Abnormal pressure detection and correction method for pressure sensing device
By correcting abnormal pressures in the pressure sensor of the closed-loop tank using the sliding window method and secondary detection method, the problem of misjudgment caused by external interference was solved, and accurate detection and correction of the working status of the closed-loop tank was achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHANJIANG BRANCH OF CHINA NATIONAL OFFSHORE OIL CORP
- Filing Date
- 2023-08-14
- Publication Date
- 2026-06-23
AI Technical Summary
In traditional methods, the pressure change curve of the pressure sensor in a closed-loop tank is easily affected by external factors, leading to misjudgment of the tank's working status. Existing technologies cannot accurately distinguish between pressure surges caused by external interference and abnormalities inside the tank.
The sliding window method and the secondary detection method are adopted. By setting the sliding time window and quantile interval, suspected outliers are initially detected, and the outliers are confirmed by short-term sequences and difference thresholds to ensure the accuracy of detection.
It effectively corrects abnormal pressure values caused by external interference, improves the detection accuracy of pressure sensing equipment, avoids misjudgment, and ensures accurate judgment of the working status of closed-loop tanks.
Smart Images
Figure CN117109805B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of offshore oil exploration and drilling engineering technology, and more specifically, to a method for detecting and correcting abnormal pressure in a pressure sensing device. Background Technology
[0002] The discharge system of offshore oil and gas fixed platforms is an essential auxiliary system for ensuring the safe production of oil and gas on the platform. This system primarily collects and treats fluids discharged from production equipment, as well as rainwater and oily wastewater from the deck. Offshore platform discharge systems are divided into open discharge and closed discharge systems. The closed discharge system collects and treats pressurized gas-liquid mixtures discharged from production and utility systems. A complete closed discharge system typically consists of a pressure relief valve, a closed discharge collection and manifold, a separator, a closed discharge tank (i.e., a closed discharge tank), and a closed discharge pump. The closed discharge tank is the key equipment in the closed discharge system; its main function is to separate the liquid within the tank from the gas, liquid, and oil. Its proper functioning is crucial to the safe production of oil and gas on the offshore platform. The pressure value of the closed discharge tank is one of the important indicators for measuring its normal operation; this pressure value is usually collected by the closed discharge tank pressure sensor.
[0003] When a closed-loop drainage tank is operating normally, its pressure change curve generally shows a stable state. When the tank is not operating normally, its pressure change curve generally shows a slow upward or downward trend. Traditionally, the normal operation of a closed-loop drainage tank is determined by manually observing its pressure change curve. However, due to a series of environmental factors, such as strong winds at sea, sudden start-up and shutdown of the closed-loop system, personnel movement on the platform, vibrations caused by the operation of other equipment and machinery, and electromagnetic and power frequency interference from the tank itself, the pressure data of the closed-loop drainage tank may experience abrupt changes at certain times. This can cause the pressure change curve to rise or fall sharply within a certain period. Due to subjective human factors, some staff may mistakenly attribute pressure changes caused by external interference to abnormalities within the closed-loop drainage tank itself, leading to misjudgments of its operating status. Summary of the Invention
[0004] To address the potential for misjudgment in existing technologies that rely on manual assessment of abnormal pressure data in closed-loop tanks, this invention provides a method for detecting and correcting abnormal pressure in pressure sensing devices. This method can detect abnormal pressure values caused by external interference and correct them to normal values, thus avoiding misjudgments of the operating status of pressure sensing devices due to external interference.
[0005] To solve the above-mentioned technical problems, the technical solution provided by the present invention is as follows:
[0006] A method for detecting and correcting abnormal pressure in a pressure sensing device includes the following steps:
[0007] Step S1: Receive pressure and time data, convert each pressure value into a time series format, and convert the time corresponding to each pressure value into a timestamp index format. Use the time corresponding to the pressure value as the time index value of the pressure value; that is, the index value of each pressure value is the time value corresponding to the pressure.
[0008] Step S2: Set the data processing step size, and then segment the closed-loop tank pressure data to obtain multiple time series data segments. Each time series data segment includes n time data points. Since the pressure data of offshore closed-loop tanks does not have a fixed sampling rate, and data is collected on average every few seconds to tens of seconds, in order to facilitate data management, the data processing unit should be the number of data points, for example, 100 time data points of the entered data as a time step, rather than the unit of time, for example, 10 seconds as a time step.
[0009] Step S3: Use the sliding window method to perform preliminary detection on each time series data segment, set the window size and quantile interval of the sliding time window, and put the time data points into the sliding time window according to the window size of the sliding time window; the window size of the sliding time window determines the number of time data points in the sliding time window;
[0010] Step S4: Calculate the first median value of the pressure values corresponding to the time data points contained in the sliding time window; take the pressure values corresponding to the time data points after the sliding time window as the current values, calculate the difference between the current values and the first median value, and slide the sliding time window to obtain the difference sequence x1; for example, use a sliding time window with a window size of 10 time data points to slide on each time series data. The current value after the sliding time window refers to the value of the time point immediately following the sliding time window. If the first time window contains the 1st to 10th time data points in the current time series data segment, then the current value is the 11th time data point in the current time series data segment. If the next sliding time window contains the 2nd to 11th time data points in the current time series data segment, then the current value is the 12th time data point in the current time series data segment, and so on. s1 = [the difference between the 11th time data point and the median value in the 1st time window, the difference between the 12th time data point and the median value in the 2nd time window, ...];
[0011] Step S5: Set the normal range interval for the current time series data segment, compare the difference sequence x1 with the normal range interval, and determine whether all values in the difference sequence x1 fall within the normal range interval. If yes, end the process; otherwise, determine the pressure value corresponding to the value outside the normal range interval as a suspected outlier. That is, the value outside the normal range interval is the difference between the suspected outlier and the first median value of the corresponding sliding time window, and the time data point corresponding to the suspected outlier is a suspected outlier point. Then, execute the next step to perform a second detection on the suspected outlier.
[0012] Step S6: Select 'a' consecutive time data points from the time series data segment as short-term series x2, and determine the maximum and minimum pressure values corresponding to the time data points in short-term series x2; determine whether the suspected outlier is equal to the maximum or minimum pressure value corresponding to the time data points in short-term series x2. If yes, the suspected outlier is still a suspected outlier, and proceed to the next step; if no, the suspected outlier is a normal value, and the process ends.
[0013] Step S7: Set a maximum difference threshold, determine the second median value of the pressure value corresponding to the time data point in the short-term series x2, and determine whether the difference between the suspected outlier and the second median value is greater than or equal to the maximum difference threshold. If so, the suspected outlier is a true outlier, and the suspected outlier is assigned the second median value to correct it to a normal value, thus ensuring the smoothness and continuity of the short-term series. If not, the suspected outlier is a normal value, and the process ends. After the pressure outlier is detected a second time through steps S6 and S7, the entire detection and correction process is complete. Finally, the pressure values corresponding to each time data point in each time series data segment are spliced together to obtain the corrected closed-loop tank pressure data.
[0014] Preferably, in step S1, the timestamp index is accurate to the year, month, day, hour, minute, and second, with the hour recorded in 24-hour format. This ensures that the time data point corresponding to each pressure value is unique and there will be no duplicate time data.
[0015] Preferably, in step S2, the data is segmented by iterating through the data, and the specific process is as follows:
[0016] The first data point of the pressure data is taken, up to the nth data point, forming the first data segment, where n represents the data processing step size. Then, the (n+1)th data point of the original pressure data is taken, up to the 2nth data point, forming the second data segment, and so on. Specifically:
[0017] Time series data segment 1 = {the 1st time data point, the 2nd time data point, ..., the nth time data point};
[0018] Time series data segment 2 = {the (n+1)th time data point, the (n+2)th time data point, ..., the 2nth time data point};
[0019] The time series data segment k = {the (k-1)*n+1th time data point, the (k-1)*n+2th time data point, ..., the k*nth time data point}.
[0020] Preferably, in step S2, data that is insufficient to form a time series data segment is discarded and not processed, so that the data processing in subsequent steps can be carried out in an orderly and effective manner.
[0021] Preferably, in step S2, after segmenting the closed-loop tank pressure data, the validity of the time data points within the time series data segment is checked to determine that each time data point in the time series data segment is arranged in chronological order, that is, the earlier time data points are placed at the beginning and the later time data points are placed at the end, thereby improving the accuracy of outlier detection.
[0022] Preferably, in step S3, the window size and quantile multiples of the sliding time window are fixed values, meaning they are manually set. Generally, if the automatically set sliding time window size and quantile multiples are not properly configured, outliers may be missed during actual detection (outliers exist but are not detected), or misjudged (normal values are identified as outliers). Therefore, to improve the accuracy of outlier detection, the window size and quantile multiples of the sliding time window need to be set manually and reasonably, so that in step S5, more misjudgments are preferable to fewer missed ones. Even if there are many misjudged outliers in step S5, the misjudged data points can be identified and corrected in steps S6 and S7, thus comprehensively improving the accuracy of outlier identification.
[0023] Preferably, in step S4, the difference between the current value and the first median value is the absolute value. This way, the suspected outliers detected in subsequent steps could be outliers causing a sudden increase in pressure or outliers causing a sudden decrease in pressure. Therefore, abnormal upward and downward pressure trends can be detected, making the detection results more intuitive and accurate.
[0024] Preferably, in step S5, the normal range interval is [Q1-c*IQR, Q3+c*IQR], where Q1 and Q3 are the lower quartile and upper quartile of the pressure value corresponding to the time data point in the current time series data segment, respectively. That is, after arranging the pressure values corresponding to all time data points in the current time series data segment from smallest to largest, the pressure values at the 25th and 75th quartiles are the pressure values. c is the quartile distance of the sliding time window in step S3, and IQR is the interquartile distance, which has the value IQR = Q3-Q1.
[0025] Preferably, in step S6, the value of 'a' is 5, meaning each short-term sequence x2 contains 5 time data points. Selecting an odd number of time data points for the short-term sequence x2 facilitates the determination of the second median value, ensuring that the second median value corresponds to the pressure value of a certain time data point in the short-term sequence x2. If the short-term sequence x2 contains three time data points, it is difficult to identify whether the pressure value corresponding to that point is a true outlier due to the small number of time data points; if the number of time data points is too large, it will cause data waiting delays. Therefore, to quickly and accurately determine true outliers, it is more appropriate for each short-term sequence x2 to contain 5 time data points.
[0026] Preferably, in step S6, the three time data points preceding the suspected anomaly, the suspected anomaly itself, and the one time data point following the suspected anomaly in the time series data segment are selected as the short-term sequence x2; when the suspected anomaly is among the top three positions in the time series data segment, the first five time data points in the time series data segment are selected as the short-term sequence x2; when the suspected anomaly is among the last positions in the time series data segment, the last five time data points in the time series data segment are selected as the short-term sequence x2.
[0027] In practical applications, in order to more accurately apply this invention to offshore oil and gas production sites, the pressure data used may vary depending on the different environments or equipment in different regions. This comprehensive detection method can be applied to new equipment to detect and correct outliers in closed-loop tank equipment (including other sensor equipment) on site.
[0028] The beneficial effects of this invention are as follows: This method applies anomaly detection and correction to sensor equipment on offshore oil production platforms, including but not limited to closed-loop tanks, and can detect and correct abnormal pressure values caused by external interference, thereby avoiding misjudgments of the cause of abnormal pressure due to external factors. This method is easy to automate, adopts a combination of preliminary detection and secondary detection, reduces the false negative rate of anomalies by setting reasonable parameters, and reduces the false positive rate of anomalies by using secondary detection judgment, which can greatly improve the accuracy of anomaly detection. Attached Figure Description
[0029] Figure 1 This is a flowchart of an abnormal pressure detection and correction method for a pressure sensing device;
[0030] Figure 2 It is a section of the original pressure change curve of a closed-loop tank on an oilfield production platform;
[0031] Figure 3 This is a preliminary test result using an abnormal pressure detection and correction method based on a pressure sensing device.
[0032] Figure 4 This is a corrected pressure change curve. Detailed Implementation
[0033] The technical solution of the present invention will be further described in detail below through specific embodiments and in conjunction with the accompanying drawings:
[0034] Example 1
[0035] like Figure 1 The method for detecting and correcting abnormal pressure in a pressure sensing device, which uses pressure data from a closed-loop tank pressure sensor as the processing object, includes the following steps:
[0036] Step S1: Receive pressure and time data from the closed-loop tank, convert each pressure value into a time series format, and convert the time corresponding to each pressure value into a timestamp index format. Use the time corresponding to the pressure value as the time index value of the pressure value; that is, the index value of each pressure value is the time value corresponding to the pressure being generated. In this embodiment, there are a total of 25,000 pressure values.
[0037] Step S2: Set the data processing step size, then segment the closed-loop tank pressure data to obtain multiple time series data segments, each time series data segment including n time data points; in this embodiment, the data processing step size is 100 points, and each time series data segment includes 100 time data points. Since the pressure data of offshore closed-loop tanks does not have a fixed sampling rate, and data is collected on average every few seconds to tens of seconds, in order to facilitate data management, the data processing unit needs to be the number of data points, for example, using 100 time data points of the entered data as a time step, rather than using a unit of time, for example, using 10 seconds as a time step;
[0038] Step S3: Use the sliding window method to perform preliminary detection on each time series data segment. Set the window size and quantile interval of the sliding time window, and put the time data points into the sliding time window according to the window size of the sliding time window; the window size of the sliding time window determines the number of time data points in the sliding time window.
[0039] Step S4: Calculate the first median value of the pressure values corresponding to the time data points contained within the sliding time window; take the pressure values corresponding to the time data points after the sliding time window as the current values, calculate the difference between the current values and the first median value, slide the sliding time window to obtain the difference sequence x1; for example, use a sliding time window with a window size of 10 time data points to slide over each time series data. The current value after the sliding time window refers to the value of the time point immediately following the sliding time window. If the first time window contains the 1st to 10th time data points in the current time series data segment, then the current value is the 11th time data point in the current time series data segment. If the next sliding time window contains the 2nd to 11th time data points in the current time series data segment, then the current value is the 12th time data point in the current time series data segment, and so on. s1 = [the difference between the 11th time data point and the median value in the 1st time window, the difference between the 12th time data point and the median value in the 2nd time window, ...];
[0040] Step S5: Set the normal range interval for the current time series data segment, compare the difference sequence x1 with the normal range interval, and determine whether all values in the difference sequence x1 fall within the normal range interval. If yes, end the process; otherwise, determine the pressure values corresponding to the values outside the normal range interval as suspected outliers. That is, the values outside the normal range interval are the difference between the suspected outlier and the first median value of the corresponding sliding time window. The time data point corresponding to the suspected outlier is the suspected outlier point. Then, proceed to the next step to perform a second detection on the suspected outliers.
[0041] Step S6: Select 'a' consecutive time data points in the time series data segment as the short-term series x2, and determine the maximum and minimum values of the pressure values corresponding to the time data points in the short-term series x2; determine whether the suspected outlier is equal to the maximum or minimum value of the pressure values corresponding to the time data points in the short-term series x2. If yes, it means that the suspected outlier is still a suspected outlier, and proceed to the next step; if no, it means that the suspected outlier is a normal value, and end the process.
[0042] Step S7: Set the maximum difference threshold, which is 0.5 in this embodiment; determine the second median value of the pressure value corresponding to the time data point in the short-term series x2; determine whether the difference between the suspected outlier and the second median value is greater than or equal to the maximum difference threshold of 0.5. If so, it means that the suspected outlier is a real outlier, and the suspected outlier is assigned the second median value to correct the suspected outlier to a normal value, so as to ensure the smoothness and continuity of the short-term series. If not, it means that the suspected outlier is a normal value, and the process ends. After the pressure outlier is detected twice in steps S6 and S7, the entire detection and correction process is completed. Finally, the pressure values corresponding to each time data point in each time series data segment are spliced together to obtain the corrected closed-loop tank pressure data.
[0043] It should be noted that the first median and the second median are actually both median values, just from different sources. The first and second are used only to distinguish between the two and have no other special meaning.
[0044] Verification of the implementation results of this embodiment: Appendix Figure 2 This is a section of the original pressure change curve of a closed-loop tank on an oilfield production platform, from... Figure 2 The sudden drop in pressure around 23:02:05 indicates the presence of a genuine anomaly at that point. (Attached) Figure 3 This is a detection result diagram after the preliminary detection in this invention. Figure 3 As can be seen, the initial detection classified both true outliers and some normal values as suspected outliers. Figure 3 The pressure values corresponding to the points traversed by the four light-colored vertical lines are all suspected anomalies. After a second test, Figure 3 After the first suspected outlier in the graph was determined to be a true outlier, it was corrected. Furthermore, the following three suspected outliers in the graph were determined to be normal values, resulting in the following: Figure 4 The pressure change curve shown is... Figure 4 As can be seen, after correcting for the true outliers, the pressure curve becomes smoother; at the same time, Figure 4 It can be seen that even if false detections occur in the initial test, the falsely detected suspected abnormal values can be judged as normal values in the second test.
[0045] The beneficial effects of this embodiment are as follows: This method applies anomaly detection and correction to sensor equipment on offshore oil production platforms, including but not limited to closed-loop tanks, and can detect and correct abnormal pressure values caused by external interference, thereby avoiding misjudgment of the cause of abnormal pressure due to external factors; This method is easy to automate, adopts a combination of preliminary detection and secondary detection, and can greatly improve the accuracy of anomaly detection through rigorous mathematical analysis logic.
[0046] Example 2
[0047] This embodiment further supplements steps S1 to S4 based on embodiment 1. In step S1, the timestamp index is accurate to the year, month, day, hour, minute, and second, with the hour recorded in 24-hour format. This ensures that the time data point corresponding to each pressure value is unique and there will be no duplicate times.
[0048] Furthermore, in step S2, the data is segmented by iterating through the data, as follows:
[0049] The first data point of the pressure data is taken, up to the nth data point, forming the first data segment, where n represents the data processing step size. Then, the (n+1)th data point of the original pressure data is taken, up to the 2nth data point, forming the second data segment, and so on. Since the data processing step size is 100 data points, the specific process of data segmentation is as follows:
[0050] Time series data segment 1 = {1st time data point, 2nd time data point, ..., 100th time data point};
[0051] Time series data segment 2 = {101st time data point, 102nd time data point, ..., 200th time data point};
[0052] The time series data segment k = {the (k-1)*100+1th time data point, the (k-1)*100+2nd time data point, ..., the k*100th time data point}.
[0053] Furthermore, in step S2, data that is insufficient to form a time series data segment is discarded and not processed, so that the data processing in subsequent steps can be carried out in an orderly and effective manner.
[0054] Furthermore, in step S2, after segmenting the closed-loop tank pressure data, the validity of the time data points within the time series data segment is checked to ensure that each time data point in the time series data segment is arranged in chronological order. That is, in each time series data segment, the earlier time data points are listed first, and the later time data points are listed last, thereby improving the accuracy of outlier detection.
[0055] Furthermore, in step S3, the window size and quantile multiples of the sliding time window are fixed values, meaning they are manually set. Generally, if the automatically set sliding time window size and quantile multiples are not properly configured, outliers may be missed during actual detection (outliers exist but are not detected), or they may be misjudged (normal values are identified as outliers during detection). Therefore, to improve the accuracy of outlier detection, the sliding time window size and quantile multiples need to be set manually and reasonably, so that in step S5, it is better to have more misjudgments than missed ones. Even if there are many misjudged outliers in step S5, the misjudged data points can be identified and corrected in steps S6 and S7, thus comprehensively improving the accuracy of outlier identification.
[0056] Furthermore, in step S4, the difference between the current value and the first median value is the absolute value. Thus, the suspected outliers detected in subsequent steps could be those causing a sudden increase in pressure or those causing a sudden decrease in pressure. Therefore, abnormal upward and downward pressure trends can be detected, making the detection results more intuitive and accurate.
[0057] Other features, working principles, and beneficial effects of this embodiment are the same as those of Embodiment 1.
[0058] Example 3
[0059] This implementation further supplements steps S5 to S6 based on Example 2. Specifically, in step S5, the normal range interval is [Q1 - c * IQR, Q3 + c * IQR], where Q1 and Q3 are the lower and upper quartiles of the pressure values corresponding to the time data points in the current time series data segment, respectively. That is, after arranging all the pressure values corresponding to the time data points in the current time series data segment from smallest to largest, the pressure values at the 25th and 75th quartiles are considered. c is the quartile distance of the sliding time window in step S3, and IQR is the interquartile range, with a value of IQR = Q3 - Q1.
[0060] Further, in step S6, 'a' is set to 5, meaning each short-term sequence x2 contains 5 time data points. Selecting an odd number of time data points for the short-term sequence x2 facilitates the determination of the second median value, ensuring that the second median value corresponds to the pressure value at a specific time data point within the short-term sequence x2. If the short-term sequence x2 contains only three time data points, the limited number makes it difficult to identify whether a point represents a true outlier; conversely, too many time data points would cause data delays. Therefore, to quickly and accurately determine true outliers, it is more appropriate for each short-term sequence x2 to contain 5 time data points.
[0061] Further, in step S6, the three time data points before the suspected anomaly, the suspected anomaly itself, and the one time data point after the suspected anomaly in the time series data segment are selected as short-term sequence x2; when the suspected anomaly is in the top three positions in the time series data segment, the first five time data points in the time series data segment are selected as short-term sequence x2; when the suspected anomaly is in the last position in the time series data segment, the last five time data points in the time series data segment are selected as short-term sequence x2.
[0062] Other features, working principles, and beneficial effects of this embodiment are the same as those of Embodiment 2.
[0063] It should be noted that, in the appendix Figure 1 The steps shown in the flowchart can be executed in a computer system, such as a set of computer-executable instructions. Although the logical order is shown in the flowchart, in some cases, the steps described may be performed in a different order than that shown here, such as the order in which the window size of the sliding time window and its quantile multiple are set.
[0064] The present invention has been further described above with reference to specific embodiments. However, it should be understood that the specific description herein should not be construed as limiting the nature and scope of the present invention. Various modifications made by those skilled in the art to the above embodiments after reading this specification, such as modifications to parameters such as the maximum difference threshold or the data processing step size, are all within the scope of protection of the present invention.
Claims
1. A method for detecting and correcting abnormal pressure in a pressure sensing device, characterized in that, Includes the following steps: Step S1: Receive pressure and time data, convert each pressure value into a time series format, and convert the time corresponding to each pressure value into a timestamp index format, using the time corresponding to the pressure value as the time index value of the pressure value; Step S2: Set the data processing step size, and then segment the closed-loop tank pressure data to obtain multiple time series data segments, each of which includes n time data points; Step S3: Set the window size and quantile multiple of the sliding time window, and place the time data points into the sliding time window according to the window size of the sliding time window; Step S4: Calculate the first median value of the pressure values corresponding to the time data points contained in the sliding time window; take the pressure values corresponding to the time data points after the sliding time window as the current values, calculate the difference between the current values and the first median value, slide the sliding time window, and obtain the difference sequence x1; Step S5: Set the normal range interval for the current time series data segment, compare the difference sequence x1 with the normal range interval, and determine whether all values in the difference sequence x1 fall within the normal range interval. If yes, end the process; otherwise, determine the pressure value corresponding to the value falling outside the normal range interval as a suspected abnormal value and proceed to the next step. Step S6: Select 'a' consecutive time data points from the time series data segment as short-term series x2, and determine the maximum and minimum pressure values corresponding to the time data points in short-term series x2; determine whether the suspected outlier is equal to the maximum or minimum pressure value corresponding to the time data points in short-term series x2. If yes, the suspected outlier is still a suspected outlier, and proceed to the next step; if no, the suspected outlier is a normal value, and the process ends. Step S7: Set the maximum difference threshold, determine the second median value of the pressure value corresponding to the time data point in the short-term sequence x2, and determine whether the difference between the suspected outlier and the second median value is greater than or equal to the maximum difference threshold. If yes, it means that the suspected outlier is a real outlier, and the suspected outlier is assigned the second median value. If no, it means that the suspected outlier is a normal value, and the process ends.
2. The abnormal pressure detection and correction method for a pressure sensing device according to claim 1, characterized in that, In step S1, the timestamp index is accurate to the year, month, day, hour, minute, and second, with the hour recorded in 24-hour format.
3. The abnormal pressure detection and correction method for a pressure sensing device according to claim 1, characterized in that, In step S2, the data is segmented by iterating through the data. The specific process is as follows: The first data point of the pressure data is taken, up to the nth data point, forming the first data segment, where n represents the data processing step size. Then, the (n+1)th data point of the original pressure data is taken, up to the 2nth data point, forming the second data segment, and so on. Specifically: Time series data segment 1 = {the 1st time data point, the 2nd time data point, ..., the nth time data point}; Time series data segment 2 = {the (n+1)th time data point, the (n+2)th time data point, ..., the 2nth time data point}; The time series data segment k = {the (k-1)*n+1th time data point, the (k-1)*n+2th time data point, ..., the k*nth time data point}.
4. The abnormal pressure detection and correction method for a pressure sensing device according to claim 1, characterized in that, In step S2, data that is insufficient to form a time series data segment is discarded and not processed.
5. The abnormal pressure detection and correction method for a pressure sensing device according to claim 1, characterized in that, In step S2, after segmenting the closed-loop tank pressure data, the validity of the time data points within the time series data segment is checked to determine that each time data point in the time series data segment is arranged in chronological order.
6. The abnormal pressure detection and correction method for a pressure sensing device according to claim 1, characterized in that, In step S3, the window size and quantile distance of the sliding time window are fixed values.
7. The abnormal pressure detection and correction method for a pressure sensing device according to claim 1, characterized in that, In step S4, the difference between the current value and the first median value is the absolute value.
8. The abnormal pressure detection and correction method of a pressure sensing device according to claim 1, characterized in that, In step S5, the normal range interval is [Q1-c*IQR, Q3+c*IQR], where Q1 and Q3 are the lower quartile and upper quartile of the pressure value corresponding to the time data point in the current time series data segment, respectively, c is the quartile distance of the sliding time window, and IQR is the interquartile distance, which has the value IQR=Q3-Q1.
9. A method for detecting and correcting abnormal pressure in a pressure sensing device according to any one of claims 1 to 8, characterized in that, In step S6, the value of 'a' is 5.
10. The abnormal pressure detection and correction method of a pressure sensing device according to claim 9, characterized in that, In step S6, the three time data points preceding the suspected outlier, the suspected outlier itself, and the one time data point following the suspected outlier in the time series data segment are selected as the short-term sequence x2; when the suspected outlier ranks among the top three in the time series data segment, the first five time data points in the time series data segment are selected as the short-term sequence x2; when the suspected outlier ranks last in the time series data segment, the last five time data points in the time series data segment are selected as the short-term sequence x2.