A crude oil long-distance pipeline-oriented intelligent energy-saving regulation and control and carbon footprint optimization method

Abstract

This invention discloses an intelligent energy-saving control and carbon footprint optimization method for long-distance crude oil pipelines, aiming to solve the problem that traditional control strategies are difficult to coordinately reduce energy consumption and carbon emissions. The method includes: constructing a dynamic model coupling pressure-flow-temperature three fields; integrating distributed sensor data and grid carbon and electricity information to form a multi-dimensional input; establishing a weighted joint function with comprehensive energy consumption per unit transport volume and total carbon footprint as dual objectives, and dynamically adjusting the weights based on grid carbon intensity and electricity price; employing an improved sequential quadratic programming algorithm for rolling time-domain optimization to control pump station motor speed, furnace status, and valve opening in real time; and introducing a closed-loop feedback correction mechanism to identify and update key model parameters online. Through the above technical solutions, this application achieves coordinated optimization of energy consumption and carbon emissions, improves robustness under complex operating conditions, enhances real-time adaptive capabilities, and supports refined energy efficiency management and carbon footprint accounting.

Description

Technical Field

[0001] This invention belongs to the field of oil and gas storage and transportation technology. Specifically, it relates to an intelligent energy-saving control and carbon footprint optimization method for long-distance crude oil pipelines. Background Technology

[0002] As a vital part of the nation's energy supply, the energy efficiency and carbon emission levels of long-distance crude oil pipeline systems are receiving increasing attention. Long-distance crude oil pipelines are characterized by long transport distances, numerous pumping stations, and complex operating conditions. Traditional control strategies often rely on experience-based settings or static optimization models, making it difficult to simultaneously achieve the dual objectives of minimizing energy consumption and controlling carbon footprint. Especially in real-world operating scenarios with frequent load fluctuations, significant changes in ambient temperature, or large differences in oil properties, the system needs to dynamically coordinate multi-dimensional control variables such as power allocation at each pumping station, furnace start-up and shutdown, and valve openings while ensuring safe transport. This places higher demands on the real-time performance, adaptability, and low-carbon orientation of the control algorithms.

[0003] However, existing energy-saving control methods typically focus solely on electricity or fuel consumption as a single optimization objective, neglecting the spatiotemporal heterogeneity of carbon emission factors under different energy structures. This leads to a disconnect between energy-saving effects and carbon reduction outcomes. Furthermore, the fluid dynamics and thermodynamic processes of pipeline systems are highly coupled, and traditional models lack the ability to accurately model the multi-field co-evolution mechanism of pressure, flow, and temperature, making it difficult to support refined energy efficiency management. In addition, current control strategies often employ offline optimization or periodic rescheduling mechanisms, which cannot effectively respond to fluctuations in grid carbon intensity, peak and off-peak electricity prices, or sudden operational disturbances, resulting in significant redundant energy consumption and hidden carbon emissions during dynamic system operation. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent energy-saving control and carbon footprint optimization method for long-distance crude oil pipelines. It mainly addresses the problem that existing energy-saving control methods usually only take electricity consumption or fuel consumption as a single optimization target, ignoring the spatiotemporal heterogeneity of carbon emission factors under different energy structures, which leads to a disconnect between energy-saving effects and carbon reduction results.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A method for intelligent energy-saving control and carbon footprint optimization for long-distance crude oil pipelines includes the following steps:

[0007] S1, based on the fluid dynamics and thermodynamics of long-distance crude oil pipelines, introduces the nonlinear relationship between oil density and oil viscosity with temperature, and constructs a dynamic mathematical model of pressure-flow-temperature three-field coupling by combining the spatial distribution function of ambient temperature along the pipeline.

[0008] S2, acquires multi-source operating data and external carbon electricity information to form a multi-dimensional input vector containing system state variables and external disturbance factors;

[0009] S3 generates a weighted joint objective function with the dual objectives of minimizing the comprehensive energy consumption per unit of transport and minimizing the carbon footprint across the entire supply chain;

[0010] S4, perform rolling time-domain optimization solution, under the premise of meeting the minimum delivery pressure, maximum allowable temperature drop and equipment operation constraints, perform rolling optimization on the motor speed of each pump station, the start and stop status of the heating furnace and the opening of the regulating valve in the next scheduling cycle;

[0011] S5. Implement a closed-loop feedback correction mechanism, analyze the deviation between the actual operating data and the model prediction, update the thermal conductivity coefficient and friction resistance coefficient using the recursive least squares method, and restart the optimization solution process.

[0012] Furthermore, in step S1, the fluid dynamics characteristics are described by a momentum equation that considers pressure gradient, gravity, and frictional resistance; the thermodynamic characteristics are described by an energy equation that considers changes in internal energy, pressure work, frictional heat generation, and heat exchange with the environment.

[0013] Furthermore, in step S1, the nonlinear relationship between the oil density and temperature is derived from the API gravity, and the expression is: ,in Reference temperature The density below, The coefficient of volume expansion;

[0014] The nonlinear relationship between the viscosity of the oil and temperature is accurately described using the Andrade empirical formula:

[0015]

[0016] Wherein, A and B are constants calibrated based on the API gravity of crude oil, which is determined by laboratory multi-point viscosity measurements and least squares fitting.

[0017] Furthermore, the multi-source operational data is acquired through a distributed sensor network, which includes a pressure transmitter, a PT100 temperature sensor, and a Coriolis mass flow meter; the external carbon electricity information includes the real-time carbon emission intensity of the power grid and time-of-use electricity prices; all raw data undergoes sliding window median filtering at edge nodes and is normalized to zero mean, forming a multidimensional input vector of dimension M.

[0018] Furthermore, the specific process of step S3 is as follows:

[0019] S31, Define the normalized energy consumption index E:

[0020]

[0021] in, Let j be the real-time power of the motor in the j-th pump station. The thermal power of the l-th heating furnace;

[0022] Total transport volume;

[0023] S32, defines the normalized carbon emission index C:

[0024]

[0025] in, For the power consumption of pump station j, The instantaneous flow rate of natural gas to the heating furnace. The carbon emission coefficient for natural gas;

[0026] S33, Construct the weighted joint objective function:

[0027]

[0028] in Weight The electricity price will be dynamically adjusted based on the real-time carbon emission intensity of the power grid and the time-of-use electricity price.

[0029] Furthermore, in step S4, the optimization solution adopts an improved sequential quadratic programming algorithm, which decomposes the problem into J+1 subproblems using a pipeline chain topology structure, and performs parallel calculations through an ADMM parallel solver; the output optimal control sequence is converted into analog or switching signals by the instruction issuing and execution module.

[0030] Furthermore, in step S4, the optimization solution employs a proximal policy optimization (PPO) algorithm based on deep reinforcement learning; state space With multidimensional input vectors Consistency; Action Space The variables are a continuous-discrete mixture, including motor speed, furnace start / stop status, and control valve opening; reward function. ,in To constrain violations of indicator functions.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] (1) The invention constructs a dynamic mathematical model that couples pressure, flow rate, and temperature, and introduces the nonlinear relationship between oil density and viscosity with temperature and the spatial distribution function of ambient temperature. This accurately characterizes the fluid dynamics and thermodynamic coupling characteristics of long-distance crude oil pipelines, solving the problem of insufficient modeling of multi-field co-evolution mechanisms in traditional models. Combined with the multi-dimensional input of distributed sensor data and power grid carbon electricity information, it provides a foundation for accurate perception and prediction of pipeline operating status, supports refined energy efficiency management, and effectively improves the adaptability to complex operating conditions (such as load fluctuations, ambient temperature changes, and differences in oil properties).

[0033] (2) This invention establishes a weighted joint function with the dual objectives of comprehensive energy consumption per unit transport volume and carbon footprint across the entire pipeline. The weights are dynamically adjusted based on the real-time carbon intensity of the power grid and electricity prices. An improved sequential quadratic programming algorithm or a deep reinforcement learning PPO algorithm is used for rolling time-domain optimization, achieving coordinated control of energy consumption and carbon emissions. Compared to traditional single-objective energy-saving strategies, this method effectively solves the problem of the disconnect between energy-saving and carbon reduction effects while ensuring the safety of crude oil transportation (meeting constraints such as minimum transport pressure and maximum allowable temperature drop). It significantly reduces the comprehensive energy consumption and carbon footprint of pipeline operation, adapting to the low-carbon operation requirements under the "dual-carbon" strategy.

[0034] (3) This invention introduces a closed-loop feedback correction mechanism, which identifies and updates key model parameters such as thermal conductivity coefficient and friction resistance coefficient online through recursive least squares method, solving the problem that traditional offline or periodic rescheduling strategies cannot respond to dynamic changes in operating conditions in real time. The system can dynamically adjust the pump station motor speed, heating furnace status and valve opening based on the deviation between actual operating data and model predictions, enhancing the real-time adaptive capability to fluctuations in grid carbon intensity, changes in electricity price peaks and valleys and sudden operating condition disturbances, effectively reducing redundant energy consumption and carbon emissions caused by operating condition fluctuations, and improving the robustness and stability of system operation. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of the overall technical solution architecture proposed in this invention;

[0036] Figure 2 This is a schematic diagram of the core principle framework of the multi-objective collaborative optimization objective function and the rolling time-domain optimization solution in this invention. Detailed Implementation

[0037] The present invention will be further described below with reference to the accompanying drawings and embodiments. The embodiments of the present invention include, but are not limited to, the following embodiments.

[0038] Example 1

[0039] like Figure 1As shown, this invention discloses an intelligent energy-saving control and carbon footprint optimization method for long-distance crude oil pipelines. First, a multi-physics coupled dynamic model is constructed. This model is based on the fluid dynamics and thermodynamics of long-distance crude oil pipelines, introducing the nonlinear relationship between oil density and viscosity with temperature, and combining it with the spatial distribution function of ambient temperature along the pipeline. The fluid dynamics characteristics are described by a momentum equation considering pressure gradient, gravity, and frictional resistance, i.e.:

[0040]

[0041] Where t is time, Let be the density of crude oil, x be the distance along the pipeline, D be the inner diameter of the pipeline, u be the average fluid velocity, p be the fluid pressure, and g be the acceleration due to gravity. Let θ be the angle between the pipe and the horizontal direction, and f be the Darcy friction coefficient, which is related to the Reynolds number.

[0042] Thermodynamic properties are described by an energy equation that considers changes in internal energy, pressure work, frictional heat generation, and heat exchange with the environment, namely:

[0043]

[0044] Where T is the temperature of the crude oil. The specific heat capacity of crude oil at constant pressure. The equivalent thermal conductivity of the pipe insulation layer plus the soil. The outer diameter of the insulation layer, Let represent the ambient temperature distributed along the pipeline; the first term on the right represents the work done by the pressure change; the second term represents the heat generated by friction (converted from the frictional resistance in the momentum equation); and the third term represents the heat exchange between the pipeline and the external environment. The expression for the spatial distribution function of ambient temperature along the pipeline is:

[0045]

[0046] In the formula, The starting ambient temperature, The endpoint ambient temperature is L, and the total length of the pipeline is L.

[0047] In terms of physical property modeling, the relationship between crude oil density ρ and temperature T is derived from the API gravity, and the expression is: ,in Reference temperature The density below, This is the coefficient of volumetric expansion. Viscosity. With temperature The nonlinear dependency is accurately characterized using Andrade's empirical formula:

[0048]

[0049] Wherein, A and B are constants calibrated based on the API gravity of crude oil, which is determined by laboratory multi-point viscosity measurements and least squares fitting.

[0050] like Figure 2 As shown, after completing the above modeling, the system calculates the heat loss of each pipe section in real time. With pressure drop Heat loss calculations are based on Newton's law of cooling. ,in Let be the overall heat transfer coefficient of the i-th pipe segment, and D be the outer diameter of the pipe. The average temperature of the crude oil inside the pipe. This corresponds to the ambient temperature at that location. Pressure drop. Calculated using the Darcy-Weisbach formula, ,in The friction factor is obtained by iteratively solving the Colebrook-White equation. Let be the pipe segment length and v be the flow velocity. All calculation results are refreshed every 10 seconds and used as input state variables for subsequent optimization modules.

[0051] In this embodiment, a distributed sensor network is deployed along the entire pipeline. Pressure transmitters are installed at pump station outlets and key nodes at 2-kilometer intervals, with an accuracy class of 0.075% and a range covering 0 to 10 MPa. Temperature sensors, using armored platinum resistance thermometers (PT100), are deployed at 1-kilometer intervals at the pipeline burial depth, with a temperature measurement range of -40°C to 150°C and an accuracy of ±0.5°C. Each pump station inlet and outlet is equipped with a Coriolis high-precision mass flow meter, with a measurement accuracy of ±0.1% and a sampling frequency of 1Hz. All sensor data is uploaded to the central control server via industrial Ethernet (IEEE 802.3 standard) at 100ms intervals. Data packets are encapsulated using the Modbus TCP protocol and include a timestamp, device ID, measured value, and checksum.

[0052] Simultaneously, the system accesses real-time carbon emission intensity data from the power grid via a power dispatch data interface. This data, published by the provincial power trading center, is updated every 15 minutes and is expressed in grams of carbon dioxide per kilowatt-hour (GCO2 / kWh). Time-of-use electricity price information is also acquired synchronously, including the electricity value for peak, mid-peak, flat, and off-peak periods, expressed in yuan per kilowatt-hour. These external information and internal operational data together constitute a multi-dimensional input vector. Where M is the feature dimension, typically 128. This vector contains the following fields: outlet pressure of each pumping station. ( (J represents the total number of pump stations) and the temperature at each monitoring point. ( ), flow rate of each pumping station Instantaneous flow rate of natural gas for heating furnace ( ), regulating valve opening ( Current grid carbon intensity Current electricity price Ambient temperature distribution Data such as crude oil API gravity, etc. Before entering the optimization module, all data is processed by the data cleaning unit at the edge node to remove abnormal jump points and perform zero-mean normalization.

[0053] Subsequently, the normalized energy consumption index E is defined, and its calculation method is as follows: ,in The real-time power of the motor at the j-th pump station is calculated from the voltage, current, and power factor. The thermal power of the l-th heating furnace is determined by the natural gas flow rate. Multiply by the calorific value (taken as 35.2 MJ / m³). 3 And take into account thermal efficiency (typical value 92%) conversion; Total transport volume, in tons. E is scaled to the [0,1] range based on the maximum and minimum values ​​of historical operating data.

[0054] Normalized carbon emission index C is defined as ,in The energy consumption of pump station j is (kWh). The carbon emission factor for natural gas is set at 1.9 kg CO2 / m³. 3 C is also normalized to [0,1]. Based on this, a weighted joint objective function is constructed:

[0055]

[0056] Where α and β are dynamic weighting coefficients, satisfying .

[0057] The dynamic weight adjustment strategy is implemented as follows: a preset carbon intensity threshold CI is set. high =600gCO2 / kWh, CI low =300gCO2 / kWh; Electricity pricing periods are divided into peak hours (e.g., 8:00-12:00, 18:00-22:00), off-peak hours (e.g., 0:00-8:00), and others. When CI grid CI high Furthermore, during peak electricity price periods, β takes the first preset value of 0.8; when CI grid <CI low Furthermore, during periods of low electricity prices, β takes the second preset value of 0.3; under other operating conditions, β is determined by linear interpolation: if CI grid ∈[CI lowCI high ], then β = 0.3 + (CI grid -CI low ) / (CI grid -CI low )×0.5; If it is at its peak but CI grid ≤CI low If β = 0.5, then β = 0.5; if it is at a low point but CI grid ≥CI high If β = 0.6, then this strategy ensures priority carbon reduction during periods of high carbon and high electricity prices, and focuses on energy conservation during periods of low carbon and low electricity prices, thus achieving spatiotemporal adaptability of the objective function.

[0058] In the rolling time-domain optimization solution, the optimization time window is initially set to 2 hours, with a rolling step size of 10 minutes. This means that the optimal control sequence for the next 2 hours is re-solved every 10 minutes based on the latest state. Optimization variables include: the motor speed of each pump station. (j=1,...,J), with values ​​ranging from 60% to 100% of the rated speed; furnace start-up and shutdown status. 0 indicates shutdown, 1 indicates operation; regulating valve opening ∈[0,1]. Constraints fall into three categories: physical constraints, safety constraints, and equipment operation constraints.

[0059] The physical constraints are provided by the multiphysics model, which requires the pressure at any time and any location. (Typically 2.5 MPa), crude oil temperature (Based on the freezing point setting, such as 35℃). Safety constraints include a maximum allowable temperature drop rate dT / dt ≤ 2℃ / h to prevent thermal stress damage to the pipeline. The specific equipment operation constraints are: the rate of change of pump station motor speed |d / dt|≤0.5% / s, avoid mechanical impact; the start-up and shutdown interval of the heating furnace should not be less than 30 minutes to prevent frequent start-ups and shutdowns from damaging the burner; the change range of the regulating valve opening |Δ |≤5% / step, ensuring a smooth transition.

[0060] An improved Sequential Quadratic Programming (SQP) algorithm is employed for optimization. This algorithm introduces sparse matrix processing techniques, reducing the storage and computational complexity of the Hessian matrix and the constrained Jacobian matrix from... Down to Where N is the total number of optimization variables (typically 200). Specifically, utilizing the chain topology of the pipeline system, the large-scale nonlinear programming problem is decomposed into J+1 subproblems (J pump station subproblems + 1 global coordination subproblem), and solved in parallel using the Alternating Direction Multiplier Method (ADMM). The computation time for a single optimization is controlled within 8 seconds, meeting the real-time requirement of a 10-minute rolling cycle. The solution results are the optimal trajectories of each control variable in the next scheduling cycle, which are converted into specific motor inverter setpoints, furnace ignition signals, and electric regulating valve control currents by the instruction issuance module.

[0061] In the closed-loop feedback correction mechanism, the system collects actual operating data every 10 seconds and calculates the deviation between this data and the predicted values ​​from the multiphysics model at the same time. (Pressure deviation) Temperature deviation The preset pressure deviation threshold is 0.15 MPa, and the temperature deviation threshold is 1.5°C. At any monitoring point... >0.15MPa or When the temperature is above 1.5°C, the online model parameter identification module is triggered.

[0062] This module uses the recursive least squares (RLS) method to update key parameters. The parameters to be identified include: the overall heat transfer coefficient of each pipe section. With frictional resistance coefficient Darcy friction factor = The state vector of the RLS algorithm The observation equation is constructed from the measured-predicted residuals of heat loss and pressure drop. Forgetting factor. We set the value to 0.98 to balance the weights of historical data and new information. The parameter update formula is:

[0063]

[0064] in Here is the gain matrix. For the residual vector, This is the regression vector. After each identification, the new parameters are immediately injected into the multiphysics dynamic model for the next round of prediction and optimization.

[0065] In addition, the online model parameter identification module runs automatically every 30 minutes, even if no deviation exceeds the limit. If the rate of change of five consecutive identification results (i.e., ||θ(k)-θ(k-1)|| / ||θ(k-1)||) is less than 0.5%, the parameters are frozen and updated, entering steady-state mode to avoid over-adjustment due to measurement noise. The frozen state remains until the next deviation exceeds the limit event or manual reset occurs.

[0066] The hardware architecture for implementing the optimization method in this embodiment is deployed in a cloud-edge collaborative computing architecture. Edge nodes are deployed on local servers at each pump station, with a hardware configuration of Intel Xeon Silver 4310 CPU, 32GB RAM, and 1TB SSD, running a real-time operating system (RTOS). Their responsibilities include: sensor data preprocessing (filtering, compression, anomaly detection), rapid feedback control (parameter updates and local PID adjustment), and 10-second status reporting. The cloud platform is deployed in an Alibaba Cloud private cloud environment, configured with a 64-core CPU, 256GB RAM, and a GPU accelerator card, running containerized microservices. Its responsibilities include: global rolling optimization (step 4), long-term policy learning (optimizing the adjustment rules of weights α and β based on reinforcement learning), and historical data analysis. Both exchange data via MQTT over TLS protocol encrypted with the national cryptographic standard SM4. The measured average end-to-end communication latency is 120 milliseconds, less than the preset threshold of 200 milliseconds.

[0067] To verify the effectiveness of this embodiment, a long-distance crude oil pipeline with a total length of 850 kilometers and 12 pumping stations was selected for actual testing. The pipeline is designed to transport 20 million tons per year, with crude oil having an API gravity of 32° and a pour point of 32°C. The experimental period was 30 consecutive days, during which the carbon intensity of the power grid fluctuated between 280 and 650 gCO2 / kWh. After applying this method, the average daily comprehensive energy consumption was 4.82 kWh / ton, a reduction of 11.7% compared to the baseline dispatch scheme; the average daily carbon footprint was 2.91 kgCO2 / ton, a reduction of 18.3%. In an extreme scenario where the carbon intensity suddenly increased to 620 gCO2 / kWh and the ambient temperature plummeted to -15°C, the system started the backup heater 2 hours in advance and adjusted the pump configuration to a high-head, low-flow mode, successfully maintaining a 100% completion rate for the transportation task. The overall pressure qualification rate (≥2.5 MPa) reached 99.2%, and the temperature qualification rate (≥35°C) reached 98.7%. The energy efficiency assessment unit identified a total of 7 anomalies, including decreased efficiency of 2 pumps, internal leakage of 3 valves, and drift of 2 temperature sensors. After maintenance, the energy efficiency of the relevant pipe sections improved by an average of 9.4%.

[0068] Example 2

[0069] In another specific implementation, the rolling time-domain optimization solution employs an alternative algorithm—the Proximal Policy Optimization (PPO) algorithm based on deep reinforcement learning. In this embodiment, the optimization time window remains 2 hours, but the rolling step size is shortened to 5 minutes to adapt to a more dynamic power grid environment. The state space s contains all the multidimensional input vectors X obtained in step 2, and the action space a is a continuous-discrete hybrid variable of the speed of each pump station, the state of the heating furnace, and the valve opening. The reward function r is directly defined as... ,in To constrain violations of the indicator function: if the pressure or temperature exceeds the limit, apply an immediate penalty of -100.

[0070] The PPO algorithm's neural network architecture comprises two fully connected networks sharing a common bottom layer: the policy network π(a|s) outputs the action mean and variance, and the value network V(s) estimates the state value. The network has a 128-dimensional input layer and three hidden layers, each with 512 neurons, using ReLU activation. Training data comes from a digital twin simulation platform, which generates 100,000 hours of running trajectories based on the high-fidelity model from step 1. Training is performed on a cloud-based GPU cluster, employing experience replay and importance sampling, with 1000 batches updated per round. During deployment, the policy network is quantized and compressed using TensorRT before being loaded onto edge nodes, keeping inference latency within 200 milliseconds.

[0071] This embodiment was tested on the same 850 km pipeline, with an average daily comprehensive energy consumption of 4.88 kWh / ton and a carbon footprint of 2.95 kg CO2 / ton, slightly higher than Embodiment 1. However, it performed better in scenarios with high-frequency fluctuations in grid carbon intensity (changes exceeding 200 g CO2 / kWh within 15 minutes), and optimized command response latency was reduced by 40%. In addition, the PPO strategy demonstrated stronger fault self-healing capabilities: when simulating a sudden shutdown of Pump Station #7, the system completed the commissioning of the backup pump and full-line rebalancing within 3 minutes, while Embodiment 1 required 4.5 minutes.

[0072] The above embodiments are merely one of the preferred embodiments of the present invention and should not be used to limit the scope of protection of the present invention. Any modifications or refinements made to the main design concept and spirit of the present invention that are not of substantial significance, but solve the same technical problem as the present invention, should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent energy-saving control and carbon footprint optimization for long-distance crude oil pipelines, characterized in that, Includes the following steps: S1, based on the fluid dynamics and thermodynamics of long-distance crude oil pipelines, introduces the nonlinear relationship between oil density and oil viscosity with temperature, and constructs a dynamic mathematical model of pressure-flow-temperature three-field coupling by combining the spatial distribution function of ambient temperature along the pipeline. S2, acquires multi-source operating data and external carbon electricity information to form a multi-dimensional input vector containing system state variables and external disturbance factors; S3 generates a weighted joint objective function with the dual objectives of minimizing the comprehensive energy consumption per unit of transport and minimizing the carbon footprint across the entire supply chain; S4, perform rolling time-domain optimization solution, under the premise of meeting the minimum delivery pressure, maximum allowable temperature drop and equipment operation constraints, perform rolling optimization on the motor speed of each pump station, the start and stop status of the heating furnace and the opening of the regulating valve in the next scheduling cycle; S5. Implement a closed-loop feedback correction mechanism, analyze the deviation between the actual operating data and the model prediction, update the thermal conductivity coefficient and friction resistance coefficient using the recursive least squares method, and restart the optimization solution process.

2. The intelligent energy-saving control and carbon footprint optimization method for long-distance crude oil pipelines according to claim 1, characterized in that, In step S1, the fluid dynamics properties are described by a momentum equation that takes into account pressure gradient, gravity, and frictional resistance; the thermodynamic properties are described by an energy equation that takes into account changes in internal energy, pressure work, frictional heat generation, and heat exchange with the environment.

3. The intelligent energy-saving control and carbon footprint optimization method for long-distance crude oil pipelines according to claim 1, characterized in that, In step S1, the nonlinear relationship between the oil density and temperature is derived from the API gravity, and the expression is: ,in Reference temperature The density below, The coefficient of volume expansion; The nonlinear relationship between the viscosity of the oil and temperature is accurately described using the Andrade empirical formula: Wherein, A and B are constants calibrated based on the API gravity of crude oil, which is determined by laboratory multi-point viscosity measurements and least squares fitting.

4. The intelligent energy-saving control and carbon footprint optimization method for long-distance crude oil pipelines according to claim 3, characterized in that, The multi-source operational data is acquired through a distributed sensor network, which includes a pressure transmitter, a PT100 temperature sensor, and a Coriolis mass flow meter. The external carbon electricity information includes the real-time carbon emission intensity of the power grid and time-of-use electricity prices. All raw data undergoes sliding window median filtering at edge nodes and is normalized to zero mean, forming a multidimensional input vector of dimension M.

5. The intelligent energy-saving control and carbon footprint optimization method for long-distance crude oil pipelines according to claim 4, characterized in that, The specific process of step S3 is as follows: S31, Define the normalized energy consumption index E: in, Let j be the real-time power of the motor in the j-th pump station. The thermal power of the l-th heating furnace; Total transport volume; S32, defines the normalized carbon emission index C: in, For the power consumption of pump station j, The instantaneous flow rate of natural gas to the heating furnace. The carbon emission coefficient for natural gas; S33, Construct the weighted joint objective function: in Weight The electricity price will be dynamically adjusted based on the real-time carbon emission intensity of the power grid and the time-of-use electricity price.

6. The intelligent energy-saving control and carbon footprint optimization method for long-distance crude oil pipelines according to claim 1, characterized in that, In step S4, the optimization solution adopts an improved sequential quadratic programming algorithm, which decomposes the problem into J+1 subproblems using a pipeline chain topology and performs parallel calculations through an ADMM parallel solver; the output optimal control sequence is converted into analog or switching signals by the instruction issuing and execution module.

7. The intelligent energy-saving control and carbon footprint optimization method for long-distance crude oil pipelines according to claim 1, characterized in that, In step S4, the optimization solution employs the Proximal Policy Optimization (PPO) algorithm based on deep reinforcement learning; state space With multidimensional input vectors Consistency; Action Space The variables are a continuous-discrete mixture, including motor speed, furnace start / stop status, and control valve opening; reward function. ,in To constrain violations of indicator functions.

Images