Method for improving ton barrel management precision and transportation efficiency based on tail gas liquid distribution

Abstract

The invention discloses a method for improving ton barrel management precision and transportation efficiency based on tail gas liquid distribution, and relates to the technical field of logistics distribution optimization. Through a system integration technology, a network search technology, an information sensor technology, an RFID identification technology and an Internet of Things technology, ton barrel liquid levels, vehicle GPS positions, traffic road conditions and weather data of all gas stations are collected; and 2, tail gas liquid demand prediction: based on historical consumption data and external factors, adopting an LSTM neural network to predict the tail gas liquid demand of each station in the next three days. According to the method, the liquid level and the road condition are monitored in real time based on the data of the internet of things, the LSTM is utilized to predict the demand, the improved ant colony algorithm is adopted to generate the dynamic path, meanwhile, the capacity distribution of the ton barrels is optimized through the greedy algorithm, after implementation, the logistics cost can be remarkably reduced, denitration is avoided, the resource utilization rate is increased, and the method is suitable for urban multi-node logistics distribution scenes.

Description

Technical Field

[0001] The present invention relates to the technical field of logistics distribution optimization, and specifically to a method for improving the management accuracy and transportation efficiency of ton barrels based on tail gas liquid distribution. Background Art

[0002] Society is experiencing unprecedented rapid development. Along with the acceleration of the urbanization process and the significant improvement of residents' living standards, the vehicle ownership has shown a blowout growth trend. All kinds of motor vehicles have become an indispensable part of people's daily lives, greatly facilitating people's travel and logistics transportation. Behind this phenomenon, there is also an environmental pressure that cannot be ignored - the problem of vehicle exhaust emissions has become increasingly prominent and has become one of the important factors affecting air quality.

[0003] In this context, the demand for tail gas treatment liquid has also increased sharply. This kind of special liquid can effectively reduce the nitrogen oxide emissions in vehicle exhaust gas and is a key link in implementing more stringent environmental protection standards and promoting the development of green transportation. However, there are numerous gas stations in a city, and each station needs to receive tail gas liquid rationing regularly. However, the geographical locations of each station are scattered and the demand varies greatly. Traditional fixed-route distribution is prone to low efficiency and high vehicle empty running rate. Moreover, due to demand fluctuations or distribution delays, some gas stations may run out of tail gas liquid inventory, affecting normal operations. The management of ton barrel equipment for receiving and storing tail gas liquid is extensive, often resulting in redundant or insufficient capacity, leading to frequent distribution or resource waste.

[0004] Existing distribution schemes are mostly based on fixed cycles and routes and cannot respond in real time to dynamic factors such as traffic congestion and demand changes. Relying on experience to allocate the distribution volume, there is a lack of a data-driven ton barrel capacity optimization strategy. Route planning and ton barrel management are separated, and a closed-loop system with the optimal overall cost has not been formed.

[0005] Based on this, a method for improving the management accuracy and transportation efficiency of ton barrels based on tail gas liquid distribution is now provided, which can eliminate the drawbacks of existing tail gas liquid distribution methods. Summary of the Invention

[0006] The purpose of the present invention is to provide a method for improving the management accuracy and transportation efficiency of ton barrels based on tail gas liquid distribution to solve the problems of the modern product in the background art.

[0007] To achieve the above purpose, the present invention provides the following technical solutions:

[0008] A method for improving the management accuracy and transportation efficiency of ton barrels based on tail gas liquid distribution, comprising the following steps:

[0009] Step 1: Multi-channel data collection:

[0010] Using the self-developed automated logistics information collection program, through system integration technology, network search technology, information sensor technology, RFID identification technology, and Internet of Things technology, collect the liquid level of tons of barrels at each gas station, vehicle GPS location, traffic conditions, and weather data;

[0011] Step 2: Prediction of tail gas liquid demand:

[0012] Based on historical consumption data and external factors, use the LSTM neural network to predict the tail gas liquid demand of each station in the next 3 days;

[0013] Step 3: Dynamic route planning for tail gas:

[0014] Apply the improved ant colony algorithm, combine real-time traffic conditions and vehicle load constraints, introduce a time window penalty function, generate a dynamic delivery route, and plan the global optimal delivery route and vehicle and station allocation plan;

[0015] Step 4: Intelligent scheduling of tons of barrels:

[0016] According to the predicted demand and the current liquid level, dynamically adjust the delivery volume and optimize the capacity allocation of tons of barrels, establish a capacity optimization strategy, use the greedy algorithm to dynamically adjust the allocation of tons of barrel resources among multiple stations, and when the liquid level is lower than the safety threshold, trigger an emergency replenishment task and insert it into the dynamic route;

[0017] Step 5: Collaborative optimization, according to the cost target measurement function: α·transportation cost + β·out-of-stock penalty cost + γ·idle cost of tons of barrels, dynamically optimize the weight coefficients α, β, and γ through the particle swarm algorithm.

[0018] Based on the above technical solutions, the present invention also provides the following optional technical solutions:

[0019] In an optional solution: The tail gas demand prediction in Step 2 includes the following steps:

[0020] Step S1.1: Data collection and preprocessing:

[0021] Collect gas station operation data, external environment data, policy and event data, fill in missing values using time series interpolation or data from neighboring stations, and identify outliers based on the isolation forest algorithm and mark them in combination with manual review;

[0022] Step S1.2: Feature engineering and model construction:

[0023] Extract time series features, spatial features, encode events, use the LSTM-ARIMA hybrid model, generate a prediction interval based on Monte Carlo Dropout, and achieve collaborative prediction of station demand through a multi-task learning framework;

[0024] Step S1.3: Prediction of tail gas liquid demand: Output the predicted values for the next 3 days through the LSTM-ARIMA hybrid model.

[0025] In an alternative solution: The dynamic route planning of the tail gas in step three includes the following steps:

[0026] Step S2.1: Generate a gas station demand form:

[0027] Combined with the tail gas liquid distribution demand, generate a distribution order including the stations with demand, geographical locations, distribution times, and tail gas liquid distribution volumes.

[0028] Step S2.2: Collect traffic condition data:

[0029] Obtain the real-time road congestion index, traffic incidents, and estimated travel time through the map API.

[0030] Step S2.3: Obtain vehicle status information:

[0031] Use the on-vehicle GPS to obtain the real-time position, speed, and current load of the vehicle, and clarify the maximum load capacity of the vehicle.

[0032] Step S2.4: Obtain external environment data:

[0033] Collect event information such as weather, temperature, precipitation probability, traffic restrictions, and environmental protection inspections.

[0034] Step S2.5: Divide the gas station demand priorities:

[0035] According to the liquid level safety threshold, mark the liquid level < 20% as an urgent task, and the liquid level < 50% as a high-priority task. The urgent task is forcibly inserted into the current distribution plan, and the high-priority task optimizes the subsequent path.

[0036] Step S2.6: Real-time dynamic information update and multi-objective update function:

[0037] When traffic congestion or an urgent task is detected, locally update the pheromone concentration of the affected path, and optimize the path through the multi-objective update function Minimize = ω1 * Ttotal + ω2 * Cfuel + ω3 * Pdelay, where Ttotal is the total driving time, Cfuel is the fuel cost, Pdelay is the task delay penalty, and the weights ω1, ω2, ω3 are calibrated through historical data regression.

[0038] Step S2.7: Transportation route selection algorithm:

[0039] Use the improved ant colony algorithm to calculate the theoretical shortest path, and the ants follow the probability formula:

[0040]

[0041] Select the next node j, where τij is the pheromone concentration of the edge (i, j).

[0042] ηij = 1 / (Tij·Qreq,j), where Tij is the travel time and Qreq,j is the demand of the station, with α = 1 and β = 2.

[0043] In an alternative solution: The intelligent scheduling of the ton barrels in step four includes the following steps:

[0044] Step S3.1: Real-time liquid level monitoring:

[0045] Collect the quantity of ton barrels and the current liquid level data at each gas station through the Internet of Things liquid level sensor at a sampling frequency of once per hour.

[0046] Step S3.2: Ton barrel demand prediction:

[0047] Obtain the demand for tail gas liquid in the next 72 hours output by the intelligent calculation module.

[0048] Step S3.3: Define the constraints:

[0049] Consider the vehicle load limit, the vehicle distribution capacity in the area, and the rules for cross-station ton barrel transfer.

[0050] Step S3.4: Generate real-time scheduling instructions:

[0051] Generate a daily delivery plan according to the predicted demand, allocate it to the vehicle and bind the dynamic route planning result. When the available capacity of the ton barrel < the demand for tail gas liquid at the gas station, trigger the cross-station ton barrel transfer instruction.

[0052] Step S3.5: Establish a ton barrel sharing mechanism:

[0053] For adjacent station groups with frequent transfers, dynamically merge them into a "virtual shared ton barrel pool", allowing capacity sharing, and by default, allowing borrowing 20% of the capacity of adjacent stations.

[0054] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0055] 1. The present invention monitors the liquid level and road conditions in real time based on Internet of Things data, uses LSTM to predict demand and adopts an improved ant colony algorithm to generate dynamic paths. At the same time, it optimizes the ton barrel capacity allocation through the greedy algorithm. After implementation, it can significantly reduce logistics costs, avoid out-of-stock situations and improve resource utilization rate, and is applicable to the urban multi-node logistics distribution scenario.

[0056] 2. The present invention effectively reduces the waste of ton barrel resources, improves the dynamic utilization efficiency of resources, and can more fully meet the storage requirements of tail gas liquid in gas stations under the same number of ton barrels, reducing the problems of frequent distribution or idleness caused by unreasonable ton barrel capacity.

[0057] 3. By reasonably planning the distribution route and allocating the distribution volume, the present invention enables the vehicle to make more full use of its load capacity during transportation, reduces the situation of the vehicle running empty or with insufficient load, and improves the transportation efficiency of the vehicle.

[0058] 4. By optimizing the route planning, the present invention reduces the empty driving mileage of the vehicle and fuel consumption; reasonably schedules the ton barrels, reduces unnecessary distribution times, and reduces the labor cost and vehicle depreciation cost, thereby effectively reducing the logistics cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0060] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0061] In one embodiment, as Figure 1 shown, a method for improving the management accuracy and transportation efficiency of ton barrels based on tail gas liquid distribution includes the following steps:

[0062] Step 1: Multi-channel data collection:

[0063] Using a self-developed automated logistics information collection program, through system integration technology, network search technology, information sensor technology, RFID identification technology, and Internet of Things technology, collect the ton barrel liquid level, vehicle GPS position, traffic conditions, and weather data of each gas station;

[0064] Step 2: Prediction of tail gas liquid demand:

[0065] Based on historical consumption data and external factors, use the LSTM neural network to predict the tail gas liquid demand of each station in the next 3 days;

[0066] Step 3: Dynamic route planning for tail gas:

[0067] Apply an improved ant colony algorithm, combine real-time traffic conditions and vehicle load constraints, introduce a time window penalty function, generate a dynamic distribution path, and plan the global optimal distribution route and vehicle and station allocation plan;

[0068] Step 4: Intelligent scheduling of ton barrels:

[0069] Dynamically adjust the distribution volume according to the predicted demand and the current liquid level, optimize the capacity allocation of ton barrels, establish a capacity optimization strategy, and use the greedy algorithm to dynamically adjust the allocation of ton barrel resources among multiple sites. When the liquid level is lower than the safety threshold, trigger an emergency replenishment task and insert a dynamic route;

[0070] Step Five: Collaborative Optimization. According to the cost objective measurement function: α·transportation cost + β·stockout penalty cost + γ·ton barrel idle cost, dynamically optimize the weight coefficients α, β, and γ through the particle swarm optimization algorithm.

[0071] In one embodiment, the tail gas demand prediction in the second step includes the following steps:

[0072] Step S1.1: Data collection and preprocessing:

[0073] Collect gas station operation data, external environment data, policy and event data. For missing values, use time series interpolation or data from neighboring sites to fill them. For outliers, identify them based on the isolation forest algorithm and mark them in combination with manual review;

[0074] Step S1.2: Feature engineering and model construction:

[0075] Extract time series features, spatial features, encode events, use the LSTM-ARIMA hybrid model, generate a prediction interval based on Monte Carlo Dropout, and achieve collaborative prediction of site demands through a multi-task learning framework;

[0076] Step S1.3: Tail gas liquid demand prediction: Output the predicted values for the next 3 days through the LSTM-ARIMA hybrid model.

[0077] In one embodiment, the tail gas dynamic route planning in the third step includes the following steps:

[0078] Step S2.1: Generate a gas station demand form:

[0079] Combine the tail gas liquid distribution demand to generate a distribution order including the sites with demand, geographical locations, distribution times, and tail gas liquid distribution volumes;

[0080] Step S2.2: Collect traffic condition data:

[0081] Obtain the real-time road congestion index, traffic events, and estimated travel time through the map API;

[0082] Step S2.3: Obtain vehicle status information:

[0083] Use the in-vehicle GPS to obtain the real-time position, speed, current load of the vehicle, and clarify the maximum load capacity of the vehicle;

[0084] Step S2.4: Obtain external environment data:

[0085] Collect event information such as weather, temperature, precipitation probability, traffic restrictions, and environmental protection inspections;

[0086] Step S2.5: Divide the priority of gas station demand:

[0087] According to the liquid level safety threshold, mark the liquid level less than 20% as an urgent task, and mark the liquid level less than 50% as a high-priority task. The urgent task is forced to be inserted into the current distribution plan, and the high-priority task optimizes the subsequent path;

[0088] Step S2.6: Real-time dynamic information update and multi-objective update function:

[0089] When traffic congestion or an urgent task is detected, locally update the pheromone concentration of the affected path, and optimize the path through the multi-objective update function Minimize = ω1*Ttotal + ω2*Cfuel + ω3*Pdelay, where Ttotal is the total driving time, Cfuel is the fuel cost, Pdelay is the task delay penalty, and the weights ω1, ω2, ω3 are calibrated through historical data regression;

[0090] Step S2.7: Transportation path selection algorithm:

[0091] Use the improved ant colony algorithm to calculate the theoretical shortest path, and the ants follow the probability formula:

[0092]

[0093] Select the next node j, where τij is the pheromone concentration of the edge (i,j),

[0094] ηij = 1 / (Tij·Qreq,j), Tij is the passing time, Qreq,j is the demand of the station, α = 1, β = 2.

[0095] In one embodiment, the intelligent scheduling of the ton barrels in step four includes the following steps:

[0096] Step S3.1: Real-time liquid level monitoring:

[0097] Collect the number of ton barrels and the current liquid level data of each gas station through the Internet of Things liquid level sensor at a sampling frequency of once per hour;

[0098] Step S3.2: Ton barrel demand prediction:

[0099] Obtain the tail gas liquid demand in the next 72 hours output by the intelligent calculation module;

[0100] Step S3.3: Define the constraints:

[0101] Consider vehicle load limits, regional vehicle distribution capabilities, and cross-site ton barrel transfer rules;

[0102] Step S3.4: Generate real-time scheduling instructions:

[0103] Generate a daily delivery plan based on the predicted demand, allocate it to the vehicle, and bind the dynamic route planning result. When the remaining capacity of the ton barrel is less than the demand for tail gas liquid at the gas station, trigger a cross-site ton barrel transfer instruction;

[0104] Step S3.5: Establish a ton barrel sharing mechanism:

[0105] For adjacent site groups with frequent transfers, dynamically merge them into a "virtual shared ton barrel pool", allowing capacity sharing. By default, 20% of the capacity of adjacent sites can be borrowed.

[0106] The above embodiments disclose a method for improving the management accuracy and transportation efficiency of ton barrels based on tail gas liquid distribution. The specific process is as follows:.

[0107] 1. Multi-channel data collection

[0108] Self-develop an automated logistics information collection program. Through system integration technology, network search technology, information sensor technology, RFID identification technology, and Internet of Things technology, this program realizes the automated collection of freight data in various aspects and full scenarios, including but not limited to the ton barrel liquid level of each gas station, vehicle GPS position, traffic conditions, and weather data. etc.

[0109] 2. Large model for tail gas liquid demand prediction

[0110] Based on historical consumption data and external factors (holidays, weather), use the LSTM neural network to predict the tail gas liquid demand of each site in the next 3 days.

[0111] Based on the intelligent calculation method of multi-source data fusion and machine learning, through dynamic modeling and real-time feedback, achieve high-precision and strong robustness of demand prediction.

[0112] (1) Data collection and preprocessing:

[0113] S1: Gas station operation data: historical tail gas liquid consumption, ton barrel liquid level (Internet of Things sensor), frequency of refueling vehicles.

[0114] S2: External environment data: weather, natural events, holiday calendar, traffic congestion index, regional economic activities.

[0115] S3: Policy and event data: information such as environmental protection regulation adjustments, traffic restriction policies, and large-scale events.

[0116] (2) Data cleaning:

[0117] S1: Missing value filling: Use time series interpolation or data from neighboring stations for filling.

[0118] S2: Outlier handling: Identify abnormal consumption records based on the Isolation Forest algorithm and combine with manual review for marking.

[0119] (3) Feature engineering and model construction:

[0120] S1: Temporal features: Lag variables (consumption in the previous 7 days), rolling statistics (mean and variance in the past 30 days).

[0121] S2: Spatial features: Economic level of the area where the station is located, density of the surrounding road network, correlation of demand among adjacent stations.

[0122] S3: Event encoding: Convert policy adjustments and holidays into one-hot encoding (0 / 1 flags).

[0123] S4: LSTM-ARIMA hybrid model: LSTM neural network captures non-linear relationships, a hybrid prediction architecture that combines LSTM neural network and ARIMA model; a method for generating prediction intervals based on Monte Carlo Dropout; a collaborative prediction mechanism for station demand under a multi-task learning framework.

[0124] (4) Tail gas liquid demand prediction: Output predicted values for the next 3 days through the LSTM-ARIMA hybrid model

[0125] 3. Tail gas dynamic route planning module

[0126] Adopt an improved ant colony algorithm, combine real-time traffic conditions and vehicle load constraints to generate dynamic distribution routes. The dynamic programming algorithm considers key factors such as vehicle location, gas station demand, traffic congestion index, and maximum vehicle load. Introduce a time window penalty function to prioritize the distribution of stations approaching out-of-stock. At the same time, dynamically update the route weights in combination with real-time traffic data to avoid congested sections. Calculate the global optimal distribution route and vehicle-station allocation plan through the planning algorithm to ensure distribution timeliness and reduce empty driving mileage. While meeting the tail gas liquid demand of gas stations, reduce the comprehensive transportation cost.

[0127] (1) Generation of gas station demand list:

[0128] Combine the tail gas liquid distribution demand to automatically generate distribution orders. The order information includes the stations with demand for goods, geographical locations, distribution times, tail gas liquid distribution volumes, etc.

[0129] (2) Collection of traffic condition data:

[0130] Obtain real-time road congestion index, traffic events, and estimated travel time through the map API.

[0131] (3)Idle vehicle status information:

[0132] On-vehicle GPS provides real-time vehicle location, speed, current load (monitored by weight sensors), and the vehicle's maximum load capacity.

[0133] (4)External environment data:

[0134] Event information such as weather, temperature, precipitation probability, traffic restrictions, and environmental protection inspections.

[0135] (5)Priority division of gas station demand

[0136] Based on the liquid level safety threshold, a liquid level < 20% is marked as an urgent task, and a liquid level < 50% is marked as a high-priority task to generate dynamic priority tags. Urgent tasks are forced to be inserted into the current delivery plan, and high-priority tasks optimize the subsequent path.

[0137] (6)Real-time dynamic information update, multi-objective update function

[0138] When traffic congestion (travel time increases by more than 30%) or an urgent task is detected, immediately locally update the pheromone concentration of the affected path.

[0139] Minimize = ω1 * Ttotal + ω2 * Cfuel + ω3 * Pdelay

[0140] Ttotal: Total travel time (including congestion delay).

[0141] Cfuel: Fuel cost (positively correlated with mileage and load).

[0142] Pdelay: Task delay penalty (Pdelay exponentially increases in case of urgent task timeout).

[0143] Weights ω1, ω2, ω3: Calibrated by historical data regression.

[0144] (7)Transportation path selection algorithm:

[0145] Use the improved ant colony algorithm (ACO) to calculate the theoretical shortest path. Starting from the center of the tail gas liquid plant, with a weight of 0 and the path pheromone initialized to a uniform distribution. Ants select the next node j with probability Pij Select the next node j:

[0146]

[0147] τij: Pheromone concentration of edge (i, j).

[0148] ηij: Heuristic factor, taking 1 / (Tij·Qreq,j), where Tij is the travel time and Qreq,j is the demand at the station.

[0149] α = 1, β = 2 (empirical parameters)

[0150] 3. Intelligent scheduling method for ton barrels

[0151] Dynamically adjust the delivery volume and optimize the capacity allocation of ton barrels according to the predicted demand and current liquid level. Establish a capacity optimization strategy, calculate the optimal delivery volume for each station according to the predicted demand and liquid level threshold (safety stock), and avoid over - delivery or under - delivery. Use the greedy algorithm to dynamically adjust the allocation of ton barrel resources among multiple stations and balance the capacity surplus and deficit between regions. When the liquid level is lower than the safety threshold, trigger an emergency replenishment task and insert it into the dynamic route.

[0152] (1) Real - time liquid level monitoring: Real - time collect the quantity of ton barrels and current liquid level data (unit: 10 ml) at each gas station through the Internet of Things liquid level sensor, and the sampling frequency is once per hour.

[0153] (2) Ton barrel demand prediction: The future 72 - hour demand for tail gas liquid output by the intelligent calculation module

[0154] (3) Other constraint conditions: Vehicle load limit, the load of large transport vehicles is 30 tons, and the load of medium - sized transport vehicles is 10 tons. Regional vehicle delivery capacity, the maximum number of vehicles that can be scheduled daily in the region. Rules for cross - station ton barrel transfer, only allowing transfer between adjacent regional stations.

[0155] (4) Real - time scheduling instruction generation: Generate a daily delivery plan according to the predicted demand, allocate it to vehicles and bind the dynamic route planning result. When the remaining capacity of the ton barrel < the demand for tail gas liquid at the gas station, trigger a cross - station ton barrel transfer instruction.

[0156] (5) Ton barrel sharing mechanism: Dynamically merge frequently transferred adjacent station groups into a "virtual shared ton barrel pool", allowing capacity sharing, and by default, allowing borrowing 20% of the capacity of adjacent stations.

[0157] 4. Collaborative optimization mechanism

[0158] Cost target measurement function: α·transportation cost+β·stock - out penalty cost+γ·ton barrel idle cost. Among them, the weight coefficients α, β, and γ are dynamically optimized through the particle swarm algorithm.

[0159] 5. Effect measurement

[0160] (1) Measurement of ton barrel utilization rate

[0161]

[0162] Parameter description:

[0163] N: Total number of gas stations for delivery

[0164] V i (t) : The actual liquid capacity of the i-th gas station's ton barrel at time t.

[0165] C i : The total capacity of the i-th gas station's ton barrel.

[0166] T: The statistical period (e.g., 1 month).

[0167] Improvement effect:

[0168] By integrating to calculate the average utilization rate of each ton barrel during the period, and finally taking the average value of all stations to reflect the dynamic resource utilization efficiency. After implementation, the utilization rate of ton barrels has increased from 60% to 85%.

[0169] (2) Measurement of vehicle utilization rate

[0170]

[0171] Parameter description:

[0172] M: The total number of distribution vehicles

[0173] Q k : The actual load of the k-th vehicle.

[0174] Q max ,k: The maximum load of the k-th vehicle.

[0175] Improvement effect:

[0176] After actual measurement, the load utilization rate of vehicles has increased by 13%.

[0177] (3) Calculation formula for reducing logistics costs

[0178] Unit cost = Total transportation cost (fuel cost + labor cost + vehicle depreciation) / Total distribution volume (tons).

[0179] Improvement effect:

[0180] The cost per trip has decreased by 3% - 5%.

[0181] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. A method for improving the management accuracy and transportation efficiency of ton barrels based on tail gas liquid distribution, characterized in that: The following steps are involved: Step 1: Multi-channel data collection: Using the self-developed automated logistics information collection program, through system integration technology, network search technology, information sensor technology, RFID identification technology and Internet of Things technology, we collect the liquid level of each gas station's barrel, vehicle GPS location, traffic conditions and weather data; Step 2: Tail gas liquid demand forecast: Based on historical consumption data and external factors, the LSTM neural network is used to predict the tail gas liquid demand of each station in the next three days; Step 3: Dynamic exhaust route planning: Using the improved ant colony algorithm, combined with real-time traffic conditions and vehicle load constraints, the time window penalty function is introduced to generate dynamic delivery paths, plan the global optimal delivery route and vehicle and station allocation plan; Step 4: Intelligent scheduling of ton barrels: According to the predicted demand and current liquid level, the distribution volume is dynamically adjusted and the allocation of barrel capacity is optimized. A capacity optimization strategy is established, and a greedy algorithm is used to dynamically adjust the allocation of barrel resources between multiple sites. When the liquid level is lower than the safety threshold, an emergency replenishment task is triggered and a dynamic route is inserted. Step 5: Collaborative optimization, based on the cost target calculation function: α·transportation cost + β·out-of-stock penalty cost + γ·ton barrel idle cost, dynamically optimize the weight coefficients α, β, and γ through the particle swarm algorithm.

2. The method for improving the management accuracy and transportation efficiency of ton barrels based on tail gas liquid distribution according to claim 1 is characterized in that: The exhaust gas demand prediction in step 2 includes the following steps: Step S1.1: Data acquisition and preprocessing: Collect gas station operation data, external environment data, policy and event data, use time series interpolation or neighboring station data to fill missing values, identify outliers based on the isolation forest algorithm and combine manual review and marking; Step S1.2: Feature engineering and model building: Extract temporal and spatial features, encode events, use LSTM-ARIMA hybrid model, generate prediction intervals based on Monte Carlo Dropout, and achieve collaborative prediction of site demand through a multi-task learning framework; Step S1.3: Tail gas liquid demand forecast: Output the forecast value for the next 3 days through the LSTM-ARIMA hybrid model.

3. The method for improving the management accuracy and transportation efficiency of ton barrels based on tail gas liquid distribution according to claim 1 is characterized in that: The tail gas dynamic route planning in step 3 includes the following steps: Step S2.1: Generate a gas station demand order: Combined with the tail gas liquid distribution demand, a distribution order is generated that includes the required goods site, geographical location, distribution time, and tail gas liquid distribution quantity; Step S2.2: Collecting traffic data: Obtain real-time road congestion index, traffic events and estimated travel time through map API; Step S2.3: Get vehicle status information: Use the vehicle's onboard GPS to obtain the vehicle's real-time location, speed, and current load capacity, and determine the vehicle's maximum load capacity; Step S2.4: Obtain external environment data: Collect information on weather, temperature, precipitation probability, traffic restrictions, environmental inspections, and other events; Step S2.5: Prioritize gas station requirements: According to the liquid level safety threshold, liquid levels less than 20% are marked as urgent tasks, and liquid levels less than 50% are marked as high-priority tasks. Urgent tasks are forcibly inserted into the current delivery plan, and high-priority tasks optimize subsequent paths; Step S2.6: Real-time dynamic information update and multi-objective update function: When traffic congestion or an emergency task is detected, the pheromone concentration of the affected path is locally updated, and the path is optimized through the multi-objective update function Minimize = ω1*Ttotal+ω2*Cfuel+ω3*Pdelay, where Ttotal is the total driving time, Cfuel is the fuel cost, Pdelay is the task delay penalty, and the weights ω1, ω2, and ω3 are calibrated through historical data regression; Step S2.7: Transportation path selection algorithm: The improved ant colony algorithm is used to calculate the theoretical shortest path. Ants follow the probability formula: Select the next node j, where τij is the pheromone concentration of edge (i, j), ηij=1 / (Tij·Qreq,j), Tij is the travel time, Qreq,j is the demand of the station, α=1, β=2.

4. The method for improving the management accuracy and transportation efficiency of ton barrels based on tail gas liquid distribution according to claim 1 is characterized in that: The intelligent scheduling of ton barrels in step 4 includes the following steps: Step S3.1: Real-time liquid level monitoring: The IoT liquid level sensor collects the number of barrels and current liquid level data at each gas station at a sampling frequency of once per hour; Step S3.2: Ton barrel demand forecast: Obtain the exhaust liquid demand for the next 72 hours output by the intelligent measurement module; Step S3.3: Specify the constraints: Consider vehicle load restrictions, regional vehicle delivery capabilities, and cross-site tonnage transfer rules; Step S3.4: Generate real-time scheduling instructions: Generate a daily delivery plan based on forecast demand, assign it to vehicles and bind it to dynamic route planning results. When the free capacity of a ton barrel is less than the exhaust liquid volume required by the gas station, trigger a cross-site ton barrel transfer instruction; Step S3.5: Establish a bucket sharing mechanism: Frequently transferred adjacent site groups are dynamically merged into a "virtual shared bucket pool" to allow capacity sharing. By default, 20% of the capacity of adjacent sites can be borrowed.

Images