Travel comprehensive cost optimization method, system and equipment based on multi-modal large model

Abstract

The invention relates to the technical field of automobile travel service, discloses a travel comprehensive cost optimization method, system and equipment based on a multi-modal large model, and aims to solve the problems of poor travel experience and high comprehensive cost caused by difficulty in considering charging site selection, accommodation hotel screening and travel cost management and control during long-distance travel. The method comprises the steps of collecting travel associated data based on travel demand information; constructing a comprehensive cost function including charging cost, hotel cost and travel time cost, and defining a weight coefficient based on user preference learning; calculating charging cost, hotel cost and travel time cost in the comprehensive cost function according to travel association data; and based on a calculation result of the comprehensive cost function, screening, sorting and optimizing the multiple groups of charging-hotel-path combination schemes by adopting a decision inference engine, generating an optimal electric travel scheme, and outputting the optimal electric travel scheme to a vehicle-mounted terminal for a vehicle owner to select, so that the comprehensive cost optimization and travel experience improvement of the long-distance travel of the vehicle are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automobile travel service, and in particular to a travel comprehensive cost optimization method, system and device based on a multi-modal large model. BACKGROUND

[0002] In recent years, the new energy vehicle industry has developed rapidly, and the number of vehicles in use has continued to rise, from 4.92 million in 2020 to 36.89 million in 2025, with the proportion of motor vehicles breaking through 10%. The proportion of new energy electric vehicle registrations in the first half of 2025 is close to 45% of the total number of new vehicle registrations. New energy vehicles have gradually entered ordinary families and become an important means of transportation for daily and long-distance travel. With the increasing popularity of new energy vehicles, the demand for long-distance travel is increasing, but the current long-distance travel of new energy vehicles still faces many pain points, which restricts the travel experience and cost control of vehicle owners.

[0003] In the process of long-distance travel of new energy vehicles, the owner needs to consider charging and accommodation requirements at the same time, and the matching degree, price fluctuation, location distribution of charging resources and accommodation resources, and the time loss caused by path selection will all affect the comprehensive cost and experience of travel. In the prior art, new energy vehicle travel-related services are mostly single-dimensional services, such as charging pile query services that only provide charging pile locations and real-time availability, without associating with surrounding accommodation resources; hotel reservation services that only focus on accommodation conditions and prices, without considering distance adaptation with charging piles; path planning services that mostly target the shortest distance or time, without incorporating charging and accommodation needs, resulting in the need for vehicle owners to manually switch between multiple service platforms to screen charging piles, hotels and paths, making the decision-making process cumbersome and difficult to balance charging costs, accommodation costs and time costs, which can easily lead to long waiting times for charging piles, excessive distance between hotels and charging piles requiring detours, high accommodation prices or poor experiences, and ultimately result in increased comprehensive travel costs and poor travel experiences.

[0004] Some existing technologies attempt to integrate charging and path planning functions, but are based on static electricity prices and fixed path parameters for planning, without considering the influence of temporal and spatial dynamic fluctuations in electricity prices, charging pile load rate changes, hotel price adjustments, holidays, weather and other environmental factors on travel, resulting in insufficient flexibility in planning results and an inability to adapt to dynamically changing travel scenarios. At the same time, existing technologies lack adaptation to user preferences, with fixed weights uniformly used to calculate travel costs, which cannot meet the differentiated needs of different users for charging costs, accommodation experiences and travel times, resulting in insufficient targeting of optimization solutions and an inability to truly achieve comprehensive cost optimization and travel experience improvement.

[0005] In addition, the processing efficiency of multi-dimensional travel data in the prior art is low, the multi-type data modalities such as charging piles, hotels, paths and environment are greatly different, and it is difficult to efficiently fuse and analyze, resulting in slow scheme generation and unable to respond to the travel decision-making needs of vehicle owners in time; and the scheme screening optimization only uses single sorting or simple calculation method, without combining multi-objective optimization algorithm, and the optimal scheme in terms of comprehensive cost and adaptability cannot be accurately screened, further affecting the travel experience of vehicle owners.

[0006] In view of the problems existing in the prior art, there is an urgent need for an electric travel comprehensive cost optimization method capable of fusing multi-modal data, adapting to dynamic travel scenarios and taking into account user preferences. SUMMARY

[0007] In order to overcome the defects of single long-distance travel service, poor dynamic adaptability, insufficient user preference adaptation and difficult comprehensive cost control of new energy vehicles in the prior art, the present application provides a travel comprehensive cost optimization method and system based on a multi-modal large model, which realizes the coordinated planning of charging, accommodation and path, automatically generates an optimal travel scheme in terms of comprehensive cost, reduces the difficulty of travel decision-making of vehicle owners, and automatically matches the charging, accommodation and path schemes that meet the demand without the need for vehicle owners to manually screen multiple types of travel-related resources, greatly reducing the difficulty of travel decision-making, saving the comprehensive cost of travel, and adapting to the long-distance travel scenario demand under the background of popularization of new energy vehicles.

[0008] To achieve the above-mentioned purposes, the present application adopts the following technical solutions:

[0009] In a first aspect, the present application provides a travel comprehensive cost optimization method based on a multi-modal large model, which comprises:

[0010] obtaining travel demand information of a vehicle owner;

[0011] collecting travel-related data based on the travel demand information;

[0012] constructing a comprehensive cost function taking into account charging cost, hotel cost and travel time cost, and defining a weight coefficient based on user preference learning;

[0013] calculating the charging cost, hotel cost and travel time cost in the comprehensive cost function according to the travel-related data;

[0014] based on the calculation results of the comprehensive cost function, using a decision reasoning engine to screen, sort and optimize multiple charging-hotel-path combination schemes, generating an optimal electric travel scheme, and outputting the optimal electric travel scheme to a vehicle terminal for selection by the vehicle owner.

[0015] Optionally, the obtaining of the travel demand information of the vehicle owner comprises: receiving voice travel demand input by the vehicle owner through the vehicle terminal to the multi-modal large model, performing voice recognition and semantic analysis on the voice travel demand, extracting target city information, and simultaneously obtaining vehicle current state information uploaded by the vehicle terminal in synchronization; wherein the vehicle current state information comprises vehicle remaining power information and vehicle current position information.

[0016] The vehicle remaining power information and the vehicle current position information are associated and bound with the target city information to determine the travel demand information.

[0017] Optionally, the collecting of the travel-related data based on the travel demand information comprises: calling a charging pile data interface, a hotel data interface, a map service interface and a space-time environment data interface through the large model cloud service to respectively collect charging pile-related data, hotel-related data, path-related data and space-time environment-related data related to the travel demand information, performing format standardization processing on the collected multiple types of data, and eliminating invalid redundant data to obtain processed travel-related data.

[0018] The travel-related data comprises charging pile-related data, hotel-related data, space-time environment-related data and path-related data.

[0019] The charging pile-related data comprises charging pile position information, charging pile real-time load rate and charging pile charging standard; the hotel-related data comprises hotel position information, hotel charging information, hotel evaluation text and hotel facility picture.

[0020] The space-time environment-related data comprises weather information and holiday information of the target city and the regions along the way.

[0021] The path-related data comprises multiple candidate path information from the current position of the vehicle to the target city, distance information of each candidate path and facility distribution information along the way.

[0022] Optionally, the construction of a comprehensive cost function taking into account charging cost, hotel cost and travel time cost, and the definition of a weight coefficient based on user preference learning comprise: defining a weight coefficient with charging cost, hotel cost and travel time cost as dimensions through a multi-modal large model in combination with charging pile-related data, hotel-related data, space-time environment-related data and path-related data, and constructing a comprehensive cost function through the following formula:

[0023] Total_Cost = α × Charging_Cost + β × Hotel_Cost + γ × Travel_Cost.

[0024] Wherein, a is the charging cost weight coefficient, β is the hotel cost weight coefficient, γ is the travel time cost weight coefficient, a+β+γ=1, Charging_Cost is the charging cost, Hotel_Cost is the hotel cost, and Travel_Cost is the travel time cost.

[0025] The historical travel decision data and preference feedback data of the user are collected, the multi-modal large model is used to learn the attention priority of the user to the charging cost, the hotel cost and the travel time cost, the values of a, β and γ are dynamically adjusted, and the weight coefficients suitable for the current user are generated.

[0026] Optionally, the calculation of the charging cost comprises: obtaining historical electricity price data, charging pile peak and valley period data, charging pile load rate data in charging pile associated data, and weather information, holiday information in space-time environment associated data as training data; a space-time price prediction model is constructed through a TimeSformer model, and the space-time price prediction model is trained using the training data until the model prediction accuracy meets a preset threshold;

[0027] The real-time load rate of the charging pile, the current weather information and the current holiday information are input into the trained space-time price prediction model, and the dynamic electricity price of the charging pile and the real-time parking fee of the charging pile are predicted; the vehicle required charging quantity is calculated based on the vehicle remaining electric quantity and the target driving distance, and the charging cost is determined based on the predicted dynamic electricity price of the charging pile, the required charging quantity and the real-time parking fee of the charging pile.

[0028] Optionally, the calculation of the hotel cost comprises: obtaining hotel location information, hotel charging information, hotel evaluation text and hotel facility pictures in hotel associated data; the hotel evaluation text is analyzed by a multi-modal large model to extract hotel experience related features; the visual feature extraction is performed on the hotel facility pictures to evaluate the hotel facility adaptation degree;

[0029] The hotel emergency coefficient is set in combination with the distance relationship between the hotel location and the charging pile location; the hotel emergency coefficient is negatively correlated with the distance from the hotel to the charging pile;

[0030] The hotel dynamic price is output based on the space-time price prediction model, and the hotel cost is calculated in combination with the hotel emergency coefficient.

[0031] Optionally, the calculation of the travel time cost comprises: the charging pile location,

[0032] The hotel location is fused into the path planning network as a path node, and a plurality of groups of candidate paths containing charging nodes and accommodation nodes are generated based on the total length of the path, the detour distance and the node resource adaptation degree as constraint conditions, and a path planning engine is constructed;

[0033] The input path association data is analyzed by a path planning engine to analyze a detour distance corresponding to each candidate path; the detour distance is an extra driving distance of the candidate path deviating from a direct path;

[0034] A time value coefficient is set, and a travel time cost is calculated based on the detour distance, a preset driving speed, and the time value coefficient.

[0035] Optionally, the screening, sorting, and optimization of the multiple groups of charging-hotel-path combination schemes by the decision reasoning engine include: substituting the calculated charging cost, hotel cost, and travel time cost into a comprehensive cost function to obtain initial values of comprehensive costs corresponding to the multiple groups of charging-hotel-path combination schemes; pre-screening the multiple groups of combination schemes by spatial indexing to eliminate combination schemes that do not meet preset spatial constraint conditions; wherein the spatial constraint conditions include distance constraints of charging piles and hotels and direction constraints of paths and target cities.

[0036] The combination schemes after the pre-screening are coarsely sorted, the approximate values of the comprehensive costs of the combination schemes are calculated, and the combination schemes are prioritized based on the approximate values; the combination schemes after the coarse sorting are finely optimized, the combination schemes are modeled and solved by using a discrete optimization algorithm, and the solving results are secondarily optimized by using a continuous optimization algorithm to obtain accurate values of the comprehensive costs of the combination schemes; the combination schemes are finally sorted based on the accurate values of the comprehensive costs, and a combination scheme with the optimal accurate value of the comprehensive cost is selected as an optimal travel comprehensive scheme.

[0037] In a second aspect, the present application provides a travel comprehensive cost optimization system based on a multi-modal large model, the system comprising:

[0038] An information acquisition module is configured to acquire travel demand information of a vehicle owner.

[0039] A data acquisition module is configured to acquire travel association data based on the travel demand information.

[0040] A construction module is configured to construct a comprehensive cost function taking into account charging costs, hotel costs, and travel time costs, and to define weight coefficients based on user preference learning.

[0041] A cost calculation module is configured to calculate the charging costs, hotel costs, and travel time costs in the comprehensive cost function according to the travel association data.

[0042] A travel scheme formulation module is configured to screen, sort, and optimize multiple groups of charging-hotel-path combination schemes by using a decision reasoning engine based on the calculation results of the comprehensive cost function, to generate an optimal electric travel scheme, and to output the optimal electric travel scheme to a vehicle-mounted terminal for selection by the vehicle owner.

[0043] In a third aspect, the present application provides an electronic device, comprising:

[0044] at least one processor; and

[0045] a memory in communication with the at least one processor; wherein

[0046] the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the method of any one of the first aspect.

[0047] Compared with the closest prior art, the present application has the beneficial effects of:

[0048] The travel comprehensive cost optimization method, system and device based on a multi-modal large model proposed in the present application can integrate three core travel demands of charging, accommodation and path, integrate multi-dimensional data through a multi-modal large model, and generate an integrated comprehensive travel scheme, without the need for the vehicle owner to manually switch between multiple service platforms, greatly reducing the difficulty of travel decision-making and improving decision-making efficiency. Secondly, the TimeSformer model is used to construct a spatio-temporal price prediction model, dynamic factors such as weather, holidays, and charging pile load rate are integrated, fluctuating data such as electricity prices and hotel prices are accurately predicted, and a graph neural network is combined to plan a path in real time, so that the scheme dynamically adapts to changes in the travel scenario, avoiding the poor adaptability problem caused by static planning.

[0049] The present application learns user historical travel preferences through a multi-modal large model, dynamically adjusts the weight coefficients of the comprehensive cost function, generates a targeted optimization scheme, meets the differentiated needs of different users for cost, time and experience, and improves the adaptability of the scheme and user satisfaction.

[0050] Combined with spatial indexing and multi-objective optimization algorithms, the present application not only reduces invalid calculation amount and improves scheme generation speed, but also ensures the accuracy of comprehensive cost calculation, and filters out the truly optimal travel scheme. By accurately predicting dynamic prices, optimizing paths to reduce detours, and matching charging and accommodation resources, the present application can effectively reduce charging costs, hotel costs and time loss conversion costs, achieve comprehensive cost optimization for long-distance travel of new energy vehicles, improve travel experience, adapt to the demand for long-distance travel scenarios after the popularization of new energy vehicles, and has a wide application prospect. BRIEF DESCRIPTION OF DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the specific embodiments or prior art of the present application, the drawings needed in the specific embodiments or prior art description will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, each element or part is not necessarily drawn according to the actual scale.

[0052] Figure 1 is a flow chart of a travel comprehensive cost optimization method based on a multi-modal large model provided by the present application;

[0053] Figure 2 is a flow chart of the overall architecture of the electric travel comprehensive cost optimization method based on a multi-modal large model of the present application;

[0054] Figure 3 is a data interaction architecture diagram of the present application;

[0055] Figure 4 is a decision reasoning engine optimization flow chart of the present application;

[0056] Figure 5 is a scheme generation and output flow chart of the present application;

[0057] Figure 6 is a system structure schematic diagram of a travel comprehensive cost optimization method based on a multi-modal large model provided by the present application;

[0058] Figure 7 is an internal structure diagram of an electronic device provided by the present application. DETAILED DESCRIPTION

[0059] The embodiments of the technical solutions of the present application will be described in detail below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present application, and therefore only serve as examples, and cannot limit the protection scope of the present application.

[0060] It should be noted that, unless otherwise specified, the technical terms or scientific terms used in the present application should be understood as the usual meaning understood by the skilled person in the field to which the present application belongs.

[0061] The present application provides a travel comprehensive cost optimization method, system and device based on a multi-modal large model. It is suitable for the collaborative planning and comprehensive cost control of charging, accommodation and path in the long-distance travel scenario of new energy vehicle owners.

[0062] Please refer to Figure 1 The present application provides a travel comprehensive cost optimization method based on a multi-modal large model, which specifically includes the following steps:

[0063] S101 obtaining travel demand information of the vehicle owner;

[0064] S102 collecting travel associated data based on the travel demand information;

[0065] S103 constructing a comprehensive cost function taking into account charging cost, hotel cost and travel time cost, and defining a weight coefficient based on user preference learning;

[0066] S104 calculates the charging cost, hotel cost and travel time cost in the comprehensive cost function according to the travel associated data;

[0067] S105, based on the calculation result of the comprehensive cost function, uses a decision-making reasoning engine to screen, sort and optimize a plurality of charging-hotel-path combination schemes, generates an optimal electric travel scheme, and outputs the optimal electric travel scheme to a vehicle terminal for selection by the vehicle owner.

[0068] In the above step S101, the obtaining of the travel demand information of the vehicle owner includes: receiving a voice travel demand input by the vehicle owner through the vehicle terminal to the multi-modal large model, performing voice recognition and semantic analysis on the voice travel demand, extracting target city information, and simultaneously obtaining vehicle current state information uploaded by the vehicle terminal; wherein the vehicle current state information includes vehicle remaining power information and vehicle current location information;

[0069] The vehicle remaining power information, the vehicle current location information and the target city information are associated and bound to determine the travel demand information.

[0070] In the above step S102, the collecting of the travel associated data based on the travel demand information includes: calling charging pile data interface, hotel data interface, map service interface and space-time environment data interface through the large model cloud service to collect charging pile associated data, hotel associated data, path associated data and space-time environment associated data related to the travel demand information, performing format standardization processing on the collected multiple types of data, eliminating invalid redundant data, and obtaining processed travel associated data;

[0071] The travel associated data includes charging pile associated data, hotel associated data, space-time environment associated data and path associated data.

[0072] The charging pile associated data includes charging pile location information, charging pile real-time load rate and charging pile charging standard; the hotel associated data includes hotel location information, hotel charging information, hotel evaluation text and hotel facility picture;

[0073] The space-time environment associated data includes weather information and holiday information of the target city and the regions along the way.

[0074] The path associated data includes multiple candidate path information from the vehicle current location to the target city, distance information of each candidate path and facility distribution information along the way.

[0075] In the step S103, the comprehensive cost function taking into account the charging cost, the hotel cost and the travel time cost is constructed, and the weight coefficient based on user preference learning is defined, including: defining the weight coefficient in the dimension of the charging cost, the hotel cost and the travel time cost through a multi-modal large model in combination with charging pile associated data, hotel associated data, time and space environment associated data and path associated data, and constructing the comprehensive cost function through the following formula:

[0076] Total_Cost = a * Charging_Cost + b * Hotel_Cost + g * Travel_Cost;

[0077] Wherein, a is the charging cost weight coefficient, b is the hotel cost weight coefficient, g is the travel time cost weight coefficient, a+b+g = 1, Charging_Cost is the charging cost, Hotel_Cost is the hotel cost, and Travel_Cost is the travel time cost.

[0078] The user historical travel decision data and preference feedback data are collected, the multi-modal large model is used to learn the attention priority of the user to the charging cost, the hotel cost and the travel time cost, the values of a, b and g are dynamically adjusted, and the weight coefficient suitable for the current user is generated.

[0079] In the step S104, the charging cost is calculated, including: obtaining historical electricity price data, charging pile peak and valley period data and charging pile load rate data in the charging pile associated data, and weather information and holiday information in the time and space environment associated data as training data; a time and space price prediction model is constructed through a TimeSformer model, the training data is used to train the time and space price prediction model until the prediction accuracy of the model meets a preset threshold;

[0080] The real-time load rate of the charging pile, the current weather information and the current holiday information are input into the trained time and space price prediction model, the dynamic electricity price of the charging pile and the real-time parking fee of the charging pile are predicted, the required charging amount of the vehicle is calculated in combination with the remaining electric quantity and the target driving distance of the vehicle, and the charging cost is determined based on the predicted dynamic electricity price of the charging pile, the required charging amount and the real-time parking fee of the charging pile.

[0081] Further, the hotel cost is calculated, including: obtaining hotel location information, hotel charging information, hotel evaluation text and hotel facility pictures in the hotel associated data; the multi-modal large model is used to perform semantic analysis on the hotel evaluation text to extract hotel experience related features; the visual feature extraction is performed on the hotel facility pictures to evaluate the hotel facility adaptation degree;

[0082] The hotel emergency coefficient is set in combination with the distance relationship between the hotel location and the charging pile location; the hotel emergency coefficient is negatively correlated with the distance from the hotel to the charging pile.

[0083] The hotel dynamic price is output based on the space-time price prediction model, and the hotel cost is calculated in combination with the hotel emergency coefficient.

[0084] Further, the travel time cost includes: charging pile position,

[0085] The hotel location is integrated into the path planning network as a path node, and a plurality of groups of candidate paths containing charging nodes and accommodation nodes are generated under the constraint conditions of path total length, detour distance and node resource adaptation degree, and a path planning engine is constructed;

[0086] The path association data is input, and the detour distance corresponding to each candidate path is analyzed by the path planning engine; the detour distance is the additional driving distance of the candidate path deviating from the direct path;

[0087] The time value coefficient is set, and the travel time cost is calculated based on the detour distance, the preset driving speed and the time value coefficient.

[0088] In the step S105, the decision reasoning engine is used to screen, sort and optimize the plurality of groups of charging-hotel-path combination schemes, including: substituting the calculated charging cost, hotel cost and travel time cost into the comprehensive cost function to obtain the initial value of the comprehensive cost corresponding to the plurality of groups of charging-hotel-path combination schemes; pre-screening the plurality of groups of combination schemes by spatial indexing, and eliminating the combination schemes that do not meet the preset spatial constraint conditions; wherein the spatial constraint conditions include the distance constraint between the charging pile and the hotel, and the direction constraint between the path and the target city.

[0089] The combination schemes after pre-screening are roughly sorted, the approximate values of the comprehensive costs of the combination schemes are calculated, and the combination schemes are prioritized based on the approximate values; the combination schemes after rough sorting are finely optimized, the combination schemes are modeled and solved by using a discrete optimization algorithm, and the solution results are secondarily optimized by using a continuous optimization algorithm to obtain the accurate values of the comprehensive costs of the combination schemes; the combination schemes are finally sorted based on the accurate values of the comprehensive costs, and the combination scheme with the optimal accurate value of the comprehensive cost is selected as the optimal travel comprehensive scheme.

[0090] The specific process of outputting the optimal electric travel comprehensive scheme to the vehicle terminal is: the optimal electric travel comprehensive scheme is disassembled into charging pile information, hotel information, path planning information and comprehensive cost detail information, and is arranged into a visual travel scheme in a preset format, and is transmitted to the vehicle terminal, and the scheme content is displayed through the vehicle screen, and the demand adjustment of the vehicle owner based on the scheme is supported, and the optimal electric travel comprehensive scheme is updated after receiving the adjustment instruction of the vehicle owner. The vehicle owner can intuitively understand the details of the scheme, and the demand adjustment can further adapt to the temporary demand of the vehicle owner, and the flexibility of the scheme and the user satisfaction are improved.

[0091] Embodiment 1: In order to further illustrate the above scheme, this embodiment 1 sets the following implementation scenario: a new energy vehicle owner drives a vehicle from the current city A to the target city B, and needs to generate a long-distance travel comprehensive scheme including charging piles, hotels and paths, and gives consideration to travel cost and experience. Figure 2 In this embodiment, the owner inputs the travel demand and various limiting conditions through the vehicle terminal, the system obtains three types of core data of hotel data, map service data and charging pile data, processes them through the space-time price prediction module, the accommodation cost and experience analysis module and the path planning engine, inputs them into the multi-objective optimizer and the decision reasoning engine, and finally outputs the user interaction interface to display the optimal scheme.

[0092] Specific implementation steps:

[0093] 1. Obtain travel demand information:

[0094] The owner inputs the voice travel demand "plan a long-distance travel scheme from city A to city B" through the vehicle terminal to the multi-modal large model. After receiving the voice information, the multi-modal large model converts the voice into text through the voice recognition module, and then extracts the target city B information through the semantic analysis module. At the same time, the vehicle terminal synchronously uploads the current state information of the vehicle, including the remaining battery capacity of 50% and the current position of a certain road section in city A. The target city B information, vehicle remaining battery capacity information and current position information are associated and bound to generate a structured travel demand data set.

[0095] 2. Collect travel-related data

[0096] Based on the structured travel demand data set, the multi-modal large model calls various data interfaces to collect data through the large model cloud service:

[0097] Call the charging pile data interface to collect charging pile-related data along the way from city A to city B and within city B, including the specific location of each charging pile, real-time load rate, current charging standard, historical electricity price curve and peak and valley period record;

[0098] Call the hotel data interface to collect hotel-related data close to the charging pile along the way and within city B, including hotel location, real-time charging price, user review text, hotel lobby, guest room, parking lot and other facility pictures;

[0099] Call the map service interface to collect multiple candidate path information from city A to city B, including the total length of each path, the road condition along the way, and the distribution position of charging piles and hotels along the way;

[0100] Call the space-time environment data interface to collect real-time weather information and current holiday information along the way from city A to city B and within city B.

[0101] The collected multi-type data is subjected to format standardization processing, and invalid data (such as charging piles with unrecognized locations, and hotels with insufficient evaluation quantities) are eliminated to obtain regularized travel-related data.

[0102] In the above embodiments, Figure 3 The interaction relationship between the vehicle-mounted terminal and the large model cloud service is shown, the large model cloud service integrates hotel information and charging pile information, and data transmission and service response are realized through load balancing and export routing.

[0103] 3. Build a comprehensive cost function

[0104] Define charging cost (Charging_Cost), hotel cost (Hotel_Cost), and travel time cost (Travel_Cost) as the core dimensions of the comprehensive cost, and initially configure the weight coefficients α = 0.5, β = 0.3, and γ = 0.2;

[0105] Build a comprehensive cost function: Total_Cost = α × Charging_Cost + β × Hotel_Cost + γ × Travel_Cost;

[0106] Where, α + β + γ = 1;

[0107] Collect the vehicle owner's historical travel data, including past charging pile price preferences, hotel consumption grades, and path selection preferences for speed, etc. Learn user preferences through multi-modal large models to find that the vehicle owner pays more attention to charging cost, and adjust the weight coefficients to α = 0.6, β = 0.2, and γ = 0.2 to generate a comprehensive cost function that adapts to the vehicle owner.

[0108] 4. Calculate the cost of each type

[0109] 4.1 Charging cost calculation

[0110] Extract historical electricity price data, peak and valley period data, and load rate data from the regularized charging pile-related data, as well as weather and non-holiday information from the spatiotemporal environment-related data, as training data;

[0111] Based on the TimeSformer model, build a spatiotemporal price prediction model, input the training data for model training, and constantly adjust the model parameters during the training process until the prediction accuracy reaches more than 90%;

[0112] Input the real-time charging pile load rate, current good weather, and non-holiday information into the trained spatiotemporal price prediction model to predict the dynamic electricity price and real-time parking fee of each charging pile;

[0113] Combined with the remaining battery level of the vehicle 50%, the driving distance from City A to City B 300km, and the vehicle's power consumption per 100km 15kWh, the required charging amount is calculated as (300x15-50%x total capacity of vehicle battery)=22.5kWh. Combined with the predicted dynamic charging price and real-time parking fee, the charging cost corresponding to each charging pile is calculated.

[0114] 4.2 Hotel cost calculation

[0115] Extract the hotel location, real-time charging price, user review text, and facility pictures from the normalized hotel association data.

[0116] The multi-modal large model text processing module performs semantic analysis on the user review text to extract experience features such as hotel hygiene, service, and soundproofing. The image processing module extracts visual features from hotel facility pictures to evaluate the age and completeness of facilities, and comprehensively determines the suitability of hotel facilities.

[0117] Based on the distance between the hotel and each charging pile, set the hotel emergency coefficient. The emergency coefficient is 1.0 within 1km, 1.1 within 1-3km, and 1.2 above 3km (the farther the distance, the higher the emergency coefficient, reflecting the added cost of detouring).

[0118] Based on the spatio-temporal price prediction model, output the dynamic price of each hotel, and combined with the corresponding hotel emergency coefficient, calculate the hotel cost Hotel_Cost corresponding to each hotel.

[0119] 4.3 Travel time cost calculation

[0120] Based on graph neural networks, build a path planning engine, input the candidate path information from City A to City B, the distance of each path, and the distribution information of charging piles and hotels along the way, and integrate charging piles and hotels as path nodes into the planning network.

[0121] With the total length of the path, the detour distance, and the node resource adaptation degree as constraints, generate 5 groups of candidate paths containing charging nodes and accommodation nodes, and analyze the detour distance of each group of paths from the direct path.

[0122] Set the time value coefficient (reflecting the value of the vehicle owner per unit time), based on the detour distance, the preset average driving speed, and the time value coefficient, calculate the travel time cost Travel_Cost corresponding to each group of paths.

[0123] 5、Optimization scheme

[0124] Figure 4The specific optimization process of the display decision reasoning engine is shown from submitting data to adopting R-tree spatial index, random forest agent model, CP-SAT algorithm, continuous optimization algorithm, and finally returning the ranking result.

[0125] Specifically, the charging cost of each charging pile, the hotel cost, and the path travel time cost are substituted into the comprehensive cost function of the adapted vehicle owner to obtain the initial values of the comprehensive costs of 200 charging-hotel-path combination schemes.

[0126] An R-tree spatial index technology is used to construct a spatial index, and spatial constraint conditions (charging pile and hotel distance ≤ 3 km, path direction and city A to city B direct direction deviation ≤ 15°) are set to pre-screen 200 schemes, and 50 schemes that do not meet the constraints are removed, leaving 150 schemes.

[0127] A random forest agent model is constructed, and the dimension parameters and initial values of the comprehensive costs of the 150 schemes are input to train the model to learn the correlation between the parameters and the costs. The approximate values of the comprehensive costs of each scheme are calculated by the model, sorted from low to high, and the top 30 schemes are selected for the calculation phase.

[0128] The 30 schemes are optimized in the calculation phase. First, the CP-SAT algorithm is used to model and solve the discrete parameters such as charging pile selection and hotel selection to determine the preliminary optimization scheme. Then, the model predictive control algorithm is used to dynamically optimize the continuous parameters such as path detour distance and charging time to obtain the accurate values of the comprehensive costs of each scheme.

[0129] The combination scheme with the lowest accurate value of the comprehensive cost is selected as the optimal electric travel comprehensive scheme, which includes the adapted charging pile location, adapted hotel information, optimal path planning, and various cost details.

[0130] 6. Output scheme

[0131] As Figure 5 The optimal electric travel comprehensive scheme is disassembled into charging pile location, charging standard, hotel location, room price, path navigation information, and comprehensive cost details, and is arranged into visual interface data and transmitted to the vehicle terminal for display to the vehicle owner through the vehicle screen. After the vehicle owner views it, the vehicle owner may request an adjustment of the hotel rating to be ≥ 4.5 points. After receiving the adjustment instruction, the system repeats steps 2-5 to collect hotel data (selecting hotels with a rating ≥ 4.5 points), calculate costs, and optimize and update the optimal travel scheme for the vehicle owner to confirm and select.

[0132] Based on the same inventive concept, the embodiments of the present application also provide a multimodal large model-based travel comprehensive cost optimization system for implementing the multimodal large model-based travel comprehensive cost optimization method described above. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme described in the above-mentioned embodiment method, so the specific limitations in one or more multimodal large model-based travel comprehensive cost optimization system embodiments provided below can refer to the limitations of the multimodal large model-based travel comprehensive cost optimization method described above, which will not be repeated here.

[0133] Embodiment 2: In one embodiment, the multimodal large model-based travel comprehensive cost optimization system provided by Embodiment 2 of the present application comprises: Figure 6 as shown, including: an information acquisition module 210, a data acquisition module 220, a construction module 230, a cost calculation module 240 and a travel plan making module 250, wherein:

[0134] The information acquisition module 210 is configured to acquire travel demand information of a vehicle owner.

[0135] The data acquisition module 220 is configured to acquire travel-related data based on the travel demand information.

[0136] The construction module 230 is configured to construct a comprehensive cost function taking into account charging cost, hotel cost and travel time cost, and define a weight coefficient based on user preference learning.

[0137] The cost calculation module 240 is configured to calculate the charging cost, hotel cost and travel time cost in the comprehensive cost function according to the travel-related data.

[0138] The travel plan making module 250 is configured to filter, sort and optimize a plurality of charging-hotel-path combination schemes using a decision-making reasoning engine based on the calculation result of the comprehensive cost function, generate an optimal electric travel scheme, and output the optimal electric travel scheme to a vehicle-mounted terminal for selection by the vehicle owner.

[0139] In one embodiment, an electronic device, which can be a terminal, is provided, and an internal structure diagram of the electronic device can be as shown in Figure 7As shown in the figure. The electronic device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the electronic device is used for wired or wireless communication with external terminals. Wireless communication can be achieved through WIFI, mobile cellular network, NFC (near field communication) or other technologies. The computer program is executed by the processor to implement any one of the steps S101 to S105 of the travel comprehensive cost optimization method based on the multi-modal large model. The display screen of the electronic device can be a liquid crystal display screen or an electronic ink display screen. The input device of the electronic device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the electronic device, or an external keyboard, touchpad or mouse, etc.

[0140] Those skilled in the art can understand that, Figure 7 Those skilled in the art can understand that,

[0141] Those skilled in the art can understand that,

[0142] The embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can be in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can be in 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. Figure 1 Figure 1 The embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can be in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can be in 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 embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can be in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can be in 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.

[0143] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the flow Figure 1 one or more flow or blocks Figure 1 one or more flow or blocks

[0144] These computer program instructions can 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 that are executed on the computer or other programmable apparatus provide steps for implementing the flow Figure 1 one or more flow or blocks Figure 1 one or more flow or blocks

[0145] The above merely provides an embodiment of the present application, but is not intended to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the scope of the claims of the present application.

Claims

1. A method for optimizing the overall travel cost based on a multimodal large model, characterized in that, The method includes: Obtain information on car owners' travel needs; Collect travel-related data based on travel demand information; Construct a comprehensive cost function that includes charging costs, hotel costs, and travel time costs, and define weight coefficients based on user preference learning; Based on travel-related data, calculate the charging cost, hotel cost, and travel time cost in the comprehensive cost function; Based on the calculation results of the comprehensive cost function, a decision reasoning engine is used to filter, sort and optimize multiple charging-hotel-route combination schemes to generate the optimal electric travel scheme, and output the optimal electric travel scheme to the vehicle terminal for the car owner to choose.

2. The method as described in claim 1, characterized in that, The process of obtaining the vehicle owner's travel demand information includes: receiving the vehicle owner's voice travel demand input into the multimodal big data model through the vehicle terminal, performing voice recognition and semantic analysis on the voice travel demand, extracting target city information, and simultaneously obtaining the vehicle's current status information uploaded by the vehicle terminal; wherein, the vehicle's current status information includes: vehicle remaining battery information and vehicle current location information. By associating and binding the vehicle's remaining battery power information, current vehicle location information, and target city information, travel demand information can be determined.

3. The method as described in claim 1, characterized in that, The process of collecting travel-related data based on travel demand information includes: calling charging pile data interfaces, hotel data interfaces, map service interfaces, and spatiotemporal environment data interfaces through the large model cloud service to collect charging pile related data, hotel related data, route related data, and spatiotemporal environment related data related to travel demand information, respectively, standardizing the format of the collected data, removing invalid and redundant data, and obtaining the processed travel-related data. The travel-related data includes: charging pile related data, hotel related data, spatiotemporal environment related data, and route related data; The charging pile associated data includes: charging pile location information, charging pile real-time load rate, and charging pile charging fee standard; the hotel associated data includes: hotel location information, hotel fee information, hotel review text, and hotel facility pictures. The spatiotemporal environment-related data includes: weather information and holiday information for the target city and the areas along the route; The path association data includes: information on multiple candidate paths from the vehicle's current location to the target city, distance information for each candidate path, and information on the distribution of facilities along the route.

4. The method as described in claim 1, characterized in that, The construction of a comprehensive cost function that incorporates charging costs, hotel costs, and travel time costs, and the definition of weight coefficients based on user preference learning, includes: using a multimodal large model, combining charging pile-related data, hotel-related data, spatiotemporal environment-related data, and path-related data, defining weight coefficients with charging costs, hotel costs, and travel time costs as dimensions, and constructing the comprehensive cost function using the following formula; Total_Cost=α×Charging_Cost+β×Hotel_Cost+γ×Travel_Cost; Where α is the charging cost weighting coefficient, β is the hotel cost weighting coefficient, γ is the travel time cost weighting coefficient, α+β+γ=1, Charging_Cost is the charging cost, Hotel_Cost is the hotel cost, and Travel_Cost is the travel time cost. Collect users' historical travel decision data and preference feedback data, and learn the user's priority of charging cost, hotel cost and travel time cost through a multimodal large model. Dynamically adjust the values ​​of α, β and γ to generate weight coefficients that are suitable for the current user.

5. The method as described in claim 1, characterized in that, The calculation of charging costs includes: acquiring historical electricity price data, peak and off-peak period data, and charging pile load rate data from the charging pile associated data, as well as weather information and holiday information from the spatiotemporal environment associated data, as training data; constructing a spatiotemporal price prediction model through the TimeSformer model, and training the spatiotemporal price prediction model using the training data until the model prediction accuracy meets a preset threshold. The real-time load rate of charging piles, current weather information, and current holiday information are input into the trained spatiotemporal price prediction model to predict the dynamic electricity price and real-time parking fee of charging piles. Combined with the vehicle's remaining battery power and target driving range, the required charging amount of the vehicle is calculated. Based on the predicted dynamic electricity price, required charging amount, and real-time parking fee of charging piles, the charging cost is determined.

6. The method as described in claim 5, characterized in that, The calculation of hotel costs includes: obtaining hotel location information, hotel pricing information, hotel review text, and hotel facility images from hotel-related data; performing semantic analysis on the hotel review text using a multimodal large model to extract hotel experience-related features; and extracting visual features from the hotel facility images to assess the suitability of the hotel facilities. An emergency coefficient for the hotel is set based on the distance between the hotel and the charging station; the emergency coefficient for the hotel is negatively correlated with the distance between the hotel and the charging station. The hotel's dynamic price is output based on the spatiotemporal price prediction model, and the hotel's cost is calculated by combining the hotel's urgency coefficient.

7. The method as described in claim 6, characterized in that, The calculation of travel time cost includes: the location of charging stations, The hotel location is integrated into the path planning network as a path node. With the total path length, detour distance and node resource adaptability as constraints, multiple sets of candidate paths containing charging nodes and accommodation nodes are generated to build a path planning engine. Input path association data and analyze the detour distance corresponding to each candidate path through the path planning engine; the detour distance is the additional driving distance of the candidate path deviating from the direct path; A time value coefficient is set, and the travel time cost is calculated based on the detour distance, preset travel speed, and time value coefficient.

8. The method as described in claim 1, characterized in that, The process of using a decision reasoning engine to screen, sort, and optimize multiple charging-hotel-route combination schemes includes: substituting the calculated charging cost, hotel cost, and travel time cost into a comprehensive cost function to obtain initial comprehensive cost values ​​for multiple charging-hotel-route combination schemes; pre-screening multiple combination schemes using spatial indexing to eliminate combination schemes that do not meet preset spatial constraints; wherein, the spatial constraints include distance constraints between charging piles and hotels, and directional constraints between the route and the target city; The pre-screened combination schemes are coarsely ranked, and the approximate comprehensive cost of each combination scheme is calculated. Based on the approximate value, the combination schemes are prioritized. The coarsely ranked combination schemes are then finely optimized. First, a discrete optimization algorithm is used to model and solve the combination schemes. Then, a continuous optimization algorithm is used to optimize the solution results for a second time to obtain the accurate comprehensive cost value of each combination scheme. Based on the accurate comprehensive cost value, the combination schemes are finally ranked, and the combination scheme with the best accurate comprehensive cost value is selected as the optimal comprehensive travel scheme.

9. A comprehensive travel cost optimization system based on a multimodal large model, characterized in that, The system includes: The information acquisition module is used to acquire information about the travel needs of car owners; The data acquisition module is used to collect travel-related data based on travel demand information; The module is used to construct a comprehensive cost function that includes charging costs, hotel costs, and travel time costs, and to define weight coefficients based on user preference learning. The cost calculation module is used to calculate the charging cost, hotel cost, and travel time cost in the comprehensive cost function based on travel-related data. The travel plan formulation module is used to filter, sort and optimize multiple charging-hotel-route combination schemes based on the calculation results of the comprehensive cost function, and generate the optimal electric travel plan. The optimal electric travel plan is then output to the vehicle terminal for the car owner to select.

10. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-8.

Images