Multi-stage pressure regulated steam delivery network and method of optimization thereof

Abstract

The application provides a multi-stage pressure regulation steam delivery network and an optimization method thereof, comprising: judging steam grades according to preset steam grade division standards to obtain a steam grade distribution map in a pipe network; for a steam demand order submitted by an enterprise, extracting key steam parameters, matching the key steam parameters with the steam grade distribution map in the pipe network, searching for a feasible delivery path from a steam source to the target enterprise, and selecting a target path as a delivery scheme of the order; continuously monitoring real-time parameters of a target node in the pipe network, comparing the real-time parameters with predicted parameters, judging that a delivery state is abnormal when a deviation value exceeds a preset threshold, and taking corresponding adjustment measures; constructing a pressure regulation optimization model to adjust a preset multi-stage pressure regulation method; when there is a change in an enterprise in the park, a local pipe network topology structure and a delivery path are re-planned and integrated into a delivery scheduling system for unified management.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a multi-stage pressure-regulated steam transport network and its optimization method. Background Technology

[0002] The industrial park's steam pipeline network faces a complex and ever-changing supply and demand relationship and operating environment, making efficient and precise steam dispatch a major challenge. Traditional static dispatching methods are ill-suited to the dynamic changes in enterprise steam demand, resulting in both resource waste and insufficient supply. Simultaneously, pressure losses and parameter fluctuations during long-distance transmission make it difficult for the pipeline network to consistently meet user needs at the end. Furthermore, changes in enterprises within the park introduce uncertainty into pipeline layout and dispatching. These factors collectively lead to problems such as low steam resource utilization efficiency, high energy consumption, and poor reliability. Achieving real-time perception, precise control, and dynamic optimization of the entire pipeline network in a complex and ever-changing environment is a core technical challenge that urgently needs to be addressed. In addition, enterprises may be added or removed from the park, and their steam demands may also change. Dynamically optimizing and adjusting the transmission strategy during the operation of the steam pipeline network to ensure the stability and efficiency of steam transmission is also a complex optimization problem. Summary of the Invention

[0003] This invention provides a multi-stage pressure-regulated steam transport network and its optimization method, mainly including:

[0004] Obtain steam monitoring values ​​from target nodes in the park's steam pipeline network;

[0005] The steam level is determined according to the preset steam level classification standard, and the steam level distribution map in the pipeline network is obtained.

[0006] For steam demand orders submitted by enterprises, key steam parameters are extracted and matched with the pipeline steam level distribution map to search for feasible transportation paths from the steam source to the target enterprise, and the target path is selected as the transportation plan for the order.

[0007] Based on the aforementioned transportation scheme, the steam parameters at the end of the pipeline are predicted, and the prediction results are compared with the steam demand of the target enterprise to optimize the steam supply parameters until the target demand is met.

[0008] Continuously monitor the real-time parameters of the target nodes in the pipeline network, compare the real-time parameters with the predicted parameters, and when the deviation exceeds the preset threshold, it is judged as an abnormal transport status, and corresponding adjustment measures are taken.

[0009] Based on the steam pressure change data obtained from the regulation measures, a pressure regulation optimization model is constructed, and the preset multi-stage pressure regulation method is adjusted.

[0010] When there are changes in the enterprises within the park, the topology and transportation routes of the local pipeline network are replanned and incorporated into the transportation scheduling system for unified management.

[0011] Furthermore, obtaining the steam parameters of the target node in the park's steam pipeline network includes:

[0012] Acquire pressure, temperature, and flow data of the target node; calculate the average value and standard deviation of each sensor over the most recent time period based on the acquired data of the target node; if the data point deviates from the average value by more than a preset multiple of the standard deviation, it is determined to be abnormal data; use interpolation methods to correct the abnormal data and obtain the corrected data.

[0013] Furthermore, the step of determining the steam level according to a preset steam level classification standard to obtain a steam level distribution map within the pipeline network includes:

[0014] Based on the corrected steam monitoring data and the preset steam level classification standard, the steam level of each target node is determined; the target node is spatially interpolated using the inverse distance weighting method to obtain the interpolation result; the park is gridded using the interpolation result to determine the steam level value of each grid unit; and a steam level distribution map of the park is generated based on the steam level value.

[0015] Furthermore, the step of predicting the steam parameters at the end of the pipeline based on the transportation plan, and comparing the prediction results with the steam demand of the target enterprise to optimize the steam supply parameters until the target demand is met includes:

[0016] The pipe length, diameter, insulation material parameters, and pipe roughness are obtained based on the steam transport path; pressure loss and heat loss are calculated; the predicted steam parameters at the end of the pipe are obtained based on the calculated pressure loss and heat loss; the predicted steam parameters are compared with the steam demand of the target enterprise; if the pressure error or temperature error is greater than the threshold, the gradient descent method is initiated to optimize the steam supply parameters.

[0017] Furthermore, the real-time parameters of the target nodes within the continuous monitoring network are compared with the predicted parameters. When the deviation exceeds a preset threshold, it is determined that the transportation status is abnormal, and corresponding adjustment measures are taken, including:

[0018] Collect pressure, temperature, and flow data of the target node; perform lossless compression on the acquired target node data to obtain compressed data; compare the compressed data with pre-stored prediction parameters, and calculate the average deviation value using the sliding window method to obtain pressure deviation, temperature deviation, and flow deviation; if at least one of the pressure deviation, temperature deviation, and flow deviation exceeds a preset threshold, an early warning mechanism is triggered.

[0019] Furthermore, the step of acquiring steam pressure change data based on regulation measures, constructing a pressure regulation optimization model, and adjusting the preset multi-level pressure regulation method includes:

[0020] Real-time pressure data of each node in the steam pipeline network is acquired; data cleaning and normalization are performed on the real-time pressure data to obtain a standardized pressure change dataset; hierarchical clustering is performed on the standardized pressure change dataset using dynamic time regularization distance as a similarity metric to divide the pressure change data into multiple levels, resulting in a multi-level pressure change dataset; based on the multi-level pressure change dataset, a long short-term memory network is constructed as a pressure regulation optimization model.

[0021] Furthermore, when there are changes in enterprises within the park, the topology and transmission routes of the local pipeline network are replanned and incorporated into the transmission scheduling system for unified management, including:

[0022] The system acquires geographic information data of the park's steam pipeline network; establishes a digital twin model based on the geographic information data; calculates the optimal transmission path and pressure distribution scheme using the Dijkstra algorithm; receives access signals from newly added enterprises; re-plans the local pipeline network structure using the Steiner tree algorithm based on the access signals from newly added enterprises; acquires steam demand data from newly added enterprises; and predicts future steam consumption using the ARIMA model.

[0023] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0024] This invention discloses a multi-stage pressure-regulated steam transmission network and its optimization method. The method involves installing sensors at key nodes in the pipeline network to collect data in real time, establishing a steam level distribution map, and matching the optimal transmission path based on enterprise order requirements. After determining the path, the invention calculates losses during transmission, predicts pipeline end parameters, and adjusts upstream parameters as needed to meet demand. During transmission, the invention continuously monitors real-time parameters, triggering an early warning and implementing adjustment measures when deviations exceed thresholds. By constructing a pressure regulation optimization model, the invention can continuously improve regulation technology. Furthermore, the invention can flexibly adjust the transmission scheduling system according to changes in enterprises within the industrial park, achieving dynamic optimization of the pipeline network. This method significantly improves the operational efficiency and reliability of the industrial park's steam pipeline network, achieving precise scheduling of steam resources and energy conservation and emission reduction. Attached Figure Description

[0025] Figure 1 This is a flowchart of the multi-stage pressure-regulated steam delivery network and its optimization method according to the present invention.

[0026] Figure 2 This is a schematic diagram of the multi-stage pressure-regulated steam delivery network and its optimization method according to the present invention. Detailed Implementation

[0027] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0028] like Figure 1-2 The multi-stage pressure-regulated steam delivery network and its optimization method in this embodiment may specifically include:

[0029] S101. Install pressure, temperature, and flow sensors at the target nodes of the steam pipeline network in the park to collect steam monitoring values ​​of the target nodes in the pipeline network in real time. Transmit the collected data to the data processing center, establish steam level classification standards in advance, determine the steam level of each point according to the steam level classification standards, and obtain the steam level distribution map in the pipeline network.

[0030] The process involves acquiring pressure, temperature, and flow data from target nodes, collected by sensors in the steam pipeline network. The average value and standard deviation of each sensor over the most recent time period are calculated based on this data. Data points deviating from the average value by more than a preset multiple of the standard deviation are considered abnormal. An interpolation algorithm is used to correct the abnormal data, resulting in corrected data. The steam level of each target node is determined based on the corrected steam monitoring data and a preset steam level classification standard. Spatial interpolation is performed on the target nodes using an inverse distance weighting method to obtain the interpolation result. The interpolation result is used to grid the entire area, determining the steam level value for each grid cell. A steam level distribution map of the entire area is generated based on the steam level values, with different colors representing different steam levels.

[0031] For example, pressure, temperature, and flow rate data of target nodes in the park's steam pipeline network are collected. The collected real-time data is transmitted to the data processing center via a data transmission network. The data from each node is classified according to a pre-established steam level classification standard, generating a steam level distribution map within the pipeline network. For data collected by each sensor, the average value and standard deviation over the most recent 10 minutes are calculated. If a data point deviates from the average value by more than three times the standard deviation, it is marked as abnormal data. For abnormal data points, linear interpolation is used to correct them, ensuring data continuity and reliability. Based on preset steam level classification standards, such as pressure range, temperature range, and flow rate range, the steam level of each node is determined. If the pressure is within 0.6-0.8 MPa, the temperature is within 150-180℃, and the flow rate is within 50-100 t / h, it is determined to be medium-pressure steam. For high-pressure and low-pressure steam, corresponding parameter ranges are set for determination. Based on the geographical location information of the target node and the steam level determination results, spatial interpolation is performed using an inverse distance weighting method. For locations in the pipeline network where no sensors are installed, the steam level at that location is calculated based on the steam levels and distances of known surrounding nodes. The closer the known node, the greater its influence weight. Using interpolation results, the entire park is gridded, with each grid cell assigned a corresponding steam level value. Different colors are used to represent different steam levels, generating a heat map of the steam level distribution across the entire park. High-pressure steam areas are represented in red, medium-pressure in yellow, and low-pressure in green. Color gradients visually represent continuous changes in steam levels, intuitively displaying the steam quality status of different areas within the pipeline network. Pressure, temperature, and flow sensors are installed at target nodes in the park's steam pipeline network, with a sampling frequency set to once every 10 seconds. The collected data is transmitted to the data processing center via industrial Ethernet at a transmission rate of 100Mbps. After receiving the real-time data, the data processing center first performs data preprocessing. Using a 10-minute time window, the average value and standard deviation of each sensor are calculated. For example, if a pressure sensor collects 60 data points within 10 minutes, the calculated average pressure is 0.7 MPa, and the standard deviation is 0.05 MPa. If the pressure value of a data point exceeds the range of 0.7 ± 0.15 MPa, it is marked as abnormal data. For abnormal data, linear interpolation is used for correction. The corrected data proceeds to the next processing step along with the normal data. The steam level of each node is determined according to the preset steam level classification standards. High-pressure steam is defined as pressure greater than 0.8 MPa, temperature higher than 180℃, and flow rate greater than 100 t / h; medium-pressure steam is defined as pressure between 0.6 and 0.8 MPa, temperature between 150 and 180℃, and flow rate between 50 and 100 t / h; low-pressure steam is defined as pressure less than 0.6 MPa, temperature less than 150℃, and flow rate less than 50 t / h. For locations in the pipeline network where no sensors are installed, spatial interpolation is performed using an inverse distance weighting method.Suppose there is an unknown point P, surrounded by three known points A, B, and C, at distances of 10m, 15m, and 20m respectively. The steam levels at points A, B, and C are high-pressure, medium-pressure, and low-pressure, respectively. When calculating the steam level at point P, the weight of point A is 1 / 10, point B is 1 / 15, and point C is 1 / 20. The steam level at point P is calculated based on these weights. The entire area is then gridded with a grid size of 5m × 5m. Each grid cell is assigned a corresponding steam level value based on the interpolation result. Finally, a steam level distribution heatmap is generated, using red, yellow, and green to represent high-pressure, medium-pressure, and low-pressure steam regions, respectively. Gradient color areas represent transitional zones between steam levels, visually displaying the steam mass distribution in different areas of the pipeline network.

[0032] S102. For the steam demand order submitted by the enterprise, extract the key steam parameters of the target in the order. The key steam parameters include grade, pressure and temperature. Match the key steam parameters with the pipeline steam grade distribution map to search for feasible transportation paths from the steam source to the target enterprise, calculate the transportation efficiency and energy consumption of each path, and select the target path as the transportation plan for the order.

[0033] The process involves: acquiring steam demand orders; extracting key steam parameters from these orders, including steam grade, pressure, and temperature requirements; performing a matching search on the pipeline steam grade distribution map based on these parameters, organizing pipeline nodes by geographical location using an R-tree spatial index structure; determining the steam source nodes and the location nodes of the target enterprise within the pipeline network if the query range meets preset conditions; searching for multiple feasible transport paths using the Dijkstra algorithm based on the steam source nodes and the target enterprise node, with the weights of each feasible transport path calculated from pipeline length, pressure loss rate, and temperature decay rate; estimating the transport efficiency and energy consumption of each feasible transport path using a thermodynamic calculation model that includes pipeline friction loss and heat loss calculations; calculating a comprehensive score for each feasible transport path based on the transport efficiency and energy consumption; determining if the comprehensive score is the highest, and if so, identifying the corresponding feasible transport path as the preliminary transport plan for the order.

[0034] For example, key steam parameters, including steam grade, pressure, and temperature requirements, are extracted from steam demand orders submitted by enterprises. Regular expressions are used to match numerical values ​​and units in the order text, such as "pressure: 0.8MPa" and "temperature: 180℃". The extracted parameters are stored as structured data in key-value pair format. Based on the extracted key steam parameters, a matching search is performed on the pipeline steam grade distribution map, using an R-tree spatial index structure to organize pipeline nodes by geographical location. Based on the extracted parameters, a query range is set, such as pressure ±0.1MPa and temperature ±10℃, to quickly locate steam source nodes that meet the requirements, and simultaneously determine the location node of the target enterprise in the pipeline network. Using shortest path algorithms in graph theory, such as Dijkstra's algorithm, multiple feasible transport paths from the steam source node to the target enterprise node are searched, considering factors such as pipeline length, pressure loss, and temperature decay. When applying the Dijkstra algorithm, pipe length, pressure loss, and temperature decay are converted into edge weights. The weight calculation formula is: Weight = Pipe Length * (1 + Pressure Loss Rate + Temperature Decay Rate). A comprehensive score for each path is calculated by considering multiple factors. For the multiple transport paths found, a thermodynamic calculation model is used to estimate the transport efficiency and energy consumption of each path. The thermodynamic calculation model includes: pipe friction loss calculation using the Darcy-Weisbach equation; heat loss calculation, considering the pipe insulation layer, using the heat conduction formula; transport efficiency calculation, output steam energy divided by input steam energy; and energy consumption calculation, considering pump power consumption and heat loss. Taking into account factors such as pipe friction and heat loss, the path with the highest comprehensive score is selected as the final transport solution for the order. After the enterprise submits a steam demand order, the regular expression "(\d+(.\d+)?)([MPa|℃])" is used to match the numerical values ​​and units in the order text. For example, from the text "Requirement: Medium-pressure steam, Pressure: 0.8MPa, Temperature: 180℃", the pressure of 0.8MPa and the temperature of 180℃ are extracted. The extracted parameters are stored as key-value pairs {"pressure":0.8,"temperature":180}. Next, a matching search is performed on the pipeline steam level distribution map. An R-tree spatial index structure is used to organize the pipeline nodes, with each node containing location coordinates and steam parameters. The query range is set to pressure ±0.1MPa and temperature ±10℃, i.e., 0.7-0.9MPa and 170-190℃. The R-tree quickly locates steam source nodes that meet the conditions, while simultaneously identifying the target enterprise node. Then, the Dijkstra algorithm is applied to search for feasible delivery paths. Pipeline length, pressure loss, and temperature decay are converted into edge weights. Assuming a pipeline is 100 meters long, has a pressure loss rate of 0.01 MPa / 100m, and a temperature decay rate of 1℃ / 100m, the weight of this edge is calculated as 100*(1+0.01+0.01)=102. The algorithm traverses all possible paths, calculates the total weight, and finds the paths with the minimum weights. Finally, thermodynamic calculations are performed on the searched paths.The Darcy-Weisbach equation is used to calculate pipeline friction loss. Assuming a pipe diameter of 100 mm, a flow velocity of 20 m / s, and a friction coefficient of 0.02, the pressure loss of 100 meters of pipeline is approximately 0.04 MPa. Heat loss is calculated considering the pipeline insulation layer. Assuming an insulation layer thickness of 50 mm, a thermal conductivity of 0.04 W / (m·K), and an ambient temperature of 20℃, the temperature loss of 100 meters of pipeline is approximately 2℃. The transport efficiency is calculated as the output steam energy divided by the input steam energy, considering pressure and temperature losses, and assuming an efficiency of 95%. Energy consumption calculations include pump power consumption and heat loss. Assuming a pump power of 5 kW, a heat loss of 2 kW, and a total energy consumption of 7 kW, the route with the highest overall score is selected as the final transport scheme.

[0035] Based on the initial transportation plan, a set of enterprises with overlapping routes is identified. For overlapping route segments, it is determined whether the pressure and temperature requirements of the enterprises involved match. If they match, the original plan is maintained; otherwise, a dynamic programming algorithm is used to search for alternative routes until the needs of all enterprises are met, forming the final transportation plan. This includes: obtaining the steam demand parameters of each enterprise, including pressure, temperature, and flow requirements, based on the initial transportation plan; traversing the paths in the initial transportation plan using graph theory algorithms to identify the set of enterprises with overlapping routes; extracting the pressure and temperature demand parameters of the enterprises involved in the overlapping route segments and determining whether the needs of each enterprise match through similarity calculation; if the matching degree of the demand parameters of the enterprises involved in the overlapping route segments is greater than the parameter matching degree threshold, the original transportation plan remains unchanged. If the matching degree of the demand parameters of the enterprises involved in the overlapping path segment is less than the parameter matching degree threshold, the path segment is marked as needing to be replanned. For the overlapping path segments that need to be replanned, a dynamic programming algorithm is used to search for alternative paths, considering factors such as pipeline length, pipe diameter, and flow constraints, resulting in multiple candidate paths. The candidate paths are evaluated and scored, and the path with the best score is selected as the alternative to the overlapping path segment, taking into account indicators such as pipeline laying cost, transportation efficiency, and energy consumption. The replanned path replaces the overlapping path segments in the original transportation plan to form a new transportation plan. It is determined whether the new transportation plan meets the steam demand of all enterprises. If it does, it is determined as the final transportation plan; if not, the similarity calculation is iteratively executed until the new transportation plan is formed, until the demand of all enterprises is met or the maximum number of iterations is reached. For example, the steam demand parameters of each enterprise are first obtained based on the preliminary transportation plan. For example, enterprise A needs steam with a pressure of 2MPa, a temperature of 300℃, and a flow rate of 5t / h, while enterprise B needs steam with a pressure of 0MPa, a temperature of 280℃, and a flow rate of 3t / h. Then, Dijkstra's algorithm was used to traverse the paths in the initial transportation plan, identifying a 20km overlap between companies A, B, and C. For this overlapping path segment, the pressure and temperature requirements of the three companies were extracted, and the matching degree of each company's requirements was determined using cosine similarity calculation. If the matching degree was greater than a threshold of 9, the original plan was maintained; otherwise, the path segment was marked as needing to be replanned. Next, a dynamic programming algorithm was used to search for alternative paths. Under the constraints of a pipeline length not exceeding 25km, a pipe diameter of DN200, and a flow rate of 10t / h, two alternative paths were obtained. The alternative paths were evaluated and scored, and considering factors such as pipeline laying cost, transportation efficiency, and pump power consumption, the path with a total score of 85 was selected as the alternative plan. The replanned path replaced the overlapping segment in the original plan, forming a new transportation plan. Finally, it was determined whether the new plan met the requirements of all companies. If it did, it was determined as the final plan; otherwise, the above steps were iteratively executed until the requirements were met or the upper limit of 10 iterations was reached, forming a complete multi-company steam transportation path optimization plan.

[0036] S103. After determining the steam transmission path, calculate the pressure loss and heat loss during the transmission process based on the length, diameter, and insulation material parameters of the pipeline, and predict the steam parameters at the end of the pipeline. If the predicted parameters meet the steam demand of the target enterprise, the transmission plan is executed directly; if the predicted parameters do not meet the demand, adjust the steam supply parameters of the upstream steam source, recalculate and optimize the transmission plan until the target demand is met.

[0037] The pipeline length, diameter, insulation material parameters, and pipeline roughness are obtained based on the steam transport path. Pressure loss and heat loss are calculated. Predicted steam parameters at the pipeline end are obtained based on the calculated pressure and heat losses. These predicted steam parameters are compared with the target enterprise's steam demand. If the pressure error or temperature error between the predicted parameters and the demand exceeds a threshold, the gradient descent method is initiated to optimize the steam supply parameters. In the gradient descent method, the objective function is defined as the sum of the squared differences between the predicted parameters and the demand parameters. The partial derivatives of the objective function with respect to each steam supply parameter are calculated, and the steam supply parameters of the upstream steam source are adjusted along the gradient direction. The adjusted steam supply parameters are substituted into the Darcy-Weisbach equation and the heat conduction formula to recalculate the pipeline end parameters. If the objective function value is less than a preset value or the number of iterations reaches a preset number, the optimal steam supply parameters are output.

[0038] For example, based on the determined steam transport path, the pipe length, diameter, insulation material parameters, and pipe roughness are obtained from the database, and the pressure loss is calculated using the Darcy-Weisbach equation.

[0039]

[0040] In the calculation formula, ΔP represents the pressure loss in the pipeline, f is the friction factor, L is the pipe length, D is the pipe diameter, ρ is the steam density, and v is the flow velocity. Heat loss is calculated using the heat conduction formula.

[0041]

[0042] Let Q represent the heat loss per unit time, Δt represent the time interval, k represent the conduction coefficient, A represent the pipe surface area, T1 represent the steam temperature, T2 represent the ambient temperature, and d represent the insulation layer thickness. The predicted steam parameters at the end of the pipe are obtained based on the calculation results. These predicted steam parameters are compared with the steam demand of the target enterprise, and parameter matching thresholds are set: pressure error ±0.05 MPa and temperature error ±5℃. If the predicted parameters are within the threshold range, the demand is met, the delivery plan is marked as executable, and stored in the plan database. If the predicted parameters do not meet the demand, the gradient descent method is initiated to optimize the steam supply parameters. The objective function is defined as the sum of the squared differences between the predicted parameters and the demand parameters. The partial derivatives of the objective function with respect to each steam supply parameter are calculated, and the steam supply parameters of the upstream steam source, including pressure, temperature, and flow rate, are automatically adjusted along the gradient direction. An iterative calculation method is used, with a maximum of 100 iterations and a convergence threshold of the objective function value being less than 0.01. The adjusted steam supply parameters were substituted into the Darcy-Weisbach equation and the heat conduction formula to recalculate the pipeline end parameters and compare them with the target demand. After each iteration, the termination condition was checked. If it was met, the optimal steam supply parameters were output; otherwise, the iteration continued until the optimal combination of steam supply parameters that met the demand was found or the maximum number of iterations was reached. After determining the steam transport path, pipeline parameters were extracted from the database, including a length of 1000 meters, a diameter of 200 millimeters, a thermal conductivity of 0.04 W / (m·K) for the insulation material, a thickness of 50 millimeters for the insulation layer, and a pipe roughness of 0.045 millimeters. The pressure loss was calculated using the Darcy-Weisbach equation, assuming a steam density of 5 kg / m³ and a flow velocity of 20 m / s. The calculated friction factor f was 0.015, and the final pressure loss was 0.15 MPa. The heat loss was calculated using the heat conduction formula, assuming an initial steam temperature of 200℃ and an ambient temperature of 20℃. The calculated heat loss was 50 kW. Based on these calculations, the predicted steam parameters at the pipeline terminal are 0.85 MPa pressure and 185℃ temperature. The target company's steam demand is 0.8 ± 0.05 MPa pressure and 180 ± 5℃ temperature. Comparing the predicted and demand parameters, the pressure is found to be within the threshold range, but the temperature is slightly higher. Therefore, the gradient descent method is initiated to optimize the steam supply parameters. The objective function is defined as (0.85 - 0.8)² + (185 - 180)² = 25.0025. The partial derivatives of the objective function with respect to pressure and temperature are calculated to determine the adjustment direction. The first iteration adjusts the steam supply parameters to reduce the pressure by 0.02 MPa and the temperature by 2℃. Substituting the new parameters into the calculation model, new terminal prediction parameters are obtained: 0.83 MPa pressure and 183℃ temperature. The objective function value decreases to 9.49, which is less than the set convergence threshold of 0.01, but still not optimal. Iterative calculations continue, with fine-tuning of the steam supply parameters and reassessment of the terminal situation each time.After 15 iterations, the optimal steam supply parameters were finally determined to be a pressure of 0.95 MPa and a temperature of 188℃. At this point, the predicted terminal parameters fully met the target company's needs. This set of optimal parameters was stored in the database as the final delivery scheme.

[0043] S104. During the steam transmission process, the real-time parameters of the target nodes in the pipeline network are continuously monitored. The real-time parameters are compared with the predicted parameters, and the deviation value is calculated. When the deviation value exceeds the preset threshold, it is judged as an abnormal transmission status, and an early warning is automatically triggered. According to the magnitude of the deviation and the location of the abnormality, corresponding adjustment measures are taken. The adjustment measures include adjusting the valve opening and differential pressure distribution, and increasing or decreasing the steam supply from the steam source.

[0044] Pressure, temperature, and flow data of target nodes in a distributed sensor network are collected in real time. The collected data is losslessly compressed to obtain compressed data. The compressed data is compared with pre-stored prediction parameters, and an average deviation value is calculated using a sliding window algorithm to obtain pressure deviation, temperature deviation, and flow deviation. It is determined whether the pressure deviation, temperature deviation, and flow deviation exceed preset thresholds. If at least one of these deviations exceeds the preset threshold, an early warning mechanism is triggered. An anomaly classification result is generated based on the early warning mechanism, including anomaly type and severity. An adjustment command is automatically generated based on the anomaly classification result, used to adjust the opening of a specified valve. After executing the adjustment command, changes in the pressure, temperature, and flow data are continuously monitored.

[0045] For example, pressure, temperature, and flow data of target nodes within the pipeline network are collected in real time through a distributed sensor network. Huffman coding is used to losslessly compress the collected data, and the compressed data is transmitted to the central control system. After receiving the real-time data, the central control system compares it with pre-stored prediction parameters and uses a sliding window algorithm to calculate the average deviation value for the most recent hour, obtaining the pressure deviation, temperature deviation, and flow deviation. When calculating the deviation value, the relative error formula is used: Deviation rate = (Measured value - Predicted value) / Predicted value * 100%. The deviation rate is calculated separately for pressure, temperature, and flow. Anomalies are judged based on preset thresholds. If the deviation value exceeds the threshold, an early warning mechanism is triggered. The early warning mechanism uses a decision tree to classify anomaly types. The root node of the decision tree represents the deviation type (pressure / temperature / flow), the second layer represents the deviation magnitude (high / medium / low), the third layer represents the anomaly duration (short / medium / long), and the leaf nodes represent the specific anomaly type and severity, determining the anomaly location and severity. Based on the anomaly classification results, automatic adjustment commands are generated, including adjusting the opening of specified valves, redistributing pressure differentials across pipe sections, and sending control signals to the steam source to increase or decrease steam supply. Depending on the anomaly type and severity, a corresponding adjustment scheme is selected from a pre-set adjustment rule library. For example, for a moderately high pressure with a moderate duration, a command to reduce the upstream valve opening by 5% is generated. After executing the adjustment command, parameter changes are continuously monitored for 30 minutes. If the deviation does not improve, a secondary adjustment is triggered, achieving automatic adjustment and optimization of the steam transport process. During steam transport, a distributed sensor network collects pressure, temperature, and flow data of target nodes within the pipeline network every 5 seconds. Taking a certain node as an example, the collected raw data is: pressure 0.82 MPa, temperature 178℃, and flow rate 50 t / h. This data is compressed using Huffman coding, achieving a compression rate of 40%, and the compressed data is transmitted to the central control system. After receiving the data, it is decompressed and compared with predicted parameters, including pressure 0.80 MPa, temperature 180℃, and flow rate 48 t / h. A 60-minute sliding window was used to calculate the average deviation rate: pressure deviation 2.5%, temperature deviation -1.1%, and flow rate deviation 4.2%. The preset abnormal thresholds were ±5% for pressure, ±3% for temperature, and ±10% for flow rate. At this point, temperature and flow rate were within the normal range, but pressure was close to the abnormal threshold. Decision tree analysis showed that the pressure deviation type was "moderately high," with a duration of "medium," ranging from 30 to 60 minutes. Based on the regulation rule base, an automatic regulation command was generated: reduce the opening of upstream valve A by 3%. After executing the command, monitoring continued for 30 minutes, finding that the pressure dropped to 0.81 MPa and the deviation rate decreased to 1.25%, indicating effective regulation. If the pressure still exceeds 0.83 MPa after regulation, a secondary regulation will be triggered, further reducing the valve opening or decreasing the steam supply.

[0046] S105. Based on the regulation measures, analyze the adjustment of valve opening and differential pressure distribution, obtain steam pressure change data, divide the pressure change data into multiple levels, construct a pressure regulation optimization model, use the multi-level pressure change data as model input, obtain multi-level pressure regulation optimization results, and adjust the preset multi-level pressure regulation method.

[0047] Real-time pressure data of each node in the steam pipeline network is acquired by a data acquisition device. The real-time pressure data is then cleaned and normalized to obtain a standardized pressure change dataset. Data cleaning uses the 3σ principle to identify outliers, and normalization uses the Min-Max method to scale the data to the [0, 1] interval. Using dynamic time warping distance as a similarity metric, hierarchical clustering is performed on the standardized pressure change dataset, dividing the pressure change data into five levels: extremely low, low, medium, high, and extremely high, resulting in a multi-level pressure change dataset. Based on this multi-level pressure change dataset, a long short-term memory (LSTM) network is constructed as a pressure regulation optimization model. The LSM network uses past pressure change data as input features and future pressure predictions as output targets. The preset multi-level pressure regulation method is improved based on the output of the pressure regulation optimization model. If the predicted pressure exceeds a preset threshold, the opening of relevant valves is adjusted.

[0048] For example, based on the adjustment records of valve opening and differential pressure distribution, a data acquisition device is used to obtain real-time pressure data of each node in the steam pipeline network. Through data cleaning and normalization, a standardized pressure change dataset is generated. The data cleaning process uses the 3σ principle to identify and handle outliers, and the normalization uses the Min-Max method to scale the data to the [0,1] interval. Using dynamic time-warped distance as a similarity metric, combined with hierarchical clustering, the pressure change data is divided into five levels: extremely low, low, medium, high, and extremely high, resulting in a multi-level pressure change dataset. Based on this multi-level pressure change dataset, a long short-term memory (LSTM) network is constructed as a pressure regulation optimization model. The pressure change data of the past 24 hours is used as input features, and the predicted pressure value for the next hour is used as the output target. The model parameters are trained and optimized using cross-validation. The trained LSTM network is applied to new pressure change data to obtain the multi-level pressure regulation optimization results. Based on the optimization results, a rule-based decision system is designed. If the predicted pressure will exceed a preset threshold, the opening of the relevant valves is adjusted in advance. Simultaneously, a feedback mechanism was established to compare the errors between predicted and actual values, and the model was retrained weekly to improve accuracy. Based on the output of the decision system, the preset multi-level pressure regulation technology was improved, the valve opening control strategy and differential pressure distribution scheme were adjusted, and the regulation technology parameter library was updated. During the steam pipeline network optimization process, the data acquisition device collected pressure data from 100 nodes every 5 minutes. The original data range was 0.5-1.2 MPa, and 0.2% outliers were removed after processing according to the 3σ principle. Min-Max normalization was used to map the data to the [0,1] interval, such as converting 0.8 MPa to 0.5. For the processed 24-hour data, the similarity of the 2880 data points was calculated using dynamic time warping distance, with a maximum window width of 60, to obtain the similarity matrix. Hierarchical clustering was adopted, with the number of clusters set to 5, and the pressure changes were divided into five levels: extremely low (0-0.2), low (0.2-0.4), medium (0.4-0.6), high (0.6-0.8), and extremely high (0.8-1.0). A Long Short-Term Memory (LSTM) network was constructed, with an input layer of 24 neurons corresponding to 24 hours of data, a hidden layer containing 128 LSTM units, and an output layer of 1 neuron to predict the pressure for the next hour. Training was performed using data from the past 30 days, with a batch size of 64, a learning rate of 0.001, and 100 epochs. Applying the trained model to the input of the most recent 24 hours of data (e.g., 0.75, 0.78, 0.82, etc.), the predicted pressure for the next hour was 0.85, a normalized value. Based on the prediction, the decision system determined that the pressure was about to exceed the threshold of 0.8 and automatically generated instructions: reduce the opening of upstream valve A by 5% and increase the pressure differential distribution in branch B by 2%. After the adjustments were executed, the actual measured pressure one hour later was 0.83, with an error of 2.35% compared to the predicted value of 0.85.The prediction error was recorded. After accumulating a week's worth of data, the model was retrained using 10,080 new data points. The model's prediction accuracy improved from 94.5% to 95.8%.

[0049] S106. When there are changes in enterprises within the park, a transportation scheduling system is constructed. When a new enterprise is connected to the steam pipeline network, the topology and transportation path of the local pipeline network are re-planned according to its geographical location and steam demand, and it is incorporated into the entire transportation scheduling system for unified management. When an enterprise leaves the pipeline network, it is removed from the scheduling system, the transportation resources it occupies are released, and the operation strategy of the surrounding pipeline network is optimized and adjusted.

[0050] The process involves: acquiring geographic information data of the park's steam pipeline network, including network topology, enterprise distribution, and steam demand data; establishing a digital twin model based on the geographic information data, with the digital twin model updating the network topology, enterprise distribution, and steam demand data in real time; calculating the optimal delivery path and pressure allocation scheme using the Dijkstra algorithm; receiving new enterprise access signals, including the geographical coordinates and steam demand parameters of the new enterprise; replanning the local pipeline network structure based on the new enterprise access signals using the Steiner tree algorithm, which calculates the optimal pipeline connection scheme from the nearest steam source to the new enterprise; acquiring steam demand data of the new enterprise and predicting future steam consumption using the ARIMA model; integrating the new enterprise prediction results with existing load predictions using a weighted average method based on the ARIMA model prediction results; receiving enterprise exit signals, including the identification information of the exiting enterprise; and re-optimizing the pressure allocation and flow control parameters of the surrounding pipeline network using a genetic algorithm, which uses real-number encoding to represent the pressure and flow of each pipe segment.

[0051] For example, a digital twin model of the park's steam pipeline network based on a geographic information system is constructed. This model is updated in real time with pipeline topology, enterprise distribution, and steam demand data. The optimal delivery path and pressure distribution scheme are calculated using Dijkstra's algorithm, considering pipeline length, diameter, and pressure loss factors. When a new enterprise is detected joining the network, its geographic coordinates and steam consumption parameters are automatically acquired. The local pipeline network structure is replanned using a Steiner tree algorithm considering capacity constraints, and the optimal pipeline connection scheme from the nearest steam source to the new enterprise is calculated. Simultaneously, pipeline pressure balance calculations are performed to ensure that the new connection does not affect the stability of the existing system. The steam demand data of the new enterprise is input into the load forecasting model. The ARIMA model is used to predict the steam consumption for the next week, and a weighted average method is used to integrate the new enterprise's forecast with the existing load forecast. Based on the forecast results, an adaptive control algorithm is used to dynamically adjust the valve opening and pump speed of relevant pipeline sections, optimizing delivery capacity and pressure distribution. If an enterprise leaves the pipeline network, a resource release procedure is initiated to remove the enterprise from the scheduling database. A genetic algorithm is then used to re-optimize the pressure distribution and flow control parameters of the surrounding pipeline network. The specific implementation of the genetic algorithm includes using real-number encoding to represent the pressure and flow rate of each pipe segment, a fitness function considering total energy consumption and pressure balance, tournament selection for selection, arithmetic crossover for crossover, and Gaussian mutation for mutation. Population size and iteration count are set to obtain optimization results, enabling dynamic adjustment of the pipeline network operation strategy. During the dynamic scheduling of the park's steam pipeline network, the digital twin model updates data from 100 nodes in real time, including the steam demand of 50 enterprises and pressure and flow rate information of 50 key pipeline points. The Dijkstra algorithm recalculates the optimal path at 5-minute intervals, considering factors such as pipeline length (total length 50 km), diameter (200-500 mm), and pressure loss (0.01-0.05 MPa per km). When a new enterprise A's access signal is detected, its geographical coordinates (X:120.5, Y:30.8) and steam parameters (pressure 0.8 MPa, flow rate 5 tons / hour) are automatically obtained. The Steiner tree algorithm calculates the optimal connection scheme, adding a pipeline length of 1.2 km and a diameter of 300 mm. Pipeline pressure balance calculations show that the connection point pressure dropped from 0.85 MPa to 0.82 MPa, still within a safe range. The ARIMA model (parameters p=1, d=1, q=1) predicts that Company A's average daily steam consumption for the next seven days will be 4.8, 5.2, 5.1, 4.9, 5.0, 5.3, and 5.1 tons / hour. A weighted average method is used to integrate this result with existing load forecasts, with a weighting ratio of 1:9. Based on this, the adaptive control algorithm adjusts the openings of the three regulating valves in the relevant pipeline section to 75%, 80%, and 85%, respectively, and adjusts the speeds of the two variable frequency pumps to 45Hz and 48Hz. When Company B exits the pipeline network, a genetic algorithm optimizes the surrounding pipeline network parameters, with a population size of 100 and 500 iterations.The final optimization plan involves closing the upstream valve of the original company B, resetting the pressure of the three adjacent nodes to 0.79MPa, 0.81MPa and 0.78MPa, and adjusting the flow rates to decrease by 2%, increase by 3% and remain unchanged, respectively.

[0052] The above description is merely a specific implementation of this specification. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the scope of protection of this specification is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this specification, and these modifications or substitutions should all be covered within the scope of protection of this specification.

Claims

1. A multi-stage pressure-regulated steam transport network and its optimization method, characterized in that, The method includes: Obtain steam monitoring values ​​from target nodes in the park's steam pipeline network; The steam level is determined according to a preset steam level classification standard, and a steam level distribution map within the pipeline network is obtained. Specifically, this includes: determining the steam level of each target node based on the corrected steam monitoring values ​​and the preset steam level classification standard; performing spatial interpolation on the target nodes using the inverse distance weighting method to obtain the interpolation result; using the interpolation result to perform gridding processing on the park to determine the steam level value of each grid unit; and generating the steam level distribution map of the park based on the steam level value. For steam demand orders submitted by enterprises, key steam parameters are extracted and matched with the steam level distribution map in the pipeline network. Feasible transportation paths from the steam source to the target enterprise are searched, and the target path is selected as the transportation plan for the order. The steam parameters at the end of the pipeline are predicted according to the described transportation plan, and the prediction results are compared with the steam demand of the target enterprise to optimize the steam supply parameters until the target demand is met. Specifically, this includes: obtaining the pipeline length, diameter, insulation material parameters, and pipeline roughness according to the described transportation path; calculating pressure loss and heat loss; obtaining the predicted steam parameters at the end of the pipeline based on the calculated pressure loss and heat loss; comparing the predicted steam parameters with the steam demand of the target enterprise, and if the pressure error or temperature error is greater than the threshold, then initiating the gradient descent method to optimize the steam supply parameters. Continuously monitor the real-time parameters of the target nodes in the pipeline network, compare the real-time parameters with the predicted parameters, and when the deviation exceeds the preset threshold, it is judged as an abnormal transport status, and corresponding adjustment measures are taken. Based on the steam pressure change data obtained from the regulation measures, a pressure regulation optimization model is constructed, and the preset multi-level pressure regulation method is adjusted. Specifically, this includes: obtaining real-time pressure data of each node in the steam pipeline network; performing data cleaning and normalization on the real-time pressure data to obtain a standardized pressure change dataset; using dynamic time regularization distance as a similarity metric, performing hierarchical clustering on the standardized pressure change dataset to divide the pressure change data into multiple levels, obtaining a multi-level pressure change dataset; and constructing a long short-term memory network as the pressure regulation optimization model based on the multi-level pressure change dataset. When there are changes in enterprises within the park, the topology and transmission routes of the local pipeline network are replanned and incorporated into the transmission scheduling system for unified management. Specifically, this includes: acquiring geographic information data of the park's steam pipeline network; establishing a digital twin model based on the geographic information data; calculating the optimal transmission route and pressure distribution scheme using the Dijkstra algorithm; receiving access signals from new enterprises; replanning the local pipeline network structure using the Steiner tree algorithm based on the access signals from new enterprises; acquiring steam demand data from new enterprises; and predicting future steam consumption using the ARIMA model.

2. The multi-stage pressure-regulated steam transport network and its optimization method as described in claim 1, characterized in that, The acquisition of steam monitoring values ​​for target nodes in the park's steam pipeline network includes: Acquire pressure, temperature, and flow data of the target node; Based on the pressure, temperature, and flow data of the target node, calculate the average value and standard deviation of each sensor in the most recent time period. If the data point deviates from the average value by more than a preset multiple of the standard deviation, it is judged as abnormal data. An interpolation method was used to correct the abnormal data, resulting in corrected steam monitoring values.

3. The multi-stage pressure-regulated steam transport network and its optimization method as described in claim 1, characterized in that, The system continuously monitors the real-time parameters of target nodes within the pipeline network, compares these real-time parameters with predicted parameters, and determines an abnormal transport status when the deviation exceeds a preset threshold, taking corresponding adjustment measures, including: Collect pressure, temperature, and flow data at the target node; The pressure data, temperature data, and flow data of the target node are losslessly compressed to obtain compressed data. The compressed data is compared with the pre-stored prediction parameters, and the average deviation value is calculated using the sliding window method to obtain the pressure deviation, temperature deviation, and flow deviation. If at least one of the pressure deviation, temperature deviation, and flow deviation exceeds a preset threshold, an early warning mechanism is triggered.

Images