizes the coordinated optimization of the transportation network and the power network, and improves the operating efficiency of the overall system.

[0055] 3) In terms of adapting to complex and changeable actual scenarios, the user preference dynamic update mechanism constructed by the present invention is of great significance. As users' driving habits, travel needs and external environment (such as electricity price policy adjustments, traffic conditions) change, user preferences will change accordingly. Informer-GAN can capture these changes in real time and update the preference weight vector in a timely manner. For example, when a certain area causes traffic control and charging station demand changes due to large-scale events, the system can quickly adjust the charging path planning and charging resource allocation strategy to ensure that in complex and changeable situations, it can still meet the personalized needs of users and maintain the efficient and stable operation of the system.

[0056] 4) The Informer-GAN of the present invention excels in improving model prediction accuracy and computational efficiency. Compared to traditional methods, it can more accurately capture and predict the dynamic changes in user preferences in charging scenarios, providing a reliable decision-making basis for transportation-power network optimization. At the same time, its efficient computing performance makes real-time optimization and scheduling of large-scale networks possible, adapting to the large scale and high real-time requirements of modern urban transportation-power networks, providing strong technical support for the coordinated optimization of intelligent transportation systems and power dispatching systems, and effectively promoting the sustainable development of urban transportation-energy systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0058] Figure 1 is a schematic diagram of a method for optimizing the spatiotemporal coupling of transportation and power networks based on user preference weighting according to an embodiment of the present invention;

[0059] Figure 2 is a flow chart of a method for optimizing the spatiotemporal coupling of transportation and power networks based on user preference weighting according to an embodiment of the present invention;

[0060] Figure 3 This is a flowchart of solving a typical non-dominated sorting genetic algorithm II in a transportation-power network spatiotemporal coupling optimization method based on user preference empowerment according to an embodiment of the present invention. DETAILED DESCRIPTION

[0061] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention. They are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. By referring to these contents, ordinary technicians in this field should be able to understand other possible implementation methods and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0062] According to an embodiment of the present invention, a method and system for optimizing the spatiotemporal coupling of transportation and power networks based on user preference weighting are provided.

[0063] As electric vehicle penetration continues to climb in major cities, the coupling relationship between charging infrastructure load and transportation network traffic is becoming increasingly complex, necessitating the establishment of an efficient collaborative management mechanism. Electric vehicle users' charging station selection and route planning behaviors are influenced by multiple factors, including charging cost, time consumption, distance, and personalized needs. This paper constructs a spatiotemporal coupled optimization model for transportation-power networks based on the precise quantification and dynamic weighting of user preferences, achieving a synergistic improvement in system performance and user experience.

[0064] First, through questionnaire surveys, hardware perception and data mining and other multi-source channels, the charging and travel behavior data of electric vehicle users in the region are collected, and based on the improved Informer-GAN architecture, the user preferences are accurately quantified, and a preference weight vector with spatiotemporal characteristics is generated. At the same time, a dynamic update mechanism for user preferences is constructed to ensure the model's real-time response to personalized needs. Secondly, a weighted directed graph is used to construct a transportation network model, in which nodes represent transportation hubs, edges represent road connections, and weights include attributes such as distance and travel time; a distribution network model is established, in which nodes correspond to charging facilities and edges correspond to transmission lines, and attributes cover capacity constraints and load distribution; through the spatiotemporal mapping relationship, a dynamic coupling model of the transportation-power network is constructed to achieve the coordinated evolution of the dual network states. Finally, a multi-objective optimization framework oriented towards maximizing system efficiency and improving user satisfaction is established, comprehensively considering decision variables such as user preference weights, charging costs, and time consumption, and designing an adaptive optimization algorithm to achieve dynamic optimization of charging scheduling and path planning. This method effectively alleviates the problems of traffic congestion and charging load peaks through the precise modeling of user preferences and the coupled optimization of dual networks, and promotes the coordinated development of urban transportation-energy systems (its basic principles are as follows Figure 1 shown).

[0065] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Figure 2 As shown, according to one embodiment of the present invention, a method for optimizing the spatiotemporal coupling of transportation and power networks based on user preference weighting is provided, comprising the following steps:

[0066] S1. Collecting charging and travel behavior preference data of electric vehicle users in a predetermined area and preprocessing the collected behavior preference data;

[0067] Among them, electric vehicle user preference data collection includes:

[0068] Charging behavior data, including charging time, frequency, location, duration, and amount, are collected in real time through networked charging piles. Driving behavior data, including daily mileage, driving time, and speed, are collected using the on-board OBD (On-Board Diagnostics) interface and sensor system. GPS equipment and navigation systems are used to record users' precise spatiotemporal trajectories. Environmental data, including real-time traffic on the road network, load status at charging stations, and real-time electricity prices, are simultaneously acquired. The multi-source data are spatiotemporally aligned and exceptions are processed to establish a unified data fusion mechanism to achieve spatiotemporal consistency of the data.

[0069] Data preprocessing and feature extraction include:

[0070] The collected multi-source heterogeneous data is preprocessed, including outlier identification and removal based on statistical methods, missing data repair using a time series interpolation algorithm, and noise suppression based on wavelet transforms. The preprocessed data is input into an improved Informer encoder for feature extraction. The encoder, through the optimized design of an attention mechanism and a multi-head self-attention layer, efficiently captures long-range spatiotemporal dependencies such as charging patterns, load changes, and location distribution. Based on feature extraction, a generative adversarial network with an information enhancement mechanism is constructed, in which the generator is responsible for feature enhancement and the discriminator is used for feature verification. The representational capabilities of features are improved through adversarial training. Finally, an adaptive normalization algorithm is used to normalize the features to generate a standard feature vector set. This method overcomes the technical shortcomings of existing technologies in processing high-dimensional spatiotemporal data, such as high computational complexity and low feature extraction accuracy, significantly improving the efficiency and accuracy of feature extraction and laying the foundation for subsequent user modeling and network optimization.

[0071] S2, based on the information-enhanced generative adversarial network deep learning architecture, models and quantifies electric vehicle user preferences and generates preference weight vectors with spatiotemporal characteristics;

[0072] Informer-GAN is a new deep learning architecture whose innovation lies in the organic fusion of Informer's efficient long-sequence processing mechanism and GAN's generative adversarial characteristics. In this model, the generator adopts Informer's encoder-decoder structure, reducing computational complexity through a probabilistic sparse self-attention mechanism, while introducing a multi-layer distillation mechanism to achieve hierarchical extraction of key information. On the discriminator side, a multi-layer deep neural network structure is designed to improve the authenticity and accuracy of generated samples through adversarial learning strategies. This structural design is very suitable for user preference modeling in transportation-power coupled networks. It can not only accurately capture the spatiotemporal distribution characteristics of charging behavior patterns and travel choices, but also significantly improve computational efficiency while maintaining high prediction accuracy, providing strong technical support for real-time optimization and scheduling of large-scale networks.

[0073] The Informer-GAN-based user intelligent weighting method aims to accurately model and weight the behavior patterns of electric vehicle users through deep learning technology. Through intelligent weighting, the system can better meet the personalized needs of users and improve the efficiency of coordinated optimization of transportation and power networks.

[0074] Specifically, the information-enhanced generative adversarial network deep learning architecture models and quantifies electric vehicle user preferences, and generates a preference weight vector with spatiotemporal characteristics, including the following steps:

[0075] S21. Take the multi-source heterogeneous time series data stream of electric vehicle users' charging behavior data, driving behavior data, traffic network data, and power load data as input data, and use the improved Informer encoder to extract key time series features from the input data;

[0076] Informer time series feature extraction includes:

[0077] The system receives multi-source heterogeneous time series data streams X = {x1, x2, ..., x}, including user charging behavior data (charging time, frequency, location), driving behavior data (mileage, speed distribution), traffic network data (real-time road conditions, congestion status), and power load data (charging facility load, electricity price fluctuations). T}, where each time step x t It contains a charging behavior feature vector, a driving behavior feature vector, a traffic network state vector, and a power load feature vector. The input data is processed using an Informer encoder. An improved Informer encoder is used to process the input data. The encoder uses a probabilistic sparse attention mechanism to efficiently capture temporal dependencies in long sequence data. Specifically, the encoder extracts key temporal features from the data stream by calculating sparse attention weights between the query matrix Q, the key matrix K, and the value matrix V. The feature representation H output by the encoder contains temporal feature information of user behavior, providing input for the subsequent generative adversarial network.

[0078] S22. Construct a generative adversarial network with information enhancement function, wherein the generator is used to receive feature representation and random noise as input, generate user preference features through multi-layer nonlinear transformation, and the discriminator is used to distinguish the generated preference features from real user preference samples;

[0079] Generative adversarial network construction includes:

[0080] After obtaining the temporal feature representation H, a generative adversarial network with information enhancement function is constructed. Among them, the generator G wReceive feature representation H and random noise z~N(0,1) as input, where N represents the random noise obeying the standard normal distribution with mean 0 and variance 1, and generate user preference features through multi-layer nonlinear transformation; discriminator D w The generator is responsible for distinguishing the generated preference features from the real user preference samples. Through the adversarial training process, the generator's feature generation ability is continuously optimized, improving the authenticity and diversity of the generated samples.

[0081] S23. Use the adaptive mapping layer to convert the preference features output by the generator into a standardized preference vector, and dynamically adjust the preference vector based on the real-time behavior data of the electric vehicle.

[0082] The generation and update of dynamic preference vectors include:

[0083] First, the features output by the generator are converted into a standardized preference vector through an adaptive mapping layer. This vector contains the user's preference weights in terms of charging time, location selection, route planning, etc.; second, a dynamic update module based on time decay is constructed to adjust the preference vector in real time according to the latest user behavior data, ensuring that the preference model can respond to changes in user needs in a timely manner.

[0084] In addition, this embodiment also includes the verification and optimization of the preference model:

[0085] To ensure the accuracy of generated preferences, this paper employs a multi-objective evaluation mechanism to validate and optimize the model. Specifically, a comprehensive evaluation index system encompassing user satisfaction, prediction accuracy, and timeliness is established; model performance is evaluated through cross-validation; and model parameters are dynamically adjusted based on the validation results to continuously improve the quality of preference generation.

[0086] The preference vector generated by quantifying user preferences includes weights for charging time, location selection, and route planning. These weights can be incorporated into the charging guidance algorithm as decision variables during charging route planning. Based on the user's preference vector, a charging route that best meets their needs can be planned. This also optimizes the allocation of charging resources across the transportation-power network, reducing unnecessary traffic and charging costs. For example, if a user prefers to charge at a charging station close to home with lower electricity prices, route planning will prioritize charging stations that are closer to the user's current location and have electricity prices that meet their preferences. By quantifying and dynamically updating user preferences, the system can proactively predict charging demand in different areas and at different times. For charging stations with high user interest, some users can be directed to other, less-loaded charging stations in advance, avoiding congestion and load peaks at charging stations in that area. This improves charging station utilization efficiency, alleviates surrounding traffic congestion, and achieves coordinated optimization of the transportation-power network.

[0087] Through the above technical solution, accurate modeling and dynamic updating of user preferences are achieved, providing a reliable decision-making basis for subsequent transportation-power network optimization.

[0088] S3. Use weighted directed graphs to build a transportation network model and a distribution network model. Based on the time-space mapping relationship, a transportation-power coupling network model is constructed.

[0089] In order to achieve efficient coordinated optimization of the transportation-power network, build an optimization model that can comprehensively consider multiple objectives such as charging path costs, traffic congestion levels and grid operation costs, accurately reflect the actual network operation status and achieve improved system efficiency and user satisfaction, it is necessary to model the transportation-power coupling network. Through the basic data and constraints such as network structure, node and edge attributes, and interaction relationships provided by it, it supports the construction of objective functions and implementation of optimization algorithms of multi-objective optimization models.

[0090] The coupled power-transportation network consists of the distribution network layer, the transportation network layer, and the coupling layer, encompassing four entities: electric vehicles, charging stations, the transportation network, and the distribution network. Its mathematical model must comprehensively consider the characteristics of both the transportation and power networks and effectively integrate them. At the distribution network layer, charging demand and distribution network structure influence power flow distribution. The distribution network dynamically optimizes power flow based on node load and charging load forecasts and transmits updated marginal price signals to charging stations. Charging station prices and status influence users' charging decisions and route selection. At the transportation network layer, the state of the transportation network is related to user behavior. In an intelligent connected environment, demand information from the OD pair between the charging request point and the target charging station is used for real-time route selection. A traffic equilibrium model is used for route allocation and delayed reward backtracking, bringing the transportation network toward dynamic equilibrium. Changes in the transportation network are transmitted to the coupling layer, further influencing users' travel decisions (such as travel time, waiting time, charging time, and cost). Through this interaction, the coupled power-transportation network approaches equilibrium.

[0091] The method of constructing a traffic network model using a weighted directed graph and establishing a distribution network model, and constructing a traffic-power coupling network model based on a time-space mapping relationship includes the following steps:

[0092] S31. Construct a transportation network model using a weighted directed graph, wherein nodes in the transportation network model represent transportation hubs, edges represent road connections, and weights include attributes of distance and travel time;

[0093] S32. Establish a distribution network model, where nodes in the distribution network model correspond to charging facilities, edges correspond to transmission lines, and attributes include capacity constraints and load distribution;

[0094] S33. Based on the transportation network model and the distribution network model, a transportation-power coupling network model is constructed in combination with the time-space mapping relationship to achieve the coordinated evolution of the transportation and power network states.

[0095] Specifically, in order to further characterize the interactions within the coupled operation network, a traffic-power coupled network model is established, which is expressed as follows:

[0096] G={G PN , G TN , G PTN}

[0097] In the formula, G, G PN , G TN , G PTN They represent the coupled network model, distribution network model, transportation network model and coupled layer model respectively;

[0098] Specifically, the distribution network model is represented by graph theory as follows:

[0099]

[0100] Where N PN 、E PN 、F PN They represent the node set, line set and flow set of the distribution network respectively, i P represents the i-th node in the distribution network, n represents the total number of nodes in the distribution network, represents the section between nodes i and j in the distribution network, Represents the power flow of the line connecting node i and node j in the distribution network;

[0101] The transportation network model is represented by graph theory as follows:

[0102]

[0103] Where N TN 、E TN 、F TN 、T TN They represent the node set, road section set, road section flow set and road section travel time set of the traffic network respectively, i T represents the i-th node in the transportation network, m represents the total number of nodes in the transportation network, represents the road segment between node i and node j in the traffic network, Indicates road section traffic flow, Indicates road section The travel time;

[0104] The expression of the coupled layer model is:

[0105]

[0106] Where, E PTN Represents the set of virtual connection lines between the distribution network and the transportation network, representing the node location of the charging station in the two networks, represents the segment between node i and node j in the coupled network, F PTN Represents the charging station operating status set, Represents the lines in the coupled network Virtual trends and traffic conditions.

[0107] S4. Construct a multi-objective optimization model. Based on the preference weights of electric vehicle users, use the optimization algorithm to solve the multi-objective optimization model to achieve dynamic optimization of charging scheduling and path planning.

[0108] The multi-objective optimization model is constructed, and based on the electric vehicle user preference weights, an optimization algorithm is used to solve the multi-objective optimization model to achieve dynamic optimization of charging scheduling and path planning, which includes the following steps:

[0109] S41. Establish a multi-objective optimization framework oriented towards maximizing system effectiveness and improving user satisfaction, and obtain a multi-objective optimization model;

[0110] Specifically, the objective function of the multi-objective optimization model is expressed as:

[0111] min(αC user +βC congestion +γC grid )

[0112]

[0113] Where C user represents the charging path cost of electric vehicle users, C congestion Indicates the degree of traffic congestion, and uses the square of traffic flow to reflect the nonlinear growth of congestion. grid represents the operating cost of the power grid, α, β, and γ represent the weight coefficients of the charging path cost of electric vehicle users, the weight coefficient of the traffic congestion degree, and the weight coefficient of the power grid operating cost, respectively. They are used to balance the importance of different objectives and can be flexibly adjusted according to actual application scenarios and needs. M represents the total number of electric vehicles participating in the optimization, (i, j) represents the edge connected by the two endpoints i and j in the traffic-power coupling network, and R n represents the charging path of the nth electric vehicle user, c ij represents the power transmission cost on edge (i, j), f ij represents the traffic flow on edge (i, j), p k represents the electricity price of the kth charging station, q n，krepresents the charging amount of the nth electric vehicle user at the kth charging station, E t represents the set of all edges in the transportation network, u ij represents the traffic capacity on edge (i, j), that is, the maximum number of vehicles that the road section can accommodate in unit time, E e represents the set of all edges in the power network, P ij represents the transmission power on edge (i, j).

[0114] S42, determining an adaptive optimization algorithm based on decision variables of preference weights, charging costs, and time consumption of electric vehicle users;

[0115] The preference weight vector generated by the user preference weighting module represents the user's preferences for charging route, charging time, and charging station selection. Charging cost includes the user's charging fee at each charging station, calculated by multiplying the user's charging amount at each station by the electricity price at that station. Time consumption includes the travel time from the user's current location to the charging station and the charging time.

[0116] S43. Using an adaptive optimization algorithm to solve a multi-objective optimization model to achieve comprehensive optimization of charging path costs, traffic congestion levels, and grid operation costs;

[0117] Specifically, in this embodiment, a typical non-dominated sorting genetic algorithm II (NSGA-II) is used to solve the problem. The process is as follows: Figure 3 shown.

[0118] According to another embodiment of the present invention, a transportation-power network spatiotemporal coupling optimization system based on user preference empowerment is provided, which includes a behavior preference data acquisition module, a preference weight vector generation module, a transportation power network modeling module, and a multi-objective optimization solution module;

[0119] The behavior preference data collection module is used to collect charging and travel behavior preference data of electric vehicle users in a predetermined area and pre-process the collected behavior preference data;

[0120] The preference weight vector generation module is used to model and quantify electric vehicle user preferences based on an information-enhanced generative adversarial network deep learning architecture, and generate a preference weight vector with spatiotemporal characteristics;

[0121] The transportation and power network modeling module is used to construct a transportation network model using a weighted directed graph, establish a distribution network model, and construct a transportation-power coupling network model based on a time-space mapping relationship;

[0122] The multi-objective optimization solution module is used to construct a multi-objective optimization model, and based on the electric vehicle user preference weights, uses an optimization algorithm to solve the multi-objective optimization model to achieve dynamic optimization of charging scheduling and path planning.

[0123] In summary, with the help of the above technical solutions of the present invention, user preferences are modeled and quantified through a deep learning architecture based on information enhancement generative adversarial networks (Informer-GAN), thereby improving personalization and adaptability and meeting the needs of different users. At the same time, a dynamic coupling model of transportation and power networks is constructed to comprehensively consider the interaction between the two and improve the overall system efficiency. Through dynamic optimization algorithms, electric vehicle charging paths are intelligently planned to achieve real-time optimization of traffic flows and charging station queues to adapt to changes in actual applications. Through these innovations, the present invention overcomes the shortcomings of the existing technologies, achieves efficient coupling optimization of transportation and power networks, and improves the overall performance of the system and user satisfaction. Among similar methods, traditional algorithms based on collaborative filtering algorithms and matrix decomposition methods are difficult to handle the sparsity and temporal nature of data; deep learning methods based on fully connected neural networks and recurrent neural networks have low training efficiency, are prone to overfitting, and have poor interpretability; and traditional GANs and conditional GANs (CGANs) often exhibit instability in data generation and are difficult to handle complex scenarios. The core advantage of Informer-GAN over traditional methods lies in: through the information enhancement mechanism and improved adversarial training architecture, it achieves deep fusion of multi-source data and dynamic feature extraction, which not only solves the training instability problem of traditional GAN, but also can more accurately capture and predict the dynamic preference changes of users in charging scenarios.

[0124] In addition, from the perspective of coordinated optimization of transportation and power networks, the present invention quantifies user preferences and assigns weights through Informer-GAN, which can more effectively manage traffic flows and charging station queues. It can predict charging demand in different areas and time periods in advance, and guide some users to charging stations with lower loads in advance for popular charging stations. This not only alleviates traffic congestion around popular charging stations and reduces waiting time for vehicles in queues, but also avoids load peaks at charging stations and improves the utilization efficiency of charging facilities. At the same time, the optimized charging route planning reduces the invalid mileage of vehicles, reduces traffic energy consumption, realizes the coordinated optimization of the transportation network and the power network, and improves the operating efficiency of the overall system.

[0125] In addition, the dynamic update mechanism of user preferences constructed by the present invention is of great significance in adapting to complex and changeable actual scenarios. As users' driving habits, travel needs and external environment (such as electricity price policy adjustments, traffic conditions) change, user preferences will change accordingly. Informer-GAN can capture these changes in real time and update the preference weight vector in a timely manner. For example, when a certain area causes changes in traffic control and charging station demand due to large-scale events, the system can quickly adjust the charging path planning and charging resource allocation strategy to ensure that in complex and changeable situations, it can still meet the personalized needs of users and maintain the efficient and stable operation of the system.

[0126] Furthermore, the Informer-GAN proposed in this paper excels in improving model prediction accuracy and computational efficiency. Compared to traditional methods, it can more accurately capture and predict dynamic changes in user preferences in charging scenarios, providing a reliable decision-making basis for transportation-power network optimization. At the same time, its efficient computing performance enables real-time optimization and scheduling of large-scale networks, adapting to the large scale and high real-time requirements of modern urban transportation-power networks. It provides strong technical support for the coordinated optimization of intelligent transportation systems and power dispatching systems, and effectively promotes the sustainable development of urban transportation-energy systems.

[0127] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for optimizing the spatiotemporal coupling of transportation and power networks based on user preference weighting, characterized in that: The following steps are involved: S1. Collecting charging and travel behavior preference data of electric vehicle users in a predetermined area and preprocessing the collected behavior preference data; S2, based on the information-enhanced generative adversarial network deep learning architecture, models and quantifies electric vehicle user preferences and generates preference weight vectors with spatiotemporal characteristics; S3. Use weighted directed graphs to build a transportation network model and a distribution network model. Based on the time-space mapping relationship, a transportation-power coupling network model is constructed. S4. Construct a multi-objective optimization model. Based on the preference weights of electric vehicle users, use the optimization algorithm to solve the multi-objective optimization model to achieve dynamic optimization of charging scheduling and path planning.

2. The method for optimizing the spatiotemporal coupling of transportation and power networks based on user preference weighting according to claim 1 is characterized in that: The charging and travel behavior preference data includes charging behavior data, driving behavior data, traffic network data and power load data; Among them, charging behavior data includes charging time, charging frequency, charging location, charging duration and charging amount; driving behavior data includes daily mileage, driving time and driving speed; traffic network data includes real-time road conditions and congestion status; power load data includes charging facility load and electricity price fluctuations.

3. The method for optimizing the spatiotemporal coupling of transportation and power networks based on user preference weighting according to claim 1 is characterized in that: The information-enhanced generative adversarial network deep learning architecture is used to model and quantify electric vehicle user preferences and generate a preference weight vector with spatiotemporal characteristics, including the following steps: S21. Take the multi-source heterogeneous time series data stream of electric vehicle users' charging behavior data, driving behavior data, traffic network data, and power load data as input data, and use the improved Informer encoder to extract key time series features from the input data; S22. Construct a generative adversarial network with information enhancement function, wherein the generator is used to receive feature representation and random noise as input, generate user preference features through multi-layer nonlinear transformation, and the discriminator is used to distinguish the generated preference features from real user preference samples; S23. Use the adaptive mapping layer to convert the preference features output by the generator into a standardized preference vector, and dynamically adjust the preference vector based on the real-time behavior data of the electric vehicle.

4. The method for optimizing the spatiotemporal coupling of transportation and power networks based on user preference weighting according to claim 3 is characterized in that: Each time step in the multi-source heterogeneous time series data stream contains a charging behavior feature vector, a driving behavior feature vector, a traffic network state vector and a power load feature vector.

5. The method for optimizing the spatiotemporal coupling of transportation and power networks based on user preference weighting according to claim 3 is characterized in that: The improved Informer encoder is an encoder optimized by the attention mechanism and multi-head self-attention layer, which is used to efficiently capture the long-range spatiotemporal dependencies of charging mode, load changes, and location distribution.

6. The method for optimizing the spatiotemporal coupling of transportation and power networks based on user preference weighting according to claim 1, characterized in that: The method of constructing a traffic network model using a weighted directed graph and establishing a distribution network model, and constructing a traffic-power coupling network model based on a time-space mapping relationship includes the following steps: S31. Construct a transportation network model using a weighted directed graph, wherein nodes in the transportation network model represent transportation hubs, edges represent road connections, and weights include attributes of distance and travel time; S32. Establish a distribution network model, where nodes in the distribution network model correspond to charging facilities, edges correspond to transmission lines, and attributes include capacity constraints and load distribution; S33. Based on the transportation network model and the distribution network model, a transportation-power coupling network model is constructed in combination with the time-space mapping relationship to achieve the coordinated evolution of the transportation and power network states.

7. The method for optimizing the spatiotemporal coupling of transportation and power networks based on user preference weighting according to claim 6, characterized in that: The expression of the traffic-power coupling network model is: G={G PN ,G TN ,G PTN } The expression of the distribution network model is: The expression of the traffic network model is: The expression of the coupled layer model is: In the formula, G, G PN , G TN , G PTN They represent the coupled network model, distribution network model, transportation network model and coupled layer model respectively, NP N 、E PN 、F PN They represent the node set, line set and flow set of the distribution network respectively, i P represents the i-th node in the distribution network, n represents the total number of nodes in the distribution network, represents the section between nodes i and j in the distribution network, represents the line flow connecting node i and node j in the distribution network, N TN 、E TN 、F TN 、T TN They represent the node set, road section set, road section flow set and road section travel time set of the traffic network respectively, i T represents the i-th node in the transportation network, m represents the total number of nodes in the transportation network, represents the road segment between node i and node j in the traffic network, Indicates road section traffic flow, Indicates road section The travel time, E PTN represents the set of virtual connection lines between the distribution network and the transportation network, represents the segment between node i and node j in the coupled network, F PTN Represents the charging station operating status set, Represents the lines in the coupled network Virtual trends and traffic conditions.

8. The method for optimizing the spatiotemporal coupling of transportation and power networks based on user preference weighting according to claim 1 is characterized in that: The multi-objective optimization model is constructed, and based on the electric vehicle user preference weights, an optimization algorithm is used to solve the multi-objective optimization model to achieve dynamic optimization of charging scheduling and path planning, which includes the following steps: S41. Establish a multi-objective optimization framework oriented towards maximizing system effectiveness and improving user satisfaction, and obtain a multi-objective optimization model; S42, determining an adaptive optimization algorithm based on decision variables of preference weights, charging costs, and time consumption of electric vehicle users; S43. Use an adaptive optimization algorithm to solve a multi-objective optimization model to achieve comprehensive optimization of charging path costs, traffic congestion levels, and grid operation costs.

9. The method for optimizing the spatiotemporal coupling of transportation and power networks based on user preference weighting according to claim 8, characterized in that: The objective function of the multi-objective optimization model is expressed as: min(αC user +βC congestion +γC grid ) Where C user represents the charging path cost of electric vehicle users, C congestion Indicates the degree of traffic congestion, C grid represents the operating cost of the power grid, α, β, and γ represent the weight coefficients of the charging path cost of electric vehicle users, the weight coefficients of the traffic congestion degree, and the weight coefficients of the power grid operating cost, respectively. M represents the total number of electric vehicles participating in the optimization, (i, j) represents the edge connected by the two endpoints i and j in the traffic-power coupling network, and R n represents the charging path of the nth electric vehicle user, c ij represents the power transmission cost on edge (i, j), f ij represents the traffic flow on edge (i, j), p k represents the electricity price of the kth charging station, q n，k represents the charging amount of the nth electric vehicle user at the kth charging station, E t represents the set of all edges in the transportation network, u ij represents the traffic capacity on edge (i, j), E e represents the set of all edges in the power network, P ij represents the transmission power on edge (i, j).

10. A transportation-power network spatiotemporal coupling optimization system based on user preference weighting, used to implement the steps of the transportation-power network spatiotemporal coupling optimization method based on user preference weighting according to any one of claims 1 to 9, characterized in that: The system includes a behavior preference data acquisition module, a preference weight vector generation module, a traffic and power network modeling module, and a multi-objective optimization solution module; The behavior preference data collection module is used to collect charging and travel behavior preference data of electric vehicle users in a predetermined area and pre-process the collected behavior preference data; The preference weight vector generation module is used to model and quantify electric vehicle user preferences based on an information-enhanced generative adversarial network deep learning architecture, and generate a preference weight vector with spatiotemporal characteristics; The transportation and power network modeling module is used to construct a transportation network model using a weighted directed graph, establish a distribution network model, and construct a transportation-power coupling network model based on a time-space mapping relationship; The multi-objective optimization solution module is used to construct a multi-objective optimization model, and based on the electric vehicle user preference weights, uses an optimization algorithm to solve the multi-objective optimization model to achieve dynamic optimization of charging scheduling and path planning.

Images