CN-122015936-A - Oil monitoring intelligent calibration method based on multi-source data fusion

CN122015936ACN 122015936 ACN122015936 ACN 122015936ACN-122015936-A

Abstract

The invention relates to the technical field of oil calibration, and discloses an intelligent oil monitoring calibration method based on multi-source data fusion, which is used for solving the problem that data in the operation process is not suitable for updating a calibration model when oil calibration is carried out, and comprises the steps of acquiring historical oil monitoring data, historical reference monitoring data and historical deviation data set, constructing an oil calibration model, acquiring monitoring deviation of a current sampling period, inputting the monitoring deviation into the oil calibration model, obtaining deviation correction quantity of a current sampling period, carrying out calibration treatment on oil monitoring data of the current sampling period to obtain calibrated oil monitoring data, obtaining update effectiveness judgment parameters for updating judgment of an oil calibration model, calculating to obtain a data reliability index, judging whether the current calibrated oil monitoring data can be used for updating the oil calibration model or not according to the data reliability index, and effectively improving stability and reliability of an oil monitoring calibration process.

Inventors

XUE XIAOHAN
GUO ZONGKUI
HU WEIZHENG
LI MIAO
GUO ZONGHAO
YIN LINGXIANG
ZHAO YANJUN
MA GUOQI
DU JIANGNAN
Mu Shuaili
LI YONGGANG
WANG QUNYING
SHAO XUEZAN
ZHANG GUORAN

Assignees

卡松科技股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260130

Claims (10)

1. The intelligent oil monitoring calibration method based on multi-source data fusion is characterized by comprising the following steps of: Step1, oil monitoring data are collected through a multi-source sensor group; Step 2, carrying out reference measurement on the oil liquid state through a reference sensor, and acquiring reference monitoring data in a preset calibration period; Step 3, forming historical oil monitoring data and corresponding historical reference monitoring data in a plurality of historical sampling periods based on the oil monitoring data obtained in the step 1 and the reference monitoring data obtained in the step 2, calculating monitoring deviation between the historical oil monitoring data and the corresponding historical reference monitoring data for each historical sampling period to obtain a historical deviation data set, and carrying out learning processing on the historical oil monitoring data, the historical reference monitoring data and the historical deviation data set to construct an oil calibration model for generating deviation correction; Step 4, based on the oil monitoring data obtained in the step 1 and the reference monitoring data obtained in the step 2, determining oil monitoring data and reference monitoring data in a current sampling period, calculating to obtain current monitoring deviation according to the oil monitoring data and the reference monitoring data, inputting the current monitoring deviation into an oil calibration model to obtain a deviation correction corresponding to the current sampling period, and performing calibration processing on the oil monitoring data in the current sampling period according to the deviation correction to obtain calibrated oil monitoring data; Step 5, acquiring update effectiveness judgment parameters for updating judgment of the oil liquid calibration model, wherein the update effectiveness judgment parameters comprise monitoring deviation before and after calibration, control intervention records and monitoring deviation in each sampling period, calculating according to the update effectiveness judgment parameters to obtain a data reliability index, and judging whether the current calibrated oil liquid monitoring data can be used for updating the oil liquid calibration model according to the data reliability index; Step 6, if the current calibrated oil monitoring data is judged to be available for updating the oil calibration model, executing the step 7, and if the current calibrated oil monitoring data is judged to be unavailable for updating the oil calibration model, not updating the oil calibration model; And 7, taking the calibrated oil monitoring data and the corresponding reference monitoring data as new historical calibration data for updating an oil calibration model.
2. The intelligent oil monitoring calibration method based on multi-source data fusion according to claim 1, wherein the step of obtaining the historical deviation data set is as follows: in a plurality of historical sampling periods, respectively acquiring historical oil monitoring data corresponding to each historical sampling period and historical reference monitoring data corresponding to the historical sampling period; Comparing and calculating historical oil monitoring data at the same sampling time point with historical reference monitoring data aiming at each historical sampling period to obtain deviation values corresponding to all the sampling time points, and summarizing a plurality of deviation values obtained in the historical sampling period to obtain monitoring deviation; and forming a historical deviation data set by the monitoring deviation corresponding to each historical sampling period.
3. The intelligent oil monitoring calibration method based on multi-source data fusion according to claim 1, wherein the oil calibration model obtaining step comprises the following steps: based on the obtained historical oil monitoring data, historical reference monitoring data and historical deviation data sets, extracting monitoring deviation corresponding to each historical sampling period according to each historical sampling period, and arranging according to the time sequence of the historical sampling periods to form a historical monitoring deviation sequence; Constructing a deviation direction identification generation rule based on the historical monitoring deviation sequence; taking absolute values of all monitoring deviations in the historical monitoring deviation sequence to obtain a historical deviation amplitude sequence, and extracting deviation amplitude distribution characteristics based on the historical deviation amplitude sequence; Constructing an amplitude mapping rule based on the deviation amplitude distribution characteristics; and integrating the deviation direction identification generation rule with the amplitude mapping rule to form an oil liquid calibration model.
4. The intelligent oil monitoring calibration method based on multi-source data fusion according to claim 3, wherein the step of constructing a deviation direction identification generation rule based on a history monitoring deviation sequence is as follows: the deviation direction is determined to be "positive deviation" when the monitored deviation is greater than zero, the deviation direction is determined to be "negative deviation" when the monitored deviation is less than zero, and the deviation direction is determined to be "zero deviation" when the monitored deviation is equal to zero.
5. The intelligent oil monitoring calibration method based on multi-source data fusion according to claim 3, wherein the step of constructing an amplitude mapping rule based on the deviation amplitude distribution characteristics is as follows: taking the deviation amplitude distribution characteristics as interval boundary values of the amplitude intervals, and dividing the value range of the deviation amplitude to obtain at least two continuous deviation amplitude intervals; for any deviation amplitude, determining a deviation amplitude section to which the deviation amplitude belongs, and calculating a ratio based on a section boundary value of the deviation amplitude and the deviation amplitude section to which the deviation amplitude belongs to obtain a relative position ratio; and determining a corrected amplitude value corresponding to the deviation amplitude according to the relative position ratio and a preset amplitude mapping relation, so as to form an amplitude mapping rule.
6. The intelligent oil monitoring calibration method based on multi-source data fusion according to claim 1, wherein the data reliability index obtaining step comprises the following steps: in the current sampling period, acquiring monitoring deviation of oil monitoring data before and after calibration at each sampling time point, and calculating to obtain a deviation convergence stability influence coefficient according to the monitoring deviation of the oil monitoring data before and after calibration at each sampling time point; In the current sampling period, acquiring a control intervention record triggered by the calibrated oil monitoring data, and calculating according to the control intervention record to obtain a control intervention disturbance influence coefficient; acquiring a candidate updating data set and a verification data set in a current sampling period, updating a current oil calibration model according to the candidate updating data set to obtain a test updating oil calibration model, and calculating to obtain a predicted disturbance response influence coefficient based on the verification data set by combining the current oil calibration model and the test updating oil calibration model; and carrying out normalization processing on the deviation convergence stability influence coefficient, the control intervention disturbance influence coefficient and the predicted disturbance response influence coefficient, and calculating to obtain a data reliability index according to the deviation convergence stability influence coefficient, the control intervention disturbance influence coefficient and the predicted disturbance response influence coefficient after normalization processing.
7. The intelligent oil monitoring calibration method based on multi-source data fusion according to claim 6, wherein the obtaining step of the deviation convergence stability influence coefficient is as follows: In the current sampling period, acquiring monitoring deviation of each sampling time point of oil monitoring data before calibration to obtain a pre-calibration monitoring deviation sequence in the current sampling period, and simultaneously acquiring monitoring deviation of each sampling time point of calibrated oil monitoring data to obtain a post-calibration monitoring deviation sequence in the current sampling period; taking absolute values of all monitoring deviations in the pre-calibration monitoring deviation sequence and the post-calibration monitoring deviation sequence respectively to obtain a pre-calibration deviation amplitude sequence and a post-calibration deviation amplitude sequence, and calculating variation of the deviation amplitudes of adjacent sampling time points of the pre-calibration deviation amplitude sequence and the post-calibration deviation amplitude sequence respectively to obtain a pre-calibration adjacent deviation amplitude variation sequence and a post-calibration adjacent deviation amplitude variation sequence; Respectively carrying out symbol judgment on the adjacent deviation amplitude variation sequence before calibration and the adjacent deviation amplitude variation sequence after calibration to obtain a variation direction identification sequence before calibration and a variation direction identification sequence after calibration, wherein the corresponding variation direction is identified as a divergent direction when the deviation amplitude variation is greater than 0, the corresponding variation direction is identified as a convergent direction when the deviation amplitude variation is less than 0, and the corresponding variation direction is identified as a stable direction when the deviation amplitude variation is equal to 0; based on the pre-calibration change direction identification sequence and the post-calibration change direction identification sequence, counting the number of intervals with consistent change direction identifications in each adjacent sampling interval, recording the number as the consistent direction intervals, acquiring the number of sampling time points in the current sampling period, and calculating to obtain the pre-calibration and post-calibration deviation convergence consistency parameter according to the consistent direction intervals and the number of sampling time points; Screening adjacent data sets meeting the condition that two adjacent data are not 0 in the calibrated deviation amplitude sequence, and carrying out ratio calculation on each screened adjacent data to obtain a deviation amplitude ratio sequence, so as to obtain a deviation ratio median of the deviation amplitude ratio sequence and a deviation ratio median absolute deviation; Dividing the median absolute deviation of the deviation ratio by the median of the deviation ratio when the median of the deviation ratio is not 0, obtaining a calibrated convergence fluctuation ratio, and recording the calibrated convergence fluctuation ratio as 0 when the median of the deviation ratio is 0; And calculating according to the deviation convergence consistency parameter before and after calibration and the convergence fluctuation ratio after calibration to obtain a deviation convergence stability influence coefficient.
8. The intelligent oil monitoring calibration method based on multi-source data fusion according to claim 6, wherein the step of obtaining the control intervention disturbance influence coefficient is as follows: In the current sampling period, acquiring a control intervention record triggered by a control system based on the calibrated oil monitoring data, obtaining a trigger time point sequence, and counting the trigger times; acquiring the number of sampling time points in the current sampling period, dividing the triggering times by the number of sampling time points to obtain the intervention frequency of the calibrated data triggering control; calculating event intervals between two adjacent intervention triggers based on the trigger time point sequence to obtain an event interval sequence; Obtaining the absolute deviation between the median of the event interval and the median of the event interval sequence, dividing the absolute deviation of the median of the event interval by the median of the event interval to obtain the interference clustering degree when the median of the event interval is not 0, and recording the interference clustering degree as 0 when the median of the event interval is 0; When the triggering interval of two adjacent interventions is 1, judging that the two interventions belong to the same continuous intervention section, dividing the two interventions into a plurality of continuous intervention sections according to the triggering interval, counting the sum of the number of sampling points covered by all the continuous intervention sections, marking the sum as the number of the intervention coverage points, dividing the number of the intervention coverage points by the number of sampling time points, and obtaining the continuous intervention occupancy rate; and calculating to obtain the control intervention disturbance influence coefficient according to the intervention frequency, the intervention clustering degree and the intervention duration occupation degree.
9. The intelligent oil monitoring calibration method based on multi-source data fusion according to claim 6, wherein the step of obtaining the predicted disturbance response influence coefficient is as follows: Acquiring a current oil calibration model, acquiring calibrated oil monitoring data corresponding to a current sampling period and corresponding reference monitoring data, taking the calibrated oil monitoring data and the corresponding reference monitoring data as candidate updating data sets, and acquiring a historical calibration data set as a verification data set; performing one-time trial updating on the current oil calibration model based on the candidate updating data set to obtain a trial updated oil calibration model; Respectively calculating historical monitoring deviation of each piece of historical calibration data in the verification data set, and respectively inputting the historical monitoring deviation into a current oil calibration model and a trial updated oil calibration model to obtain a current model deviation correction sequence and a trial updated model deviation correction sequence; Based on the current model deviation correction sequence and the trial updating model deviation correction sequence, carrying out difference calculation on the output of the two models corresponding to the same piece of verification data to obtain a deviation correction disturbance sequence, obtaining the corrected disturbance median of the deviation correction disturbance sequence, obtaining the median of the absolute value of each deviation correction in the current model deviation correction sequence, and recording the median as the current model output amplitude scale standard; When the current model output amplitude scale reference is not 0, dividing the corrected disturbance median by the current model output amplitude scale reference to obtain a predicted disturbance response influence coefficient, and when the current model output amplitude scale reference is 0, marking the predicted disturbance response influence coefficient as 0.
10. The intelligent oil monitoring calibration method based on multi-source data fusion of claim 1, wherein the step of judging whether the current calibrated oil monitoring data can be used for updating an oil calibration model according to the data reliability index is as follows: And comparing the data reliability index with a reliability threshold, if the data reliability index is larger than or equal to the reliability threshold, judging that the current calibrated oil monitoring data can be used for updating the oil calibration model, and if the data reliability index is smaller than the reliability threshold, judging that the current calibrated oil monitoring data cannot be used for updating the oil calibration model.

Description

Oil monitoring intelligent calibration method based on multi-source data fusion Technical Field The invention relates to the technical field of oil liquid calibration, in particular to an oil liquid monitoring intelligent calibration method based on multi-source data fusion. Background With the development of industrial equipment to high reliability and long-period running, oil liquid state monitoring gradually becomes an important component in equipment running state sensing and maintenance management. In the running process of the equipment, state parameters such as temperature, pressure, conductivity and the like of the oil liquid can change along with running working conditions, environmental conditions and service time, and related monitoring data not only influence the judging result of the running state of the equipment, but also relate to the safety and stability of the equipment in the long-term running process. Therefore, how to perform continuous, stable and adaptive calibration processing on oil monitoring data has become a research direction of continuous attention in the oil monitoring field. In the prior art, oil monitoring systems based on multi-source sensor acquisition have been widely used. The system generally collects multidimensional monitoring data in the oil operation process, and combines a reference measurement result to calibrate the oil monitoring data so as to reduce the influence of sensor errors, environmental changes and operation condition fluctuation on the monitoring result. The method combines multi-source data acquisition and calibration processing, so that the automation level and measurement accuracy of the oil monitoring system are improved to a certain extent. In the existing oil monitoring and calibrating process, a result obtained by multiple times of calibration is usually introduced into a calibration model as historical data, and the calibration model is updated by continuously accumulating the historical calibration data so that the calibration model is gradually adapted to the change of the oil operation working condition. The method is widely adopted in practical application, and the basic idea is to improve the adaptability of the model to complex operation environments by continuously introducing new calibration results. However, the above technology has at least the following technical problems: In the actual operation process, the oil operation working condition has obvious dynamic and uncertainty characteristics, and the calibration results formed in different time periods have large differences in stability, background formation and representativeness. When all calibration results are introduced into the calibration model as effective historical samples without distinction in the model updating process, part of calibration results formed under abnormal working conditions or short-time fluctuation conditions may adversely affect the long-term calibration effect of the model, thereby reducing the stability and reliability of the oil monitoring calibration process. Disclosure of Invention In order to overcome the defects in the prior art, the invention provides an intelligent oil monitoring calibration method based on multi-source data fusion, which aims to solve the problems in the background art. In order to achieve the above purpose, the present invention provides the following technical solutions: An intelligent oil monitoring calibration method based on multi-source data fusion comprises the following steps of 1, collecting oil monitoring data through a multi-source sensor group, 2, carrying out reference measurement on an oil state through a reference sensor, obtaining reference monitoring data in a preset calibration period, 3, forming historical oil monitoring data and corresponding historical reference monitoring data in a plurality of historical sampling periods based on the oil monitoring data obtained in the step 1 and the reference monitoring data obtained in the step 2, calculating monitoring deviation between the historical oil monitoring data and the corresponding historical reference monitoring data for each historical sampling period to obtain a historical deviation data set, carrying out learning processing on the historical oil monitoring data, the historical reference monitoring data and the historical deviation data set to form an oil calibration model for generating deviation correction, 4, determining the oil monitoring data and the reference monitoring data in the current sampling period based on the oil monitoring data obtained in the step 1 and the reference monitoring data obtained in the step 2, calculating the current monitoring deviation according to the oil monitoring data obtained in the step 2, inputting the current monitoring deviation into the oil calibration model to obtain a deviation corresponding to the current sampling period, carrying out updating the correction quantity according to the deviation, carrying out upd