Search

CN-121996996-A - Data-driven normal pressure tower top operation health state online assessment method

CN121996996ACN 121996996 ACN121996996 ACN 121996996ACN-121996996-A

Abstract

The invention discloses a data-driven normal pressure tower top operation health state online assessment method. According to the method, historical data of operation sensitive parameters of the atmospheric tower are obtained, abnormal value elimination, missing value filling and data alignment processing are carried out, discrete time sequences of the operation sensitive parameters of the atmospheric tower are obtained, sensitive parameter relative factors are calculated, weights of the operation sensitive parameters of the atmospheric tower are set, an operation health state assessment model of the atmospheric tower is built based on weighted Euclidean distances, after the historical data are utilized for training, online test data of the operation sensitive parameters of the atmospheric tower are input into the trained operation health state assessment model of the atmospheric tower, the operation health state value of the atmospheric tower is calculated in real time by utilizing the operation health state assessment model of the atmospheric tower, and the operation health state of the atmospheric tower is determined. The invention realizes the accurate on-line evaluation of the operation health state of the normal pressure tower top, and is beneficial to guaranteeing the long-period safe and stable operation of the atmospheric and vacuum device.

Inventors

  • LIU XIAOJIN
  • ZHANG YANLING
  • PAN LONG
  • QU DINGRONG
  • HAN LEI

Assignees

  • 中国石油化工股份有限公司
  • 中石化安全工程研究院有限公司

Dates

Publication Date
20260508
Application Date
20241104

Claims (16)

  1. 1. The online evaluation method for the operation health state of the normal pressure tower top based on data driving is characterized by comprising the following steps of: step 1, acquiring historical data of operation sensitive parameters of an atmospheric tower; Step 2, cleaning historical data of the operation sensitive parameters of the atmospheric tower, removing abnormal values in the historical data of the operation sensitive parameters of the atmospheric tower, and obtaining the historical data of the operation sensitive parameters of the atmospheric tower after the abnormal values are removed; Step 3, carrying out missing value supplementing treatment on the atmospheric tower operation sensitive parameter historical data with the abnormal values removed, and filling missing values in the atmospheric tower operation sensitive parameter historical data to obtain atmospheric tower operation sensitive parameter historical data with the missing values supplemented; Step 4, performing discrete processing on the history data of the operation sensitive parameters of the atmospheric towers after the missing value supplementation, and aligning the history data of the operation sensitive parameters of each atmospheric tower to obtain a discrete time sequence of the operation sensitive parameters of each atmospheric tower; Step 5, calculating sensitive parameter relative factors corresponding to historical data of the operation sensitive parameters of each atmospheric tower based on discrete time sequences of the operation sensitive parameters of each atmospheric tower; step 6, determining the weight of each atmospheric tower operation sensitive parameter based on the influence length of each atmospheric tower operation sensitive parameter on the operation health of the atmospheric tower top; step 7, constructing and training an atmospheric tower top operation health state assessment model based on the weighted Euclidean distance LOF to obtain a trained atmospheric tower top operation health state assessment model; step 8, acquiring online test data of operation sensitive parameters of each atmospheric tower, and aligning the online test data of the operation sensitive parameters of each atmospheric tower; step 9, calculating sensitive parameter relative factors corresponding to the online test data of the operation sensitive parameters of each atmospheric tower; And 10, inputting online test data of the operation sensitive parameters of the atmospheric tower into a trained model for evaluating the operation health state of the atmospheric tower, calculating to obtain an operation health state value of the atmospheric tower by combining the relative factors of the sensitive parameters corresponding to the online test data of the operation sensitive parameters of the atmospheric tower and the preset weight of the operation sensitive parameters of the atmospheric tower, determining the operation health state of the atmospheric tower, and drawing an operation health state trend chart of the atmospheric tower.
  2. 2. The online evaluation method for the operation health state of the normal pressure tower based on data driving according to claim 1, wherein the historical data of the operation sensitive parameters of the normal pressure tower and the online sensitive parameters of the operation health of the normal pressure tower are obtained from an enterprise DCS database, a LIMS system and a corrosion monitoring system, and the historical data of the operation sensitive parameters of the normal pressure tower and the online sensitive parameters of the operation health of the normal pressure tower comprise historical data of water injection quantity deviation of the normal pressure tower, iron ion content of the sulfur-containing sewage, chloride ion content of the sulfur-containing sewage, PH value of the sulfur-containing sewage, water dew point temperature difference, ammonium salt crystallization temperature difference, corrosion online monitoring data, shutdown overhaul data and fixed-point thickness measurement data; Wherein, the The water injection quantity deviation of the normal pressure tower is the difference between the actual water injection quantity and the theoretical water injection quantity, the water dew point temperature difference is the difference between the dew point temperature and the tower top temperature of the normal pressure tower, and the ammonium salt crystallization temperature difference is the difference between the ammonium salt crystallization temperature and the tower top temperature of the normal pressure tower.
  3. 3. The online evaluation method for the operation health status of the normal pressure tower top based on the data driving according to claim 1, wherein in the step 2, the outlier rejection method comprises a 3δ method, a quartile method, a Z-sore method and a sliding window median filtering method.
  4. 4. The online evaluation method for the operation health status of the normal pressure tower top based on the data driving according to claim 1, wherein in the step 3, the missing value supplementing method comprises a cubic spline interpolation method, a median interpolation method, a mean interpolation method and an adjacent value interpolation method.
  5. 5. The online evaluation method for the operation health status of the atmospheric tower top based on the data driving according to claim 1, wherein the step 4 comprises the following steps: Step 4.1, based on the historical data of the atmospheric tower operation sensitive parameters, sequencing the atmospheric tower operation sensitive parameters according to the sequence of the data acquisition interval time from small to large, acquiring the atmospheric tower operation sensitive parameter with the shortest acquisition interval time in all the atmospheric tower operation sensitive parameters, and determining the shortest acquisition interval time T 1 ; And 4.2, performing discrete processing on the historical data of the operation sensitive parameters of each atmospheric tower according to the shortest acquisition interval time T 1 , and adjusting the sampling interval time of the historical data of the operation sensitive parameters of each atmospheric tower to the shortest acquisition interval time T 1 to obtain the discrete time sequence data of the operation sensitive parameters of each atmospheric tower.
  6. 6. The online evaluation method of the operation health status of the atmospheric tower top based on the data driving according to claim 5, wherein in the step 4.2, when the historical data of the operation sensitive parameters of the atmospheric tower are subjected to discrete processing, the operation sensitive parameters of the atmospheric tower corresponding to the shortest acquisition interval time T 1 do not need to be subjected to discrete processing, and the operation sensitive parameters of the atmospheric tower corresponding to the acquisition interval time being longer than the shortest acquisition interval time T 1 are subjected to discrete processing according to the historical data of the operation sensitive parameters of the atmospheric tower, the operation sensitive parameter data of the atmospheric tower acquired last time is obtained as a supplementary value to be subjected to discrete processing, and the sampling interval time of the operation sensitive parameter historical data of the atmospheric tower is adjusted to the shortest acquisition interval time, so that the discrete time sequence data of the operation sensitive parameters of the atmospheric tower is obtained.
  7. 7. The online evaluation method for the operation health status of the atmospheric tower top based on the data driving according to claim 1, wherein the step 5 comprises the following steps: step 5.1, respectively aiming at each atmospheric tower operation sensitive parameter, calculating the average value of the atmospheric tower operation sensitive parameters by utilizing the historical data of the atmospheric tower operation sensitive parameters; and 5.2, calculating sensitive parameter relative factors corresponding to each historical data of the atmospheric tower operation sensitive parameters according to the average value of the atmospheric tower operation sensitive parameters and the historical data of the atmospheric tower operation sensitive parameters.
  8. 8. The online evaluation method of the operation health status of the atmospheric tower top based on the data driving according to claim 7, wherein in the step 5, the average value of the operation sensitivity parameter R of the atmospheric tower is calculated based on the history data of the operation sensitivity parameter R of the atmospheric tower, as shown in the formula (1): Wherein Rmeas is the average value of the normal pressure tower operation sensitive parameters R, i is the serial number, n is the total number of the history data of the normal pressure tower operation sensitive parameters R, and R i is the ith acquired data in the history data of the normal pressure tower operation sensitive parameters R; calculating a sensitive parameter relative factor corresponding to each historical data of the atmospheric tower operation sensitive parameter by using the average value of the atmospheric tower operation sensitive parameter R and combining all the historical data of the atmospheric tower operation sensitive parameter, wherein the sensitive parameter relative factor is shown in a formula (2): Wherein R i ' is a sensitive parameter relative factor corresponding to the ith acquired data in the history data of the normal pressure tower operation sensitive parameter R.
  9. 9. The online evaluation method for the operation health status of the atmospheric tower top based on the data driving according to claim 1, wherein the step 7 comprises the following steps: Step 7.1, selecting atmospheric tower operation sensitive parameter data from the atmospheric tower operation sensitive parameter historical data as training data of an atmospheric tower top operation health state evaluation model, and constructing a training data set U, wherein x i is the ith training data in the training data set U, and m is the total number of training data in the training data set U; step 7.2, calculating weighted Euclidean distance between each training data in the training data set U; Step 7.3, calculating a kth weighted distance of each training data in the training data set U; Step 7.4, calculating a kth distance neighborhood of each training data in the training data set U; step 7.5, calculating the kth weighted reachable distance of each training data in the training data set U; Step 7.6, calculating the kth weighted local reachable density of each training data in the training data set U; And 7.7, calculating weighted local outlier factors of all training data in the training data set U to obtain a trained normal-pressure tower top operation health state evaluation model based on the weighted Euclidean distance LOF, calculating the weighted local outlier factors by using the normal-pressure tower top operation health state evaluation model based on the weighted Euclidean distance LOF and using the weighted local outlier factors as an operation health state value of the normal-pressure tower top, and determining the normal-pressure tower top operation health state.
  10. 10. The online evaluation method of the operating health status of the atmospheric tower top based on data driving according to claim 9, wherein in the step 7.2, the weighted euclidean distance calculation formula is: Wherein d (·) is a weighted Euclidean distance value, x 1 、x 2 is training data in a training data set U, j is a sequence number, M is a dimension of history data of the atmospheric tower operation sensitive parameters and is used for representing class numbers of the atmospheric tower operation sensitive parameters, x 1j is a data value of the j-th atmospheric tower operation sensitive parameters corresponding to the training data x 1 , x 2j is a data value of the j-th atmospheric tower operation sensitive parameters corresponding to the training data x 2 , and omega j is the weight of the j-th atmospheric tower operation sensitive parameters.
  11. 11. The online evaluation method of the operating health status of the atmospheric tower top based on data driving according to claim 9, wherein in the step 7.3, the k-th weighted distance calculation process of the training data is as follows: Aiming at training data x 1 in a training data set U, outwards diffusing by taking the training data x 1 as a circle center, and determining kth training data nearest to the training data x 1 ; the coverage area of the diffusion process in the training data set U satisfies the following conditions: At least k training data are present in the training data set U excluding the training data x 1 , such that d (x 1 ,o′)≤d(x 1 , o), where d (x 1 , o ') is a weighted euclidean distance between the training data x 1 and the training data o ', d (x 1 , o) is a weighted euclidean distance between the training data x 1 and the training data o, and o ' o is the training data in the training data set U excluding the training data x 1 ; At the same time, at most k-1 training data are present in the training data set U excluding training data x 1 , such that d (x 1 ,o′)<d(x 1 , o).
  12. 12. The online evaluation method of the operating health status of the atmospheric tower top based on data driving according to claim 11, wherein in the step 7.4, the kth distance neighborhood of the training data in the training data set U is a set of all training data in the kth weighted distance neighborhood of the training data, as shown in the formula (4): N k (x 1 )＝{o′∣d(x 1 ,o′)≤d k (x 1 )} (4) Where x 1 is the training data in training data set U, N k (·) is the kth distance neighborhood of the training data, and d k (·) is the kth weighted distance of the training data.
  13. 13. The method for online evaluation of operational health of a data-driven atmospheric tower top according to claim 12, wherein in said step 7.5, the kth weighted reachable distance of the training data is: d k-reach (o,x 1 )＝max{d(o,x 1 ),d k (o)} (5) Where o and x 1 are both training data in the training data set U, max is a maximum function, d k-reach (o,x 1 ) is the kth weighted reachable distance of the training data x 1 , and is used to represent the maximum value of the weighted euclidean distance between the training data o and the training data x 1 and the kth weighted distance of the training data o.
  14. 14. The online evaluation method of the operating health status of the atmospheric tower top based on data driving according to claim 13, wherein in the step 7.6, a k-weighted local reachable density calculation formula of the training data is: Where d k-ρ (·) is the kth weighted local reachable density of the training data.
  15. 15. The online evaluation method of the operating health status of the atmospheric tower top based on data driving according to claim 14, wherein in the step 7.7, the weighted local outlier factor of the training data is calculated by the formula: Where LOF k (·) is a weighted local outlier; And the weighted local outlier factor is used for representing the outlier degree of the training data, and when the weighted local outlier factor is greater than 1.5, the normal pressure overhead operation health state is judged to be a sub-health state.
  16. 16. The online evaluation method of the operation health status of the atmospheric tower top based on the data driving according to claim 1, wherein in the step 9, a calculation formula of a sensitive parameter relative to a factor corresponding to online test data of the operation sensitive parameter of the atmospheric tower is: Wherein T u ' is a sensitive parameter relative factor corresponding to the ith acquired data in the online test data of the operation sensitive parameters of the atmospheric tower, T i is the ith acquired data in the online test data of the sensitive parameters of the pressure tower, and Rmean is the average value of the operation sensitive parameters R of the atmospheric tower.

Description

Data-driven normal pressure tower top operation health state online assessment method Technical Field The invention relates to the technical field of pressure tower top operation state evaluation, in particular to a data-driven normal pressure tower top operation health state online evaluation method. Background Along with the increasing trend of global crude oil inferior quality and the use of auxiliary agents in the crude oil extraction process, the content of harmful substances such as salt, organic chloride and the like in crude oil entering an atmospheric and vacuum device is increased, so that the problems of low-temperature corrosion, ammonium salt scaling and the like frequently occur at the top of the atmospheric and vacuum device of a refinery are caused, and the stable operation of the refinery is seriously influenced. The atmospheric and vacuum distillation device is used as a first set of device for processing crude oil in petroleum refineries of petrochemical enterprises, and the long-period safe and healthy operation capability directly influences the healthy operation of subsequent devices such as catalytic cracking, coking and the like. Therefore, how to accurately evaluate the operation health state of the normal pressure tower top has important significance for guiding enterprises to develop safe production, reduce cost and increase efficiency. In recent years, only a few studies on a method for evaluating the health of the operating state of an atmospheric tower top are conducted on the mechanism of a specific phenomenon such as low-temperature corrosion and ammonium salt crystallization of the atmospheric tower top. The invention of China patent CN 105629930A discloses a real-time prediction method for atmospheric pressure overhead dew point corrosion, which comprises the steps of firstly calculating water dew point temperature based on flash evaporation of gas-hydrocarbon-water three-phase thermodynamic equilibrium, and then judging the risk probability of system dew point corrosion by comparing the difference between the system DCS operation temperature parameter and the calculated dew point temperature, thereby realizing the prediction of the real-time risk probability of atmospheric pressure overhead dew point corrosion. The invention patent CN 107220705A discloses a method for predicting the atmospheric pressure tower top dry point of an atmospheric and vacuum device, which is used for predicting the atmospheric pressure tower top dry point by logging in a prediction system, wherein the prediction system is arranged on a server, and the server is respectively connected with a real-time database system and a server of a LIMS system through network wires. The invention discloses a method and a device for testing the scale formation rate of the top of an atmospheric distillation tower of an oil refinery, wherein the device for testing the scale formation rate of the top of the atmospheric distillation tower of the oil refinery sequentially comprises a heating kettle, a glass filler tower, a tower head and an air cooler from bottom to top, wherein a pressure gauge and a material inlet are arranged on the heating kettle, a collector is arranged at the upper middle part of the tower head, a plurality of side lines of the glass filler tower body are respectively connected with a sampling collecting bottle and control the sampling amount through a stop valve, an injection port and a thermocouple are arranged on the air cooler, the glass filler tower and the tower head are provided with heat insulation layers, the collector is arranged at the middle upper part in the tower head and consists of cross supports, test pieces, fillers and curved fine transistors, the tower head section is radially provided with a cross support perpendicular to an axis and is used for supporting and hanging a scale collection groove, the filler is placed in the scale collection groove, the scale collection groove is connected with the curved fine transistors through threads, materials to be tested are added into the tower kettle, then the upper middle part of the tower head is provided with the scale collector in a detection area, and simultaneously, the method for testing the top of the atmospheric distillation tower can be quantitatively analyzed by the weight of the material to be tested, but the weight of the scale formation rate of the top of the tower can be quantitatively analyzed by the sample is not changed when the weight of the sample is required to be tested, but the weight of the scale formation rate is not changed. In the prior art, the health state of the operation of the atmospheric tower top is researched from the angles of atmospheric tower top dew point corrosion risk prediction, scaling rate prediction, dry point prediction and the like, but the influence of various factors on the health state of the operation of the atmospheric tower top is not comprehensively considered, and an online operat