Fire power plant pipeline pressure monitoring method and system based on digital twinning
By generating continuous pressure cloud maps using digital twin technology and combining them with historical time-series data, the problem of difficulty in anomaly identification caused by the limited number of sensor points in pipeline pressure monitoring of thermal power plants has been solved, enabling accurate location and real-time monitoring of local anomalies.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NORTHWEST THIRD ELECTRIC POWER CONSTR CO LTD
- Filing Date
- 2026-04-24
- Publication Date
- 2026-07-21
AI Technical Summary
Existing methods for monitoring pipeline pressure in thermal power plants rely on a limited number of point sensors, making it difficult to accurately identify abnormal situations, especially when there are regional pressure anomalies between sensors.
Digital twin technology is used to generate continuous pressure cloud maps. The data is divided into sub-regions by interpolation algorithm. The product of internal consistency coefficient and spatial smoothness coefficient is calculated as the smoothness. Combined with historical time series data, the degree of anomaly is judged, so as to achieve accurate positioning and monitoring of the pressure inspection area.
This has enabled an upgrade from point monitoring to area monitoring, significantly improving the accuracy of identifying local anomalies in pipelines, reducing false alarms, and enhancing the real-time performance and accuracy of monitoring.
Smart Images
Figure CN122084190B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing. More specifically, this invention relates to a method and system for monitoring pipeline pressure in thermal power plants based on digital twins. Background Technology
[0002] As the core hub of energy supply, thermal power plants' pipeline systems (including main steam pipelines and reheat steam pipelines) are constantly subjected to high temperature, high pressure, and frequent peak-shaving conditions, making them prone to safety accidents such as stress concentration, fatigue cracks, and even ruptures. Therefore, real-time monitoring of pipeline pressure is of great significance.
[0003] Currently, the common method for monitoring pipeline pressure in thermal power plants is to install multiple pressure sensors at key points in the pipeline (such as elbows, welds, and tees), and then use the pressure data collected by each sensor to determine the pipeline's operating status. This method can obtain pressure values at each installation point, providing data support for pipeline safety assessments.
[0004] However, due to limitations such as installation conditions, sensors are mostly deployed at limited locations, and the data collected by each sensor is independent and presented in isolation. When regional pressure anomalies occur in the pipeline (such as distortion of pressure distribution in a certain pipe section), a single sensor may not be able to capture the abnormal features, making it difficult to accurately identify the anomaly. Summary of the Invention
[0005] To address the limitations of existing monitoring methods, which make it difficult to accurately identify abnormal situations, this invention provides solutions in the following aspects.
[0006] In the first aspect, a digital twin-based method for monitoring pipeline pressure in thermal power plants includes: Real-time acquisition of pipeline pressure data from thermal power plants, followed by preprocessing of the acquired data to obtain pressure data sequences; Pressure cloud maps are generated at each sampling time based on the pressure data sequence through interpolation; the pressure cloud maps are divided into several sub-regions; for each sub-region, the internal consistency coefficient and the spatial smoothness coefficient are calculated, and the product of the internal consistency coefficient and the spatial smoothness coefficient is used as the smoothness of the sub-region. Based on the smoothness, the pressure inspection area is selected from all sub-regions, and the degree of abnormality of the pressure inspection area is calculated; Pipeline pressure monitoring is performed based on the degree of abnormality in the pressure testing area.
[0007] Optionally, after generating the pressure contour map, the following steps are also included: Calculate the pressure variance of all pressure data in the pressure contour plot; The reciprocal of the pressure variance of all pressure data in the pressure cloud map is used as the overall smoothness of the pressure cloud map.
[0008] Optionally, the calculation of the internal consistency coefficient includes: Calculate the square of the difference between each pressure data point in the sub-region and the mean of all pressure data points in the sub-region. Iterate through all pressure data points in the sub-region, multiply the square of each calculated difference by the overall smoothness of the target cloud map obtained above, sum the results of the multiplication, and then divide by the total number of pressure data points in the sub-region to obtain the weighted dispersion of the sub-region. By using the weighted dispersion as the exponent of the exponential function, the negative exponential function value with the natural constant e as the base is calculated to obtain the internal consistency coefficient of the sub-region.
[0009] Optionally, the calculation of the spatial smoothing coefficient includes: Calculate the gradient value of each pressure data point within the sub-region; The reciprocal of the sum of the gradient values of all pressure data within a subregion is used as the spatial smoothing coefficient of the subregion.
[0010] Optionally, the step of filtering the pressure-to-inspection area from all sub-regions based on smoothness includes: A preset screening threshold is used to mark sub-regions with a smoothness level less than the screening threshold as pressure testing areas.
[0011] Optionally, the calculation of the degree of abnormality includes: Obtain several historical sampling times prior to the current sampling time to determine a historical window; obtain the indexes of the current sampling time and all historical sampling times within the historical window; Calculate the index difference between the current sampling time and any historical sampling time within the historical window, and use the reciprocal of the index difference as the temporal weight coefficient of that historical sampling time; Calculate the absolute value of the difference between the pressure at the current sampling time and the pressure at the historical sampling time for each pressure data in the pressure testing area, and then normalize it before averaging it across all pressure data in the pressure testing area to obtain the average pressure difference at that historical sampling time. The weighted difference is obtained by multiplying the time-series weighting coefficient by the average pressure difference. The average of the weighted differences over all historical sampling times is then calculated to obtain the degree of anomaly in the pressure testing area.
[0012] Optionally, the step of performing pipeline pressure monitoring based on the degree of anomaly in the pressure-to-be-inspected area includes: A preset abnormal threshold is set. For each pressure inspection area, if the corresponding abnormality level is greater than the abnormal threshold, it is determined that there is a pressure abnormality in the pressure inspection area, it is marked as a pressure abnormality area, and an alarm record containing location information is generated.
[0013] Secondly, a digital twin-based pipeline pressure monitoring system for thermal power plants includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the digital twin-based pipeline pressure monitoring method for thermal power plants described in any one of the claims is implemented.
[0014] The present invention has the following beneficial effects: 1. Traditional methods rely on a limited number of point sensors. When an anomaly occurs between two sensors, the data from a single sensor may not show any anomalies, leading to missed detections. This invention uses spatial interpolation to generate a continuous pressure cloud map from discrete pressure data and divides the cloud map into several sub-regions. By calculating the product of the internal consistency coefficient and the spatial smoothness coefficient of each sub-region as the smoothness level, it can directly locate local areas where pressure gradients change abruptly, achieving a substantial improvement from "point monitoring" to "area monitoring".
[0015] 2. Valves, elbows, and other components naturally exhibit significant pressure gradients under normal operating conditions. Relying solely on spatial unevenness can easily lead to false alarms. This invention, after identifying the pressure areas to be inspected, further incorporates historical time-series data: calculating the pressure difference between the current sampling time and each historical sampling time, and then weighting the time intervals for a weighted average to determine the degree of anomaly. The degree of anomaly only increases when the current pressure distribution significantly deviates from its historical normal state, while long-standing high-gradient areas are not misjudged, significantly improving monitoring accuracy. Attached Figure Description
[0016] Figure 1 This is a flowchart of steps S1-S3 in the digital twin-based pipeline pressure monitoring method for thermal power plants according to an embodiment of the present invention. Detailed Implementation
[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.
[0018] Reference Figure 1 The method for monitoring pipeline pressure in thermal power plants based on digital twins includes steps S1-S3, as detailed below: S1: Real-time acquisition of pipeline pressure data from thermal power plants, and preprocessing of the acquired data.
[0019] Firstly, high-precision pressure sensors are installed in stress concentration areas such as elbows, welds, tees, and valve inlets / outlets in the pipeline system of thermal power plants. Additionally, one pressure sensor can be installed every ten meters along straight pipe sections. For straight pipe sections less than 10 meters long, one pressure sensor is installed at the midpoint. Data from each pressure sensor is collected in real time at a sampling frequency of 10Hz to obtain the time-series pressure data for each sensor.
[0020] All the collected pipeline pressure data are preprocessed to remove obvious outliers and imputate missing values. For example, pipeline pressure data that exceeds the pressure sensor's range or exceeds the pipeline's design pressure by 1.5 times are removed.
[0021] Through the above deployment and acquisition operations, a discrete sensor pressure data sequence with timestamps is obtained.
[0022] S2: Interpolate the pipeline pressure data to obtain a pressure cloud map, divide the pressure cloud map into regions to obtain several sub-regions; calculate the smoothness of each sub-region, filter out the pressure inspection areas based on the smoothness, and calculate the anomaly degree of the pressure inspection areas.
[0023] The discrete pressure data obtained in S1 above can only reflect pressure data at a limited number of points, and cannot directly present the continuous pressure distribution of the entire pipeline, nor can it identify local anomalies from a spatial distribution perspective. Therefore, it is necessary to further convert the discrete pressure data into a continuous pressure field, so as to extract quantitative indicators reflecting the degree of local anomalies from the pressure field.
[0024] First, a continuous pressure field is generated using the existing Kriging interpolation algorithm, exemplarily. For example, a spherical model is selected as the variogram model for Kriging interpolation, and the nugget effect is set to 0.01.
[0025] The pressure field is mapped as a blue-green-red gradient: blue represents pressure ≤10MPa, green represents 10 MPa -25 MPa, and red represents ≥25 MPa. This gradient is then overlaid onto a pre-built 3D pipeline model to obtain a pressure contour map. For example, a 3D model of the pipeline system is created based on the system's design drawings, 3D laser scanning data, or historical BIM (Building Information Modeling). This model includes information on the pipeline's direction, diameter, wall thickness, bend angles, tee locations, and valve layout.
[0026] Next, the pressure cloud map is analyzed. Local areas of drastic pressure changes in the map may correspond to pipeline stress concentration or abnormal events, but the degree of "drasticness" needs to be quantified and compared with the overall picture. Therefore, the pressure cloud map is tiled; for example, it is divided into K sub-regions, each with a size of 21×21 pressure data points. If the size of the pressure cloud map at the boundary is less than 21×21, the insufficient portion is discarded (without padding) to ensure that each sub-region has complete statistical significance.
[0027] Furthermore, the pressure cloud map generated at any sampling time is selected as the target cloud map. The pressure variance of all pressure data in the target cloud map is calculated. The reciprocal of the pressure variance of all pressure data in the target cloud map is taken as the overall smoothness of the target cloud map. The larger the calculated pressure variance, the smaller the overall smoothness of the target cloud map, indicating that the overall pressure distribution of the target cloud map is more uneven.
[0028] Furthermore, if a pressure anomaly exists in a sub-region of the aforementioned target cloud map, it is typically manifested as drastic spatial variations (large gradient) in the pressure values within that sub-region, and a significant deviation of each pressure value from the sub-region's mean. Therefore, it is necessary to construct an index that simultaneously considers both "spatial gradient" and "internal discrepancy" to quantify the smoothness of the sub-region.
[0029] The index of "internal discrepancy" is defined as the internal consistency coefficient. The calculation process of the internal consistency coefficient of a sub-region is as follows: First, calculate the square of the difference between each pressure data point within a sub-region and the mean of all pressure data points within that sub-region. Then, iterate through all pressure data points within the sub-region, multiplying the square of each calculated difference by the overall smoothness of the target cloud map obtained above. Summate the multiplication results and divide by the total number of pressure data points within the sub-region to obtain the weighted dispersion of the sub-region. Finally, use the weighted dispersion as the exponent of an exponential function to calculate the negative exponential function value with the natural constant e as the base, obtaining the internal consistency coefficient of the sub-region. The larger the weighted dispersion, the smaller the internal consistency coefficient.
[0030] When all pressure data within a subregion have the same value, that is, the difference between each pressure data point within the subregion and the mean of all pressure data within the subregion is 0, the internal consistency coefficient of the subregion is 1.
[0031] The index of "spatial gradient" is defined as the spatial smoothness coefficient. The calculation process of the spatial smoothness coefficient of a sub-region is as follows: For each pressure data point within a sub-region, the sum of the absolute values of the pressure differences between it and its adjacent pressure data points is calculated as the gradient value of the corresponding pressure data. The spatial smoothing coefficient of the sub-region is obtained by summing the gradient values of all pressure data and taking the negative exponent (or by summing the gradient values of all pressure data and taking the reciprocal). A larger sum of gradients indicates more drastic spatial pressure changes, and therefore a smaller spatial smoothing coefficient.
[0032] Multiplying the internal consistency coefficient and spatial smoothness coefficient obtained above, we get the smoothness of the sub-region. When the spatial variation is zero and the interior is completely uniform (i.e., the gradient sum is 0 and the weighted dispersion is 0), both are 1, and the smoothness of the sub-region is 1, indicating the smoothest. When the spatial variation is drastic or the internal dispersion is large, the smoothness of the sub-region approaches 0, indicating the least smooth.
[0033] The above calculations yield the smoothness of all sub-regions in the target cloud map. A filtering threshold is then set, for example, to 0.72 (this value can be adjusted between 0.6 and 0.8 depending on actual operating conditions). Sub-regions with smoothness levels less than the filtering threshold are marked as pressure-inspection areas. These pressure-inspection areas are the initially screened sub-regions that may exhibit pressure anomalies.
[0034] Then, the selected pressure inspection areas are further analyzed. The presence of pressure inspection areas only indicates a lack of smoothness in spatial distribution; it doesn't necessarily confirm a true pressure anomaly, as some areas may already exhibit significant pressure gradients under normal operating conditions (such as valves and elbows). Therefore, further judgment is needed by combining historical data: if the current pressure distribution of a pressure inspection area differs significantly from its historical normal pressure distribution, then that area is more likely to be a true anomaly. Based on this, the degree of anomaly for each pressure inspection area is calculated by comparing the pressure contour maps at the current and historical times.
[0035] The calculation process for the degree of abnormality in the pressure testing area mentioned above is as follows: First, a historical window is defined: the most recent L historical sampling times before the current sampling time are taken, where L is a preset positive integer, and in this embodiment, it is set to 10. Let the index of the current sampling time be t, and the index of any historical sampling time within the historical window be s, where s = t-1, t-2, ..., tL. Calculate the index difference between the current sampling time and any historical sampling time before the current sampling time. Then, use this index difference as the exponent of an exponential function and calculate the negative exponential function value with the natural constant e as the base. The result is used as the time-series weight coefficient for that historical sampling time. Alternatively, the reciprocal of the index difference between these two sampling times can be taken as the time-series weight coefficient for that historical sampling time. Since the index difference is a discrete positive integer and its minimum value is 1, the range of values for the time-series weight coefficient is... The larger the interval, the smaller the time-series weighting coefficient.
[0036] Then, the absolute value of the difference between the pressure at the current sampling time of each pressure data in the pressure test area and the pressure at any historical sampling time in the historical window is calculated, divided by the pressure range at the current time in the pressure test area for normalization, and then the average of all pressure data in the pressure test area is calculated to obtain the average pressure difference at the historical sampling time.
[0037] Finally, the calculated time-series weighting coefficient is multiplied by the average pressure difference to obtain the weighted difference at that historical sampling time. Then, the average of the weighted differences over all historical sampling times is calculated to obtain the degree of anomaly in the pressure area under investigation. The greater the degree of anomaly, the more significant the deviation of the current pressure from the historical normal state.
[0038] The above operations can be used to obtain the degree of abnormality in all pressure testing areas.
[0039] S3: Complete pipeline pressure monitoring based on the degree of abnormality in the pressure inspection area.
[0040] The degree of abnormality for each pressure testing area calculated by S2 above is a continuous quantified value, while maintenance personnel need clear abnormal location and alarm signals. Therefore, these quantified values need to be compared with preset abnormal thresholds, and areas exceeding the abnormal thresholds are identified as pressure abnormal areas, with the monitoring results output externally.
[0041] In this embodiment of the invention, the preset abnormal threshold value range is 0.8-0.95, and an exemplary value of 0.9 is used. For each pressure inspection area, if its abnormality level is greater than the abnormal threshold, it is determined that there is a pressure abnormality in the pressure inspection area, it is marked as a pressure abnormality area, and an alarm record containing location information is generated. This abnormal threshold can be adjusted according to the actual operating conditions of the pipeline: the smaller the threshold, the more sensitive the judgment, and the more alarms may be generated; the larger the threshold, the stricter the judgment, and only alarms are generated for significant abnormalities.
[0042] The system updates the monitoring status in real time. Once an area with abnormal pressure is detected, it immediately notifies the maintenance personnel and recommends that they go to the site to investigate the cause (such as localized corrosion, micro-cracks in welds, water hammer impact, etc.) and repair it in a timely manner.
[0043] The system includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement the digital twin-based pipeline pressure monitoring method for thermal power plants according to the first aspect of the present invention.
[0044] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.
[0045] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.
Claims
1. A method for monitoring pipeline pressure in thermal power plants based on digital twins, characterized in that, include: Real-time acquisition of pipeline pressure data from thermal power plants, followed by preprocessing of the acquired data to obtain pressure data sequences; Pressure cloud maps at each sampling time are generated by interpolation based on the pressure data sequence. After generating the pressure cloud map, the following steps are also included: Calculate the pressure variance of all pressure data in the pressure contour plot; The reciprocal of the pressure variance of all pressure data in the pressure cloud map is used as the overall smoothness of the pressure cloud map; the pressure cloud map is divided into several sub-regions; for each sub-region, an internal consistency coefficient and a spatial smoothness coefficient are calculated, and the product of the internal consistency coefficient and the spatial smoothness coefficient is used as the smoothness of the sub-region; the calculation of the internal consistency coefficient includes: Calculate the square of the difference between each pressure data point in the sub-region and the mean of all pressure data points in the sub-region. Iterate through all pressure data points in the sub-region, multiply the square of each calculated difference by the overall smoothness of the target cloud map obtained above, sum the results of the multiplication, and then divide by the total number of pressure data points in the sub-region to obtain the weighted dispersion of the sub-region. Using the weighted dispersion as the exponent of an exponential function, the negative exponential function value with the natural constant e as the base is calculated to obtain the internal consistency coefficient of the sub-region; the calculation of the spatial smoothness coefficient includes: For each pressure data point within a sub-region, the sum of the absolute values of the pressure differences between it and its adjacent pressure data points is calculated as the gradient value of the corresponding pressure data. The reciprocal of the sum of the gradient values of all pressure data within a sub-region is used as the spatial smoothing coefficient of the sub-region. Pressure inspection areas are selected from all sub-regions based on smoothness, and the degree of anomaly in these areas is calculated. The calculation of the anomaly degree includes: Obtain several historical sampling times prior to the current sampling time to determine a historical window; obtain the indexes of the current sampling time and all historical sampling times within the historical window; Calculate the index difference between the current sampling time and any historical sampling time within the historical window, and use the reciprocal of the index difference as the temporal weight coefficient of that historical sampling time; Calculate the absolute value of the difference between the pressure at the current sampling time and the pressure at the historical sampling time for each pressure data in the pressure testing area, and then normalize it before averaging it across all pressure data in the pressure testing area to obtain the average pressure difference at that historical sampling time. The weighted difference is obtained by multiplying the time-series weighting coefficient by the average pressure difference. The average of the weighted differences over all historical sampling times is then calculated to obtain the degree of anomaly in the pressure testing area. Pipeline pressure monitoring is performed based on the degree of abnormality in the pressure testing area.
2. The method for monitoring pipeline pressure in a thermal power plant based on digital twins according to claim 1, characterized in that, The step of filtering the pressure-to-inspection area from all sub-regions based on smoothness includes: A preset screening threshold is used to mark sub-regions with a smoothness level less than the screening threshold as pressure testing areas.
3. The method for monitoring pipeline pressure in a thermal power plant based on digital twins according to claim 1, characterized in that, The process of monitoring pipeline pressure based on the degree of abnormality in the pressure testing area includes: A preset abnormal threshold is set. For each pressure inspection area, if the corresponding abnormality level is greater than the abnormal threshold, it is determined that there is a pressure abnormality in the pressure inspection area, it is marked as a pressure abnormality area, and an alarm record containing location information is generated.
4. A pipeline pressure monitoring system for thermal power plants based on digital twins, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement the method for monitoring pipeline pressure in a thermal power plant based on digital twins according to any one of claims 1-3.