Model for reducing octane number loss in gasoline refining process
1. Background
Gasoline is the main fuel for small vehicles, and exhaust emissions from gasoline combustion have an important impact on the atmospheric environment. For this reason, countries all over the world have formulated increasingly stringent gasoline quality standards (see the table below). The focus of gasoline cleaning is to reduce the sulfur and olefin content in gasoline while maintaining its octane number as much as possible.
my country's dependence on foreign crude oil exceeds 70%, and most of it is sulfur-containing and high-sulfur crude oil in the Middle East. The heavy oil in crude oil usually accounts for 40-60%, and this part of the heavy oil (the impurity content represented by sulfur is also high) is difficult to directly use. In order to effectively use heavy oil resources, my country has vigorously developed heavy oil lightening technology with catalytic cracking as the core, which converts heavy oil into gasoline, diesel and low-carbon olefins. More than 70% of gasoline is produced by catalytic cracking, so refined gasoline More than 95% of sulfur and olefins come from catalytic cracking gasoline. Therefore, FCC gasoline must be refined to meet the gasoline quality requirements.
The octane number (expressed in RON) is the most important indicator reflecting the combustion performance of gasoline and is used as the commercial brand name of gasoline (for example, 89#, 92#, 95#). The prior art generally reduces the gasoline octane number in the process of desulfurization and olefin reduction of FCC gasoline. Each decrease in the octane number by 1 unit is equivalent to a loss of about 150 yuan/ton. Taking a 1 million tons/year catalytic cracking gasoline refinery unit as an example, if it can reduce the RON loss by 0.3 units, its economic benefit will reach 45 million yuan.
The modeling of chemical process is generally achieved through data association or mechanism modeling, and certain results have been achieved. However, due to the complexity of the refining process and the diversity of equipment, their operating variables (control variables) have a highly nonlinear and strongly coupled relationship with each other, and there are relatively few variables in the traditional data association model, and the mechanism is built The mold has high requirements for raw material analysis and does not respond in time to process optimization, so the effect is not ideal.
A petrochemical enterprise's catalytic cracking gasoline refinery desulfurization unit has been in operation for 4 years and accumulated a large amount of historical data. The average loss of gasoline product octane number is 1.37 units, while the minimum loss value of similar units is only 0.6 units. So there is more room for optimization. Participating graduate students are invited to explore the use of data mining technology to solve chemical process modeling problems.
2. The target is
based on 325 data samples collected from the catalytic cracking gasoline refining unit (each data sample has 354 operating variables), through data mining technology to establish a gasoline octane number (RON) loss prediction model, and give The optimized operating conditions of each sample ensure the desulfurization effect of gasoline products (Euro VI and China VI standards are not greater than 10μg/g, but in order to leave room for the operation of the enterprise device, this modeling requires that the product sulfur content is not greater than 5μg /g), try to reduce the gasoline octane loss to more than 30%.
Three, the problem
- Data processing: Please refer to the pre-processing results of the industrial data (see appendix 1 "325 data sample data.xlsx") in the past 4 years, and pre-process the data samples No. 285 and No. 313 according to the "Sample Determination Method" Process (for the original data, see Appendix 3 "Sample No. 285 and 313 Original Data.xlsx") and add the processed data to the corresponding sample number in Appendix 1 for the following research.
- Finding the main variables for modeling:
Since the catalytic cracking gasoline refining process is continuous, although the operating variables are sampled every 3 minutes, the measurement of the octane number (dependent variable) is more troublesome and cannot be matched only twice a week. However, according to the actual situation, it can be considered that the measured value of the octane number is the comprehensive effect of the manipulated variable within two hours before the measurement time. Therefore, the average value of the manipulated variable within two hours is taken in the pretreatment to correspond to the measured value of the octane number. This resulted in 325 samples (see Annex 1).
The establishment of the octane loss reduction model involves variables including 7 raw material properties, 2 spent adsorbent properties, 2 regenerated adsorbent properties, 2 product properties and other variables, as well as another 354 operating variables (a total of 367 variables). Engineering technology The method of dimensionality reduction first and then modeling is often used in applications, which helps to ignore secondary factors and discover and analyze the main variables and factors that affect the model. Therefore, please use the 325 sample data provided (see appendix 1) to filter out the main modeling variables from the 367 operating variables through dimensionality reduction to make them as representative and independent as possible (for ease of engineering applications) , It is recommended that the main variables after dimensionality reduction be less than 30), and please explain in detail the selection process and rationality of the main variables of the modeling. (Hint: Please consider the octane number of the raw material as one of the modeling variables). - Establish an octane number (RON) loss prediction model: using the above samples and modeling main variables, establish an octane number (RON) loss prediction model through data mining technology, and perform model verification.
- Optimization of the main variable operation plan: It is required to use your model to obtain 325 data samples (see appendix 4 "325 data sample data.xlsx") under the premise that the product sulfur content is not greater than 5μg/g, the octane number (RON) The optimized operating conditions of the main variables corresponding to the samples whose loss reduction is greater than 30% (the properties of raw materials, spent adsorbents, and regenerated adsorbents remain unchanged during the optimization process, subject to their data in the sample).
- Visual display of the model: In order to stabilize the production of industrial plants, the optimized main operating variables (ie: the main variables in question 2) can only be gradually adjusted in place. Please check the sample No. 133 (raw material properties, standby adsorbent and regeneration The property data of the adsorbent remains unchanged, subject to the data in the sample), and the corresponding gasoline octane number and sulfur content change trajectory during the optimization and adjustment of its main operating variables are displayed graphically. (For each main operating variable, the allowable adjustment range value Δ is shown in Appendix 4 "354 operating variable information.xlsx").
Attachment:
Attachment 1: 325 sample data.xlsx
Attachment 2: Sample determination method.docx
Attachment 3: No. 285 and 313 sample original data.xlsx
Attachment 4: 354 operating variable information.xlsx