Cold chain transportation path optimization method and system based on AI prediction

Abstract

The invention provides a cold-chain transportation path optimization method and system based on AI prediction, and the method comprises the steps: firstly obtaining a sample cold-chain transportation order data set which comprises a transportation node sequence, a cargo fresh-keeping requirement parameter, an environment temperature fluctuation record and a transportation aging completion state; then extracting a combined transportation characteristic set containing time sensitivity, environmental stability and cargo deterioration risk characteristics, training a cold chain demand prediction model according to the combined transportation characteristic set, obtaining a cargo fresh-keeping demand prediction value and path congestion probability distribution, then generating a plurality of candidate transportation paths, and carrying out fresh-keeping reliability and aging achievement rate scoring; and finally, dynamically adjusting a current transportation path node sequence according to a score weighting result, and updating model input data in real time to optimize a subsequent path decision. According to the method, multi-dimensional data and the AI technology are utilized, the scientificity and reliability of cold chain transportation path planning are effectively improved, the cargo deterioration risk is reduced, and the transportation efficiency is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and specifically, to a cold chain transportation route optimization method and system based on AI prediction. Background Art

[0002] With the rapid development of the cold chain logistics industry, cold chain transportation plays a vital role in ensuring the quality of temperature-sensitive goods such as fresh food and medicine. The particularity of cold chain transportation is that it not only ensures that the goods are delivered on time, but also maintains a suitable temperature environment throughout the process to prevent the goods from spoiling. However, the existing cold chain transportation path planning methods have exposed many limitations when dealing with complex and changeable actual transportation scenarios, and innovative technical means are urgently needed to solve them.

[0003] In the traditional cold chain transportation route planning process, most methods only consider basic factors such as distance and transportation time to determine the optimal route. These methods are often based on simple empirical rules or mathematical models, ignoring the key factor of the preservation requirements of the goods themselves in cold chain transportation. For example, for different types of fresh food or medicine, the required preservation temperature range, the degree of temperature fluctuation that can be tolerated, and the risk factor of deterioration are all different, but traditional methods do not take these differences into account in route planning, resulting in the deterioration of goods due to improper temperature control during actual transportation, which has caused huge economic losses to enterprises. Summary of the invention

[0004] In view of this, the purpose of this application is to provide a cold chain transportation route optimization method and system based on AI prediction.

[0005] According to the first aspect of the present application, a cold chain transportation route optimization method based on AI prediction is provided, the method comprising: Acquire a sample cold chain transport order data set, wherein the sample cold chain transport order data set includes a transport node sequence, cargo preservation requirement parameters, ambient temperature fluctuation records, and transport timeliness completion status of multiple sample orders; Extracting transport features from the sample cold chain transport order data set to generate a combined transport feature set corresponding to each sample order, wherein the combined transport feature set includes a time-sensitive feature, an environmental stability feature, and a cargo deterioration risk feature; Based on the combined transport feature set, a cold chain demand forecasting model is called to generate a forecast value of the cargo preservation demand in the target area within a preset time range, and based on sample traffic flow data and weather event records in the target area, a path congestion probability distribution is generated; According to the predicted value of the cargo preservation demand and the probability distribution of route congestion, multiple candidate transportation routes from the starting point to the end point are generated, and each candidate transportation route is scored for preservation reliability and timeliness achievement rate; Based on the weighted results of the freshness-keeping reliability score and the timeliness achievement rate score, the node sequence of the current transportation path is dynamically adjusted, and the input data of the cold chain demand prediction model is updated in real time to optimize subsequent path decisions.

[0006] According to the second aspect of the present application, a cold chain transportation service system is provided, which includes a machine-readable storage medium and a processor, wherein the machine-readable storage medium stores machine-executable instructions, and when the processor executes the machine-executable instructions, the cold chain transportation service system implements the aforementioned cold chain transportation path optimization method based on AI prediction.

[0007] According to the third aspect of the present application, a computer-readable storage medium is provided, in which computer-executable instructions are stored. When the computer-executable instructions are executed, the aforementioned cold chain transportation path optimization method based on AI prediction is implemented.

[0008] According to any one of the above aspects, the technical effect of the present application is: The embodiment of the present application can comprehensively capture the key information in the cold chain transportation process by acquiring a sample cold chain transportation order data set that includes a transportation node sequence, cargo preservation requirement parameters, ambient temperature fluctuation records, and transportation time completion status, so that subsequent analysis and model training are no longer limited to a single dimension, but are based on multi-dimensional, multi-variable real-world scenario data.

[0009] When extracting transportation features from the sample cold chain transportation order data set, a combined transportation feature set including time-sensitive features, environmental stability features, and cargo deterioration risk features is generated, breaking the previous limitation of focusing on a single factor and comprehensively and comprehensively describing the complex characteristics of cold chain transportation. The time-sensitive feature ensures that the transportation timeliness is guaranteed, the environmental stability feature takes into account the impact of the external environment on the preservation of goods, and the cargo deterioration risk feature is directly related to the quality and safety of goods. The combination of the three provides a more realistic, deeper and broader feature input for subsequent model training, greatly improving the model's understanding and prediction capabilities for cold chain transportation scenarios.

[0010] Based on this combined transport feature set, the cold chain demand forecasting model is called to generate accurate forecast values ​​of cargo preservation demand and path congestion probability distribution in the target area within a preset time range. Unlike traditional simple estimation or empirical methods, this cold chain demand forecasting model is based on a large amount of actual data and in-depth feature analysis, which enables intelligent and accurate forecasts of cargo preservation demand and path congestion. The cargo preservation demand forecast value provides a scientific basis for the temperature control strategy during transportation, and the path congestion probability distribution provides early warning of possible traffic conditions, helping transportation planners to prepare for response in advance, greatly improving the foresight and scientific nature of transportation decisions.

[0011] According to the predicted value of the goods preservation demand and the probability distribution of path congestion, multiple candidate transportation routes are generated, and each candidate transportation route is scored for preservation reliability and timeliness. Traditional path planning often only focuses on a single goal, such as the shortest path or the fastest arrival time. This method takes into account both the reliability of goods preservation and the timeliness achievement rate, fully considers the particularity of cold chain transportation, and ensures that the goods are delivered to the destination in a good state of preservation within the specified time. This not only improves the transportation quality of goods and reduces the economic losses caused by deterioration, but also enhances customer satisfaction and improves the competitiveness of enterprises in the cold chain transportation market.

[0012] Finally, based on the weighted results of the freshness reliability score and the timeliness achievement rate score, the node sequence of the current transportation path is dynamically adjusted, and the input data of the cold chain demand forecasting model is updated in real time to optimize subsequent path decisions, so that the transportation path can continuously optimize itself according to the actual situation and adapt to the ever-changing transportation environment. In the face of sudden weather changes, traffic congestion and other conditions, it can respond quickly and adjust the path in time to ensure that the transportation process always remains efficient and reliable. At the same time, real-time updating of the model input data enables the cold chain demand forecasting model to continuously learn new information, further improving the accuracy of the forecast and the adaptability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0014] Figure 1 A schematic diagram of the process of the cold chain transportation path optimization method based on AI prediction provided in an embodiment of the present application is shown; Figure 2A schematic diagram of the component structure of a cold chain transportation service system provided in an embodiment of the present application for implementing the above-mentioned cold chain transportation path optimization method based on AI prediction is shown. DETAILED DESCRIPTION

[0015] The embodiments of the present application are described below in conjunction with the drawings in the present application. It should be understood that the implementation methods described below in conjunction with the drawings are exemplary descriptions for explaining the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions of the embodiments of the present application.

[0016] It will be understood by those skilled in the art that, unless specifically stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the terms "including" and "comprising" used in the embodiments of the present application refer to that the corresponding features can be implemented as the presented features, information, data, steps, operations, elements and / or components, but do not exclude the implementation as other features, information, data, steps, operations, elements, components and / or combinations thereof supported by the technical field. It should be understood that when an element is "connected" or "coupled" to another element, the element may be directly connected or coupled to the other element, or may refer to the element and the other element establishing a connection relationship through an intermediate element. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling, and the term "and / or" used herein indicates at least one of the items defined by the term, such as "A and / or B" may be implemented as "A", or as "B", or as "A and B".

[0017] In order to make the purpose, technical solution and advantages of the present application clearer, the implementation mode of the present application will be further described in detail with reference to the accompanying drawings. The technical solution of the embodiment of the present application and the technical effect produced by the technical solution of the present application are explained below by describing several exemplary implementation modes. It should be pointed out that the following implementation modes can refer to, draw lessons from or combine with each other, and the same terms, similar features and similar implementation steps in different implementation modes are not described repeatedly.

[0018] Figure 1 The flowchart of the cold chain transportation path optimization method and system based on AI prediction provided by the embodiment of the present application is shown. It should be understood that in other embodiments, the order of some steps in the cold chain transportation path optimization method based on AI prediction of this embodiment can be shared with each other according to actual needs, or some steps can be omitted or maintained. The detailed steps of the cold chain transportation path optimization method based on AI prediction include: Step S110, obtaining a sample cold chain transport order data set, wherein the sample cold chain transport order data set includes a transport node sequence, cargo preservation requirement parameters, ambient temperature fluctuation records, and transport time completion status of multiple sample orders.

[0019] In this embodiment, in the fresh cold chain transportation scenario, take the transportation of fruits (such as strawberries) and beef as an example. For the transportation order of strawberries, the transportation node sequence may include starting from the picking farm in the place of origin, passing through the local collection center, regional cold chain warehouse, transit distribution center, and finally arriving at various retail supermarkets in the city. In terms of the parameters required for the preservation of goods, strawberries are a kind of fruit that is extremely easy to rot and need to be stored in an environment with low temperature and suitable humidity. The preservation temperature requirement may be set at 0-4 degrees Celsius, and the humidity is maintained at about 85-90%. The total time during the entire transportation process cannot exceed 48 hours, otherwise serious deterioration will occur. For the transportation of beef, the transportation node sequence may start from the slaughterhouse, pass through the meat processing plant, the large cold chain logistics hub, and then to the meat wholesalers in various places, and finally arrive at the restaurant or retail outlet. Compared with strawberries, the preservation requirements of beef are required to be below -18 degrees Celsius to ensure the freshness of the meat and prevent bacterial growth. Although the transportation time is relatively long, it is also necessary to complete the transportation within the prescribed 72 hours to ensure the quality of the beef.

[0020] In terms of ambient temperature fluctuation records, during the transportation of strawberries, if the refrigeration equipment of the transport vehicle may be aged or the ambient temperature is too high (for example, during the hot summer period, the ambient temperature reaches above 35 degrees Celsius), the temperature in the carriage may fluctuate from the standard temperature of 0-4 degrees Celsius to 5-6 degrees Celsius, and this fluctuation lasts for 1-2 hours. For beef transportation, if the transport vehicle passes through a mountainous section from a meat processing plant to a cold chain logistics hub, the temperature fluctuates from -18 degrees Celsius to -10 degrees Celsius for about 30 minutes due to changes in altitude and a temporary failure of the refrigeration equipment when passing through a mountainous section.

[0021] The transportation time completion status refers to whether the transportation is completed within the specified time during the actual transportation process. For example, a batch of strawberries was transported, but due to temporary road control at a certain transportation node, the transportation time was extended and it ultimately failed to reach the retail supermarket within 48 hours. In this case, the transportation time completion status is incomplete. However, a batch of beef was transported, which passed smoothly at each transportation node and arrived at the restaurant within 72 hours as required. The transportation time completion status is completed. By collecting relevant information on multiple such strawberry and beef transportation orders, including transportation node sequences, cargo preservation requirement parameters, ambient temperature fluctuation records, and transportation time completion status, a sample cold chain transportation order dataset is formed.

[0022] Step S120, extracting transport features from the sample cold chain transport order data set to generate a combined transport feature set corresponding to each sample order, wherein the combined transport feature set includes time-sensitive features, environmental stability features, and cargo deterioration risk features.

[0023] Taking strawberry transportation as an example, the residence time data of each transportation node is extracted from the sample cold chain transportation order data set. Assuming that the planned residence time of strawberries from the picking farm to the collection center should be 2 hours, but due to the low sorting efficiency of the collection center, the actual residence time reached 4 hours. Regarding the number of times the ambient temperature deviates from the threshold, during the transportation from the regional cold chain warehouse to the transit distribution center, the temperature in the carriage deviated from the threshold of 0-4 degrees Celsius 3 times due to the failure of the sensor of the refrigeration equipment. In terms of the integrity record of the cargo packaging, if part of the packaging of the strawberries is squeezed and deformed during transportation, the internal strawberries will be damaged to a certain extent. The time-sensitive feature is determined based on the difference between the residence time data and the preset cargo preservation time limit. Since it stayed at the collection center for 2 more hours, it is close to half of the 48-hour preservation time limit of the strawberries, which indicates that the risk level of exceeding the preservation time limit in the transportation node sequence is high.

[0024] When calculating the environmental stability characteristics, the number of times the ambient temperature deviates from the threshold and the percentage of deviation duration are used. In the above transportation from the regional cold chain warehouse to the transit distribution center, the temperature deviated from the threshold three times, and each deviation lasted about 30 minutes. The total transportation time was 4 hours, and the deviation duration accounted for (3×30)÷(4×60)=37.5%. This percentage shows that the temperature fluctuation during transportation has a greater cumulative impact on the quality of strawberries.

[0025] For the risk characteristics of cargo deterioration, the cargo packaging integrity record and the deterioration rate parameters corresponding to the cargo type are combined. Strawberries themselves deteriorate at a fast rate, and because some of the packaging is damaged, the strawberries are exposed to non-ideal environments, which greatly increases the contribution of strawberries to the deterioration of cargo quality at different nodes in the transportation route.

[0026] For beef transportation, the planned stay time from the slaughterhouse to the meat processing plant is 3 hours, and the actual stay time is 5 hours. In the transportation from the large cold chain logistics hub to the meat wholesaler, due to the influence of high temperature outside, the compartment temperature deviated from the -18 degrees Celsius threshold twice, and each deviation lasted for 15 minutes. The transportation time was 5 hours, and the deviation duration accounted for (2×15)÷(5×60)=10%. If the beef packaging is scratched, combined with the spoilage rate of beef (relatively slower than strawberries but will also deteriorate in adverse environments), it will also increase the risk characteristics of cargo deterioration. Finally, the time-sensitive characteristics, environmental stability characteristics, and cargo deterioration risk characteristics of strawberry and beef transportation orders are normalized to generate their respective combined transportation feature sets.

[0027] Step S130, calling the cold chain demand prediction model based on the combined transport feature set, generating a forecast value of the cargo preservation demand in the target area within a preset time range, and generating a path congestion probability distribution based on sample traffic flow data and weather event records in the target area.

[0028] Let's continue with the strawberry and beef transportation scenario. Input the combined transportation feature set of strawberry transportation into the time series convolutional neural network. Assume that the transportation nodes include picking farm A, collection center B, regional cold chain warehouse C, transit distribution center D, and retail supermarket E. The time series convolutional neural network extracts the dynamic association pattern between transportation nodes by analyzing the transportation data between these nodes, such as the transportation from A to B, B to C, etc. in different time periods. For example, it is detected that the transportation from B to C is prone to transportation delays every Monday morning because the goods at the collection center B are shipped in batches at this time, and the vehicle scheduling is tight. After being processed by the time series convolutional neural network, the initial path demand prediction result is output.

[0029] Then, the long short-term memory network is used to make time-dependent corrections to the initial path demand forecast results. Since strawberries have strict preservation time limits, the long short-term memory network takes into account that the freshness of strawberries will decrease over time and adjusts the initial path demand forecast results. For example, if the initial forecast shows that the preservation resource demand at a certain node is low, but considering the cumulative effect of the transportation time, the preservation resource demand at that node is increased after correction, and a revised cargo preservation demand forecast value is generated. This value includes the distribution of preservation resource demand at different transportation nodes in the transportation path. For example, more refrigeration resources are needed at the transit distribution center D to maintain the freshness of the strawberries.

[0030] For the target area (such as a city and its surrounding areas), a path congestion probability prediction sub-model is constructed based on the sample traffic flow data and weather event records of the area. If the city often experiences heavy rain in summer, according to historical data, waterlogging on roads during heavy rain will cause traffic paralysis on certain sections of the road. By analyzing the correlation between weather events and traffic flow changes, for example, if it is detected that the probability of road congestion from regional cold chain warehouse C to transit distribution center D will increase by 30% when it rains, the path congestion probability distribution for future time periods can be output.

[0031] The same process is used for beef transportation. The combined transport feature set is input into the time series convolutional neural network to analyze the dynamic association patterns between nodes such as slaughterhouses, meat processing plants, and cold chain logistics hubs, and output the initial path demand forecast results. The long short-term memory network makes time-dependency corrections based on the preservation characteristics of beef (such as a relatively slow deterioration rate at -18 degrees Celsius but still affected by time), and obtains the cargo preservation demand forecast value that includes the distribution of preservation resource demand at each transportation node. When constructing the path congestion probability prediction sub-model, considering that winter snowfall may affect transportation from meat processing plants to cold chain logistics hubs, the relationship between snowfall and traffic flow is analyzed to obtain the corresponding path congestion probability distribution. Finally, the revised cargo preservation demand forecast values ​​for strawberry and beef transportation are integrated with the path congestion probability distribution to generate a joint output of the cold chain demand prediction model. This joint output can identify the preservation resource occupancy rate and congestion delay risk of different candidate transportation paths. For example, for a candidate path for transporting strawberries, the preservation resource occupancy rate may reach 80%, and the congestion delay risk is 20%, while for a candidate path for transporting beef, the preservation resource occupancy rate is 60%, and the congestion delay risk is 15%.

[0032] Step S140, generating a plurality of candidate transport routes from the starting point to the end point according to the predicted value of the goods preservation demand and the probability distribution of route congestion, and performing a preservation reliability score and a timeliness achievement rate score on each candidate transport route.

[0033] Take strawberry transportation as an example, with the starting and ending points being the picking farm at the origin and the urban retail supermarket. The expected total transportation time of each candidate transportation path is calculated based on the node sequence length of the transportation path, the sample average speed, and the path congestion probability distribution. Assume that there is a candidate transportation path that is picking farm-collection center-transit distribution center-retail supermarket, and the node sequence length of this path is 3 nodes. In terms of the sample average speed, if the average speed from the picking farm to the collection center is 60 km / h and the distance is 120 km, the average speed from the collection center to the transit distribution center is 50 km / h and the distance is 100 km, and the average speed from the transit distribution center to the retail supermarket is 40 km / h and the distance is 80 km. Combined with the path congestion probability distribution, if the congestion probability from the collection center to the transit distribution center is 30%, this will cause the speed to decrease, and the expected total transportation time is calculated.

[0034] The expected total transportation time is compared with the cargo freshness limit to determine the timeliness achievement rate score. Since the freshness limit of strawberries is 48 hours, if the calculated expected total transportation time is 40 hours, then the timeliness achievement rate score is higher, because the negative deviation of the expected total transportation time relative to the freshness limit is smaller. This score can be an inverse proportional function, such as 1-(40-48)÷48=1-(-1 / 6)=7 / 6 (this is just an example calculation method).

[0035] Based on the distribution of fresh-keeping resource demand at the corresponding node in the forecast value of the goods' fresh-keeping demand, combined with the current available refrigeration equipment capacity and energy supply status, the fresh-keeping reliability score is calculated. For example, in the transit distribution center, according to the forecast value of the goods' fresh-keeping demand, 50% of the refrigeration equipment capacity is required to maintain the freshness of the strawberries, and the current available refrigeration equipment capacity of the transit distribution center is 80%, and the energy supply status is good, so the resource matching degree in the fresh-keeping reliability score is high. At the same time, considering the temperature control capability of the refrigeration equipment, if the temperature control accuracy is high at 0-4 degrees Celsius, then the fresh-keeping reliability score will also be high. This score is the weighted average of the node resource matching degree and the temperature control capability.

[0036] For beef transportation, the starting and ending points are from the slaughterhouse to the restaurant. Similarly, the expected total transportation time of the candidate transportation routes is calculated, such as the slaughterhouse-meat processing plant-cold chain logistics hub-meat wholesaler-restaurant route. The expected total transportation time is calculated based on the distance of each segment, the average speed, and the probability distribution of path congestion. The timeliness achievement rate score is determined by comparing with the 72-hour fresh-keeping time limit of beef. According to the distribution of fresh-keeping resource requirements of each node (such as the cold chain logistics hub) in the predicted value of the goods' fresh-keeping demand, the fresh-keeping reliability score is calculated in combination with the capacity of the refrigeration equipment and the energy supply status. Finally, the fresh-keeping reliability score and the timeliness achievement rate score of each candidate transportation route for strawberry and beef transportation are dynamically weighted, and the comprehensive priority ranking of each candidate transportation route is generated. For example, for a candidate strawberry transportation route, the weight of the fresh-keeping reliability score is 0.6, and the weight of the timeliness achievement rate score is 0.4. The priority among all candidate routes is obtained by comprehensive calculation.

[0037] Step S150, based on the weighted results of the freshness-keeping reliability score and the timeliness achievement rate score, dynamically adjust the node sequence of the current transportation path, and update the input data of the cold chain demand prediction model in real time to optimize subsequent path decisions.

[0038] Taking strawberry transportation as an example, real-time temperature data, traffic status update information and abnormal operation events of refrigeration equipment at each node are monitored during transportation. The IoT sensors deployed on the transport vehicles collect real-time temperature distribution data inside the carriage. If the sensor detects that the temperature distribution inside the carriage is uneven and the local temperature reaches 6 degrees Celsius during transportation from the collection center to the transit distribution center, this abnormal temperature fluctuation pattern is identified. The real-time traffic data interface of the traffic management system is obtained to learn that there is a road construction event on the road section from the transit distribution center to the retail supermarket. This is the traffic status update information. Monitor the compressor operating frequency, refrigerant pressure and energy consumption rate of the refrigeration equipment, and generate an equipment health status assessment report. If the refrigerant pressure is detected to be lower than the normal level, this is an abnormal operation event of the refrigeration equipment.

[0039] When the abnormal temperature fluctuation duration corresponding to the abnormal temperature fluctuation pattern in the carriage exceeds the preset tolerance threshold (for example, 15 minutes), the stay time limit of the subsequent nodes is adjusted according to the continuous deviation of the real-time temperature data from the preset temperature threshold. For example, the original plan is to stay at the transit distribution center for 1 hour. Due to temperature fluctuations, in order to ensure the freshness of the strawberries, the stay time is shortened to 30 minutes, and the freshness reliability score is recalculated.

[0040] If it is detected that the traffic status update information indicates that the congestion probability of the path from the transit distribution center to the retail supermarket exceeds the preset risk threshold (for example, 50%), the regeneration process of the candidate transportation path is triggered. According to the risk node location (the section from the transit distribution center to the retail supermarket) and the impact radius (for example, within 10 kilometers) in the path congestion probability distribution corresponding to the traffic status update information, the geographical boundary range of the detour area is determined. Within this range, alternative nodes with refrigerated facilities are screened, such as a small cold chain warehouse is detected within the detour range and has sufficient refrigeration equipment. The compatibility data of the refrigeration equipment type of the alternative node and the current transportation of strawberries (for example, whether the temperature control range is suitable for strawberry preservation) are extracted, and combined with the resource demand distribution in the cargo preservation demand forecast value, the alternative nodes with qualified compatibility data are matched to generate emergency path branches, such as the detour main path, temporary transit path, etc. According to the real-time collected strawberry temperature stability index, the branches that exceed the remaining fresh-keeping time constraint are excluded from the emergency path branches to generate a valid emergency path subset. The subset of effective emergency paths is input into the cold chain demand forecasting model, and re-sorted according to the dynamic weights of the freshness-keeping reliability score and the timeliness achievement rate score, and the target emergency path with the highest priority is output.

[0041] Feedback the adjusted node sequence (such as adding new replacement nodes), real-time temperature data and traffic status update information to the cold chain demand forecasting model to iteratively optimize the accuracy of the predicted value of the cargo preservation demand. For example, the actual transportation results of the target emergency path (including new nodes and path information) are analyzed with the predicted preservation reliability score to generate a path decision deviation index. If it is detected that the preservation resource demand at a certain node in the actual transportation is higher than the predicted value, adjust the convolution kernel size of the time series convolutional neural network according to the path decision deviation index to enhance the local feature extraction ability of sudden congestion events. Based on the strawberry quality feedback data after the execution of the target emergency path, optimize the forgetting gate threshold of the time-dependent correction in the long-term and short-term memory network. In a preset period (such as every day), the adjusted node sequence, real-time temperature data, traffic status update information and abnormal operation events of refrigeration equipment are integrated into an incremental training data set by timestamp, and the incremental training data set is injected into the cold chain demand forecasting model through an online learning algorithm to update the model weight information of the cold chain demand forecasting model and retain the high-value patterns in the historical path decision.

[0042] The same process applies to beef transportation. Real-time temperature data, traffic status update information, and abnormal operation of refrigeration equipment are monitored during transportation. When abnormal temperature fluctuations or traffic congestion risks occur, adjustments are made according to the above process, such as adjusting the length of stay at subsequent nodes, regenerating candidate transportation routes, and optimizing the cold chain demand forecasting model based on feedback data, to ensure that the beef can reach the destination within the specified preservation conditions and transportation time.

[0043] Based on the above steps, the embodiment of the present application can comprehensively capture the key information in the cold chain transportation process by acquiring a sample cold chain transportation order data set that includes a transportation node sequence, cargo preservation requirement parameters, ambient temperature fluctuation records, and transportation time completion status, so that subsequent analysis and model training are no longer limited to a single dimension, but are based on multi-dimensional, multi-variable real-world scenario data.

[0044] When extracting transportation features from the sample cold chain transportation order data set, a combined transportation feature set including time-sensitive features, environmental stability features, and cargo deterioration risk features is generated, breaking the previous limitation of focusing on a single factor and comprehensively and comprehensively describing the complex characteristics of cold chain transportation. The time-sensitive feature ensures that the transportation timeliness is guaranteed, the environmental stability feature takes into account the impact of the external environment on the preservation of goods, and the cargo deterioration risk feature is directly related to the quality and safety of goods. The combination of the three provides a more realistic, deeper and broader feature input for subsequent model training, greatly improving the model's understanding and prediction capabilities for cold chain transportation scenarios.

[0045] Based on this combined transport feature set, the cold chain demand forecasting model is called to generate accurate forecast values ​​of cargo preservation demand and path congestion probability distribution in the target area within a preset time range. Unlike traditional simple estimation or empirical methods, this cold chain demand forecasting model is based on a large amount of actual data and in-depth feature analysis, which enables intelligent and accurate forecasts of cargo preservation demand and path congestion. The cargo preservation demand forecast value provides a scientific basis for the temperature control strategy during transportation, and the path congestion probability distribution provides early warning of possible traffic conditions, helping transportation planners to prepare for response in advance, greatly improving the foresight and scientific nature of transportation decisions.

[0046] According to the predicted value of the goods preservation demand and the probability distribution of path congestion, multiple candidate transportation routes are generated, and each candidate transportation route is scored for preservation reliability and timeliness. Traditional path planning often only focuses on a single goal, such as the shortest path or the fastest arrival time. This method takes into account both the reliability of goods preservation and the timeliness achievement rate, fully considers the particularity of cold chain transportation, and ensures that the goods are delivered to the destination in a good state of preservation within the specified time. This not only improves the transportation quality of goods and reduces the economic losses caused by deterioration, but also enhances customer satisfaction and improves the competitiveness of enterprises in the cold chain transportation market.

[0047] Finally, based on the weighted results of the freshness reliability score and the timeliness achievement rate score, the node sequence of the current transportation path is dynamically adjusted, and the input data of the cold chain demand forecasting model is updated in real time to optimize subsequent path decisions, so that the transportation path can continuously optimize itself according to the actual situation and adapt to the ever-changing transportation environment. In the face of sudden weather changes, traffic congestion and other conditions, it can respond quickly and adjust the path in time to ensure that the transportation process always remains efficient and reliable. At the same time, real-time updating of the model input data enables the cold chain demand forecasting model to continuously learn new information, further improving the accuracy of the forecast and the adaptability of the model.

[0048] In a possible implementation, step S120 includes: Step S121, extracting the dwell time data of each transport node, the number of times the ambient temperature deviates from the threshold, and the cargo packaging integrity record from the sample cold chain transport order data set.

[0049] In this embodiment, for strawberry transportation, for example, at the transportation node from the picking farm to the collection center, the planned stay time is 2 hours, but the actual stay time reaches 3.5 hours; from the collection center to the regional cold chain warehouse, the planned stay is 1.5 hours, and the actual stay is 2.5 hours. In terms of the number of times the ambient temperature deviates from the threshold, during the transportation from the regional cold chain warehouse to the transit distribution center, due to the brief failure of the refrigeration equipment, the temperature in the carriage deviated from the threshold of 0-4 degrees Celsius by 2 times. The cargo packaging integrity record shows that the packaging of some strawberries was squeezed during transportation, and 10% of the packaging was damaged.

[0050] Step S122, determining the time-sensitive feature according to the difference between the dwelling time data and the preset cargo preservation time limit, wherein the time-sensitive feature is used to identify the risk level of exceeding the cargo preservation time limit in the transport node sequence.

[0051] For example, the shelf life of strawberries is limited to 48 hours. They stay an extra 1.5 hours from the picking farm to the collection center, and an extra hour from the collection center to the regional cold chain warehouse. This means that a lot of shelf life has been consumed in the early stages of transportation, and the risk level of exceeding the shelf life limit in the transportation node sequence is at a high level. This high risk level indicates that more stringent time control is required in the subsequent transportation links to prevent the strawberries from deteriorating due to excessive time.

[0052] Step S123, calculating the environmental stability characteristic according to the number of times the ambient temperature deviates from the threshold value and the proportion of the deviation duration, wherein the environmental stability characteristic is used to characterize the cumulative impact of temperature fluctuations on the quality of the goods during transportation.

[0053] For example, during the transportation from the regional cold chain warehouse to the transit distribution center, the temperature deviated from the threshold twice, and each deviation lasted about 20 minutes. The total transportation time was 3 hours, and the deviation duration accounted for (2×20)÷(3×60)≈22.2%. This percentage reflects that the cumulative impact of temperature fluctuations on the quality of strawberries during transportation is more obvious, which may accelerate the deterioration of strawberries, because strawberries are very sensitive to temperature, and even short temperature fluctuations may affect their freshness and quality.

[0054] Step S124, combining the cargo packaging integrity record and the deterioration rate parameter corresponding to the cargo type, generating the cargo deterioration risk feature, wherein the cargo deterioration risk feature is used to quantify the contribution of different nodes in the transportation path to the deterioration of cargo quality.

[0055] For example, strawberries are perishable fruits that deteriorate quickly. In addition, 10% of the packaging is damaged, which increases the possibility of strawberries being exposed to the external environment, further increasing the contribution of different nodes in the transportation route to the decline in cargo quality. For example, in the transit distribution center, if the damaged strawberries are not handled in time, they will deteriorate significantly faster during the subsequent transportation to the retail supermarket due to temperature fluctuations and transportation time, thus affecting the quality of the entire batch of strawberries.

[0056] For beef transportation, the planned stay from the slaughterhouse to the meat processing plant was 3 hours, but the actual stay was 4 hours; from the meat processing plant to the cold chain logistics hub, the planned stay was 2 hours, but the actual stay was 2.5 hours. During the transportation from the cold chain logistics hub to the meat wholesaler, due to changes in the external ambient temperature and minor malfunctions in the refrigeration equipment, the temperature in the carriage deviated from the threshold of -18 degrees Celsius once, which lasted for 10 minutes. In terms of the packaging integrity of the beef, 5% of the packaging was slightly scratched.

[0057] The shelf life limit of beef is 72 hours, and it stays for an extra hour from the slaughterhouse to the meat processing plant, which increases the risk level of exceeding the shelf life limit, but the risk level is lower than that of strawberries. When calculating the environmental stability characteristics, the transportation time from the cold chain logistics hub to the meat wholesaler is 4 hours, the temperature deviates from the threshold once, and each deviation lasts for 10 minutes. The deviation duration accounts for (1×10)÷(4×60)≈4.2%, indicating that the cumulative impact of temperature fluctuations on beef quality is relatively small. Combined with the deterioration rate of beef (slower than strawberries) and the 5% damage of the packaging, although the overall risk of cargo deterioration is lower than that of strawberries, the damage of the packaging will still increase the contribution of different nodes in the transportation path to the decline in cargo quality to a certain extent, especially during long-term transportation, the beef at the damaged part may be contaminated by microorganisms and deteriorate.

[0058] Step S125, normalizing the time-sensitive features, environmental stability features, and cargo deterioration risk features to generate the combined transport feature set.

[0059] Finally, this embodiment normalizes the time-sensitive characteristics, environmental stability characteristics, and cargo deterioration risk characteristics of strawberry and beef transportation to generate their own combined transportation feature sets. This normalization process is to unify the feature values ​​of different magnitudes and ranges into a standard range so that effective data processing and analysis can be performed in subsequent operations such as cold chain demand forecasting model training. For example, the time-sensitive feature value in strawberry transportation may be between 0.6-0.8 (hypothesis), the environmental stability feature value is between 0.5-0.7, and the cargo deterioration risk feature value is between 0.7-0.9; for beef transportation, the time-sensitive feature value may be between 0.3-0.5, the environmental stability feature value is between 0.2-0.4, and the cargo deterioration risk feature value is between 0.4-0.6. These combined transportation feature sets will serve as an important data basis for subsequent operations, accurately reflecting the various characteristics and risk factors of different fresh products during transportation.

[0060] In a possible implementation, step S130 includes: Step S131, input the combined transport feature set into a temporal convolutional neural network, extract the dynamic association pattern between transport nodes, and output the initial path demand prediction result.

[0061] In the fresh cold chain transportation scenario of this embodiment, the transportation of strawberries and beef is used as an example for explanation.

[0062] Strawberry transportation: The time-sensitive eigenvalue reflects the risk level of exceeding the shelf life limit at the transportation node. For example, in the transportation of strawberries, from the picking farm A to the collection center B, if the transportation is usually carried out in the morning, the roads are unobstructed, the staff is efficient, and the transportation time is stable, the time-sensitive eigenvalue of this node is low, indicating that the risk of exceeding the shelf life limit at this stage is small. However, from the collection center B to the regional cold chain warehouse C, if the shipment volume of the collection center increases significantly, resulting in longer vehicle loading and dispatching time and transportation delays, the time-sensitive eigenvalue will increase, which means that the risk of exceeding the shelf life limit increases.

[0063] This time-sensitive characteristic value directly affects the estimated transportation time and fresh-keeping resource demand in the initial route demand forecast results. When the initial route demand forecast results analyze the transportation from B to C, due to the increase in the time-sensitive characteristic value, the forecast will take into account the possible impact of transportation delays on the freshness of strawberries, and thus estimate that the transportation time will increase. In order to ensure the freshness of strawberries during the extended transportation time, it will be preliminarily estimated that the corresponding fresh-keeping resources need to be increased, such as increasing the refrigeration power or increasing the refrigeration time, to slow down the deterioration of strawberries.

[0064] Beef transportation: In beef transportation, from slaughterhouse F to meat processing plant G, the transportation time is relatively stable due to the relatively fixed workflow of the slaughterhouse, and the time-sensitive feature value is low. However, during holidays, the shipment volume from meat processing plant G to cold chain logistics hub H increases, the waiting time of transportation vehicles is extended, and the time-sensitive feature value increases.

[0065] For the initial route demand forecast results, when considering the transportation from G to H, the increase in the time-sensitive feature value makes the forecast aware of the possible delay in transportation, and then estimates the increase in transportation time. At the same time, in order to ensure that the quality of beef is not greatly affected during the extended transportation time, the demand for fresh-keeping resources will be adjusted in a preliminary estimate, such as more precise temperature control or additional fresh-keeping packaging materials may be required to maintain the freshness of the beef.

[0066] For the environmental stability characteristic, strawberry transportation: The environmental stability characteristic value is calculated based on the number of times the ambient temperature deviates from the suitable temperature range for strawberry preservation (assuming it is 0-5 degrees Celsius) and the percentage of the deviation duration, and is used to characterize the cumulative impact of temperature fluctuations on strawberry quality during transportation. If during the transportation process from the regional cold chain warehouse C to the transit distribution center D, the temperature sensor records that the temperature frequently deviates from the suitable range, and the deviation duration accounts for a large proportion, then the environmental stability characteristic value is low, indicating that temperature fluctuations have a great impact on the quality of strawberries.

[0067] The initial route demand forecast results will be adjusted accordingly based on this environmental stability characteristic value. Since large temperature fluctuations may accelerate the deterioration of strawberries, the forecast will consider increasing refrigeration resources during this transportation stage or using more advanced temperature control equipment to stabilize the transportation environment temperature and ensure the freshness of the strawberries. At the same time, in the estimation of transportation time, the time changes caused by the possible need for additional measures due to temperature issues may also be considered.

[0068] Beef transportation: For beef transportation, the environmental stability characteristic value is calculated based on the temperature deviation from the suitable temperature range for beef preservation (assuming -18-15 degrees Celsius). For example, during the transportation from cold chain logistics hub H to meat wholesaler I, if the number of times the temperature deviates from the suitable range is small and the deviation duration is small, the environmental stability characteristic value is relatively high, indicating that temperature fluctuations have little impact on beef quality.

[0069] In the initial route demand forecast results, considering the good environmental stability, the preliminary estimate of the demand for fresh-keeping resources may be relatively stable and will not be adjusted significantly. However, if the environmental stability characteristic value is low, that is, the temperature fluctuates greatly, it is predicted that there may be an increase in the demand for temperature regulation equipment, such as equipment with stronger refrigeration capacity or more precise temperature monitoring instruments, to ensure the quality of beef during transportation.

[0070] For the risk characteristics of cargo deterioration, strawberry transportation: the cargo deterioration risk characteristic value is generated by combining the strawberry packaging integrity record and the high deterioration rate parameter of the strawberry itself, which is used to quantify the contribution of different nodes in the transportation path to the deterioration of strawberry quality. Assuming that at the transit distribution center D, some strawberry packaging is found to be squeezed and damaged, and the strawberry itself deteriorates quickly, then the cargo deterioration risk characteristic value at this node is high.

[0071] The initial route demand forecast results focus on this situation and will focus on increasing the protection measures and preservation resources for strawberries at this node and in the subsequent transportation stages. The demand for replacing packaging materials may increase, or the use of preservatives and other preservation substances may increase to reduce the risk of goods spoilage. At the same time, in terms of transportation time estimation, the time consumption caused by dealing with the risk of goods spoilage may also be considered.

[0072] Beef transportation: In beef transportation, the cargo spoilage risk characteristic value is also generated by combining the beef packaging integrity and the relatively low spoilage rate parameters. For example, at meat processing plant G, if the beef packaging is damaged and not handled in time, and considering the beef spoilage rate, the cargo spoilage risk characteristic value at this node will increase.

[0073] The initial route demand forecast results will adjust the demand for fresh-keeping resources based on this increased risk characteristic value of goods spoilage. It may increase the secondary packaging of beef or strengthen the fresh-keeping treatment, such as increasing the injection of fresh-keeping gas. In the transportation time estimation, the time spent on operations related to handling the risk of goods spoilage will also be considered to ensure that the quality of beef is not greatly affected during subsequent transportation.

[0074] Let's first look at strawberry transportation. Its transportation nodes include picking farm A, collection center B, regional cold chain warehouse C, transit distribution center D and retail supermarket E. The combined transportation feature set covers the time-sensitive feature values, environmental stability feature values ​​and cargo deterioration risk feature values ​​of each node. In terms of time-sensitive feature values, for example, if the transportation from A to B is in the morning, the transportation time is relatively stable due to the small number of vehicles on the road and the high efficiency of the staff, which makes the time-sensitive feature value relatively low, indicating that the risk of exceeding the shelf life limit is low; but for transportation from B to C, if the collection center has a large shipment volume, the vehicle loading and scheduling time increases, resulting in transportation delays, and the time-sensitive feature value will increase. In terms of environmental stability feature values, for example, in some transportation sections, the temperature sensor records the number of times the temperature deviates from the suitable temperature range for strawberry preservation (assuming 0-5 degrees Celsius) and the proportion of deviation duration, and the environmental stability feature value is calculated based on this. If the number of deviations is large and the duration is large, the feature value is low, which means that the temperature fluctuation has a large cumulative impact on the quality of strawberries. The cargo deterioration risk characteristic value is generated by combining the strawberry packaging integrity record (such as whether there is any squeezing damage, etc.) and the strawberry's own deterioration rate parameter (since strawberries are easy to deteriorate, the deterioration rate is relatively high).

[0075] The combined transport feature set of strawberry transport is input into the time series convolutional neural network. For the first convolutional layer of the network, the convolution kernel size is assumed to be 7×3, where 7 represents the window size on the transport node sequence (time dimension), 3 matches the dimension of the three-dimensional feature vector (time sensitivity, environmental stability, and cargo deterioration risk feature value) of each node, and the step size is set to 1. The convolution kernel slides on the combined transport feature set. For example, starting from the first node A, the three-dimensional feature vectors of the seven nodes from A to A+6 will be multiplied and added through a pre-set weight matrix to capture local dynamic association patterns such as the impact of the change in transport time from A to B on the environmental stability and cargo deterioration risk of subsequent nodes. After the convolution operation, the nonlinear transformation is performed through the ReLU activation function. Multiple convolution kernels with different weight matrices work in parallel. Assuming that the first convolutional layer has 15 convolution kernels, a feature map with the same number of nodes but a feature dimension of 15 is obtained after the operation. Then enter the pooling layer, use maximum pooling, pooling window size 3×1, step size 3, take the maximum value in each pooling window, reduce the amount of data while retaining important features, and obtain a new feature map. After multiple convolutional layers and pooling layers are processed alternately, the initial path demand prediction result is output.

[0076] This result includes the estimated transportation time, possible delays, and a preliminary estimate of the demand for fresh-keeping resources at each node. For example, the transportation from C to D is predicted to take a long time and has a high risk of temperature fluctuations (reflected by the low environmental stability characteristic value). It is initially estimated that 10% more refrigeration resources will be needed to maintain the freshness of the strawberries. This is a comprehensive consideration of the dynamic relationship between the time sensitivity, environmental stability, and cargo deterioration risk characteristic values ​​of each node during the transportation process.

[0077] Looking at beef transportation, the transportation nodes include slaughterhouse F, meat processing plant G, cold chain logistics hub H, meat wholesaler I and restaurant J. Its combined transportation feature set is also composed of the time sensitivity, environmental stability and cargo deterioration risk feature values ​​of each node. In terms of time-sensitive feature values, the transportation from F to G is relatively stable due to the fixed workflow of the slaughterhouse, and the feature value is low; but during holidays, the transportation from G to H has a long waiting time for vehicles due to the increase in shipments from the processing plant, and the feature value increases. The environmental stability feature value is calculated based on the temperature deviation from the suitable temperature range for beef preservation (assuming -18--15 degrees Celsius). The cargo deterioration risk feature value is generated by combining the beef packaging integrity and beef deterioration rate parameters (relatively low deterioration rate of strawberries).

[0078] The combined transport feature set of beef transportation is input into the time series convolutional neural network, which extracts dynamic association patterns between transportation nodes in a similar way. For example, the relationship between the transportation time from F to G is stable, but the transportation time from G to H is affected by the shipment volume and vehicle scheduling, and finally the initial path demand forecast result of beef transportation is output, including the estimated transportation time of each node, possible delay factors and preliminary judgment of the demand for fresh-keeping resources. For example, the transportation time from H to I is expected to be 5 hours. Because the temperature fluctuation is small and the time is not long (judged from the environmental stability and time-sensitive characteristic values), the preliminary estimate of the demand for fresh-keeping resources is the same as usual.

[0079] Step S132, using a long short-term memory network to perform time-dependency correction on the initial path demand forecast result, and generate a corrected cargo preservation demand forecast value, wherein the corrected cargo preservation demand forecast value includes the distribution of preservation resource demand of different transportation nodes in the transportation path.

[0080] For strawberry transportation, due to the strict restrictions on the preservation time, the freshness decreases rapidly over time. The LSTM network takes this time dependency into account to adjust the initial prediction results. For example, in the initial prediction, the demand for preservation resources from the transit distribution center D to the retail supermarket E is estimated based on the shorter transportation time and lower temperature fluctuation risk. However, the LSTM network considers the accumulated preservation time from the picking farm A. If a lot of preservation time has been consumed by D, even if the transportation time and temperature fluctuation risk from D to E remain unchanged, additional preservation resources are required.

[0081] The input gate, forget gate, output gate and memory cell of the long short-term memory network play a role. Taking node D as an example, the input gate combines the initial path demand prediction result of the current node (such as the estimation of fresh-keeping resource demand based on the short time from D to E and the low temperature fluctuation risk) and the hidden state transmitted by the previous node C (including the information related to the fresh-keeping resource demand of node C and the previous transportation process, and affected by the eigenvalues ​​at that time), and obtains the input gate vector through the weight matrix operation to determine the degree to which the new information enters the memory cell. The forget gate also calculates the forget gate vector with the input information to determine the degree to which the old information is retained in the memory cell. If the environment in the previous transportation process is relatively stable and the demand for temperature regulation resources is stable, but the eigenvalue of the environmental stability deteriorates at node D, the forget gate may reduce the degree to which the old information of the corresponding temperature regulation resource demand is retained. The memory cell is updated in combination with the input gate and forget gate results, remembers the long-term dependence of the fresh-keeping resource demand in the transportation process, and adjusts according to the dynamic changes of each eigenvalue. The output gate calculates the output gate vector with the memory cell state to determine the output hidden state, which integrates the information of each node and the influence of the eigenvalue change.

[0082] After correction, the generated revised forecast value of cargo preservation demand contains the distribution of preservation resource demand at different transportation nodes. At picking farm A, as the starting point of transportation, the demand for preservation resources is mainly to ensure that strawberries quickly enter a suitable low-temperature environment, with a demand ratio of about 20%; at the collection center B, in addition to maintaining low temperature, sorting and packaging are also required, and the demand ratio increases to 30%; the regional cold chain warehouse C, due to the possibility of long-term storage waiting for transshipment, the demand ratio reaches 40%; the transit distribution center D, considering transportation losses and remaining preservation time, the demand ratio is increased to 50%; the retail supermarket E, to ensure freshness on the shelf, the demand ratio remains at about 30%.

[0083] For beef transportation, the LSTM network also performs time-dependent corrections. Although beef spoils slower than strawberries, long-term transportation is also affected by time. For example, if the initial forecast is from cold chain logistics hub H to meat wholesaler I, the estimate of fresh-keeping resource demand may not fully consider the cumulative transportation time starting from slaughterhouse F.

[0084] Long-term and short-term memory networks work in a similar way. Taking node H as an example, the input gate, forget gate, memory cell and output gate work together to adjust the demand for fresh-keeping resources according to the changes in the time sensitivity, environmental stability and risk characteristics of goods deterioration at each node in the transportation process. The revised forecast value of the goods' fresh-keeping demand contains the distribution of fresh-keeping resource demand at each node. At slaughterhouse F, the demand for fresh-keeping resources is mainly to ensure that the beef quickly enters a low-temperature environment of -18 degrees Celsius, with a demand ratio of about 15%; meat processing plant G, in addition to maintaining low temperature, also requires processing, and the demand ratio increases to 20%; cold chain logistics hub H, due to the possibility of long-term storage waiting for transshipment, the demand ratio reaches 25%; meat wholesaler I, considering transportation losses and remaining fresh-keeping time, the demand ratio is increased to 30%; restaurant J, to ensure freshness before processing, the demand ratio remains at about 20%.

[0085] Step S133, based on the sample traffic flow data and weather event records of the target area, a path congestion probability prediction sub-model is constructed, and the path congestion probability prediction sub-model outputs the path congestion probability distribution in the future time period by analyzing the correlation between weather events and traffic flow changes.

[0086] Then, a path congestion probability prediction submodel is constructed based on the sample traffic flow data and weather event records of the target area (such as a city and its surrounding areas). For strawberry transportation, the traffic flow data of the target area shows that the roads in the city center area will be congested during the morning and evening rush hours on weekdays. Weather event records show that the road capacity will be greatly reduced during heavy rain in summer and snow in winter. The path congestion probability prediction submodel outputs the path congestion probability distribution in the future time period by analyzing the correlation between weather events and traffic flow changes. For example, in the transportation path from the regional cold chain warehouse C to the transit distribution center D, if it passes through the city center area, the congestion probability may reach 30% during the morning rush hour (7-9 o'clock) on weekdays; if it encounters heavy rain, this congestion probability may further increase to 50%, because heavy rain will cause water accumulation on the road, slow down the speed of vehicles, and even cause traffic paralysis on local sections.

[0087] For beef transportation, a path congestion probability prediction sub-model is also constructed. Traffic flow data in the target area show that during holidays, roads leading to commercial and residential areas will be congested. Weather event records show that in foggy weather, road visibility is reduced and traffic flow is restricted. In the transportation route from meat wholesaler I to restaurant J, if it passes through the commercial area, the congestion probability may reach 20% in the afternoon of holidays (14-16 o'clock); if there is foggy weather, the congestion probability may increase to 30%, because in foggy weather, drivers will reduce their speed for safety, resulting in a decrease in road capacity.

[0088] Step S134, integrating the corrected cargo preservation demand prediction value with the path congestion probability distribution to generate a joint output of the cold chain demand prediction model, wherein the joint output is used to identify the preservation resource occupancy rate and congestion delay risk of different candidate transportation paths.

[0089] Finally, the revised cargo preservation demand forecast values ​​for strawberry and beef transportation are combined with the path congestion probability distribution to generate the joint output of the cold chain demand forecast model. This joint output is used to identify the preservation resource occupancy rate and congestion delay risk of different candidate transportation paths. For example, for a candidate transportation path ABCDE for strawberry transportation, the preservation resource occupancy rate of the entire path is determined by the distribution of preservation resource demand at each node according to the revised cargo preservation demand forecast value. If during the transportation from C to D, due to the high demand for preservation resources and the high probability of road congestion (such as 50%), the congestion delay risk of this path will increase. The preservation resource occupancy rate may reach 70%, and the congestion delay risk is 30%. For a candidate transportation path FGHIJ for beef transportation, the preservation resource occupancy rate is determined according to the revised cargo preservation demand forecast value, and combined with the path congestion probability distribution, if during the transportation from H to I, the road congestion probability is 30%, then the preservation resource occupancy rate of this path may be 60%, and the congestion delay risk is 20%. Through this joint output, a comprehensive evaluation of different candidate transportation routes can be carried out, providing an important basis for the subsequent selection of the optimal transportation route, ensuring that strawberries and beef can meet the preservation requirements during transportation while minimizing the impact of congestion and delays.

[0090] In a possible implementation, step S140 includes: Step S141, calculating the expected total transportation time of each candidate transportation path according to the node sequence length of the transportation path, the sample average travel speed and the congestion probability distribution of the path.

[0091] Step S142, comparing the expected total transportation time with the cargo preservation time limit to determine the timeliness achievement rate score, which is an inversely proportional function of the negative deviation degree of the expected total transportation time relative to the cargo preservation time limit.

[0092] Step S143, based on the distribution of preservation resource demand of the corresponding node in the cargo preservation demand prediction value, combined with the current available refrigeration equipment capacity and energy supply status, calculate the preservation reliability score, which is a weighted average of the node resource matching degree and the temperature control capability.

[0093] Step S144, dynamically weighting the freshness-keeping reliability score and the timeliness achievement rate score to generate a comprehensive priority ranking for each candidate transportation route.

[0094] For example, for strawberry transportation, the starting point and end point of the transportation are from the picking farm to the retail supermarket. First, the expected total transportation time of each candidate transportation path is calculated based on the node sequence length of the transportation path, the sample average speed, and the probability distribution of path congestion. Assume that there are the following three candidate transportation paths: Path 1 is picking farm-collection center-regional cold chain warehouse-transit distribution center-retail supermarket, and the node sequence length of this path is 5 nodes; Path 2 is picking farm-collection center-transit distribution center-retail supermarket, and the node sequence length is 4 nodes; Path 3 is picking farm-regional cold chain warehouse-retail supermarket, and the node sequence length is 3 nodes.

[0095] In terms of the average speed of samples, the average speed from the picking farm to the collection center is 60 km / h, and the distance is 120 km; the average speed from the collection center to the regional cold chain warehouse is 50 km / h, and the distance is 100 km; the average speed from the regional cold chain warehouse to the transit distribution center is 40 km / h, and the distance is 80 km; the average speed from the transit distribution center to the retail supermarket is 30 km / h, and the distance is 60 km. Combined with the probability distribution of path congestion, for example, the probability of congestion from the collection center to the regional cold chain warehouse in path 1 is 30%, which will reduce the actual speed to 35 km / h (assuming that it is calculated according to the model of the impact of congestion on speed), and the probability of congestion from the regional cold chain warehouse to the transit distribution center is 20%, and the actual speed becomes 32 km / h.

[0096] Based on the above data, the expected total transportation time of each route is calculated. Route 1: (120÷60+100÷50+80÷35+60÷32)≈2+2+2.29+1.88=8.17 hours; Route 2: (120÷60+100÷35+60÷30)≈2+2.86+2=6.86 hours; Route 3: (120÷50+60÷30)≈2.4+2=4.4 hours.

[0097] The expected total transportation time is compared with the fresh-keeping time limit of strawberries (48 hours) to determine the timeliness achievement rate score. Since the timeliness achievement rate score is an inversely proportional function of the negative deviation of the expected total transportation time relative to the fresh-keeping time limit, let the timeliness achievement rate score be S, the fresh-keeping time limit be T, and the expected total transportation time be t, then S=k÷(Tt) (k is a constant, here we assume k=100 for ease of calculation and explanation). For path one, S1=100÷(48-8.17)≈2.5; for path two, S2=100÷(48-6.86)≈2.3; for path three, S3=100÷(48-4.4)≈2.2.

[0098] Based on the distribution of fresh-keeping resource demand at the corresponding node in the predicted value of strawberry goods fresh-keeping demand, combined with the current available refrigeration equipment capacity and energy supply status, the fresh-keeping reliability score is calculated. Assuming that in path one, the predicted value of fresh-keeping resource demand at the collection center is 30%, and the current available refrigeration equipment capacity of the collection center is 50%, the energy supply status is good, and the temperature control capacity can meet the requirements of strawberries at 0-4 degrees Celsius, the node resource matching weight is 0.6, and the temperature control capacity weight is 0.4. Then the fresh-keeping reliability score of the collection center is (0.6×(50÷30)+0.4×1)=1.4 (here it is assumed that 1 means that the temperature control requirements are fully met). The fresh-keeping reliability scores of other nodes in the path are calculated in the same way, and then the weighted average is used to obtain the fresh-keeping reliability score of path one. Similarly, the fresh-keeping reliability scores of paths two and three are calculated.

[0099] Finally, the freshness reliability score and the timeliness achievement rate score are dynamically weighted to generate the comprehensive priority ranking of each candidate transportation path. Assume that the weight of the freshness reliability score is 0.6 and the weight of the timeliness achievement rate score is 0.4. For path one, the comprehensive priority is 0.6×freshness reliability score + 0.4×2.5; for path two, the comprehensive priority is 0.6×freshness reliability score + 0.4×2.3; for path three, the comprehensive priority is 0.6×freshness reliability score + 0.4×2.2. By comparing the values ​​of these comprehensive priorities, the comprehensive priority ranking of the three candidate transportation paths is determined.

[0100] For beef transportation, the starting point and end point are from the slaughterhouse to the restaurant. There are also candidate transportation routes, such as route one is slaughterhouse-meat processing plant-cold chain logistics hub-meat wholesaler-restaurant, route two is slaughterhouse-meat processing plant-meat wholesaler-restaurant, and route three is slaughterhouse-cold chain logistics hub-restaurant. According to the above calculation method for strawberry transportation, the expected total transportation time is calculated based on the distance of each section, the average sample speed, and the probability distribution of path congestion. For example, the average speed from the slaughterhouse to the meat processing plant is 50 kilometers per hour and the distance is 100 kilometers; the average speed from the meat processing plant to the cold chain logistics hub is 40 kilometers per hour and the distance is 80 kilometers, etc. The expected total transportation time of each path is calculated in combination with the congestion probability. Compared with the cargo freshness limit of beef (72 hours), the timeliness achievement rate score is determined.

[0101] Then, based on the distribution of fresh-keeping resource demand at the corresponding node in the predicted value of beef cargo fresh-keeping demand, the fresh-keeping reliability score is calculated in combination with the current available refrigeration equipment capacity and energy supply status. For example, in path 1, the predicted value of the fresh-keeping resource demand of the meat processing plant is 20%, the current available refrigeration equipment capacity is 30%, the energy supply status is normal, and the temperature control capacity meets the requirement of -18 degrees Celsius. The fresh-keeping reliability score of the node is calculated according to the weight, and then the fresh-keeping reliability score of the entire path is obtained. Finally, the fresh-keeping reliability score and the timeliness achievement rate score are dynamically weighted, and the comprehensive priority ranking of each candidate transportation route is determined, so as to provide a basis for selecting the optimal transportation route for beef transportation, ensuring that the beef can meet the fresh-keeping requirements and arrive at the destination on time during transportation.

[0102] In a possible implementation, step S150 includes: Step S151, monitor the real-time temperature data, traffic status update information and abnormal operation events of refrigeration equipment at each node during transportation.

[0103] In this embodiment, for strawberry transportation, the real-time temperature data, traffic status update information and refrigeration equipment operation abnormal events of each node need to be strictly monitored during transportation. During the transportation process from the picking farm to the collection center, the temperature data inside the carriage is collected in real time by the temperature sensor installed on the transport vehicle. Assuming that the preset temperature threshold of strawberries is 0-4 degrees Celsius, if the temperature of a certain area in the carriage is detected to reach 5 degrees Celsius for 5 minutes during transportation, this is the continuous deviation of the real-time temperature data from the preset temperature threshold. In terms of traffic status update information, the transport vehicle can obtain real-time road conditions through the connection with the traffic management system. For example, in the transportation from the collection center to the regional cold chain warehouse, a notification sent by the traffic management system is received, and the road ahead is congested due to a traffic accident. This is the traffic status update information. In terms of refrigeration equipment operation abnormal events, by monitoring the various operating parameters of the refrigeration equipment, such as detecting that the operating frequency of the refrigeration compressor is suddenly reduced, this may mean that the refrigeration equipment has a fault or is operating abnormally.

[0104] Step S152, according to the continuous deviation of the real-time temperature data from the preset temperature threshold, adjusting the stay time limit of the subsequent node and recalculating the preservation reliability score.

[0105] In the case of the above temperature deviation, since the temperature increase may accelerate the deterioration of strawberries, in order to reduce the time in transit, the subsequent node stay time limits from the collection center to the regional cold chain warehouse and from the regional cold chain warehouse to the transit distribution center need to be adjusted. The original plan to stay at the collection center for 2 hours is now adjusted to 1.5 hours. When recalculating the preservation reliability score, factors such as the resource matching degree of each node need to be re-evaluated. For example, at the collection center, due to the shortened stay time, more preservation resources may need to be provided in a short period of time, and the available refrigeration equipment capacity and energy supply status of the current collection center remain unchanged, so the preservation reliability score is recalculated according to the new situation.

[0106] Step S153: When it is detected that the traffic status update information indicates that the path congestion probability exceeds a preset risk threshold, a regeneration process of the candidate transportation path is triggered, and the path congestion probability distribution is updated based on the latest data.

[0107] Assuming that the preset risk threshold of the congestion probability of the path from the regional cold chain warehouse to the transit distribution center is 30%, if the traffic management system notifies that the congestion probability of the road section due to a traffic accident has reached 40%, the regeneration process of the candidate transportation path is triggered. When regenerating the candidate transportation path, it is necessary to comprehensively consider the current traffic conditions, the location of each node, and the distribution of refrigeration equipment. For example, the original transportation path is picking farm-collection center-regional cold chain warehouse-transit distribution center-retail supermarket. Now the regenerated candidate path may be picking farm-collection center-another regional cold chain warehouse-transit distribution center-retail supermarket. This other regional cold chain warehouse may be located near the original regional cold chain warehouse and in an area with good traffic conditions. At the same time, the path congestion probability distribution is updated based on the latest traffic data, weather data, etc. For example, the congestion probability from the collection center to another regional cold chain warehouse in the new path is estimated to be 10% based on real-time traffic flow and road conditions.

[0108] Step S154, feeding back the adjusted node sequence, real-time temperature data and traffic status update information to the cold chain demand prediction model, and iteratively optimizing the accuracy of the cargo preservation demand prediction value.

[0109] For example, a new sequence of transport path nodes (such as picking farms-collection centers-another regional cold chain warehouse-transit distribution center-retail supermarkets), real-time temperature data (such as actual temperature changes at each node during transportation), and traffic status update information (such as changes in congestion probability at each road section, etc.) can be fed back to the cold chain demand forecasting model. The model re-analyzes the dynamic association pattern between transport nodes based on these new data and revises the previous forecast value of the goods preservation demand. For example, it was previously predicted that 30% of refrigeration resources would be needed at the transit distribution center to maintain the freshness of strawberries. Based on the new data, it may be adjusted to 35% to more accurately reflect the actual transportation situation, thereby improving the accuracy of the forecast value of the goods preservation demand.

[0110] Step S155, generating a path adjustment instruction according to the prediction result after iteration, wherein the path adjustment instruction includes an operation command to add a spare node, skip a significant risk node, or enable an emergency cold storage resource.

[0111] For example, if the prediction results after iteration show that the demand for fresh-keeping resources at a certain node increases, and the current refrigeration equipment is not in good condition, such as the refrigeration effect decreases, the corresponding route adjustment instructions need to be generated. If it is detected in the transit distribution center that the refrigeration effect of the refrigeration equipment cannot meet the demand, the route adjustment instructions may include adding a new spare node, such as finding a temporary storage point with sufficient refrigeration resources near the transit distribution center as a spare node; or skipping a significant risk node. If the refrigeration equipment at a certain node fails and cannot be repaired in time, skip the node directly and transport the strawberries to the next node; or enable emergency refrigeration resources, such as adding temporary refrigeration equipment to the transport vehicle to meet the fresh-keeping needs.

[0112] The same process is used for beef transportation. During transportation, monitor the real-time temperature data of each node from the slaughterhouse to the meat processing plant, the meat processing plant to the cold chain logistics hub (the preset temperature threshold of beef is below -18 degrees Celsius), traffic status update information (such as congestion caused by road construction, traffic accidents, etc.), and abnormal operation events of refrigeration equipment (such as abnormal refrigerant pressure of refrigeration equipment, etc.). According to the deviation of the real-time temperature data from the preset temperature threshold, adjust the stay time limit of the subsequent nodes and recalculate the fresh-keeping reliability score. When the traffic status update information indicates that the path congestion probability exceeds the preset risk threshold, regenerate the candidate transportation path and update the path congestion probability distribution. Feed the adjusted relevant data back to the cold chain demand forecasting model for iterative optimization, and finally generate appropriate path adjustment instructions based on the iterative forecast results and abnormal operation events of refrigeration equipment. For example, if the refrigeration equipment fails during transportation from the meat wholesaler to the restaurant, a spare refrigerated storage point may be added, or the node may be skipped to directly transport the beef to the restaurant, and emergency refrigeration resources may be enabled to ensure the freshness of the beef. Through these operations, the freshness and transportation time of strawberries and beef can be ensured during transportation, while the transportation route decisions can be continuously optimized.

[0113] In a possible implementation, step S151 may include: Step S1511, collect the temperature distribution data inside the carriage in real time through the Internet of Things sensor deployed on the transport carrier, and identify the abnormal temperature fluctuation pattern of the temperature distribution data inside the carriage.

[0114] In the scenario of fresh cold chain transportation, taking strawberry and beef transportation as an example, during the strawberry transportation process, from the picking farm to the collection center, multiple IoT sensors are deployed on the transport vehicles. These sensors can accurately collect temperature data at different locations inside the carriage to form temperature distribution data inside the carriage. For example, the sensor collects data every 5 minutes, and the temperature data of the front, middle and rear of the carriage are recorded. Under normal circumstances, the preservation temperature requirement for strawberries is 0-4 degrees Celsius. When the data collected by the sensor shows that within a certain period of time, the temperature in the middle of the carriage gradually rises from 3 degrees Celsius to 5 degrees Celsius, and then drops to 4 degrees Celsius, this fluctuation pattern is regarded as an abnormal temperature fluctuation pattern. For beef transportation, from the slaughterhouse to the meat processing plant, the preset preservation temperature is below -18 degrees Celsius. If the sensor detects that the local temperature in the carriage rises to -10 degrees Celsius in a short period of time, this is also an abnormal temperature fluctuation pattern. The identification of this abnormal temperature fluctuation pattern is crucial for taking timely measures to protect the quality of goods.

[0115] Step S1512, obtaining a real-time traffic data interface of a traffic management system, obtaining road construction events, traffic accident reports and weather warning signals in a target area as the traffic status update information.

[0116] For example, during the transportation of strawberries, when the transport vehicle goes from the regional cold chain warehouse to the transit distribution center, the transport enterprise obtains the road construction event information of the target area by obtaining the real-time road condition data interface of the traffic management system. For example, road construction is being carried out on a main road in the city where the transit distribution center is located, resulting in the closure of some lanes, which will affect the speed of vehicles. At the same time, traffic accident reports are also obtained, such as a rear-end collision between two vehicles on a certain section of the road on the way from the collection center to the regional cold chain warehouse, causing traffic congestion. In addition, weather warning signals are also obtained. For example, during summer transportation, high temperature warning signals are received, which may affect the performance of the refrigeration equipment of the transport vehicle and increase the risk of cargo deterioration. For beef transportation, if a weather warning signal of road icing is received during the process from the meat processing plant to the cold chain logistics hub, this will affect transportation safety and transportation time. At the same time, road construction events and past traffic accident reports near the cold chain logistics hub are obtained. This information is used as traffic status update information for subsequent transportation decision adjustments.

[0117] Step S1513, monitor the compressor operating frequency, refrigerant pressure and energy consumption rate of the refrigeration equipment, generate an equipment health status assessment report, and analyze the abnormal operation events of the refrigeration equipment in the equipment health status assessment report.

[0118] For example, in the transportation of strawberries, the refrigeration equipment is installed on the transport vehicle to continuously monitor its compressor operating frequency, refrigerant pressure and energy consumption rate. For example, under normal circumstances, the compressor operating frequency is maintained at around 3,000 revolutions per minute, the refrigerant pressure is within a certain standard range, and the energy consumption rate is relatively stable. If during the transportation process, it is monitored that the compressor operating frequency suddenly drops to 2,000 revolutions per minute, the refrigerant pressure is lower than the standard lower limit, and the energy consumption rate increases abnormally, an equipment health status assessment report will be generated. By analyzing this equipment health status assessment report, abnormal operation events of the refrigeration equipment can be detected, which may be problems such as compressor failure or refrigerant leakage. For beef transportation, during the transportation process from the cold chain logistics hub to the meat wholesaler, if the compressor operating frequency of the refrigeration equipment is unstable, the refrigerant pressure fluctuates greatly, and the energy consumption rate is 20% higher than normal, these situations will be recorded in the equipment health status assessment report, and abnormal operation events of the refrigeration equipment will be analyzed, which may affect the preservation effect of beef and need to be handled in time.

[0119] Step S1514: when it is detected that the abnormal temperature fluctuation duration corresponding to the abnormal temperature fluctuation pattern exceeds a preset tolerance threshold, a first-level warning signal is generated and a local path adjustment is triggered.

[0120] For example, in the transportation of strawberries, assume that the preset tolerance threshold is 10 minutes. If during the transportation from the collection center to the regional cold chain warehouse, the abnormal temperature fluctuation pattern in the carriage lasts for 15 minutes, exceeding the preset tolerance threshold. At this time, the system will generate a first-level warning signal. This warning signal will trigger a local route adjustment. For example, the original plan was to go directly to the regional cold chain warehouse without stopping on the way, but due to abnormal temperature fluctuations, it may be decided to stop at a small refrigerated station on the way to check the status of the goods and adjust the refrigeration equipment to prevent the strawberries from deteriorating due to temperature fluctuations. For beef transportation, during the process from the slaughterhouse to the meat processing plant, if the abnormal temperature fluctuation lasts for more than 10 minutes, the first-level warning signal will be triggered, and the transportation speed may be adjusted or a short stop may be made at a nearby refrigeration facility to ensure the freshness of the beef.

[0121] Step S1515, when the received weather warning signal includes a freezing or high temperature red warning, a second level warning signal is generated and an alternative transportation path is forcibly activated.

[0122] For example, during the transportation of strawberries, if a high temperature red warning signal is received during summer transportation, this indicates that the outside temperature is extremely high, which may pose a serious threat to the preservation of strawberries. At this time, the system will generate a second-level warning signal and force the use of an alternative transportation path. For example, the original transportation route is from the picking farm-collection center-regional cold chain warehouse-transit distribution center-retail supermarket. The alternative transportation path may be from the picking farm-another collection center-alternative regional cold chain warehouse-transit distribution center-retail supermarket. This alternative path may pass through areas with relatively low temperatures or better traffic conditions to ensure the preservation of strawberries in high temperature environments. For beef transportation, if a freezing red warning signal is received during winter transportation, since freezing weather may affect road traffic and the normal operation of refrigeration equipment, the second-level warning signal will be triggered to enable an alternative transportation path, such as from the slaughterhouse-alternative meat processing plant-cold chain logistics hub-meat wholesaler-restaurant. The nodes on this alternative path may be more adapted to transportation requirements under freezing weather.

[0123] In a possible implementation, step S152 may include: Step S1521, counting the cumulative time proportion and the maximum deviation range of the temperature deviation threshold in the current transport section.

[0124] For example, in the transportation of strawberries, the preset temperature threshold for the transportation section from the regional cold chain warehouse to the transit distribution center is 0-4 degrees Celsius. Assume that during this transportation section, due to some minor faults in the refrigeration equipment, the temperature deviates from the threshold. Through the analysis of real-time temperature data, it is detected that the cumulative time of temperature deviation from the threshold is 30 minutes, and the entire transportation section is 3 hours (180 minutes), so the cumulative time accounts for 30÷180=1 / 6. In terms of the maximum deviation, the temperature reached 6 degrees Celsius, and the maximum deviation is 2 degrees Celsius relative to the maximum threshold of 4 degrees Celsius. For beef transportation, the preset temperature threshold for the transportation section from the cold chain logistics hub to the meat wholesaler is below -18 degrees Celsius. If the cumulative time accounts for 1 / 5, the maximum deviation is 8 degrees Celsius (from -18 degrees Celsius to -10 degrees Celsius). The statistics of these data provide the basis for subsequent calculations.

[0125] Step S1522, calculating the cargo mass decay acceleration coefficient according to the cumulative time proportion and the maximum deviation amplitude.

[0126] For example, the cargo quality decay acceleration coefficient can be calculated based on a special calculation formula (assuming it is an empirical formula based on the deterioration characteristics of strawberries and the impact of temperature on strawberries), using the cumulative time ratio of 1 / 6 and the maximum deviation amplitude of 2 degrees Celsius. For example, this cargo quality decay acceleration coefficient may be positively correlated with the cumulative time ratio and the maximum deviation amplitude. After calculation, the cargo quality decay acceleration coefficient is 0.3 (this is just an example value). For beef transportation, according to its corresponding calculation formula, combined with the cumulative time ratio of 1 / 5 and the maximum deviation amplitude of 8 degrees Celsius, the cargo quality decay acceleration coefficient is calculated to be 0.5 (also an example value). This coefficient reflects the degree of influence of temperature deviation on the speed of cargo quality decline.

[0127] Step S1523, dynamically shortening the maximum allowed stay time of subsequent transport nodes based on the cargo mass decay acceleration coefficient.

[0128] For example, the maximum allowed stay time at the subsequent transport node from the transit distribution center to the retail supermarket was originally planned to be 2 hours. Since the previously calculated acceleration coefficient of cargo quality decay is 0.3, according to the pre-set adjustment rules (for example, for every increase of 0.1 in the acceleration coefficient of cargo quality decay, the maximum allowed stay time at the subsequent node is shortened by 15%), the maximum allowed stay time at this node needs to be shortened by 0.3÷0.1×15%=45%, and the adjusted maximum allowed stay time is 2×(1-0.45)=1.1 hours. For beef transportation, the maximum allowed stay time at the subsequent transport node from the meat wholesaler to the restaurant was originally 1.5 hours. According to the calculated acceleration coefficient of cargo quality decay of 0.5, according to the corresponding adjustment rules (for example, for every increase of 0.1 in the acceleration coefficient of cargo quality decay, the maximum allowed stay time at the subsequent node is shortened by 20%), the adjusted maximum allowed stay time is 1.5×(1-0.5÷0.1×0.2)=0.6 hours.

[0129] Step S1524, re-evaluate the distribution of fresh-keeping resource requirements of the adjusted node sequence, and update the resource matching weight in the fresh-keeping reliability score.

[0130] For example, the adjusted node sequence is picking farm-collection center-regional cold chain warehouse-transit distribution center-retail supermarket, in which the stay time in the transit distribution center is shortened. Due to the shortened stay time, in order to ensure the fresh-keeping quality of strawberries, fresh-keeping resources need to be provided more centrally in the transit distribution center. Originally, the distribution of fresh-keeping resource demand in the transit distribution center was 30% for temperature control and 20% for air circulation, etc., which may now be adjusted to 40% for temperature control and 30% for air circulation, etc. For the resource matching weight in the fresh-keeping reliability score, the resource matching weight was originally calculated based on the planned resource allocation and available resources. Now, due to the change in the distribution of fresh-keeping resource demand, it needs to be recalculated. For example, assuming that the available refrigeration equipment capacity in the transit distribution center is 50%, the resource matching degree calculated according to the planned resource demand was 0.6 (30% ÷ 50%) before, and the resource matching degree calculated according to the adjusted resource demand is now 0.8 (40% ÷ 50%), thereby updating the resource matching weight in the fresh-keeping reliability score. For beef transportation, after the node adjustment from meat wholesalers to restaurants, the distribution of preservation resource demand is re-evaluated. For example, more resources are allocated to temperature control to cope with the preservation pressure caused by the shortened stay time, and the resource matching weight in the preservation reliability score is updated at the same time.

[0131] Step S1525, when the cargo quality decay acceleration coefficient exceeds a critical value, an emergency unloading instruction is triggered and a quality remediation plan is started at the nearest node.

[0132] For example, assume that the critical value of the cargo quality decay acceleration coefficient is 0.5. If in a certain transportation section, due to a serious failure of temperature control, the calculated cargo quality decay acceleration coefficient reaches 0.6, which exceeds the critical value. At this time, the system will trigger an emergency unloading instruction, such as at the nearest transit distribution center, and the vehicle will go there as soon as possible to unload the goods. Initiating a quality remediation plan at the transit distribution center may include rapid cooling of strawberries, checking the deterioration of strawberries, repackaging non-deteriorated strawberries and transferring them to spare refrigeration equipment for further transportation. For beef transportation, if the cargo quality decay acceleration coefficient exceeds the critical value, an emergency unloading instruction will also be triggered, and a quality remediation plan will be initiated at the nearest node (such as a meat wholesaler), such as inspecting the beef and readjusting the refrigeration environment.

[0133] In a possible implementation, step S153 may include: Step S1531, determining the geographical boundary range of the detour area according to the risk node position and influence radius in the path congestion probability distribution corresponding to the traffic status update information.

[0134] For example, during the transportation of strawberries, on the transportation route from the regional cold chain warehouse to the transit distribution center, based on the traffic status update information, it is detected that the probability of path congestion due to a traffic accident on a certain section of road exceeds the preset risk threshold (assuming it is 30%). In the probability distribution of path congestion, the risk node location is determined as the point of accident occurrence, and the impact radius is assumed to be 10 kilometers (determined based on factors such as traffic flow and road type). Then, the geographical boundary range of the detour area is a circular area with a radius of 10 kilometers centered on the point of accident occurrence. For beef transportation, on the path from the meat processing plant to the cold chain logistics hub, if the probability of congestion due to road construction exceeds the preset risk threshold, the geographical boundary range of the detour area is determined based on the risk node location (road construction section) and the impact radius (for example, 8 kilometers).

[0135] Step S1532, based on the geographic boundary range, filter alternative nodes with refrigeration facility records, and extract compatibility data between the refrigeration equipment type of the alternative nodes and the currently transported goods.

[0136] For example, within the determined geographical boundary of the detour area, alternative nodes with refrigeration facilities are screened out by querying the refrigeration facility registration database. For example, several small refrigerated warehouses and some logistics sites with refrigeration equipment are screened out. Then, the compatibility data of the refrigeration equipment types of these alternative nodes and the current transportation of strawberries are extracted. If the temperature control range of the refrigeration equipment of a small refrigerated warehouse is -5-5 degrees Celsius, and the preservation temperature requirement of strawberries is 0-4 degrees Celsius, the compatibility data needs to evaluate whether this temperature range can meet the preservation requirements of strawberries, and the compatibility may be 80% (based on certain evaluation criteria). For beef transportation, after selecting alternative nodes in the detour area, if the temperature of the refrigeration equipment of the alternative node is -20--10 degrees Celsius, the compatibility data is evaluated based on the preservation requirements of beef below -18 degrees Celsius.

[0137] Step S1533, combining the resource demand distribution in the cargo preservation demand forecast value, matching the alternative nodes with qualified compatibility data to generate emergency path branches, the emergency path branches include bypassing the main path, temporary transit path and multimodal transport connection path.

[0138] For example, according to the resource demand distribution in the forecast value of the goods preservation demand, 30% of the refrigeration resources may be required in the transit distribution center. From the alternative nodes screened previously, select those nodes whose compatibility data meet the standards. For example, if there is a small refrigerated warehouse whose refrigeration equipment temperature control range and capacity meet the preservation needs of strawberries, it can be included in the emergency path branch. The emergency path branch includes the detour main path, such as bypassing the congested road section from the regional cold chain warehouse, passing through this small refrigerated warehouse and then to the transit distribution center; the temporary transit path, if the goods need to be temporarily transferred in this small refrigerated warehouse, such as repackaging, adjusting refrigeration equipment, etc.; the multimodal transport connection path, if the transportation mode needs to be changed in this area, such as from road transportation to rail transportation, corresponding connection arrangements can also be made. For beef transportation, the resource demand distribution in the forecast value of the goods preservation demand is also combined, and the alternative nodes are matched to generate emergency path branches to ensure that the beef can be transported smoothly and kept fresh when encountering traffic congestion.

[0139] Step S1534: based on the cargo temperature stability index collected in real time, branches that exceed the remaining fresh-keeping time constraint are excluded from the emergency path branches to generate a valid emergency path subset.

[0140] For example, the temperature stability index of the goods collected in real time shows the current temperature status and remaining fresh-keeping time of the strawberries. If a certain emergency path branch is expected to arrive at the retail supermarket more than the remaining fresh-keeping time of the strawberries due to a long detour distance or too many transfer links, then this branch will be excluded. For example, although there is an emergency path branch that can bypass congested sections, it is excluded because it needs to pass through multiple temporary transfer points and the total transportation time is expected to exceed the remaining fresh-keeping time of the strawberries. After such screening, an effective emergency path subset is generated. For beef transportation, based on the temperature stability index and remaining fresh-keeping time of beef, the branches that do not meet the requirements are excluded from the emergency path branches to obtain an effective emergency path subset.

[0141] Step S1535, input the valid emergency path subset into the cold chain demand prediction model, re-sort according to the dynamic weight of the fresh-keeping reliability score and the timeliness achievement rate score, and output the target emergency path with the highest priority.

[0142] For example, a subset of valid emergency paths is input into the cold chain demand forecasting model. The cold chain demand forecasting model reorders these paths based on the dynamic weights of the previously calculated freshness reliability score (based on the matching degree of refrigeration equipment resources at each node, temperature control capability, etc.) and the timeliness achievement rate score (based on the expected transportation time and the strawberry freshness time limit). For example, one emergency path has a high freshness reliability score but a slightly lower timeliness achievement rate score, and another path has a high timeliness achievement rate score but a slightly lower freshness reliability score. The comprehensive score is calculated based on the set dynamic weights (such as the freshness reliability score weight is 0.6 and the timeliness achievement rate score weight is 0.4). After comparison, the target emergency path with the highest priority is output, and this target emergency path will be used for actual transportation adjustments. For beef transportation, the subset of valid emergency paths is also input into the cold chain demand forecasting model, and the dynamic weights of the freshness reliability score and the timeliness achievement rate score are reordered to determine the target emergency path.

[0143] In a possible implementation, step S154 may include: Step S1541, performing a difference analysis between the actual transportation result of the target emergency path and the predicted freshness preservation reliability score to generate a path decision deviation index.

[0144] For example, in the transportation of strawberries, it is assumed that the target emergency path is from regional cold chain warehouse-small refrigerated warehouse-transit distribution center-retail supermarket. In the actual transportation process, the time to arrive at each node, the actual use of fresh-keeping resources at each node, etc. are different from the data based on the predicted fresh-keeping reliability score. For example, it is predicted that 30% of refrigeration resources are needed in a small refrigerated warehouse to keep strawberries fresh, but 35% are actually used. By comparing the actual transportation results with the prediction, the deviation value of each node is calculated, and then the path decision deviation index is generated comprehensively. For beef transportation, in the actual transportation of the target emergency path, the actual use of fresh-keeping resources, transportation time, etc. are compared with the prediction, and the path decision deviation index is generated.

[0145] Step S1542, adjusting the convolution kernel size of the temporal convolutional neural network according to the path decision deviation index to enhance the local feature extraction capability of sudden congestion events.

[0146] For example, if the path decision deviation index shows that the model's ability to extract local features is insufficient when responding to sudden congestion events (such as congestion caused by traffic accidents). According to this deviation index, adjust the convolution kernel size of the time series convolutional neural network. For example, if the deviation index shows that the temperature control and transportation time prediction of nodes near congested sections are inaccurate, the convolution kernel size may be increased to better capture the local features of these nodes such as temperature changes and residence time under sudden congestion events. For beef transportation, the convolution kernel size of the time series convolutional neural network is also adjusted according to the path decision deviation index to improve the ability to respond to sudden congestion events.

[0147] Step S1543, based on the cargo quality feedback data after the execution of the target emergency path, optimizing the forgetting gate threshold of the time-dependent correction in the long short-term memory network.

[0148] For example, after the target emergency path is executed, the quality feedback data of strawberries when arriving at the retail supermarket, such as the freshness and spoilage rate of strawberries, is collected. If the spoilage rate of strawberries is detected to be higher than expected, it means that the correction of time dependency may not be accurate enough during transportation. At this time, the forget gate threshold of time dependency correction in the long short-term memory network is optimized by analyzing the feedback data on the quality of the goods. For example, if the high spoilage rate of strawberries is due to insufficient attention to time in the later stage of transportation, the forget gate threshold may be lowered, so that the long short-term memory network can further strengthen the impact of time factors on the quality of goods in subsequent predictions. For beef transportation, the forget gate threshold in the long short-term memory network is optimized based on the quality feedback data of beef when it arrives at the restaurant.

[0149] Step S1544, within a preset period, the adjusted node sequence, the real-time temperature data, the traffic status update information and the abnormal operation events of the refrigeration equipment are integrated into an incremental training data set according to timestamps.

[0150] For example, in the transportation of strawberries, assume that the preset cycle is every day. During the transportation process of one day, multiple adjusted node sequences will be generated. For example, the original transportation path node sequence when starting from the picking farm in the morning is picking farm-collection center-regional cold chain warehouse-transit distribution center-retail supermarket. During the transportation process, due to temperature fluctuations and traffic congestion, the adjusted node sequence may become picking farm-collection center-standby regional cold chain warehouse-transit distribution center-retail supermarket. At the same time, the real-time temperature data collected at each transportation node, such as the temperature at the collection center is 2 degrees Celsius, the temperature at the standby regional cold chain warehouse is 3 degrees Celsius, etc.; traffic status update information, including road construction, traffic accidents, and weather warnings; abnormal operation events of refrigeration equipment, such as the reduction of the compressor operation frequency of the refrigeration equipment during transportation, are all integrated according to the timestamp. The timestamp is accurate to the minute, which can clearly reflect the order of events. Integrating these data forms an incremental training data set. For beef transportation, the preset cycle is also one day, and the relevant data during the transportation process is integrated to form an incremental training data set, including relevant information from each node from slaughterhouses to meat processing plants, cold chain logistics hubs, etc.

[0151] Step S1545, injecting the incremental training data set into the cold chain demand prediction model through an online learning algorithm, updating the model weight information of the cold chain demand prediction model and retaining high-value patterns in historical path decisions.

[0152] For example, the online learning algorithm will adjust the parameters of the model according to the new incremental training data set. For example, for the time series convolutional neural network part, the connection weights between neurons are adjusted according to the new node sequence and temperature data to better extract the dynamic association pattern between transportation nodes. For the long short-term memory network part, the weights of the forget gate, input gate, and output gate are adjusted according to the new time series data (including the time series of traffic status update information and abnormal operation events of refrigeration equipment) to optimize the time dependency correction. While updating the model weight information, high-value patterns in historical path decisions are also retained. For example, a transportation path decision from a picking farm to a retail supermarket in a certain season has achieved good preservation effect and timeliness achievement rate in multiple transportations, and this pattern will be retained. For beef transportation, the incremental training data set is injected into the cold chain demand prediction model through the online learning algorithm, and the model weight information is also updated and the valuable historical path decision pattern is retained, thereby improving the accuracy of the model's prediction of the preservation demand and path decision for beef transportation.

[0153] Wherein, step S155 includes: Step S1551, analyzing the refrigeration resource occupancy rate and energy replenishment time window of each alternative node in the target emergency path.

[0154] For example, in the transportation of strawberries, the target emergency path is from regional cold chain warehouse-small refrigerated warehouse-transit distribution center-retail supermarket. For the alternative node of small refrigerated warehouse, analyze its refrigerated resource occupancy rate. Assuming that the total refrigerated capacity of the small refrigerated warehouse is 100 cubic meters and 30 cubic meters are currently occupied, the refrigerated resource occupancy rate is 30%. In terms of the energy replenishment time window, the energy replenishment of the small refrigerated warehouse is through electricity supply. There is a fixed power maintenance time of 2-3 am every day. This time period is the energy replenishment time window. For beef transportation, at the alternative node in the target emergency path, the refrigerated resource occupancy rate and energy replenishment time window are also analyzed. For example, in a certain alternative meat refrigerated warehouse, the refrigerated resource occupancy rate is 40%, and the energy replenishment time window is 10-11 pm.

[0155] Step S1552, calculate the feasibility probability of the carrier arriving at each alternative node based on the remaining cruising range of the transport carrier and the real-time power consumption of the refrigeration equipment.

[0156] For example, the transport carrier is a refrigerated truck with a remaining range of 200 kilometers and a real-time power consumption of 0.5 liters of diesel per kilometer (assumed). The distance from the current location to the small refrigerated warehouse is 150 kilometers. At normal driving speed, it takes 3 hours to reach the small refrigerated warehouse. During these 3 hours, the power consumption of the refrigerated equipment is 1.5 liters of diesel. Taking into account possible fluctuations in fuel consumption and other factors along the way, the feasibility probability of the carrier arriving at the small refrigerated warehouse is calculated to be 80% (based on the pre-set calculation model, taking into account factors such as the remaining range, real-time power consumption, and distance). For beef transportation, factors such as the remaining range of the transport carrier, the real-time power consumption of the refrigerated equipment, and the distance to the alternative node are used to calculate the feasibility probability of arriving at each alternative node. For example, from the current location to an alternative cold chain logistics hub, the feasibility probability is calculated to be 70% based on relevant data.

[0157] Step S1553, generating an instruction sequence matching the feasibility probability, wherein the instruction sequence includes detour turning coordinates, cooling power adjustment gradient, and an upper limit of the transit node stay time.

[0158] For example, if the feasibility probability of the carrier arriving at a small cold storage warehouse is 80%, the generated instruction sequence is as follows. The detour coordinates are to turn left at the intersection 50 kilometers away from the current location (determined by map coordinates and path planning), the refrigeration power adjustment gradient is to reduce the refrigeration power by 10% within 10 kilometers of the small cold storage warehouse (because the refrigeration environment of the small cold storage warehouse is better), and the maximum stay time at the transfer node is 1 hour (determined according to the cargo preservation needs and the operating efficiency of the small cold storage warehouse). For beef transportation, the corresponding instruction sequence is generated according to the calculated feasibility probability. For example, when approaching an alternative cold chain logistics hub, the detour coordinates, refrigeration power adjustment gradient, and transfer node stay time limit are determined to ensure that the beef can be transported smoothly and kept fresh.

[0159] Step S1554, after synchronizing the instruction sequence to the path execution terminal of the transport carrier and the refrigeration equipment controller, the execution status data of the instruction sequence is monitored in real time. When it is detected that the refrigeration power adjustment gradient does not meet the standard or the steering coordinate deviation exceeds the limit, the manual intervention instruction is triggered and the abnormal data is fed back to the cold chain demand prediction model.

[0160] For example, the path execution terminal navigates according to the detour turning coordinates, and the refrigeration equipment controller adjusts the refrigeration power according to the refrigeration power adjustment gradient. During the transportation process, the execution status data of the instruction sequence is monitored in real time by the sensor installed on the transport carrier. If it is detected that the refrigeration power adjustment gradient does not meet the standard, for example, the refrigeration power should be reduced by 10%, but it is actually only reduced by 5%, or the steering coordinate deviation exceeds the allowable range, such as deviating from the predetermined steering coordinate by 100 meters, the manual intervention instruction will be triggered. The transportation personnel will receive a notification and make manual adjustments. At the same time, abnormal data (such as substandard refrigeration power, deviated steering coordinates, etc.) will be sent back to the cold chain demand prediction model so that the model can further optimize and adjust subsequent predictions and decisions. For beef transportation, the instruction sequence is also synchronized to the relevant equipment of the transport carrier, and the execution status data is monitored in real time. When an abnormality occurs, the manual intervention instruction is triggered and the abnormal data is sent back to the cold chain demand prediction model, so as to continuously improve the preservation and transportation efficiency of beef transportation.

[0161] Figure 2 A cold chain transportation service system 100 provided in an embodiment of the present application is shown, including a processor 1001 and a memory 1003 and a program code stored on the memory 1003, and the processor 1001 executes the above program code to implement the steps of the cold chain transportation path optimization method based on AI prediction.

[0162] Figure 2The cold chain transportation service system 100 shown includes: a processor 1001 and a memory 1003. The processor 1001 and the memory 1003 are connected, such as through a bus 1002. Optionally, the cold chain transportation service system 100 may also include a transceiver 1004, which can be used for data interaction between the cold chain transportation service system and other cold chain transportation service systems, such as data transmission and / or data reception. It should be noted that the transceiver 1004 is not limited to one in actual scheduling, and the structure of the cold chain transportation service system 100 does not constitute a limitation on the embodiments of the present application.

[0163] Processor 1001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 1001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0164] The bus 1002 may include a path to transmit information between the above components. The bus 1002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 1002 may be divided into an address bus, a data bus, a control bus, and the like.

[0165] The memory 1003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), magnetic disk storage medium, other magnetic storage devices, or any other medium that can be used to have or store program code and can be read by a computer, without limitation herein.

[0166] The memory 1003 is used to store program codes for executing the embodiments of the present application, and the execution is controlled by the processor 1001. The processor 1001 is used to execute the program codes stored in the memory 1003 to implement the steps shown in the above method embodiments.

[0167] An embodiment of the present application provides a computer-readable storage medium having program code stored thereon. When the program code is executed by a processor, the steps and corresponding contents of the aforementioned method embodiment can be implemented.

[0168] It should be understood that, although each operation step is indicated by arrows in the flowchart of the embodiment of the present application, the implementation order of these steps is not limited to the order indicated by the arrows. Unless there is clear explanation in this article, in some implementation scenarios of the embodiment of the present application, the implementation steps in each flowchart can be performed in other orders based on demand. In addition, some or all of the steps in each flowchart may include multiple sub-steps or multiple stages according to the actual implementation scenario, and some or all of these sub-steps or stages may be executed at the same time, and each sub-step or stage in these sub-steps or stages may also be executed at different times respectively. In different scenarios at the execution time, the execution order of these sub-steps or stages can be flexibly configured based on demand, and the embodiment of the present application does not limit this.

[0169] The above is only an optional implementation method for some implementation scenarios of the present application. It should be pointed out that for ordinary technicians in this technical field, without departing from the technical concept of the scheme of the present application, other similar implementation methods based on the technical ideas of the present application are also within the protection scope of the embodiments of the present application.

Claims

1. A cold chain transportation route optimization method based on AI prediction, characterized in that: The method comprises: Acquire a sample cold chain transport order data set, wherein the sample cold chain transport order data set includes a transport node sequence, cargo preservation requirement parameters, ambient temperature fluctuation records, and transport timeliness completion status of multiple sample orders; Extracting transport features from the sample cold chain transport order data set to generate a combined transport feature set corresponding to each sample order, wherein the combined transport feature set includes a time-sensitive feature, an environmental stability feature, and a cargo deterioration risk feature; Based on the combined transport feature set, a cold chain demand forecasting model is called to generate a forecast value of the cargo preservation demand in the target area within a preset time range, and based on sample traffic flow data and weather event records in the target area, a path congestion probability distribution is generated; According to the predicted value of the cargo preservation demand and the probability distribution of route congestion, multiple candidate transportation routes from the starting point to the end point are generated, and each candidate transportation route is scored for preservation reliability and timeliness achievement rate; Based on the weighted results of the freshness-keeping reliability score and the timeliness achievement rate score, the node sequence of the current transportation path is dynamically adjusted, and the input data of the cold chain demand prediction model is updated in real time to optimize subsequent path decisions.

2. The cold chain transportation route optimization method based on AI prediction according to claim 1 is characterized in that: The extracting of transport features from the sample cold chain transport order data set to generate a combined transport feature set corresponding to each sample order includes: Extracting the dwell time data, the number of times the ambient temperature deviates from the threshold, and the cargo packaging integrity records of each transport node from the sample cold chain transport order data set; Determine the time-sensitive feature according to the difference between the dwelling time data and the preset cargo freshness limit, wherein the time-sensitive feature is used to identify the risk level of exceeding the freshness limit in the transport node sequence; Calculating the environmental stability feature according to the number of times the ambient temperature deviates from the threshold and the proportion of the deviation duration, wherein the environmental stability feature is used to characterize the cumulative impact of temperature fluctuations on the quality of the goods during transportation; Combining the cargo packaging integrity record and the deterioration rate parameter corresponding to the cargo type, generating the cargo deterioration risk feature, wherein the cargo deterioration risk feature is used to quantify the contribution of different nodes in the transportation path to the deterioration of cargo quality; The time-sensitive features, environmental stability features and cargo deterioration risk features are normalized to generate the combined transport feature set.

3. The cold chain transportation route optimization method based on AI prediction according to claim 2 is characterized in that: The cold chain demand prediction model is called based on the combined transport feature set to generate a forecast value of the cargo preservation demand in the target area within a preset time range, and a path congestion probability distribution is generated based on sample traffic flow data and weather event records in the target area, including: Inputting the combined transport feature set into a time series convolutional neural network, extracting dynamic correlation patterns between transport nodes, and outputting initial path demand prediction results; Using a long short-term memory network to perform time-dependency correction on the initial path demand forecast result, and generate a revised cargo preservation demand forecast value, wherein the revised cargo preservation demand forecast value includes the distribution of preservation resource demand at different transportation nodes in the transportation path; Based on the sample traffic flow data and weather event records of the target area, a path congestion probability prediction sub-model is constructed, wherein the path congestion probability prediction sub-model outputs the path congestion probability distribution in the future time period by analyzing the correlation between weather events and traffic flow changes; The corrected cargo preservation demand prediction value and the path congestion probability distribution are integrated to generate a joint output of the cold chain demand prediction model, where the joint output is used to identify the preservation resource occupancy rate and congestion delay risk of different candidate transportation paths.

4. The cold chain transportation route optimization method based on AI prediction according to claim 3 is characterized in that: According to the predicted value of the cargo preservation demand and the probability distribution of path congestion, a plurality of candidate transportation paths from the starting point to the end point are generated, and each candidate transportation path is scored for preservation reliability and timeliness achievement rate, including: Calculate the expected total transportation time of each candidate transportation route based on the node sequence length of the transportation route, the sample average travel speed and the congestion probability distribution of the route; Compare the expected total transportation time with the cargo preservation time limit to determine the timeliness achievement rate score, where the timeliness achievement rate score is an inversely proportional function of the negative deviation degree of the expected total transportation time relative to the cargo preservation time limit; Based on the distribution of fresh-keeping resource requirements of the corresponding node in the predicted value of the goods fresh-keeping demand, combined with the current available refrigeration equipment capacity and energy supply status, the fresh-keeping reliability score is calculated, and the fresh-keeping reliability score is a weighted average of the node resource matching degree and the temperature control capability; Dynamically weight the preservation reliability score and the timeliness achievement rate score to generate a comprehensive priority ranking for each candidate transportation route.

5. The cold chain transportation route optimization method based on AI prediction according to claim 4 is characterized in that: The method dynamically adjusts the node sequence of the current transportation path based on the weighted result of the freshness-keeping reliability score and the timeliness achievement rate score, and updates the input data of the cold chain demand prediction model in real time to optimize subsequent path decisions, including: Monitor real-time temperature data, traffic status updates and abnormal operation of refrigeration equipment at each node during transportation; According to the continuous deviation of the real-time temperature data from the preset temperature threshold, adjusting the stay time limit of the subsequent nodes and recalculating the freshness preservation reliability score; When it is detected that the traffic status update information indicates that the path congestion probability exceeds a preset risk threshold, a regeneration process of the candidate transportation path is triggered, and the path congestion probability distribution is updated based on the latest data; Feeding back the adjusted node sequence, real-time temperature data and traffic status update information to the cold chain demand forecasting model, iteratively optimizing the accuracy of the cargo preservation demand forecast value; A path adjustment instruction is generated according to the prediction result after the iteration, and the path adjustment instruction includes an operation command of adding a spare node, skipping a significant risk node, or activating an emergency cold storage resource.

6. The cold chain transportation route optimization method based on AI prediction according to claim 5 is characterized in that: The monitoring of real-time temperature data, traffic status update information and abnormal operation events of refrigeration equipment at each node during transportation includes: The IoT sensors deployed on the transport carrier collect the temperature distribution data inside the carriage in real time, and identify the abnormal temperature fluctuation pattern of the temperature distribution data inside the carriage; Obtaining a real-time traffic data interface of a traffic management system, obtaining road construction events, traffic accident reports and weather warning signals in a target area as the traffic status update information; Monitor the compressor operating frequency, refrigerant pressure and energy consumption rate of the refrigeration equipment, generate an equipment health status assessment report, and analyze abnormal operation events of the refrigeration equipment in the equipment health status assessment report; When it is detected that the duration of abnormal temperature fluctuation corresponding to the abnormal temperature fluctuation pattern exceeds the preset tolerance threshold, a first-level warning signal is generated and a local path adjustment is triggered; When the received weather warning signal contains a red warning for freezing or high temperature, a second-level warning signal is generated and the alternative transportation route is forcibly activated.

7. The cold chain transportation route optimization method based on AI prediction according to claim 6 is characterized in that: The step of adjusting the stay time limit of subsequent nodes and recalculating the freshness-keeping reliability score according to the continuous deviation of the real-time temperature data from the preset temperature threshold comprises: Count the cumulative time percentage and maximum deviation range of the temperature deviation threshold in the current transport section; Calculate the cargo mass decay acceleration factor based on the cumulative time proportion and the maximum deviation amplitude; Based on the cargo mass decay acceleration coefficient, dynamically shorten the maximum allowed stay time at the subsequent transportation node; Re-evaluate the distribution of fresh-keeping resource requirements of the adjusted node sequence, and update the resource matching weight in the fresh-keeping reliability score; When the cargo quality decay acceleration coefficient exceeds a critical value, an emergency unloading instruction is triggered and a quality remediation plan is started at the nearest node.

8. The cold chain transportation route optimization method based on AI prediction according to claim 7 is characterized in that: When it is detected that the traffic status update information indicates that the path congestion probability exceeds a preset risk threshold, a process of regenerating a candidate transportation path is triggered, including: Determine the geographical boundary range of the detour area according to the risk node position and impact radius in the path congestion probability distribution corresponding to the traffic status update information; Screening alternative nodes with refrigerated facilities on file based on the geographic boundary range, and extracting compatibility data between the refrigerated equipment type of the alternative nodes and the currently transported goods; Combined with the resource demand distribution in the cargo preservation demand forecast value, matching the alternative nodes with qualified compatibility data to generate emergency path branches, the emergency path branches include a detour main path, a temporary transit path and a multimodal transport connection path; According to the cargo temperature stability index collected in real time, branches exceeding the remaining fresh-keeping time constraint are excluded from the emergency path branches to generate a valid emergency path subset; The effective emergency path subset is input into the cold chain demand prediction model, and is reordered according to the dynamic weights of the freshness-keeping reliability score and the timeliness achievement rate score, and the target emergency path with the highest priority is output.

9. The cold chain transportation route optimization method based on AI prediction according to claim 8 is characterized in that: Feeding back the adjusted node sequence, real-time temperature data and traffic status update information to the cold chain demand prediction model, iteratively optimizing the accuracy of the cargo preservation demand prediction value, includes: Performing a difference analysis between the actual transportation result of the target emergency route and the predicted freshness-keeping reliability score to generate a route decision deviation index; Adjusting the convolution kernel size of the temporal convolutional neural network according to the path decision deviation index to enhance the local feature extraction capability of sudden congestion events; Based on the cargo quality feedback data after the execution of the target emergency path, optimizing the forget gate threshold of the time-dependent correction in the long short-term memory network; Integrate the adjusted node sequence, the real-time temperature data, the traffic status update information and the abnormal operation events of the refrigeration equipment into an incremental training data set according to timestamps within a preset period; Injecting the incremental training data set into the cold chain demand prediction model through an online learning algorithm, updating the model weight information of the cold chain demand prediction model and retaining high-value patterns in historical path decisions; The step of generating a path adjustment instruction according to the prediction result after iteration includes: Analyze the cold storage resource occupancy rate and energy replenishment time window of each alternative node in the target emergency path; Calculate the feasibility probability of the carrier arriving at each alternative node based on the remaining range of the transport carrier and the real-time power consumption of the refrigeration equipment; Generate an instruction sequence matching the feasibility probability, the instruction sequence including a detour turning coordinate, a cooling power adjustment gradient, and an upper limit of a stay time at a transfer node; After synchronizing the instruction sequence to the path execution terminal of the transport carrier and the refrigeration equipment controller, the execution status data of the instruction sequence is monitored in real time. When it is detected that the refrigeration power adjustment gradient does not meet the standard or the steering coordinate deviation exceeds the limit, the manual intervention instruction is triggered and the abnormal data is fed back to the cold chain demand prediction model.

10. A cold chain transportation service system, characterized in that: It includes a processor and a computer-readable storage medium, wherein the computer-readable storage medium stores machine-executable instructions, and when the machine-executable instructions are executed by the processor, the cold chain transportation path optimization method based on AI prediction described in any one of claims 1 to 9 is implemented.

Images