Green agricultural product distribution path optimization and inventory control method

Abstract

The invention provides a green agricultural product distribution path optimization and inventory control method, and the method comprises the steps: selecting a distribution path with the minimum agricultural product quality influence degree as an optimal path according to a prediction result of the agricultural product quality influence degree, and transmitting the optimized distribution path to a distribution vehicle terminal in real time; after the distribution vehicle terminal receives the optimal distribution route issued by the cloud, the distribution route is navigated in real time through a vehicle-mounted navigation system, a driver is guided to drive along the optimal route, and the distribution route is dynamically adjusted according to the real-time road condition information; meanwhile, according to the real-time position and the driving speed of the distribution vehicle, the arrival time of the agricultural products is dynamically estimated, and the estimated arrival time information is pushed to a destination customer to remind the customer to make preparation for receiving goods in advance.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for optimizing the distribution path and controlling the inventory of green agricultural products. Background Art

[0002] Agricultural products face a technical contradiction between quality loss and transportation efficiency during the distribution process. Since agricultural products are highly sensitive to environmental factors such as temperature, humidity and vibration, their quality is prone to decline during long-distance transportation. Especially under complex and changeable road conditions, the road conditions of different sections, changes in temperature and humidity, and the degree of vehicle vibration will have different degrees of impact on the quality of agricultural products. At the same time, uncertain factors such as road congestion and weather changes during the distribution process will affect the delivery timeliness, causing agricultural products to rot and deteriorate during transportation. Traditional distribution route planning mainly considers the shortest distance and the fastest time, ignoring the impact of road conditions on the quality of agricultural products, and cannot accurately predict and control the quality loss during transportation. In addition, the distribution environment data is scattered and has poor real-time performance, making it difficult to detect and handle abnormal situations during transportation in a timely manner, resulting in damage to the quality of agricultural products. These factors are interrelated, and both the distribution efficiency and the quality of agricultural products must be guaranteed, forming a contradiction between efficiency and quality during transportation. How to minimize the adverse impact of the transportation link on the quality of agricultural products while ensuring the delivery timeliness has become a core technical problem that needs to be solved in the logistics and distribution of agricultural products. Summary of the invention

[0003] The present invention provides a green agricultural product distribution path optimization and inventory control method, which mainly includes:

[0004] Collect the location coordinate data of the delivery vehicles and the road environment data along the way in real time, and use machine learning algorithms to establish a prediction model for the impact of agricultural product quality based on the historical driving trajectory data and road environment data of the delivery vehicles. By analyzing the real-time collected vehicle location coordinate data and road environment data, predict the impact of different delivery routes on agricultural product quality.

[0005] According to the prediction results of the impact of agricultural product quality, the delivery route with the least impact of agricultural product quality is selected as the optimal route, and the optimized delivery route is sent to the delivery vehicle terminal in real time;

[0006] At the delivery vehicle terminal, after receiving the optimal delivery route sent from the cloud, the delivery route is navigated in real time through the on-board navigation system, guiding the driver to drive along the optimal route, and the delivery route is dynamically adjusted according to real-time traffic information;

[0007] At the same time, the arrival time of agricultural products is dynamically estimated based on the real-time location and driving speed of the delivery vehicles, and the estimated arrival time information is pushed to the destination customers to remind them to prepare for receiving the goods in advance;

[0008] During the delivery process, the temperature and humidity indicators of the environment in which the agricultural products are located are continuously monitored. If the environmental indicators are detected to exceed the preset threshold range, the early warning mechanism is triggered, and the driver is notified in time by voice or text message to take emergency measures, which include turning on the refrigeration equipment or adjusting the temperature and humidity in the car;

[0009] After delivery is completed, the vehicle location, road conditions and agricultural product quality data collected during the delivery process are uploaded to the cloud and integrated with historical data for analysis to optimize the prediction model of the impact of agricultural product quality. The inventory strategy is adjusted by analyzing the relationship between delivery data and inventory control data.

[0010] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0011] The present invention discloses a green agricultural product distribution route optimization and inventory control method. The method collects the position and road condition data of the distribution vehicle in real time, combines the historical driving trajectory, and establishes a prediction model for the impact of agricultural product quality. According to the prediction results, the route with the least impact on the quality of agricultural products is selected as the optimal distribution route. At the same time, data such as road flatness and slope are obtained, the vibration frequency and acceleration of the vehicle are monitored, the impact of road conditions on the quality of agricultural products is predicted, and the route is replanned when necessary. In addition, according to the spoilage time of agricultural products, the quality safety threshold is determined, and the distribution time is calculated using a dynamic programming algorithm to ensure timeliness. During the distribution process, the temperature and humidity indicators are continuously monitored, and when the threshold is exceeded, the driver is notified in time to take emergency measures. After the distribution is completed, the relevant data is uploaded for model optimization and inventory strategy adjustment. The present invention realizes the dynamic optimization of agricultural product distribution routes through multi-dimensional data analysis and intelligent algorithms, effectively guarantees the quality of agricultural products, and improves distribution efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 The present invention is a flow chart of a green agricultural product distribution route optimization and inventory control method.

[0013] Figure 2 It is a schematic diagram of a green agricultural product distribution route optimization and inventory control method of the present invention.

[0014] Figure 3 This is another schematic diagram of a green agricultural product distribution route optimization and inventory control method of the present invention. DETAILED DESCRIPTION

[0015] In order to further understand the content of the present invention, the present invention is described in detail in conjunction with the accompanying drawings and embodiments. The present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are only used to explain the relevant inventions, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the invention are shown in the accompanying drawings.

[0016] like Figure 1-3 In this embodiment, a green agricultural product distribution path optimization and inventory control method may specifically include:

[0017] Step S101, collect the position coordinate data of the delivery vehicle and the road environment data along the way in real time, and use a machine learning algorithm to establish a prediction model for the degree of impact on the quality of agricultural products based on the historical driving trajectory data and road environment data of the delivery vehicle. By analyzing the real-time collected vehicle position coordinate data and road environment data, the degree of impact of different delivery routes on the quality of agricultural products is predicted.

[0018] The vehicle driving environment monitoring data is obtained by collecting the road vibration frequency value, the cabin temperature and humidity value and the vehicle acceleration change value through the vehicle-mounted sensor; based on the vehicle driving environment monitoring data, the linear regression tool is used to calculate the influence coefficient of the road flatness value, the ambient temperature value and the vehicle vibration amplitude value on the product quality change rate; the real-time position coordinate point sequence is obtained from the vehicle-mounted terminal, and the road condition environment quantitative index is obtained through convolutional neural network processing for the deviation value between the position coordinate point sequence and the planned path of the road network map; based on the road condition environment quantitative index and the influence coefficient, the moving average tool is used to calculate the product quality score value, and the road condition environment comprehensive score is obtained according to the road flatness weight, the congestion degree weight and the ambient temperature weight; it is judged whether the road condition environment comprehensive score is lower than the preset threshold lower limit value, and if the road condition environment comprehensive score is lower than the preset threshold lower limit value, a route switching instruction is triggered to the vehicle-mounted terminal.

[0019] Specifically, the vehicle-mounted sensors collect the road vibration frequency value, the cabin temperature and humidity value, and the vehicle acceleration change value every 30 seconds, and the linear regression tool is used to calculate the influence of the road flatness value, the ambient temperature value, and the vehicle vibration amplitude value on the product quality change rate, and establish a 0-100 product quality scoring system from the three dimensions of product appearance integrity, freshness, and shelf life. The real-time position coordinate point sequence is obtained from the on-board terminal of the delivery vehicle. The deviation value between the actual driving trajectory of the vehicle and the planned path of the road network map is combined with the road type score, the road condition score, the surrounding building density score, and the traffic flow value. The road condition environment data is standardized through the convolutional neural network to generate a quantitative indicator of the impact of the road condition environment on product quality. The corresponding relationship between the sampling point location coordinates and the product quality score is established based on the road condition environment monitoring data collected from the vehicle driving trajectory, and the product quality score when loading and unloading is obtained from the historical distribution record library. The comprehensive road condition environment score is calculated according to the ratio of road flatness weight 0.4, congestion weight 0.3, and ambient temperature weight 0.3. The initial quality score of the product and the initial value of the on-board sensor are recorded before the product is loaded and delivered. The moving average tool is used to smooth the collected data. When the comprehensive score of the road environment is lower than the 60-point range, it is determined to be a road section with aggravated quality loss. When the compartment temperature exceeds the preset threshold upper limit or the vehicle vibration frequency exceeds the preset threshold upper limit, the route switching instruction is triggered. The impact of road vibration frequency on the quality of agricultural products is reflected in the collision damage of fruits, the damage of vegetable leaves, and the surface abrasions of root products. The vibration frequency in the range of 3-5 Hz is most obvious for agricultural products. When the delivery vehicle is driving on a bumpy road, the sensor collects a vibration frequency of 4.2 Hz. At this time, the surface damage rate of apples reaches 12%. When it is lower than 3 Hz or higher than 5 Hz, the damage rate drops to less than 5%. Changes in compartment temperature will cause agricultural products to rot and deteriorate faster. After leafy vegetables are stored in an environment above 28 degrees for 4 hours, the water loss rate of leaves exceeds 8%, and the freshness score drops from 90 points to 75 points. When the temperature continues to rise to 32 degrees, the water loss rate of leaves reaches 15% after 4 hours, and the freshness score drops to 60 points. Real-time monitoring data from the vehicle temperature sensor showed that the temperature in the car rose to 30.5 degrees during the noon driving period and lasted for 90 minutes. The comprehensive score of road conditions reflects the degree of influence of road conditions on the loss of product quality during the delivery process. The main roads in the city are in good condition with a flatness score of 85 points, but the congestion index during peak hours in the morning and evening reaches 2.1, and the average speed of delivery vehicles is less than 15 kilometers per hour, resulting in longer delivery time and increased loss of vegetable shelf life. The congestion index of suburban roads is 0.8, and the traffic is smooth, but the road surface is damaged in some sections, with a flatness score of 65 points, which leads to increased vibration during transportation.The product quality scoring system is based on quantitative assessment of appearance integrity, freshness, and shelf life. Taking strawberry transportation as an example, the appearance integrity is scored based on the degree of fruit damage and the degree of surface fading, and the freshness is scored based on the hardness and glossiness of the fruit. When loading, the appearance integrity of the strawberries is 95 points and the freshness is 98 points. After 4 hours of delivery, the appearance integrity dropped to 82 points due to road bumps, and the high temperature environment caused the freshness to drop to 85 points. The delivery route switching rules are based on real-time monitoring data of road conditions and environment. When the comprehensive road score is lower than the 60-point range, it means that the current road condition is poor or congested. At this time, if the on-board sensor detects that the car temperature exceeds 32 degrees or the vibration frequency exceeds 4.5 Hz, the route switching command is triggered to bypass through an alternative route to avoid adverse road conditions. Records from road condition monitoring points show that the road score of urban expressways drops to 55 points during the morning rush hour. At this time, the temperature is normal but the congestion is aggravated. The system automatically plans a new route to avoid congested sections. The trend curve of agricultural product quality changes records the quality loss process throughout the distribution process. The slope of the curve reflects the speed of quality decline. When the absolute value of the slope increases, it indicates that the quality loss has intensified and countermeasures need to be taken immediately. The quality curve of the tomato distribution process shows that the quality score is 95 points when loading, and the curve drops gently to 90 points during the first 2 hours of stable transportation. The curve drops rapidly to 78 points when passing through the damaged road surface in the last 2 hours, indicating that the deterioration of road conditions has led to increased quality loss. The accumulation of historical distribution data forms a knowledge base on the relationship between road conditions and quality loss, records the degree of influence of each section on the quality of agricultural products under different seasons, time periods, and weather conditions, and assists in optimizing distribution route planning. The data shows that when driving on the viaduct section at noon in summer, the quality loss rate of fruit and vegetable products is 2.3 times that of the early morning period due to the increase in temperature, while the temperature in the tunnel section is relatively stable throughout the day, and the difference in quality loss is within 10%.

[0020] Obtain road surface flatness and slope data for different sections of the delivery route, as well as on-board sensors that monitor the vibration frequency and acceleration of the vehicle in real time during driving. Input the real-time monitoring data of road conditions into the preset association model between road conditions and agricultural product quality to predict the impact of changes in road conditions on the quality of agricultural products. If the impact exceeds the preset threshold, replan the delivery route.

[0021] Acquire the road surface flatness value, slope angle value, and concave-convex amplitude value collected by the road surface detection radar, and the road surface detection radar collects data once at a preset interval; collect vibration waveform data recorded by the vehicle-mounted three-axis acceleration sensor, perform wavelet noise reduction processing on the vibration waveform data, and calculate the predicted value of the degree of damage to agricultural products based on the noise-reduced data; obtain sampling data in the vibration waveform data where the vibration frequency exceeds the preset frequency interval value or the car body tilt angle exceeds the preset angle interval value, perform moving average smoothing processing on the sampling data, extract the acceleration peak point, deceleration valley point, and tilt angle mutation point, and generate a road condition impact score; calculate the road comprehensive score according to the road surface flatness value, slope angle value, and vibration frequency value according to a preset weight ratio; determine whether the road comprehensive score is lower than the preset score interval value or whether the vibration frequency exceeds the preset vibration frequency for a preset time, and if so, trigger a path planning instruction based on a preset agricultural product quality loss threshold.

[0022] Specifically, the road surface flatness value, slope angle value, and bump amplitude are obtained from the road surface detection radar every 5 seconds, and the vehicle's vibration waveform data in the x, y, and z directions is recorded every 0.01 seconds through the on-board three-axis acceleration sensor. After the collected data is processed by wavelet noise reduction, the predicted value of the degree of damage to leafy vegetables, fruits, and tubers is calculated based on the random forest regression tool. According to the data collected by the road surface detection radar and the three-axis acceleration sensor, when the vibration frequency exceeds the 4.2 Hz interval value or the car body tilt angle exceeds the 12 degree interval value when the vehicle passes through a bumpy section, the moving average tool is used to smooth the sampled data, and the acceleration peak point, deceleration valley point, and tilt angle mutation point are extracted from it to generate a road condition impact score. The laser rangefinder is used to scan the road surface every 2 meters to obtain the road surface void depth, crack width, and depression length. The vehicle-mounted inertial measurement device records the driving trajectory data every second, and the abnormal data that exceeds the standard deviation by 3 times is eliminated. The comprehensive road score is calculated by superimposing the road surface flatness weight of 0.4, the slope angle weight of 0.3, and the vibration frequency weight of 0.3. The agricultural product quality records are retrieved from the historical database. When the comprehensive road condition score is lower than the interval value of 65 points or the vibration frequency exceeds the interval value of 4.2 Hz for 5 seconds, the quality loss threshold is set for different agricultural product categories, with a loss threshold of 8% for leafy vegetables, 5% for fruits, and 3% for tubers. The path planning instruction is triggered on sections where the road condition score does not meet the standard. The road surface roughness value reflects the degree of unevenness of the road surface. The detection radar scans the road surface every 5 seconds to obtain the roughness index. A roughness index of 90 points indicates that the road surface is flat and smooth, which is suitable for the transportation of agricultural products. When the roughness drops to 70 points, slight potholes have appeared, and the strawberry transportation damage rate reaches 5%. When the roughness is lower than 60 points, the road surface is seriously damaged, and the strawberry damage rate exceeds 12%. The on-board three-axis acceleration sensor monitors the vibration condition during transportation in real time, with a sampling frequency of 100 Hz, that is, data is recorded every 0.01 seconds. The x-axis reflects the left and right swing of the vehicle, the y-axis represents the front and rear bumps, and the z-axis represents the up and down vibration. When driving on a damaged road, the z-axis vibration amplitude reaches 0.8g, causing the crushing rate of loaded tomatoes to rise to 8%. The slope angle value shows the degree of inclination of the road. When the slope exceeds 8 degrees, the vegetable box shakes and squeezes, and leafy vegetables are obviously damaged. The breakage rate of lettuce leaves reaches 15%. When it exceeds 12 degrees, the box shakes more severely and the breakage rate rises to 25%. When the slope is below 4 degrees, the loss of vegetable transportation is kept within 5%. The laser rangefinder scans the road surface every 2 meters to measure the depth of potholes, crack width, and depression length. Potholes with a depth of more than 3 cm will cause the vehicle to jump violently. When the crack width exceeds 2 cm, tire wear increases. When the depression length exceeds 50 cm, the vehicle needs to slow down and detour.The comprehensive road condition score is calculated by weighted calculation. Taking potato transportation as an example, the flatness of the road surface accounts for 40%, the slope angle accounts for 30%, and the vibration frequency accounts for 30%. When the comprehensive score is lower than 65 points, the potato abrasion rate exceeds 10%. When it is lower than 60 points, the abrasion rate reaches 18%. At this time, the path planning is triggered and a parallel road is selected for detour. Different agricultural products have different requirements for transportation conditions. Leafy vegetables are easily affected by vibration, resulting in leaf damage. The loss threshold is set at 8%. Fruits are prone to collision and crushing. The loss threshold is 5%. Tubers are more tolerant and the loss threshold is set at 3%. When the threshold is exceeded, it means that the current road conditions are not suitable for the transportation of such agricultural products. The vibration frequency is significantly correlated with the degree of damage to agricultural products. The frequency in the range of 3.5-4.5 Hz has the greatest impact on the damage of fruits and vegetables. Taking apples as an example, this frequency for 5 seconds will cause the collision rate to increase by 15%. When the frequency is lower than 3.0 Hz or higher than 5.0 Hz, the collision rate is maintained within 8%. Data anomaly processing uses statistical methods to calculate the mean and standard deviation of the sampled data, and eliminates outliers that exceed 3 times the standard deviation. Road surface detection data shows that during peak hours in the morning and evening, dense traffic leads to sampling interference, and the frequency of abnormal data increases. At this time, it is necessary to strengthen data cleaning efforts.

[0023] Step S102, based on the prediction result of the impact degree of agricultural product quality, select the distribution route with the least impact degree of agricultural product quality as the optimal route, and send the optimized distribution route to the distribution vehicle terminal in real time.

[0024] The road section number and condition data are obtained by using an on-board terminal, the standard deviation interval is calculated according to the road section condition data, the data exceeding the preset multiple of the standard deviation interval is eliminated, and the road section loss rate value is obtained by the temperature loss weight and the vibration loss weight; the vehicle position coordinate point is obtained according to the Beidou positioning module carried by the on-board terminal, and the road network structure data of the road around the vehicle position coordinate point is extracted from the electronic map database; the road section loss rate value is smoothed by using a moving average tool, and if the road section loss rate is lower than a preset threshold and the straight-line distance is less than a preset multiple of the shortest distance, a candidate delivery route is obtained; the road section loss rate value is read for the candidate delivery route, the weighted loss rate value is calculated according to the leaf vegetable loss threshold and the fruit loss threshold, and the weighted loss rate change trend is predicted by a recursive neural network; the candidate delivery routes are sorted according to the predicted loss rate value, and if the road section congestion index exceeds the preset threshold and the duration exceeds the preset interval, the route with the second smallest loss rate is selected from the candidate delivery routes as the new delivery route.

[0025] Specifically, the vehicle terminal collects the road section number, road condition score, congestion index value, speed value, road surface temperature value, and vibration frequency value every 30 seconds, removes abnormal data that exceeds the standard deviation by 3 times, calculates the road section loss rate value based on the temperature loss weight of 0.4, the vibration loss weight of 0.3, and the duration loss weight of 0.3, stores it in the road condition database according to the area number, and updates it every 5 minutes. The vehicle position coordinates are obtained every 10 seconds based on the vehicle-mounted Beidou positioning module, and the surrounding road network structure is extracted from the electronic map database. The moving average tool is used to smooth the road section loss rate value. In the sections with a loss rate below the preset threshold, the delivery routes with a straight-line distance less than 1.2 times the shortest distance, less than 8 turns, and less than 4 ramps are selected. For the selected delivery routes, the road loss rate values ​​are read from the road condition database, and the loss thresholds of 8% for leafy vegetables, 5% for fruits, and 3% for tubers are distinguished for calculation. The weighted loss rate values ​​are generated in combination with the road travel time ratio, and the loss rate change trend of each route within 2 hours is predicted based on the recursive neural network. The delivery routes are sorted from small to large according to the predicted loss rate values, and the route with the smallest loss rate is selected and sent to the vehicle terminal. When the road congestion index exceeds 2.0 or the road condition score is less than 65 points for 180 seconds, the route switching instruction is triggered, and the route with the second smallest loss rate and normal real-time road conditions is selected from the sorting sequence as the new delivery route. The road loss rate value reflects the degree of influence of road conditions on the quality of agricultural products. The vehicle terminal collects data every 30 seconds to monitor the road temperature, vibration, and traffic speed. When the temperature is higher than 28 degrees, the loss of leafy vegetables increases, when the vibration frequency exceeds 4 Hz, the crushing of fruits increases, and when the traffic speed is lower than 15 kilometers per hour, the delivery time is extended, resulting in quality degradation. Abnormal data processing uses statistical methods to calculate the mean and standard deviation of the sampled data, and remove data points that exceed the range of 3 times the standard deviation. The sudden change of road temperature from 25 degrees to 45 degrees, the vibration frequency from 2 Hz to 8 Hz, and the vehicle speed from 40 kilometers per hour to 5 kilometers per hour are all considered abnormal data. The Beidou positioning module provides real-time vehicle location information, with a sampling interval of 10 seconds and a positioning accuracy better than 10 meters. The driving trajectory is drawn through continuous position coordinate points to determine whether the vehicle deviates from the planned route. When the deviation distance exceeds 50 meters, a route deviation reminder is issued. The road network structure diagram shows the road connection relationship and traffic conditions, including information such as section length, turning radius, slope angle, and traffic light location. The 1.2 times straight-line distance limit prevents the route from being too circuitous, the 8-turn limit reduces the shaking of loaded goods, and the 4-section ramp limit reduces uphill losses. Different agricultural products have different sensitivities to transportation conditions. Leafy vegetables are easily affected by temperature and wilt, so the loss threshold is set at 8%. Fruits are afraid of being squeezed and broken, so the loss threshold is 5%. Tubers are relatively resistant to transportation, so the loss threshold is set at 3%. When the threshold is exceeded, the transportation mode is switched to low-temperature or shock-absorbing transportation mode.The real-time update mechanism of road condition data refreshes the road section status every 5 minutes. A road condition score below 65 points indicates poor road conditions, and a congestion index exceeding 2.0 indicates slow traffic. When the two indicators exceed the standard for 180 seconds, the route adjustment is initiated and the best alternative route is selected from the alternative routes. The prediction of the loss rate change trend is based on the law of historical data. The recursive neural network learns the change characteristics of the loss rate under different time periods, weather, and road conditions, and predicts the loss rate trend of each route in the next 2 hours to provide a decision-making basis for route selection. Route optimization is based on multi-dimensional evaluation indicators. The loss rate is the primary consideration. Combined with the requirements of delivery timeliness, transportation cost, etc., the route with the highest comprehensive score is selected as the optimal delivery route. When the road conditions change significantly, it will switch to the alternative route in time. Moving average processing eliminates the impact of data fluctuations, smoothes the continuously collected loss rate values, and the window width is set to 5 data points. After removing short-term fluctuations, it more accurately reflects the trend of road conditions and improves the accuracy of route selection.

[0026] Step S103, at the delivery vehicle terminal, after receiving the optimal delivery route sent from the cloud, the delivery route is navigated in real time through the vehicle navigation system to guide the driver to drive along the optimal route, and the delivery route is dynamically adjusted according to the real-time road condition information.

[0027] The vehicle position coordinate point sequence is obtained through the vehicle-mounted Beidou positioning module, and the moving average tool is used to obtain the smoothed position data based on the vehicle position coordinate point sequence; the road ahead traffic restriction information and lane speed limit value are obtained from the road network map database, the abnormal value is judged according to the position data, and the value outside the standard deviation interval is eliminated to obtain the valid position data; the lateral deviation distance value and the longitudinal deviation distance value are calculated by the sliding window tool according to the valid position data and the navigation route coordinate sequence, and if the lateral deviation exceeds the first distance threshold when the vehicle speed is lower than the first speed threshold, the reminder information is triggered; the long short-term memory neural network is used to predict the vehicle running trajectory according to the lateral deviation distance value and the longitudinal deviation distance value, and the deviation warning information is generated to the vehicle-mounted terminal; the real-time road condition data is read from the vehicle-mounted terminal, and if the traffic speed in the real-time road condition data is lower than the second speed threshold and the congestion index exceeds the preset threshold, the parallel road section road condition information is obtained from the road network node database to generate an alternative navigation route.

[0028] Specifically, the vehicle's real-time position coordinates are collected every 5 seconds through the vehicle-mounted Beidou positioning module. The moving average tool is used to smooth the position data according to 15 coordinate points as a window. The road traffic restrictions, construction occupation, traffic light status, and lane speed limit values ​​in the 1,000-meter area ahead are obtained from the road network map database. The collected data is judged as abnormal values, and the values ​​that deviate from the standard deviation by 3 times are eliminated. Based on the vehicle position coordinates and the navigation route coordinate sequence, the sliding window tool is used to calculate the lateral deviation distance value and the longitudinal deviation distance value. If the lateral deviation exceeds 30 meters when the vehicle speed is less than 20 kilometers per hour, a reminder is triggered. If the lateral deviation exceeds 50 meters when the vehicle speed exceeds 20 kilometers per hour, a reminder is triggered. The long short-term memory neural network is used to predict the vehicle's running trajectory and generate deviation warning information. The real-time road condition data is read from the vehicle-mounted terminal every 15 seconds, and the threshold value of the road section traffic speed, congestion index, and construction information is judged. When the road ahead has a traffic speed of less than 20 kilometers per hour or the congestion index exceeds 2.0 for 180 seconds, the parallel road section traffic condition information is extracted from the road network node database to generate an alternative navigation route. The steering warning distance is dynamically calculated based on the real-time speed of the vehicle. A warning is given 150 meters in advance when the speed is below 20 kilometers per hour, 200 meters in advance when the speed is between 20 and 40 kilometers per hour, and 250 meters in advance when the speed is above 40 kilometers per hour. The steering direction and distance information are broadcast by voice, and the lane guidance arrow is displayed on the vehicle screen. The reminder information is turned off when entering the target lane. The vehicle position data collection frequency setting reflects the accuracy requirements of the motion trajectory. The 5-second interval ensures positioning accuracy and avoids data redundancy. The position coordinates have an accuracy of 10 meters in urban environments and 5 meters in open areas in suburbs. A smooth trajectory curve is constructed through 15 consecutive coordinate points. The vehicle speed is positively correlated with the deviation warning distance. The reaction distance is short when driving at low speed, and a lateral deviation of 30 meters is sufficient to determine the deviation. When driving at high speed, the inertia is large and the reaction distance is long. The lateral deviation threshold is increased to 50 meters. Delivery vehicles usually keep below 40 kilometers per hour in urban areas, and the speed of suburban roads is controlled within 60 kilometers per hour. The real-time monitoring value of road traffic status shows the degree of congestion. The average speed during heavy traffic hours drops to 15 kilometers per hour, and the congestion index exceeds 2.0, which is equivalent to a 50% decrease in road capacity. The congestion lasting for 180 seconds indicates that the congestion is not a short-term fluctuation, and the road section has become stable due to traffic obstruction. The turning reminder distance is dynamically adjusted with the vehicle speed. The low-speed driving is broadcast 150 meters in advance to facilitate accurate grasp of the intersection position, the medium-speed is broadcast 200 meters in advance to leave a lane change buffer space, and the high-speed is broadcast 250 meters in advance to ensure turning safety. The combination of turning voice prompts and arrow guidance improves navigation accuracy. Construction road occupation information has a significant impact on navigation route planning. Road occupation construction leads to a reduction in lane width and reduced traffic capacity. The average speed of the road section drops to less than 10 kilometers per hour. Parallel sections are selected from the database for detours to avoid construction-affected areas.The long short-term memory neural network predicts the trend of driving trajectory changes, and infers the direction of movement in the next 10 seconds based on the position change sequence of the previous 30 seconds. When the predicted trajectory deviates from the navigation route by more than 45 degrees, an early warning is issued to prevent the vehicle from entering the wrong section in advance. The road condition data is updated every 15 seconds to reflect the changes in road traffic status. The speed of the viaduct during peak hours in the morning and evening is reduced from 40 kilometers per hour to 8 kilometers per hour, and the speed of parallel ground roads is maintained at 25 kilometers per hour. The navigation system automatically recommends changing to ground roads. The lane guidance arrow guidance conforms to the driver's visual habits. The arrow points to the direction of the target lane and marks the distance value. When driving on the right, the right turn arrow is displayed. When changing lanes, the arrow indicates dynamically, and the arrow prompt disappears after entering the target lane. The sliding window data processing eliminates noise interference, takes 15 seconds of sampling data to calculate the average position, and eliminates instantaneous drift points. The satellite signal in the intersection area is easily blocked, and the position drift deviation reaches 30 meters. The trajectory is kept smooth and continuous through sliding average. The road network node database records road connection relationships, including information such as road grade, travel direction, and turning restrictions. When the main road is congested, secondary roads are selected as substitutes. When the traffic capacity is similar, sections with fewer traffic lights are given priority to reduce waiting time.

[0029] Obtain spoilage time data of different agricultural products, determine the threshold of agricultural product quality and safety according to the requirements of agricultural product freshness, and use dynamic programming algorithm to calculate the estimated duration of the delivery route based on the restrictions on the type of transportation tools. If the estimated duration of the delivery route exceeds the threshold of agricultural product quality and safety, re-plan the delivery route according to the evaluation index of path timeliness and the optimization criteria to obtain the optimal path in terms of timeliness.

[0030] The quality monitoring data of leafy vegetables, fruits and tubers are obtained from the agricultural product commodity library, and a temperature fluctuation and quality loss correlation curve is established with the quality monitoring data and the temperature value collected by the car temperature sensor; the loss rate value is standardized according to the quality loss correlation curve, and the quality safety threshold is obtained through the standardized processing; the road travel time is read from the road network map database, and the road travel time is used to calculate the travel time of each section, and the section nodes that exceed the quality safety threshold are eliminated; the road condition data is smoothed using a moving average tool, and the section timeliness score is calculated according to the smoothed road condition data according to the travel time weight, temperature control weight, and road condition score weight; the sections are sorted according to the timeliness score and the section with the highest score is selected. If the section travel speed is lower than the preset speed threshold or the car temperature exceeds the preset temperature range for a preset time, the section with the second highest timeliness score is selected from the road network database for updating.

[0031] Specifically, the quality monitoring data of leafy vegetables, fruits, and tubers are obtained from the agricultural product commodity library every 15 minutes, and the temperature value is collected every 5 minutes through the car temperature sensor. According to the shelf life of 48 hours for leafy vegetables, 72 hours for fruits, and 120 hours for tubers, the random forest algorithm is used to establish the temperature fluctuation and quality loss correlation curve, and the loss rate value is standardized. Based on the conditions that the car temperature is in the range of 2-8 degrees Celsius, the cooling power is 2000 watts, and the load limit is 2000 kilograms, the road travel time is read from the road network map database, and the Dijkstra shortest path tool is used to calculate the travel time of each section. For sections that exceed the quality safety threshold by 50%, the section nodes are removed from the road network database. The real-time speed, traffic light timing, and congestion index values ​​of the roads are obtained from the road network map every 30 seconds. The timeliness score of the road section is calculated according to the weight of 0.4 for the travel time, 0.3 for the temperature control, and 0.3 for the road condition score. The road condition data is smoothed by the moving average tool to generate the evaluation results of the road section traffic status. The sections are sorted according to the timeliness score, and the sections with the highest timeliness score are selected to form the delivery route. When the speed of the section is lower than 20 kilometers per hour or the compartment temperature exceeds the range of 2-8 degrees Celsius for 180 seconds, the section with the second highest timeliness score is selected from the road network database for update to achieve dynamic optimization and adjustment of the delivery route. The agricultural product quality monitoring data reflects the degree of influence of the storage and transportation environment on the freshness of the product. The quality retention rate of lettuce stored in a 4-degree Celsius environment for 48 hours reaches 92%, and the retention rate drops to 85% when the temperature rises to 8 degrees Celsius. After exceeding the 8-degree Celsius range for 2 hours, the retention rate drops rapidly to 70%. The temperature fluctuation in the carriage directly affects the shelf life of agricultural products. The temperature sensor records data every 5 minutes. When the outside temperature is 32 degrees Celsius, the 2000-watt refrigeration unit runs for 30 minutes to reduce the temperature of the carriage to 4 degrees Celsius and maintain a constant temperature. The temperature fluctuation is controlled within the range of plus or minus 1 degree Celsius. The random forest algorithm processes multi-dimensional monitoring data and establishes the relationship between temperature, time, and quality. When strawberries are stored at 6 degrees Celsius, the shelf life is 72 hours, and the quality loss rate is 0.4% per hour. When the temperature rises to 12 degrees Celsius, the loss rate increases to 0.8% per hour. The remaining shelf life is predicted by the loss curve. The road network data is updated in real time to show changes in road conditions. The average speed of the main road during the morning rush hour drops to 15 kilometers per hour. The extended delivery time leads to greater fluctuations in the temperature of the carriage, affecting the cooling effect. At this time, detouring the secondary road increases the driving distance but shortens the delivery time and reduces the risk of quality loss. The timeliness score reflects the efficiency of the road section. The travel time accounts for 40% and reflects the transportation speed requirement. The temperature control accounts for 30% and reflects the stability of the storage and transportation environment. The road condition score accounts for 30% and predicts changes in traffic conditions. Road sections with a comprehensive score higher than 80 points will be given priority.Dynamic monitoring of road traffic status enables accurate delivery. When it is detected that the speed of the road ahead is less than 20 kilometers per hour, the extended delivery time will lead to increased quality loss, and the route will automatically switch to an alternative route to ensure that the agricultural products are delivered within the quality safety threshold time. Cold chain transportation has strict requirements on vehicle load limits. Overloading will reduce refrigeration efficiency and cause temperature rise. A 2,000-kilogram load-limited vehicle is loaded with 1,500 kilograms of strawberries, and a 500-kilogram margin is reserved to ensure the refrigeration effect. The compartment volume is 12 cubic meters to ensure that the goods are neatly stacked and evenly ventilated. The Dijkstra tool calculates the shortest path distance, selects directly connected sections from the road network nodes to form a transportation route, and eliminates sections where the delivery time exceeds the quality safety threshold by 50%, avoiding quality loss caused by transportation delays and ensuring the timeliness of agricultural product delivery. Moving average processing eliminates data fluctuation interference. Road condition monitoring data fluctuates due to changes in traffic volume. The average value of 5 data points is used to reflect the actual traffic status of the road section, improve the accuracy of road condition evaluation, and guide the optimization and adjustment of delivery routes. Route switching rules ensure a smooth and orderly delivery process. The route will be changed when the traffic speed is lower than 20 kilometers per hour or the temperature is out of range for 180 seconds. The alternative route will be selected with the highest timeliness score and normal road conditions to achieve dynamic optimization of the transportation process.

[0032] Step S104, at the same time, dynamically estimate the arrival time of the agricultural products according to the real-time position and driving speed of the delivery vehicle, and push the estimated arrival time information to the destination customer to remind the customer to prepare for receiving the goods in advance.

[0033] The position coordinate point information collected by the vehicle-mounted Beidou positioning module is received, and the position coordinate point information is smoothed by a sliding average tool to obtain a smooth position sequence; the remaining delivery distance value is obtained from the electronic map database according to the smooth position sequence, and the estimated travel time value is calculated by using the real-time road speed, traffic light timing, and road section speed limit values; the average speed, acceleration, and deceleration values ​​in the vehicle's driving trajectory are obtained, and a speed prediction model is obtained by back propagation neural network training, and the speed prediction model marks the speed value exceeding the preset multiple interval of the standard deviation as an abnormal point; the real-time congestion index, accident information, and construction road occupation data are read from the road network database. If the road speed is lower than the specified threshold or the congestion index exceeds the set threshold for a specified period of time, a parallel road section is selected from the road network database to re-plan the delivery route; according to the delivery route, when the remaining delivery time reaches the set threshold, an expected arrival reminder information is sent to the consignee, and the arrival reminder information is updated at different time intervals according to the remaining delivery time threshold.

[0034] Specifically, the vehicle position coordinates are collected every 15 seconds through the on-board Beidou positioning module, and the coordinate sequence is smoothed using a 20-point sliding average tool. The remaining delivery distance value is obtained from the electronic map database. Based on the real-time road speed, traffic light timing, and road section speed limit values, the estimated travel time is calculated according to the travel time weight of 0.6 and the road condition score weight of 0.4. The average travel speed, acceleration, and deceleration values ​​are extracted from the vehicle's driving trajectory every 10 seconds. Combined with the estimated road section travel time, the speed prediction model is trained using a back propagation neural network. The speed values ​​that exceed the standard deviation by 3 times are marked as abnormal points, and a smooth speed curve is generated using a 5-point moving average tool. The real-time congestion index, accident information, and construction road occupation data are read from the road network database every 30 seconds, and the arrival time correction value is calculated based on the smooth speed curve. When the road speed is lower than 20 kilometers per hour or the congestion index exceeds 2.0 for 180 seconds, a parallel section is selected from the road network database to re-plan the delivery route. According to the dynamically corrected arrival time, an estimated arrival reminder is sent to the recipient when the remaining delivery time is 120 minutes. The arrival time is updated every 30 minutes thereafter. When the remaining time is less than 30 minutes, accurate arrival information is pushed every 5 minutes. If the delay exceeds 10 minutes due to changes in road conditions, an update message is sent immediately. The location data collection interval setting reflects the vehicle motion accuracy requirements. The 15-second interval is sufficient to capture vehicle steering changes in urban road environments. The Beidou positioning accuracy is better than 10 meters. The 20-point sliding average processing eliminates signal jitter, and the smoothed trajectory curve is more in line with the actual driving path. The speed data extraction follows the principle of stability. The 10-second sampling interval can reflect speed changes and avoid instantaneous fluctuations. Delivery vehicles usually maintain a uniform speed of 40 kilometers per hour on the main road, with an acceleration of 0.5 meters per square second and a deceleration of 0.8 meters per square second. Sudden acceleration is marked as an abnormal point. The weight distribution of the estimated travel time reflects the degree of influence of road conditions. The travel time accounts for 60% to reflect the basic driving time, and the road condition score accounts for 40% to reflect the impact of congestion delays. The average speed during the morning rush hour on the main road drops from 40 kilometers to 15 kilometers, and the estimated arrival time increases by 45 minutes. The back propagation neural network learns the law of vehicle speed changes. The input includes historical vehicle speed, road condition index, and time period characteristics. The output predicts the speed change trend in the next 10 minutes. The model has an accuracy rate of 85% during the morning and evening peaks, providing data support for the arrival time estimation. The 30-second update frequency of road condition data meets the real-time monitoring needs. The congestion index of 2.0 indicates that the road capacity is reduced by 50%. The average speed of the delivery vehicle is less than 20 kilometers per hour for 180 seconds, which indicates that the congestion is not a short-term phenomenon and triggers the route adjustment mechanism. The hierarchical push of arrival reminder messages improves the service experience. The first reminder is sent in the remaining 120 minutes to facilitate the preparation of the consignee. After that, regular updates are made in the next 30 minutes to reflect the status changes. In the last 30 minutes, intensive push is made to ensure on-time delivery. Sudden changes in road conditions are notified in time to avoid waiting in vain.The selection of parallel sections is based on multi-dimensional evaluation indicators. The capacity of the alternative section should be equivalent to the original section, the distance should not increase by more than 20%, the number of traffic lights should be less than 8, and the delay should not exceed 15 minutes based on the original arrival time. The road condition abnormality is determined by statistical methods. The indicators such as speed, congestion index, and traffic volume are judged as abnormal when they exceed the 3 times standard deviation interval. Temporary fluctuations caused by emergencies are eliminated to ensure the accuracy of the estimated arrival time. The message push interval setting follows the principle of information value. The normal 120-minute warning maintains a 30-minute update interval, which is shortened to 5 minutes when it is close to 30 minutes, reflecting the urgency of time. Sudden changes in road conditions cause delays of more than 10 minutes and are pushed immediately to avoid affecting the delivery arrangement. Dynamic monitoring of road capacity realizes precise distribution. Information such as construction occupation, traffic control, and accident handling is included in the evaluation system. When the road section capacity drops by more than 50%, the delay time is automatically calculated and the update message is pushed to ensure that the consignee accurately grasps the arrival time.

[0035] Step S105, during the distribution process, continuously monitor the temperature and humidity indicators of the environment in which the agricultural products are located. If it is detected that the environmental indicators exceed the preset threshold range, the early warning mechanism is triggered, and the driver is promptly notified by voice or text message to take emergency measures, which include turning on the refrigeration equipment or adjusting the temperature and humidity in the car.

[0036] The temperature and humidity values ​​collected by the temperature and humidity sensor are obtained, and the moving average tool is used to obtain the smoothed temperature and humidity data according to the temperature and humidity values; the temperature and humidity distribution field is constructed according to the smoothed temperature and humidity data through the bilinear interpolation tool, and the deviation between the value of the monitoring point and the mean of the adjacent points is determined. If the deviation exceeds the preset percentage of the mean of the adjacent points, the monitoring point is marked as an abnormal point; the fluctuation value in the temperature and humidity distribution field is monitored by a vehicle-mounted terminal. If the temperature fluctuation amplitude exceeds the set threshold and lasts for a preset time, a voice reminder and SMS warning information are sent through the vehicle-mounted terminal; the power value of the refrigeration unit is read from the vehicle-mounted terminal, and the support vector regression tool is used to establish a temperature response curve according to the power value and temperature data, and the temperature change rate under different powers is obtained through step-by-step adjustment; the operating power of the refrigeration unit is adjusted according to the temperature change rate. If the temperature drop rate exceeds the preset threshold, the power of the refrigeration unit is reduced until the temperature fluctuation amplitude is lower than the preset threshold and lasts for a preset time.

[0037] Specifically, two temperature and humidity sensors on the top and four on the side walls of the carriage are used to collect temperature and humidity values ​​every 10 seconds. The collected data is smoothed using a 5-point moving average tool. The temperature and humidity thresholds of different agricultural products are read from the sensor database. When the door is opened and the temperature fluctuates by more than 3 degrees Celsius, the interference period is recorded and the data of this period is eliminated. The temperature and humidity distribution field of the carriage is constructed based on the bilinear interpolation tool, and the values ​​that exceed the mean of adjacent points by 50% are marked. When the temperature fluctuation amplitude of the monitoring point exceeds 1.5 degrees Celsius or the relative humidity fluctuation amplitude exceeds 4% for 60 seconds, the warning information is pushed through the voice reminder of the vehicle terminal and the SMS of the mobile phone, and the trend of the temperature and humidity curve is recorded. The power value of the refrigeration unit is read from the vehicle terminal every 5 seconds, and the response curve of the carriage temperature, refrigeration power, and ambient temperature is established based on the support vector regression tool. The refrigeration power is adjusted in a step-by-step manner. After starting the refrigeration power at 2000 watts for 5 minutes, it is adjusted step by step according to the gradient of 500 watts to monitor the rate of change of the carriage temperature. The operating power of the refrigeration unit and humidifier is automatically controlled according to the rate of change of temperature and humidity. When the temperature drop rate exceeds 0.5 degrees Celsius per minute or the humidity rise rate exceeds 1% per minute, the control power is reduced. When the temperature and humidity fall back to the threshold range and the fluctuation amplitude is less than 1.5 degrees Celsius or the relative humidity fluctuation amplitude is less than 4% for 120 seconds, the alarm reminder is turned off. The sensor layout scheme reflects the distribution characteristics of the temperature field in the car. The two sensors on the top monitor the sinking effect of the cold air. The temperature gradient is usually 0.3 degrees Celsius per meter. The four sensors on the side wall monitor the lateral temperature uniformity. The temperature difference is controlled within 1 degree Celsius, forming a 360-degree three-dimensional monitoring network. The sampling frequency setting is based on the dynamic characteristics of temperature changes. The 10-second interval captures temperature fluctuations and avoids data redundancy. The 5-point moving average eliminates short-term fluctuation interference. The door opening causes the temperature to rise by 5-8 degrees Celsius instantly and returns to stability within 120 seconds. The data during this period does not participate in the control calculation. Bilinear interpolation is used to construct the temperature field distribution map. The temperature value at any position in the car is calculated based on the data of 6 monitoring points. The temperature difference between adjacent points exceeding 50% usually indicates sensor failure or local refrigeration abnormality. The uniformity of the temperature field directly affects the preservation effect of agricultural products. The temperature and humidity warning threshold takes into account the tolerance of agricultural products. When the temperature of leafy vegetables rises to 8 degrees Celsius, the shelf life is shortened by 12 hours. When the temperature of fruits rises to 12 degrees Celsius, the ripening process is accelerated. Root vegetables have strong tolerance and the product quality remains stable when the temperature rises to 15 degrees Celsius. The refrigeration power is adjusted in steps to avoid system oscillation. In the startup phase, the 2000-watt power is used to quickly cool down. The car temperature usually drops by 3 degrees Celsius in 5 minutes, and then it is fine-tuned in steps of 500 watts to accurately control it within the range of 0.5 degrees Celsius above and below the target temperature. The support vector regression tool predicts the temperature change trend. The input includes three dimensions: car temperature, refrigeration power, and ambient temperature. When the ambient temperature is 35 degrees Celsius in summer, 2000 watts of refrigeration power consumes 2 degrees of electricity per hour, and the car temperature is maintained at 4 degrees Celsius.The temperature control accuracy varies with the ambient temperature fluctuation. When the ambient temperature is 25 degrees Celsius, the control accuracy reaches plus or minus 0.5 degrees Celsius, and when it is 35 degrees Celsius, it is relaxed to plus or minus 0.8 degrees Celsius. When the carriage is fully loaded, the temperature uniformity is improved by 20% compared with that when it is unloaded. Humidity regulation and temperature control affect each other. When the humidity is lower than 80%, the dehydration of agricultural products is accelerated, and when it is higher than 95%, the growth of microorganisms is promoted. When the humidifier power is 200 watts, the relative humidity increases by 0.5% per minute, and it is coordinated with refrigeration and cooling. The hierarchical push of early warning information improves the efficiency of operation and maintenance. When the temperature fluctuation exceeds 1.5 degrees Celsius, a yellow warning is triggered, and when it exceeds 3 degrees Celsius, it is upgraded to a red warning. If it is not processed for 60 seconds, it will be pushed again to ensure that abnormal working conditions are responded to in time. The adaptive adjustment of control parameters improves the stability of the system. In summer, the proportion of refrigeration in no-load operation is 80%, and when it is fully loaded, the proportion is increased to 95%. In winter, the demand for refrigeration decreases, and it switches to the temperature and humidity balance adjustment mode, which saves 30% of energy in comprehensive way.

[0038] Step S106, after the delivery is completed, the vehicle location, road conditions and agricultural product quality data collected during the delivery process are uploaded to the cloud and integrated with historical data for analysis to optimize the prediction model of the impact of agricultural product quality, and adjust the inventory strategy by analyzing the relationship between delivery data and inventory control data.

[0039] Receive the position coordinate sequence, road condition data, car compartment temperature and humidity data and product quality data collected by the vehicle terminal, the data is collected and uploaded by the vehicle terminal at preset time intervals; based on the data collected by the vehicle terminal, smooth the abnormal data through the moving average tool, and use the time series autoregression tool to obtain the road condition status and product loss rate correlation curve; read the inventory change data from the commodity inventory database, and obtain the sales forecast value through the exponential smoothing tool according to the product loss rate correlation curve and the commodity turnover speed curve; divide the inventory cycle according to the sales forecast value, calculate the fluctuation coefficient through the adaptive weighting tool, and generate the upper and lower limits of the inventory in combination with the product loss rate; obtain the inventory turnover data from the order database, if the inventory turnover speed is lower than the preset threshold or the product loss rate exceeds the preset threshold for a preset period of time, adjust the upper and lower limits of the inventory by the standard deviation step, and record the deviation value between the inventory and the actual sales.

[0040] Specifically, the vehicle terminal uploads the location coordinate sequence, road condition data, cabin temperature and humidity data, and product quality data to the cloud every 10 minutes. The abnormal data is smoothed based on the 15-point moving average tool. The time series autoregression tool is used to construct the curve associated with the road condition and product loss rate, and the loss threshold of leafy vegetables is 3%, the loss threshold of fruits is 4%, and the loss threshold of roots and tubers is 2%. The inventory change data is read from the commodity inventory database every hour, and the sales forecast value is calculated by the exponential smoothing tool in combination with the commodity turnover speed curve, product loss rate value, and time period sales record. The incremental data update is realized based on the breakpoint continuation mechanism, and the values ​​that exceed 2 times the standard deviation of the forecast interval are eliminated. The inventory cycle is dynamically divided according to the sales forecast results, and the average daily sales benchmark value is extracted from the sales database. The peak fluctuation coefficient and the off-season fluctuation coefficient are calculated through the adaptive weighting tool. Combined with the distribution radius and product loss rate of different regions, the upper and lower limits of real-time inventory are generated. Based on real-time monitoring data, the inventory warning status is judged, and the inventory turnover data for 7 consecutive days is read from the order database. When the turnover rate is lower than 0.8 or the product loss rate exceeds the preset threshold for 72 hours, the upper and lower limits of the inventory are adjusted step by step by 0.2 times the standard deviation, and the deviation between the adjusted inventory and the actual sales volume is recorded in real time. The distribution data upload frequency setting reflects the monitoring accuracy of the transportation process. The 10-minute interval ensures the real-time data and avoids the communication burden. Lettuce delivery generates 12 monitoring points in 2 hours. The 15-point moving average eliminates the abnormal values ​​caused by road condition fluctuations and accurately reflects the trend of product quality changes. The loss threshold of agricultural products varies according to the characteristics of the category. The loss threshold of leafy vegetables with fast dehydration speed is set at 3%. When the threshold is exceeded, the leaves lose water and wilt. The transportation loss threshold of fruits is 4%, mainly considering the appearance damage caused by collision and pressure. The storage stability loss threshold of roots and tubers is 2%, reflecting the weight loss after peeling. Inventory data is updated in real time to show the turnover status of goods. Data is collected every hour to reflect the trend of sales changes. When there is a breakpoint in the sales data, a request for reissue is automatically made. Exceeding the forecast interval by 2 times the standard deviation usually indicates the influence of promotional activities or weather factors, which need to be marked separately. Exponential smoothing forecast reflects the law of sales changes and gives higher weight to recent data. The daily sales benchmark of strawberries is 200 pieces, which increases by 50% during holidays and decreases by 30% in rainy and snowy weather. The forecast results guide replenishment decisions to avoid inventory backlogs or out of stock. The inventory cycle is dynamically adjusted to adapt to changes in market demand. The loss rate in the delivery radius of 5 kilometers is kept within 2%, and the loss rate increases to 4% when it is extended to 10 kilometers. The upper limit of inventory during peak periods is increased by 20% to ensure supply, and the lower limit during the off-season is reduced by 30% to control losses. Turnover speed monitoring reflects the health of inventory. An index below 0.8 indicates a backlog of goods. Exceeding the loss limit for 72 consecutive hours indicates that the inventory scale is too large. Fine-tuning by 0.2 times the standard deviation step size can ensure the correct adjustment direction and avoid drastic fluctuations.The time series autoregression tool processes the loss rate data and establishes the relationship between road conditions and quality changes. When the congestion index is 2.0, the delivery time is extended by 1 time, causing the loss rate of leafy vegetables to rise to 4.5%. At this time, the loss rate is controlled within 3% by diverting through alternative routes. The peak fluctuation coefficient quantitatively reflects the fluctuations in demand. The sales volume of leafy vegetables during the morning market is 2.5 times that of weekdays, and the sales volume of fruits increases by 80%. The consumption of fruits reaches its peak during the weekend leisure time. The inventory structure is adjusted according to the fluctuation law to ensure the accuracy of supply. The adaptive weighted calculation takes into account the impact of seasonal changes. The average daily sales volume of leafy vegetables in summer decreases by 20% compared with spring and autumn, while the sales volume of fruits increases by 40%. The root vegetables remain stable. The reasonable inventory scale is determined by automatic adjustment of weights to improve the turnover efficiency of goods. The inventory warning mechanism realizes accurate replenishment. If the sales volume exceeds the forecast upper limit for 3 consecutive days, the inventory upper limit is raised in advance. If it is lower than the forecast lower limit for 2 days, an emergency procurement order is triggered. According to the deviation between actual sales volume and forecast, the warning threshold setting is continuously optimized.

[0041] Based on the agricultural product quality data, we can obtain agricultural products of different quality grades and formulate differentiated inventory strategies. For high-quality agricultural products, we can increase the safety inventory level. For low-quality agricultural products, we can lower the reorder point and reduce inventory backlogs.

[0042] The appearance integrity score, freshness score and shelf life score of agricultural products are obtained from the quality monitoring database, and the quality score of agricultural products is obtained by weighted calculation based on the three scores; the quality score of agricultural products is graded and judged, if the quality score is higher than the first preset quality score, it is judged as a high-quality product, if the quality score is between the second preset quality score and the first preset quality score, it is judged as a standard product, if the quality score is lower than the second preset quality score, it is judged as an ordinary product, and the first preset quality score is higher than the second preset quality score; according to the agricultural product grade judgment result, the commodity sales data is obtained from the order database, and the sales change trend value is obtained by exponential smoothing operation, and the sales change trend value is calculated by long and short-term memory neural network to obtain The sales volume forecast value within a preset time period in the future; the inventory trigger value is set according to the agricultural product grade determination result, wherein the inventory trigger value of high-quality products is a first preset multiple of the sales volume forecast value, the inventory trigger value of standard products is a second preset multiple of the sales volume forecast value, and the inventory trigger value of ordinary products is a third preset multiple of the sales volume forecast value, the first preset multiple is greater than the second preset multiple, and the second preset multiple is greater than the third preset multiple; the commodity inventory is obtained from the inventory database, if the commodity inventory is lower than the inventory trigger value, the replenishment quantity is increased step by step according to the preset increase multiple of the sales volume forecast value, if the commodity inventory is higher than the inventory trigger value for a preset period of time, the promotion intensity is increased according to the inventory backlog period.

[0043] Specifically, the scores of the three dimensions of appearance integrity, freshness, and shelf life of agricultural products are read from the quality monitoring database every hour, and the weighted scoring tool is used to calculate the weight of appearance 0.4, the weight of freshness 0.4, and the weight of shelf life 0.2. The quality of agricultural products is graded based on the decision tree, and those with scores above 90 are classified as high-quality products, those with scores between 85 and 90 are classified as standard products, and those with scores below 85 are classified as ordinary products. The quality grade database is updated in real time. According to the turnover data of agricultural products of different grades, the sales trend of commodities is calculated by exponential smoothing tools, and the sales and loss data of commodities of various grades are read from the order database in real time. The dynamic inventory upper limit is calculated according to the safety inventory of 4 days for high-quality products, 3 days for standard products, and 2 days for ordinary products, and the inventory lower limit is set in combination with the shelf life of the commodities. According to the quality grade and shelf life data of commodities, the loss speed curve is obtained from the sales database, and the long short-term memory neural network is used to predict the sales trend in the next 7 days. The replenishment quantity is dynamically calculated based on the remaining time of the shelf life and the predicted sales value. The inventory trigger value of high-quality products is 3 times the average daily sales, 2.5 times for standard products, and 1.5 times for ordinary products. The frequency of inventory adjustment is differentiated according to the quality grade. The inventory of goods is read from the inventory database every hour. When the inventory of high-quality products is lower than 4 days of sales or the inventory of standard products is lower than 3 days of sales, the replenishment quantity is increased step by step according to 0.2 times the average daily sales. The inventory of ordinary products is higher than 2 days of sales for 48 hours. The promotion intensity is graded and increased according to the inventory backlog duration. The quality scoring dimension reflects the quality characteristics of agricultural products. Taking strawberries as an example, the appearance integrity evaluates the degree of damage and fading of the fruit, the freshness reflects the hardness and glossiness of the flesh, and the shelf life is based on the picking time and storage conditions. The weight distribution of the three dimensions highlights the importance of appearance and freshness. The decision tree grading standard reflects the difference in commodity value. The price of high-quality products is 30% higher than that of standard products, and the shelf life is extended by 24 hours. It is suitable for sales in high-end supermarkets. The price of ordinary products is reduced by 20%, the shelf life is shortened by 12 hours, and they are sold quickly through community stores. The setting of inventory upper limit follows the principle of differentiation. The 4-day safety inventory of high-quality products ensures stable supply, and the selling price of 15 yuan per 500 grams generates higher profits. The 3-day inventory of standard products balances supply and demand, and the selling price of 11 yuan is suitable for daily consumption. The 2-day inventory of ordinary products speeds up turnover, and the selling price of 8 yuan promotes rapid sales. Exponential smoothing prediction reflects the trend of sales changes. The sales of high-quality products reach a peak of 200 pieces on weekends and stabilize at 150 pieces on weekdays. The daily average of standard products is 250 pieces with small fluctuations. The ordinary products are 120 pieces during the morning market and 80 pieces in other periods. The prediction results guide replenishment decisions. The long short-term memory neural network predicts future sales trends. The input includes historical sales, holidays, and weather factors. The sales of high-quality products increase by 50% on holidays, and decrease by 30% in rainy and snowy weather. Standard products are less affected by seasons, and the demand for ordinary products decreases by 40% in summer. The remaining time of the shelf life affects the inventory adjustment strategy. The inventory trigger value of high-quality products 48 hours before the shelf life is increased to 4 times the daily sales, the standard products increase to 3 times after 36 hours, and the ordinary products decrease to 1 times after 24 hours, realizing dynamic optimization of inventory structure.The replenishment volume increases in a step-by-step manner to control inventory risks. When the inventory of high-quality products is insufficient, the first replenishment is 0.2 times the daily sales volume. If there is no improvement, an additional 0.4 times is added. The first replenishment of standard products is 0.3 times, and an additional 0.5 times is added. Ordinary products are fixed at 0.5 times and replenished once. Inventory backlog triggers differentiated promotions. Ordinary products with backlogs are reduced by 10% in 24 hours and 20% in 48 hours. Standard products are reduced by 15% in 36 hours. High-quality products remain at the original price and are quickly sold out through precision marketing. The loss rate monitoring guides the adjustment of inventory structure. The loss rate in high temperature seasons increases by 50%. The loss rate of high-quality products in 72 hours is 8%, standard products are 10%, and ordinary products are 15%. The replenishment rhythm is dynamically adjusted according to the loss curve. The turnover speed of goods reflects the health of inventory. The turnover days of high-quality products are controlled at 3 days, accounting for 40% of the inventory amount, 2.5 days for standard products account for 45%, and 1.5 days for ordinary products account for 15%, so as to achieve a reasonable configuration of inventory structure.

[0044] The above description is merely a preferred embodiment of one or more embodiments of the present specification and is not intended to limit one or more embodiments of the present specification. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of one or more embodiments of the present specification shall be included in the scope of protection of one or more embodiments of the present specification.

Claims

1. A green agricultural product distribution path optimization and inventory control method, characterized in that: The method comprises: Collect the location coordinate data of the delivery vehicles and the road environment data along the way in real time, and use machine learning algorithms to establish a prediction model for the impact of agricultural product quality based on the historical driving trajectory data and road environment data of the delivery vehicles. By analyzing the real-time collected vehicle location coordinate data and road environment data, predict the impact of different delivery routes on agricultural product quality. According to the prediction results of the impact of agricultural product quality, the delivery route with the least impact of agricultural product quality is selected as the optimal route, and the optimized delivery route is sent to the delivery vehicle terminal in real time; At the delivery vehicle terminal, after receiving the optimal delivery route sent from the cloud, the delivery route is navigated in real time through the on-board navigation system, guiding the driver to drive along the optimal route, and the delivery route is dynamically adjusted according to real-time traffic information; At the same time, the arrival time of agricultural products is dynamically estimated based on the real-time location and driving speed of the delivery vehicles, and the estimated arrival time information is pushed to the destination customers to remind them to prepare for receiving the goods in advance; During the delivery process, the temperature and humidity indicators of the environment in which the agricultural products are located are continuously monitored. If the environmental indicators are detected to exceed the preset threshold range, the early warning mechanism is triggered, and the driver is notified in time by voice or text message to take emergency measures, which include turning on the refrigeration equipment or adjusting the temperature and humidity in the car; After delivery is completed, the vehicle location, road conditions and agricultural product quality data collected during the delivery process are uploaded to the cloud and integrated with historical data for analysis to optimize the prediction model of the impact of agricultural product quality. The inventory strategy is adjusted by analyzing the relationship between delivery data and inventory control data.

2. The method according to claim 1, characterized in that The real-time collection of the location coordinate data of the delivery vehicle and the road environment data along the way, and the use of a machine learning algorithm to establish a prediction model of the degree of impact on the quality of agricultural products based on the historical driving trajectory data and road environment data of the delivery vehicle, and the prediction of the degree of impact of different delivery routes on the quality of agricultural products by analyzing the real-time collected vehicle location coordinate data and road environment data, including: The vehicle driving environment monitoring data is obtained by collecting the road vibration frequency value, the cabin temperature and humidity value, and the vehicle acceleration change value through the vehicle-mounted sensors; According to the vehicle driving environment monitoring data, a linear regression tool is used to calculate the influence coefficient of the road flatness value, the ambient temperature value and the vehicle vibration amplitude value on the product quality change rate; Acquire a real-time position coordinate point sequence from the vehicle terminal, and obtain a quantitative index of the road condition environment through convolutional neural network processing based on the deviation value between the position coordinate point sequence and the planned path of the road network map; Based on the road condition environment quantitative index and the influence coefficient, the moving average tool is used to calculate the product quality score value, and the comprehensive road condition environment score is obtained according to the road flatness weight, congestion weight and ambient temperature weight; Determine whether the road condition and environment comprehensive score is lower than a preset threshold value, and if the road condition and environment comprehensive score is lower than the preset threshold value, trigger a route switching instruction to the vehicle terminal; It also includes: obtaining road surface flatness data and slope data for different sections of the delivery route, as well as real-time monitoring of the vibration frequency and acceleration data of the vehicle during driving by on-board sensors, inputting the real-time monitoring data of road conditions into a preset association model between road conditions and agricultural product quality, predicting the impact of changes in road conditions on the quality of agricultural products, and replanning the delivery route if the impact exceeds a preset threshold.

3. The method according to claim 2, characterized in that The method includes obtaining the road surface flatness data and slope data of different sections of the delivery route and the vibration frequency and acceleration data of the vehicle during driving monitored in real time by the on-board sensor, inputting the real-time monitoring data of the road condition into a preset association model between the road condition and the quality of the agricultural products, predicting the impact of the change in the road condition on the quality of the agricultural products, and re-planning the delivery route if the impact exceeds a preset threshold, including: Obtaining road surface flatness values, slope angle values, and concave-convex amplitude values ​​collected by a road surface detection radar, wherein the road surface detection radar collects data once at a preset interval; Collecting vibration waveform data recorded by a vehicle-mounted triaxial acceleration sensor, performing wavelet noise reduction processing on the vibration waveform data, and calculating a predicted value of the degree of damage to agricultural products based on the noise-reduced data; Acquire sampling data in the vibration waveform data where the vibration frequency exceeds a preset frequency interval value or the vehicle compartment tilt angle exceeds a preset angle interval value, perform moving average smoothing processing on the sampling data, extract acceleration peak points, deceleration valley points, and tilt angle mutation points, and generate a road condition impact score; Calculate the comprehensive road score according to the road surface flatness value, slope angle value, and vibration frequency value according to a preset weight ratio; Determine whether the comprehensive road score is lower than a preset score interval value or whether the vibration frequency exceeds a preset vibration frequency for a preset time period. If so, trigger a path planning instruction according to a preset agricultural product quality loss threshold.

4. The method according to claim 1, characterized in that: The method of selecting the delivery route with the least impact on the quality of agricultural products as the optimal route based on the prediction result of the impact on the quality of agricultural products, and sending the optimized delivery route to the delivery vehicle terminal in real time includes: The road section number and condition data are obtained by using the vehicle terminal, the standard deviation interval is calculated according to the road section condition data, the data exceeding the preset multiple of the standard deviation interval is eliminated, and the road section loss rate value is obtained by the temperature loss weight and the vibration loss weight; Acquire the vehicle position coordinate point according to the Beidou positioning module carried by the vehicle terminal, and extract the road network structure data around the vehicle position coordinate point from the electronic map database; The moving average tool is used to smooth the value of the road section loss rate. If the road section loss rate is lower than a preset threshold and the straight-line distance is less than a preset multiple of the shortest distance, a candidate delivery route is obtained; Read the loss rate value of the road section for the candidate delivery route, calculate the weighted loss rate value according to the loss threshold of leafy vegetables and the loss threshold of fruits, and predict the change trend of the weighted loss rate through a recursive neural network; The candidate delivery routes are sorted according to the predicted loss rate values. If the road section congestion index exceeds a preset threshold and the duration exceeds a preset interval, the route with the second lowest loss rate is selected from the candidate delivery routes as a new delivery route.

5. The method according to claim 1, characterized in that After receiving the optimal delivery route sent from the cloud, the delivery vehicle terminal performs real-time navigation on the delivery route through the vehicle navigation system, guides the driver to drive along the optimal route, and dynamically adjusts the delivery route according to real-time traffic information, including: Obtaining a vehicle position coordinate point sequence through a vehicle-mounted Beidou positioning module, and obtaining smoothed position data using a moving average tool based on the vehicle position coordinate point sequence; Obtaining the road ahead traffic restriction information and lane speed limit from the road network map database, determining abnormal values ​​based on the position data, and eliminating values ​​outside the standard deviation interval to obtain valid position data; Calculate the lateral deviation distance value and the longitudinal deviation distance value using a sliding window tool according to the valid position data and the navigation route coordinate sequence, and trigger a reminder message if the lateral deviation exceeds the first distance threshold when the vehicle speed is lower than the first speed threshold; Using a long short-term memory neural network to predict the vehicle running trajectory according to the lateral deviation distance value and the longitudinal deviation distance value, and generating deviation warning information to the vehicle terminal; Reading real-time traffic data from the vehicle-mounted terminal, if the traffic speed in the real-time traffic data is lower than a second speed threshold and the congestion index exceeds a preset threshold, obtaining parallel road section traffic information from a road network node database to generate an alternative navigation route; It also includes: obtaining data on the spoilage time of different agricultural products, determining the quality and safety threshold of agricultural products based on the freshness requirements of agricultural products, and using a dynamic programming algorithm to calculate the estimated duration of the delivery route based on restrictions on the type of transportation tools. If the estimated duration of the delivery route exceeds the quality and safety threshold of agricultural products, the delivery route is replanned based on the path timeliness evaluation index and the optimization judgment criteria to obtain the path with the best timeliness.

6. The method according to claim 5, characterized in that The method of obtaining spoilage time data of different agricultural products, determining the threshold of agricultural product quality and safety according to the requirements of agricultural product freshness, and using a dynamic programming algorithm to calculate the estimated duration of the delivery route in view of the restrictions on the type of transportation tools, and if the estimated duration of the delivery route exceeds the threshold of agricultural product quality and safety, replanning the delivery route according to the evaluation index of the timeliness of the path and using the optimal evaluation criteria to obtain the optimal path in terms of timeliness includes: Acquire quality monitoring data of leafy vegetables, fruits and tubers from the agricultural product commodity library, and establish a temperature fluctuation and quality loss correlation curve with the quality monitoring data and the temperature value collected by the compartment temperature sensor; Performing standardization processing on the loss rate value according to the quality loss correlation curve, and obtaining the quality safety threshold value through the standardization processing; Reading the road travel time from the road network map database, the road travel time is used to calculate the travel time of each road section, and removing the road section nodes that exceed the quality safety threshold; The moving average tool is used to smooth the traffic data, and the timeliness score of the road section is calculated based on the smoothed traffic data according to the travel time weight, temperature control weight, and traffic score weight; The road sections are sorted according to the timeliness scores and the road section with the highest score is selected. If the traffic speed of the road section is lower than the preset speed threshold or the car temperature exceeds the preset temperature range for a preset time, the road section with the second highest timeliness score is selected from the road network database for updating.

7. The method according to claim 1, characterized in that At the same time, the arrival time of agricultural products is dynamically estimated based on the real-time location and driving speed of the delivery vehicle, and the estimated arrival time information is pushed to the destination customer to remind the customer to prepare for the delivery in advance, including: Receive the position coordinate point information collected by the vehicle-mounted Beidou positioning module, and smooth the position coordinate point information by a sliding average tool to obtain a smooth position sequence; According to the smoothed position sequence, the remaining delivery distance value is obtained from the electronic map database, and the estimated travel time value is calculated using the real-time road speed, traffic light timing and road section speed limit values; The average speed, acceleration and deceleration values ​​in the vehicle's driving trajectory are obtained, and a speed prediction model is obtained by back propagation neural network training. The speed prediction model marks the speed values ​​that exceed the preset multiple interval of the standard deviation as abnormal points; Read the real-time congestion index, accident information and construction road occupation data from the road network database. If the road speed is lower than the specified threshold or the congestion index exceeds the set threshold for a specified period of time, select parallel sections from the road network database to re-plan the delivery route; When the remaining delivery time reaches a set threshold according to the delivery route, an expected arrival reminder message is sent to the recipient customer, and the arrival reminder message is updated at different time intervals according to the remaining delivery time threshold.

8. The method according to claim 1, characterized in that During the delivery process, the temperature and humidity indicators of the environment in which the agricultural products are located are continuously monitored. If the environmental indicators are detected to exceed the preset threshold range, the early warning mechanism is triggered, and the driver is notified in time by voice or text message to take emergency measures. The emergency measures include turning on the refrigeration equipment or adjusting the temperature and humidity in the car, including: Obtain the temperature and humidity values ​​collected by the temperature and humidity sensor, and use a moving average tool to obtain smoothed temperature and humidity data based on the temperature and humidity values; According to the smoothed temperature and humidity data, a temperature and humidity distribution field is constructed by a bilinear interpolation tool to determine the deviation between the value of the monitoring point and the mean of the adjacent points. If the deviation exceeds a preset percentage of the mean of the adjacent points, the monitoring point is marked as an abnormal point. The vehicle terminal is used to monitor the fluctuation value in the temperature and humidity distribution field. If the temperature fluctuation amplitude exceeds the set threshold and lasts for a preset time, a voice reminder and SMS warning information are sent through the vehicle terminal; Read the power value of the refrigeration unit from the vehicle terminal, establish a temperature response curve using a support vector regression tool based on the power value and temperature data, and obtain the temperature change rate under different powers through step-by-step adjustment; The operating power of the refrigeration unit is adjusted according to the temperature change rate. If the temperature drop rate exceeds a preset threshold, the power of the refrigeration unit is reduced until the temperature fluctuation amplitude is lower than the preset threshold and lasts for a preset time.

9. The method according to claim 1, characterized in that: After the delivery is completed, the vehicle location, road conditions and agricultural product quality data collected during the delivery process are uploaded to the cloud and integrated with historical data for analysis to optimize the prediction model of the impact of agricultural product quality. The inventory strategy is adjusted by analyzing the relationship between the delivery data and the inventory control data, including: Receiving the position coordinate sequence, road condition data, compartment temperature and humidity data, and product quality data collected by the vehicle terminal, the data being collected and uploaded by the vehicle terminal at a preset time interval; According to the data collected by the vehicle terminal, the abnormal data is smoothed by the moving average tool, and the time series autoregression tool is used to obtain the correlation curve between road conditions and product loss rate; Reading inventory change data from a commodity inventory database, and obtaining sales forecast values ​​through an exponential smoothing tool based on the product loss rate correlation curve and the commodity turnover speed curve; Divide the inventory cycle according to the sales forecast value, calculate the fluctuation coefficient through the adaptive weighting tool, and generate the upper and lower limits of the inventory in combination with the product loss rate; Obtaining inventory turnover data from the order database, if the inventory turnover rate is lower than a preset threshold or the product loss rate exceeds the preset threshold for a preset period of time, adjusting the upper and lower limits of the inventory by the standard deviation step, and recording the deviation between the inventory and the actual sales volume; It also includes: obtaining agricultural products of different quality grades based on agricultural product quality data, formulating differentiated inventory strategies, increasing safety inventory levels for high-quality agricultural products, and lowering reorder points for low-quality agricultural products to reduce inventory backlogs.

10. The method according to claim 9, characterized in that The method of obtaining agricultural products of different quality grades based on agricultural product quality data and formulating differentiated inventory strategies, improving the safety inventory level for high-quality agricultural products and lowering the reorder point for low-quality agricultural products to reduce inventory backlogs, includes: Obtain the appearance integrity score, freshness score and shelf life score of agricultural products from the quality monitoring database, and obtain the agricultural product quality score through weighted calculation based on the three scores; The agricultural product quality score is graded and judged. If the quality score is higher than a first preset quality score, it is judged as a high-quality product. If the quality score is between a second preset quality score and the first preset quality score, it is judged as a standard product. If the quality score is lower than the second preset quality score, it is judged as an ordinary product. The first preset quality score is higher than the second preset quality score. According to the agricultural product grade determination result, the commodity sales data is obtained from the order database, the sales change trend value is obtained through exponential smoothing operation, and the sales change trend value is calculated using a long short-term memory neural network to obtain the sales forecast value within a preset time period in the future; An inventory trigger value is set according to the agricultural product grade determination result, wherein the inventory trigger value of high-quality products is a first preset multiple of the sales forecast value, the inventory trigger value of standard products is a second preset multiple of the sales forecast value, and the inventory trigger value of ordinary products is a third preset multiple of the sales forecast value, the first preset multiple is greater than the second preset multiple, and the second preset multiple is greater than the third preset multiple; The inventory level of the product is obtained from the inventory database. If the inventory level of the product is lower than the inventory trigger value, the replenishment quantity is increased step by step according to the preset increase multiple of the sales forecast value. If the inventory level of the product is higher than the inventory trigger value for a preset period of time, the promotion intensity is increased according to the inventory backlog duration.

Images