Method for adjusting air-fuel ratio of oil field gas heating furnace

Abstract

The invention provides a method for adjusting the air-fuel ratio of an oil field gas heating furnace, which is characterized by comprising the following steps of: S1, monitoring the data of each point of a PLC (Programmable Logic Controller) in real time, and storing the obtained data into a database; s2, cleaning the warehousing data, wherein the process is as follows: screening based on monitoring items and index requirements of the gas heating furnace to obtain a cleaned data set D1; screening based on the flue gas oxygen content to obtain a data set D2; screening is carried out on the basis of an Isolation Forest isolation forest algorithm, and a data set D3 is obtained; and S3, performing statistical analysis on the data according to the data set D3 so as to obtain an optimal main air door opening degree recommendation adjustment table corresponding to different gas flows under a reasonable combustion working condition, and realizing real-time adjustment of the gas flows and the air distribution ratio according to the corresponding relation. According to the air-fuel ratio relation under different flows, the air-fuel ratio recommendation value table is automatically generated, and self-updating can be carried out regularly.

Description

Technical Field

[0001] The present invention relates to a method for adjusting the air-fuel ratio of an oilfield gas heating furnace. Background Art

[0002] An oilfield heating furnace is a device for heating media such as crude oil. That is, the heating furnace uses the heat energy of fuel combustion and flue gas to conduct energy to the heated medium, so as to achieve the heating effect. Reasonable adjustment of the gas-air ratio is of great significance for ensuring the full combustion of fuel. Usually, the method for the control system to realize automatic air distribution adjustment is to divide the interval of gas flow (or the opening of the regulating valve), and set a reasonable interval of damper opening for each interval. At present, the adjustment of the air ratio under different gases is completed by the on-site technicians adjusting and measuring while adjusting according to their experience to determine the set value. The main disadvantages of this way of adjusting the air distribution are that the adjustment workload of people is large. At the same time, when the environment and working conditions change, the original settings are not easy to meet the current needs, and the manual cannot arrive in time to adjust, and only relevant technical staff can give the experience value. Therefore, it is impossible to ensure the timely update of the gas-air ratio state. The present invention analyzes the changes in historical data of gas flow, main damper opening and oxygen content, accurately locates the situation where the air distribution is not ideal, and can realize the automatic adjustment and correction of the gas-air ratio without manual participation, ensuring that the furnace efficiency is always near an optimal position. Summary of the Invention

[0003] To overcome the defects existing in the prior art, the present invention proposes a method for automatically adjusting and updating the air-fuel ratio of an oilfield gas heating furnace, which is specifically realized by the following technical means: A method for automatically adjusting and updating the air-fuel ratio of an oilfield gas heating furnace, which includes the following steps: Step S1, monitor the data of each point of the PLC in real time, and store the obtained data in the database; Step S2, clean the data stored in the database, and the process is as follows: 1) Screen based on the monitoring items and index requirements of the gas heating furnace, including: setting several screening rules based on the monitoring items and index requirements of the gas heating furnace, filtering the original data stored in the database according to the screening rules, and obtaining the cleaned data set D1; 2) Screen based on the oxygen content in the flue gas, including: Define the ratio of the actually supplied air volume V to the theoretically required air volume V0 as the excess air coefficient α, that is

[0004] According to the on-site requirements, specify the upper and lower limits [αlow, αhigh] of the excess air coefficient α. Since the approximate relationship between the oxygen content [O2] in the flue gas and the excess air α is Among them, "21" represents the percentage of oxygen content in the air. Then, based on the upper and lower limits of the excess air coefficient α, [αlow, αhigh], the upper and lower limits of the flue gas oxygen content, [O2]low and [O2]high, can be determined. Filter the dataset D1 according to the upper and lower limits of the flue gas oxygen content to obtain the dataset D2. 3) Conduct screening based on the IsolationForest algorithm, which includes two stages. In the first stage, T isolation trees are trained to form an isolation forest. In the second stage, each sample point in the dataset D2 is brought into each isolation tree in the forest, and its average height is calculated. Then, its corresponding outlier score is calculated, and data cleaning is performed according to the size order of the outlier scores to obtain the dataset D3. In step S3, statistical analysis is performed on the data according to the dataset D3 to obtain a recommended adjustment table of the optimal main air damper opening corresponding to different gas flow rates under reasonable combustion conditions. According to this corresponding relationship, real-time adjustment of the gas flow rate and air distribution ratio is achieved. There are different rated power intervals, and an upper limit of the exhaust gas temperature and a lower limit of the thermal efficiency are set for each rated power interval. The specific process of the algorithm is as follows: The first stage 11) Given the dataset X = {x1, x2,..., xN}, where N is the number of samples, and the feature dimension of each sample is d, and xi ∈ X; randomly select ψ sample points as sub-samples and put them into the root node of an isolation tree. 12) Randomly specify a dimension p from d dimensions. Within the data range of the specified dimension p at the current node, randomly generate a cut point q, and the cut point is generated between the maximum and minimum values of the data in the current dimension p. 13) The selection of this cut point generates a hyperplane that divides the data space of the current node into 2 sub-spaces: put the points less than q in the current selected dimension on the left branch of the current node, and put the points greater than or equal to q on the right branch of the current node. 14) Recursively perform steps 2) and 3) on the left and right branch nodes of the node, continuously constructing new leaf nodes until there is only one data on the leaf node or the tree has grown to the set height. 15) Loop through steps 11) to 14) until T trees are generated to form an isolation forest, and the training ends. The second stage 21) For each data point xi, let it traverse each isolation tree, calculate its path length h(xi) in each isolation tree, and then calculate the outlier score value through the following formula: where s(xi,n) is the outlier score of the sample xi. The larger the value, the greater the possibility that the sample is an outlier point; is the average value of the path length when the given number of samples is n, and is used to normalize h(xi); H(i) = lni + 0.5772156649 is the harmonic series; E(h(xi)) is the expectation of the path length of the sample xi in T isolated trees; 22) Through step 21), the outlier score value of each sample is obtained, and an ordered sort is performed according to the outlier score size, so that data screening and cleaning can be carried out according to the outlier value size; Finally, determine the outlier ratio r of the entire data set, and filter the data with the top r outlier scores, thereby obtaining the data set D3.

[0005] Preferably, in step S3, the following operations are performed: Step S31, take the gas flow rate, main air damper opening, and flue gas oxygen content in the data set D3 as the analysis targets. Based on the gas flow rate G as the grouping basis, after sorting the values according to the gas flow rate size, the data set is divided into N different working areas, that is, [G1, G2,..., GN], and each working area Gk corresponds to a gas flow rate interval; Step S32, for the data in each working area Gk, perform an aggregation index analysis on the main air damper opening. The goal is to determine a representative value of the main air damper in this working interval to minimize the mean square sum loss error, that is where xi represents the main air damper value of the i-th data, n represents the total number of data in the Gk working area, and x^ represents the representative value of the main air damper data to be solved; According to the optimization theory, when the mean square sum error is the smallest, and the representative value Ak of the main air damper corresponding to the working area Gk can be obtained from this formula; according to this method, the representative values [A1, A2,..., AN] of the main air damper corresponding to each working interval [G1, G2,..., GN] are obtained; Step S33, according to the above steps S31 and S32, obtain the optimal main air damper opening recommended adjustment table corresponding to different gas flow rates from the historical operation data, that is, the air-fuel ratio recommendation table; During on-site operation, when the gas flow rate falls into a certain working area Gk, according to this air-fuel ratio recommendation table, the main air damper automatically switches to the corresponding value Ak to achieve gas-air ratio adjustment.

[0006] Preferably, when performing gas-air ratio adjustment, fine-tuning is performed through the opening of the micro air damper to ensure that the flue gas oxygen content is always within a reasonable range.

[0007] Preferably, set an update check period T, and perform the following update operations on the air-fuel ratio recommendation adjustment table every T time: Clean the stored data, including screening based on the monitoring items and index requirements of the gas heating furnace, and screening based on the IsolationForest algorithm to obtain the dataset DS1; Divide the working areas according to the gas flow rate. For each working area Gk, count the total number of data records Ak in this working area and the number of data records where the flue gas oxygen content index deviates from the upper and lower limits [O2]low, [O2]high

[0008] Set a threshold β, and at the same time define as the proportion of abnormal data that deviates from the normal upper and lower limits of the flue gas oxygen content; when a≥β, update the corresponding main air damper opening value of the working area Gk; when a<β, there is no need to update the corresponding main air damper opening value of the working area Gk; Through the above steps, a new air-fuel ratio recommendation table is obtained, and subsequent gas-air volume synchronization adjustment is based on this.

[0009] The beneficial effects of the present invention are: Based on the algorithm model of intelligent heating furnace gas-air distribution ratio adjustment with self-learning, according to the air-fuel ratio relationship under different flow rates, there is no need for manual adjustment and measurement to determine the set value. The software automatically generates a table of air-fuel ratio recommendation values and can perform self-update regularly, and can be applied to different types of gas heating furnaces. Specific embodiments

[0010] The following further describes the solution of the present application: A method for automatic adjustment and update of the air-fuel ratio of an oilfield gas heating furnace, which includes the following steps: Step S1, monitor the data of each point of the PLC in real time and store the obtained data in the database; Step S2, clean the stored data; Since the on-site data collected by the sensors may be diverse, such as sensor failure, unstable furnace conditions on site and other factors, it is necessary to clean the relevant data before data analysis to avoid misjudgment of the analysis results caused by interference factors. This part mainly conducts the data cleaning process from three aspects: 1) Screen based on the monitoring items and index requirements of the gas heating furnace, including: Set several screening rules based on the monitoring items and index requirements of the gas heating furnace, and filter the original stored data according to the screening rules to obtain the cleaned dataset D1; See Table 1 below, for heating furnaces with different rated powers, their corresponding flue gas temperatures and thermal efficiencies There are certain upper and lower limit requirements, and the original data in the database is screened according to this screening rule. For example, the rated power is 2.0 ≤ P rated ≤ 2.5, the flue gas temperature must be less than 200 °C, and the thermal efficiency must be higher than 82%. Filter according to this rule to obtain the cleaned dataset D1;

[0011] 2) Screening based on flue gas oxygen content; During the combustion process, it is very important to correctly select and control the mixing ratio of gas flow and air. The theoretical air requirement refers to the minimum amount of air required for complete combustion of the gas. However, since it is difficult to achieve complete uniformity in the mixing of gas and air, if only the theoretical air amount is provided in the actual combustion device, it is difficult to ensure the full mixing and combustion of gas and air. Therefore, in actual operation, the actual air amount should be greater than the theoretical air requirement, that is, a part of excess air needs to be supplied: Define the ratio of the actually supplied air volume V to the theoretical air requirement V0 as the excess air coefficient α, that is

[0012] According to the site requirements, specify the upper and lower limits [α low, α high] of the excess air coefficient α. Since the approximate relationship between the flue gas oxygen content [O2] and the excess air α is where "21" represents the percentage of oxygen content in the air, then the upper and lower limits [O2] low, [O2] high of the flue gas oxygen content can be determined based on the upper and lower limits [α low, α high] of the excess air coefficient α; Screen the dataset D1 according to the upper and lower limits of this flue gas oxygen content to obtain the dataset D2; At this time, the flue gas oxygen content in the dataset D2 is within the reasonable range of good combustion conditions; 3) Screening based on the IsolationForest algorithm; The above two steps screened and cleaned the data based on prior artificial rules. Since there are limitations in considering data screening only in one feature dimension, it would be more reasonable to further clean the data by comprehensively considering the influence of each feature dimension of the dataset; The IsolationForest algorithm includes two stages. In the first stage, T isolation trees are trained to form an isolation forest; In the second stage, each sample point of the dataset D2 is brought into each isolation tree in the forest, and its average height is calculated. Then, its corresponding outlier score is calculated, and the data is cleaned according to the size order of the outlier scores to obtain the dataset D3; Specifically: The first stage 11) Given the dataset X = {x1, x2, ……, xN}, N is the number of samples, and each sample The feature dimension is d, and xi ∈ X; randomly select ψ sample points as sub-samples and put them into the root node of an isolation tree; 12) Randomly specify a dimension p from d dimensions. Within the data range of the specified dimension p at the current node, randomly generate a cut point q, and the cut point is generated between the maximum value and the minimum value of the data in the current dimension p; 13) The selection of this cut point generates a hyperplane that divides the data space of the current node into 2 sub-spaces: place the points less than q in the current selected dimension on the left branch of the current node, and place the points greater than or equal to q on the right branch of the current node; 14) Recursively perform steps 2) and 3) on the left and right branch nodes of the node, continuously constructing new leaf nodes until there is only one data on the leaf node or the tree has grown to the set height; 15) Loop through steps 11) to 14) until T trees are generated to form an isolation forest, and the training ends; The second stage 21) For each data point xi, let it traverse each isolation tree, calculate its path length h(xi) in each isolation tree, and then calculate the outlier score value through the following formula: where s(xi,n) is the outlier score of sample xi, and the larger the value, the greater the possibility that the sample is an outlier; is the average value of the path length when the given number of samples is n, and is used to normalize h(xi); H(i) = lni + 0.5772156649 is the harmonic series; E(h(xi)) is the expectation of the path length of sample xi in T isolation trees; 22) Through step 21), obtain the outlier score value of each sample, sort them in an orderly manner according to the outlier score size, so that data screening and cleaning can be performed according to the outlier value size; Finally, determine the outlier ratio r (unit: %) of the entire data set, and filter the data with the top r outlier scores, thereby obtaining the data set D3; Step S3, perform statistical analysis on the data according to the data set D3 to obtain a recommended adjustment table for the optimal main air damper opening corresponding to different gas flow rates under reasonable combustion conditions (that is, the quality of the working condition is determined by the upper and lower limits of the flue gas oxygen content). According to this corresponding relationship, realize the real-time adjustment of the gas flow rate and the air distribution ratio. The specific implementation has the following operations: Step S31: Take the gas flow rate, main air damper opening, and flue gas oxygen content in the dataset D3 as the analysis targets. Based on the gas flow rate G as the grouping criterion, after sorting the numerical values according to the gas flow rate, divide the dataset into N different working regions, namely [G1, G2, ……, GN]. Each working region Gk corresponds to a gas flow rate interval. Step S32: For the data in each working region Gk, conduct an aggregated index analysis on the main air damper opening. The goal is to determine a representative value of the main air damper in this working interval to minimize the mean squared error loss, that is where xi represents the main air damper value of the i-th data, n represents the total number of data in the Gk working region, and x^ represents the representative value of the main air damper data to be solved. According to the optimization theory, when the mean squared error is minimized. From this equation, the representative value Ak of the main air damper corresponding to the working region Gk can be obtained. According to this method, the representative values [A1, A2, ……, AN] of the main air damper corresponding to each working interval [G1, G2, ……, GN] are obtained. Step S33: According to the above steps S31 and S32, obtain the recommended adjustment table of the optimal main air damper opening corresponding to different gas flow rates from the historical operation data, that is, the air-fuel ratio recommendation table. During on-site operation, when the gas flow rate falls into a certain working region Gk, according to this air-fuel ratio recommendation table, the main air damper automatically switches to the corresponding value Ak to achieve the adjustment of the gas-air ratio. For further precise adjustment, fine-tuning is carried out through the opening of the micro air damper to ensure that the flue gas oxygen content is always within a reasonable range.

[0013] The operation of the on-site heating furnace in the oilfield is not an idealized process. As the actual operation time increases, there will be certain interference factors in the heating furnace. Adjusting blindly according to experience or the existing air-fuel ratio recommendation table will result in a deviation of the combustion furnace condition, and the direct result is that the flue gas oxygen content deviates from the normal range. Therefore, we need to let the model go through a self-learning process to complete the automatic update of the air-fuel ratio recommendation adjustment table. The specific process is as follows: 1) Set an update check period T (for example, 1 month). Every T time, perform the following update operations on the air-fuel ratio recommendation adjustment table: 2) Clean the data stored in the database, including screening based on the monitoring items and index requirements of the gas heating furnace as in step S2, and screening based on the IsolationForest algorithm to obtain the dataset DS1. 3) Divide the working regions according to the gas flow rate. For each working region Gk, count the total number of data Ak in this working region and the number of data whose flue gas oxygen content index is outside the upper and lower limits [O2]low, [O2]high.

[0014] 4) Set a threshold β, and at the same time define as the proportion of abnormal data that deviates from the normal upper and lower limits of the oxygen content in the flue gas; when a ≥ β, update the main damper opening value corresponding to the working area Gk; when a < β, there is no need to update the main damper opening value corresponding to the working area Gk; 5) Through the above steps, obtain a new air-fuel ratio recommendation table again, and based on this, perform synchronous adjustment of gas and air volume subsequently.

[0015] The above preferred implementation manners should be regarded as examples of the implementation manners of the solution of this application. All technical deductions, replacements, improvements, etc. that are identical, similar to or based on the solution of this application should be regarded as within the protection scope of this patent.

Claims

1. A method for adjusting the air-fuel ratio of an oilfield gas heating furnace, characterized in that, It includes the following steps: Step S1, real-time monitor the data of each point of the PLC, and store the obtained data in the database; Step S2, clean the data stored in the database, and the process is as follows: 1) Filter based on the monitoring items and index requirements of the gas heating furnace, including: setting several filtering rules based on the monitoring items and index requirements of the gas heating furnace, and filtering the original data stored in the database according to the filtering rules to obtain the cleaned data set D1; 2) Filter based on the oxygen content in the flue gas, including: Define the ratio of the actual supplied air volume V to the theoretical air requirement V0 as the excess air coefficient α, i.e., ; According to the on-site requirements, specify the upper and lower limits [αlow, αhigh] of the excess air coefficient α. Since the approximate relationship between the flue gas oxygen content [O2] and the excess air α is where "21" represents the percentage of oxygen content in the air, then based on the upper and lower limits [αlow, αhigh] of the excess air coefficient α, the upper and lower limits [O2]low, [O2]high of the flue gas oxygen content can be determined; Filter the data set D1 according to the upper and lower limits of the oxygen content in the flue gas to obtain the data set D2; 3) Filter based on the IsolationForest isolation forest algorithm, which includes two stages. In the first stage, T isolation trees are trained to form an isolation forest; in the second stage, each sample point of the data set D2 is brought into each isolation tree in the forest, and its average height is calculated, and then its corresponding outlier score is calculated. Sort according to the outlier score size to clean the data and obtain the data set D3; Step S3, statistically analyze the data according to the data set D3 to obtain the optimal main damper opening recommendation adjustment table corresponding to different gas flow rates under reasonable combustion conditions. According to this corresponding relationship, realize the real-time adjustment of the gas flow rate and the air distribution ratio; The step S2 further includes: setting different rated power intervals, and setting an upper limit of the exhaust gas temperature and a lower limit of the thermal efficiency for each rated power interval; In the step S2, when filtering based on the IsolationForest isolation forest algorithm, the specific process of the algorithm is as follows: The first stage 11) Given a dataset X = {x1, x2, ……, xN}, where N is the number of samples, and the feature dimension of each sample is d, and xi ∈ X; randomly select ψ sample points as sub-samples and put them into the root node of an isolation tree; 12) Randomly specify a dimension p from d dimensions. In the data range of the specified dimension p at the current node, randomly generate a cut point q, and the cut point is generated between the maximum value and the minimum value of the data of the current dimension p; 13) The selection of this cut point generates a hyperplane that divides the data space of the current node into 2 subspaces: place the points less than q under the current selected dimension on the left branch of the current node, and place the points greater than or equal to q on the right branch of the current node; 14) Recursively perform steps 2) and 3) on the left and right branch nodes of the node, and continuously construct new leaf nodes until there is only one data on the leaf node or the tree has grown to the set height; 15) Loop through steps 11) to 14) until T trees are generated to form an isolation forest and the training ends; The second stage 21) For each data point xi, let it traverse each isolation tree and calculate its path length h(xi) in each isolation tree, and then calculate the outlier score value through the following formula: Where s(xi,n) is the outlier score of the sample xi, and the larger the value, the greater the possibility that the sample is an outlier; It is the average value of the path length for a given number of samples n and is used to normalize h(xi); H(i) = lni + 0.5772156649 is the harmonic series; E(h(xi)) is the expectation of the path length of the sample xi in T isolation trees; 22) Through step 21), obtain the outlier score value of each sample, sort them in an orderly manner according to the outlier score size, so that the data can be filtered and cleaned according to the outlier size; Finally, determine the outlier proportion r of the entire data set, and filter the data with the top r outlier scores to obtain the data set D3.

2. The method for adjusting the air-fuel ratio of an oilfield gas heating furnace according to claim 1, wherein In step S3, the following operations are performed: In step S31, the gas flow rate, main air damper opening, and flue gas oxygen content in the data set D3 are used as the analysis targets. Based on the gas flow rate G as the grouping basis, after sorting the values according to the gas flow rate, the data set is divided into N different working regions, namely [G1, G2, ……, GN], and each working region Gk corresponds to a gas flow rate interval. Step S32: For the data in each working area Gk, perform an aggregated index analysis on the main air damper opening. The goal is to determine a representative value of the main air damper in this working range to minimize the sum of squared loss errors, that is , where xi represents the value of the main air damper for the i-th data, and n represents the total number of data in this G k working area, and x^ represents the size of the representative value of the main air damper data to be solved; According to the optimization theory, when the mean square error is minimized, and the representative value Ak of the main damper corresponding to the working area Gk can be obtained from this formula; according to this method, the representative values [A1, A2, ……, AN] of the main damper corresponding to each working interval [G1, G2, ……, GN] are obtained; In step S33, according to the above steps S31 and S32, the optimal main air damper opening recommended adjustment table corresponding to different gas flow rates, that is, the air-fuel ratio recommendation table, is obtained from the historical operation data. During on-site operation, when the gas flow rate falls into a certain working region Gk, according to this air-fuel ratio recommendation table, the main air damper automatically switches to the corresponding value Ak to achieve the gas-air ratio adjustment.

3. The method for adjusting the air-fuel ratio of an oilfield gas heating furnace according to claim 2, wherein When performing the gas-air ratio adjustment, fine-tuning is carried out through the micro air damper opening to ensure that the flue gas oxygen content is always within a reasonable range.

4. The method for adjusting the air-fuel ratio of an oilfield gas heating furnace according to claim 1, characterized in that Set an update check period T, and perform the following update operations on the air-fuel ratio recommended adjustment table every T time: Clean the data stored in the database, including screening based on the monitoring items and index requirements of the gas heating furnace, and screening based on the IsolationForest isolation forest algorithm to obtain the data set DS1. Divide the working area according to the gas flow. For each working area Gk, count the total number of data records Ak in this working area and the number of data records where the flue gas oxygen content index deviates from the upper and lower limits [O2]low and [O2]high. Set a threshold β, and at the same time define as the proportion of abnormal data that deviates from the normal upper and lower limits of the flue gas oxygen content; when a≥β, update the corresponding main air damper opening value for the working area Gk; when a<β, there is no need to update the corresponding main air damper opening value for the working area Gk. Through the above steps, a new air-fuel ratio recommendation table is obtained, and subsequent gas-air volume synchronous adjustment is based on this.

Images