CN-122022638-A - New energy vehicle collaborative vehicle cargo matching system and method based on big data

CN122022638ACN 122022638 ACN122022638 ACN 122022638ACN-122022638-A

Abstract

The invention discloses a big data-based new energy vehicle collaborative vehicle-cargo matching system and a big data-based new energy vehicle collaborative vehicle-cargo matching method, which relate to the technical field of new energy vehicles. The method comprises the steps of constructing a vehicle energy consumption prediction digital twin body by using a deep learning algorithm, outputting vehicle unit mileage energy consumption and required total energy consumption, constructing a fault probability model of a battery by using a multi-order learning model, defining a fault probability vector of a current vehicle based on the fault probability model, calculating a matching risk index of the current vehicle and goods by using a goods feature vector, the fault probability vector and a goods-battery risk association matrix, constructing a reachability check by using the required total energy consumption output by the digital twin body, obtaining a candidate vehicle by using a composite check mode, extracting multi-source data of the candidate vehicle, calculating a vehicle comprehensive score, and selecting the highest score as an optimal vehicle to recommend.

Inventors

DU SONGLIN
ZHANG LIYANG
LEI TAO
Sadamu Shadik
WANG BINGQUAN
WANG XIAOFENG
YU JIONG
DU XUSHENG

Assignees

杭州骋风而来数字科技有限公司
新疆丝路融创网络科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260410

Claims (10)

1. The new energy vehicle collaborative vehicle cargo matching method based on big data is characterized by comprising the following steps of: S100, deploying a vehicle-mounted terminal sensor on each new energy vehicle, and collecting battery data, vehicle states, environment data, road data and cargo information in real time; S200, constructing a vehicle energy consumption prediction digital twin body by using a deep learning algorithm, inputting vehicle state and environment data into the digital twin body, and outputting the unit mileage energy consumption and the required total energy consumption of the vehicle; s300, calculating the health degree of the battery in real time by using battery data, constructing a fault probability model of the battery by using a multi-order learning model, and defining a fault probability vector of the current vehicle based on the fault probability model; S400, constructing a cargo feature vector by using cargo information to generate a cargo-battery risk correlation matrix, calculating a matching risk index of the current vehicle and the cargo by using the cargo feature vector, the fault probability vector and the cargo-battery risk correlation matrix, and obtaining a risk upper limit by using experience distribution based on the matching risk index of the historical transportation success case; s500, constructing a reachability test based on the required total energy consumption output by the digital twin body, constructing a safety risk test by using a risk upper limit and a matching risk index of the current vehicle and cargoes, integrating to obtain a composite test mode, and obtaining a candidate vehicle by using the composite test mode; And S600, extracting multi-source data from the candidate vehicles, calculating vehicle comprehensive scores, and selecting the highest score as the optimal vehicle for recommendation.
2. The method for collaborative vehicle cargo matching of a new energy vehicle based on big data according to claim 1, wherein the defining the fault probability vector of the current vehicle based on the fault probability model in S300 is specifically: Classifying faults of the battery into thermal runaway, outage, unstable voltage and capacity dip; extracting the highest single temperature, the maximum temperature rise rate, the temperature inconsistency and the internal resistance change rate along with the temperature as feature vectors aiming at thermal runaway faults, wherein the highest single temperature represents the maximum value of all battery temperatures acquired at different positions of a battery, the maximum temperature rise rate represents the time-based derivative of the battery temperature after acquiring the battery temperature aiming at a time sequence and takes the maximum value; Extracting a current charge state, a battery available capacity, a current discharge power ratio, an average dropping rate of a voltage in a discharge process and an internal resistance increase coefficient as feature vectors aiming at power failure, wherein the current discharge power ratio represents a ratio of the current discharge power to the maximum discharge power of the battery, the average dropping rate of the voltage in the discharge process represents a derivative average value of the voltage based on time, and the internal resistance increase coefficient represents a ratio of the current direct current internal resistance of the battery to the initial direct current internal resistance; extracting a single voltage range, a single voltage variance, a voltage fluctuation range and a voltage dip time count as feature vectors for a voltage instability fault, wherein the single voltage range=maximum voltage-minimum voltage, the voltage fluctuation range represents the standard deviation of a voltage sequence, and the voltage dip time count represents the number of times that the voltage is instantaneously dropped by more than 5% within 10 minutes; Extracting a health degree descending rate, an accumulated cycle number, an accumulated throughput electric quantity and an internal resistance increasing rate as feature vectors aiming at capacity dip faults, wherein the health degree descending rate= delta SOH/[ delta ] t; forming the feature vectors of the three fault types into a high-dimensional feature vector; Predicting occurrence probabilities of four fault types respectively by utilizing a multi-order learning model, selecting a random forest as a base classifier, training a two-class model for each fault mode, and collecting feature vectors recorded by historical fault events as a training set; Aiming at each fault type, a training set training probability model is used for inputting a feature vector and outputting a probability, wherein the formula is F k =P（y k = 1|X, y k =1 represents the occurrence of the kth fault type, X represents the feature vector corresponding to the kth fault type, P represents a probability model, F k represents the occurrence probability of the kth fault, four probability models are integrated to form a fault probability model, and the fault probability vectors F= [ F 1 、f 2 、f 3 、f 4 ];f 1 、f 2 、f 3 、f 4 ] of the four fault types are output to respectively represent the occurrence probabilities of thermal runaway, outage, unstable voltage and capacity dip.
3. The method for matching new energy vehicles with vehicles and goods based on big data according to claim 1, wherein the feature vector of the goods in S400 is specifically: Extracting a cargo feature vector g= [ G 1 、g 2 、g 3 、g 4 、g 5 ],g 1 ] from cargo information, wherein the cargo feature vector g= [ G 1 、g 2 、g 3 、g 4 、g 5 ],g 1 ] represents a temperature sensitivity coefficient, and the temperature sensitivity coefficient = the reciprocal of a cargo allowable storage temperature range, and the cargo allowable storage temperature = the maximum cargo allowable temperature-the minimum cargo allowable temperature; g 2 represents a time sensitivity coefficient, wherein the time sensitivity coefficient is the inverse of the maximum delay time allowed by the goods, and the maximum delay time allowed by the goods is determined by staff according to the transportation timeliness requirement; g 3 represents a risk level index, which is quantitatively acquired by a professional; g 4 denotes the shock sensitivity = inverse of the maximum acceleration that the cargo can withstand, which is set by the professional; g 5 denotes the humidity sensitivity coefficient, humidity sensitivity coefficient = the inverse of the allowable relative humidity range for the good, which is determined by the professional according to the good storage specifications.
4. The method for collaborative vehicle-cargo matching of new energy vehicles based on big data according to claim 1, wherein the generating cargo-battery risk correlation matrix in S400 is specifically: Collecting the event of goods loss caused by battery faults in historical transportation, recording the goods characteristic vector and fault types in the event, and counting the goods loss rate of different goods characteristics under each fault type, wherein the goods loss rate represents the proportion of damaged goods to total goods when faults occur, and calculating the severity coefficient w ij of the loss of the goods with the ith goods attribute when the jth fault occurs in the battery; After normalizing the severity coefficient, constructing a cargo-battery risk correlation matrix by using the severity coefficient as the correlation weight of the cargo feature vector and the fault probability vector, wherein elements in the matrix are severity coefficient normalized values; The obtaining the risk upper limit based on the matching risk index of the historical transportation success case by using the experience distribution is specifically as follows: And respectively extracting a cargo feature vector in the cargo transportation order to be matched and a fault probability vector of the vehicle to be matched aiming at the current cargo transportation order to be matched and all vehicles to be matched, calculating a matching risk index R match , collecting all cargo transportation orders which are successfully completed in a history and have no faults, calculating the R match value of each order to form experience distribution, and defining the upper limit gamma of risk as 95% quantile of the experience distribution.
5. The method for collaborative vehicle-cargo matching of new energy vehicles based on big data according to claim 1, wherein the step S200 is characterized in that the vehicle state and the environmental data are input into a digital twin body, and the unit mileage energy consumption and the required total energy consumption of the output vehicle are specifically: The professional builds a vehicle energy consumption physical model according to the vehicle knowledge, utilizes a deep learning algorithm to combine the vehicle energy consumption physical model to build and obtain a vehicle energy consumption prediction digital twin body, inputs the current vehicle state and environment data into the digital twin body, outputs the unit mileage energy consumption of the vehicle in a future road section, And extracting a cargo loading point, a cargo unloading point and a vehicle position in the cargo transportation order as starting points, constructing a preset running path of the starting points, the cargo loading point and the cargo unloading point, sequentially predicting and outputting unit mileage energy consumption by digital twin body segments according to the preset running path of the cargo transportation order of the vehicle to be matched, obtaining each section of mileage energy consumption by multiplying each section of path length by the unit mileage energy consumption, and summing and predicting to obtain the required total energy consumption E trip of the cargo transportation order.
6. The method for matching new energy vehicles with vehicles and goods based on big data according to claim 1, wherein the step S500 of obtaining the candidate vehicles by using the composite test mode is specifically as follows: The accessibility test is specifically E trip ≤E batt x SOH x (1-alpha), wherein alpha represents a safety margin coefficient and is determined by professionals according to statistical experience, E batt represents the rated total energy of a battery in a vehicle, SOH represents the health degree of the battery, and E trip represents the required total energy consumption of a cargo transportation order; The safety risk test specifically comprises that if R match is less than or equal to gamma, the vehicle passes the safety test and enters a candidate list, if R match is more than gamma, the matching is refused, and a refusing reason is automatically generated, wherein the refusing reason is that the cargo characteristics and the fault probability of the maximum severity coefficient normalization value in the current vehicle risk association matrix are not matched, and R match is a matching risk index; s502, in the current cargo transportation order to be matched and all vehicles to be matched, firstly, carrying out preliminary screening by utilizing accessibility test, if the conditions are met, entering a safety risk test, otherwise, directly eliminating and recording the reason of 'insufficient endurance', and if the conditions are met in the safety risk test, entering a candidate vehicle queue.
7. The method for collaborative vehicle-cargo matching of new energy vehicles based on big data according to claim 6, wherein the selecting the highest score as the optimal vehicle for recommendation in S600 is specifically as follows: Extracting the cargo transportation demand time T y in the transportation timeliness demand for each candidate vehicle in the candidate vehicle queue, calculating the predicted transportation time T in by using the average speed of the vehicle and the cargo transportation distance, calculating the transportation duration deviation DeltaT=T in -T y , collecting the ratio of the successful times of the cargo transportation orders of the history corresponding vehicles to the total times as the on-time completion rate Re, calculating the vehicle comprehensive score of each candidate vehicle by using the required total energy, the transportation duration deviation and the on-time completion rate after normalization and weighting summation, And (3) after calculating the vehicle comprehensive scores of all the candidate vehicles, selecting the vehicle corresponding to the maximum value as an optimal vehicle, pushing the optimal vehicle information to the goods transportation order, and simultaneously sending the goods transportation order to an optimal vehicle driver, wherein the optimal vehicle driver selects whether to accept or not.
8. The new energy vehicle collaborative cargo matching system based on big data is characterized by comprising a data acquisition module, a digital twin body module, a fault prediction module, a data analysis module, a composite screening module, a final matching module and a feedback module; the data acquisition module is used for deploying a vehicle-mounted terminal sensor on each new energy vehicle and acquiring battery data, vehicle states, environment data, road data and cargo information in real time; The digital twin body module is used for constructing a vehicle energy consumption prediction digital twin body by using a deep learning algorithm, and inputting vehicle state and environment data into the digital twin body to output the unit mileage energy consumption and the residual endurance mileage of the vehicle; the fault prediction module is used for calculating the health degree of the battery in real time by utilizing the battery data, and constructing a fault probability model of the battery by utilizing the multi-order learning model; The data analysis module is used for constructing a cargo feature vector by utilizing cargo information, defining a fault probability vector of a current vehicle based on a fault probability model, generating a cargo-battery risk association matrix, and calculating a matching risk index of the current vehicle and the cargo by utilizing the cargo feature vector, the fault probability vector and the cargo-battery risk association matrix; The composite screening module is used for constructing reachability inspection based on the required total energy consumption output by the digital twin body, constructing safety risk inspection by utilizing the upper risk limit and the matching risk index of the current vehicle and cargoes, integrating to obtain a composite inspection mode, and obtaining a candidate vehicle by utilizing the composite inspection mode; the final matching module is used for extracting multi-source data from the candidate vehicles, calculating the comprehensive score of the vehicles, and selecting the highest score as the optimal vehicle for recommendation; And the feedback module is used for recording the matching risk index and the reject reason of the rejected vehicle after each matching is completed, successfully matching the actual energy consumption, the battery state change and whether faults occur or not of the order, and feeding back the recorded data to the digital twin body by the timing rate of the driver and the cargo state.
9. The big data-based new energy vehicle collaborative cargo matching system according to claim 8, wherein the data analysis module includes a risk correlation unit and a risk index unit; The risk association unit is used for calculating a severity coefficient after extracting the cargo feature vector, normalizing the severity coefficient and constructing a risk association matrix as association weights of the cargo feature vector and the fault probability vector; The risk index unit is used for respectively extracting a cargo characteristic vector in a cargo transportation order to be matched and a fault probability vector of the vehicle to be matched, and calculating a matching risk index.
10. The big data based new energy vehicle collaborative cargo matching system according to claim 8, wherein the composite screening module includes a reachability verification unit and a security risk verification unit; the accessibility checking unit is used for judging whether the vehicle can reach the unloading point or not by utilizing the required total energy consumption and the rated total energy of the battery; The safety risk checking unit is used for judging the matching risk index of each vehicle by using the upper risk limit, and whether the delivery of the vehicle is safe or not.

Description

New energy vehicle collaborative vehicle cargo matching system and method based on big data Technical Field The invention relates to the technical field of new energy automobiles, in particular to a new energy automobile collaborative vehicle-cargo matching system and method based on big data. Background The current logistics industry is in charge of new energy conversion and surge, new energy freight vehicles are applied in large scale in same city distribution, middle and short distance inter-city transportation due to the advantages of low carbon, environmental protection, low operation cost and the like, but the existing mode and technical system are difficult to adapt to the technical characteristics of the new energy vehicles, and meanwhile, the industry pain point of traditional vehicle-cargo matching is not solved effectively; Unlike conventional fuel vehicles, the transportation capability of new energy freight vehicles is highly dependent on battery status, but the core indexes of battery endurance, health, failure risk and the like are not included in the consideration range of the existing vehicle-cargo matching. On one hand, the existing matching mode does not accurately predict the energy consumption of the vehicle, only judges the transportation capacity according to the nominal endurance of the vehicle, ignores the influence of factors such as load, road gradient, environmental temperature and the like on the actual endurance, is extremely easy to cause the condition of insufficient endurance of the vehicle and halfway anchoring and seriously affects the transportation aging, on the other hand, the faults such as thermal runaway, outage and unstable voltage of the battery are directly related to the transportation safety of the goods, the prior art cannot quantitatively predict the fault probability of the battery, and the association relation between the fault of the battery and the loss of the goods is not established, and the economic loss is easily caused due to the goods loss caused by the sudden fault of the battery for special goods such as cold chains, dangerous chemicals, precise instruments and the like. The existing vehicle-cargo matching is dependent on modes such as manual butt joint, extensive release of an information platform and the like, a cargo owner can only select a carrier vehicle according to basic information such as vehicle types and loads, a vehicle driver can also select cargoes according to order distance and freight charges, and personalized and accurate matching of the vehicles and the cargoes cannot be achieved by both sides. Disclosure of Invention The invention aims to provide a new energy vehicle collaborative vehicle cargo matching system and method based on big data, so as to solve the problems in the prior art. In order to achieve the above purpose, the present invention provides the following technical solutions: the new energy vehicle collaborative vehicle cargo matching method based on big data comprises the following steps: S100, deploying a vehicle-mounted terminal sensor on each new energy vehicle, and collecting battery data, vehicle states, environment data, road data and cargo information in real time; Further, the battery data comprise battery cell voltage, current, temperature, state of charge (SOC), battery health (SOH) and internal resistance, the vehicle state comprises instantaneous speed, acceleration, load and total mass, the environment data comprise environment temperature, humidity and weather conditions, the road data comprise real-time GPS positions, road slopes in front, curvatures and altitudes, the goods information comprises goods types, goods values, weights, volumes and transportation aging requirements, the five kinds of core data of the battery, the vehicle state, the environment, the roads and the goods are covered, a complete and accurate original data base is provided for subsequent energy consumption prediction, fault analysis and risk matching, analysis deviation caused by data loss is avoided, and the whole-scene data requirement of new energy vehicle transportation is adapted. Cleaning all the collected original data, wherein the cleaning comprises removing abnormal values and time alignment, the abnormal values represent the original data which deviate from the historical average value of the corresponding data by three times, and the normalization processing is carried out on all the cleaned original data. And the data of different dimensions and orders of magnitude are standardized, the problem of unbalanced data weight in deep learning and model calculation is solved, and the training efficiency and the prediction accuracy of the subsequent energy consumption prediction digital twin body and fault probability model are improved. S200, constructing a vehicle energy consumption prediction digital twin body by using a deep learning algorithm, and inputting vehicle state and environment data into the digital twin body