A method for predicting oxygen consumption in silicon liquid oxidation refining based on historical data

Abstract

The present invention aims at the problem that the amount of oxygen blown into the silicon bag can only be judged by manual experience during the silicon liquid oxidation refining process, and the appropriate amount of oxygen cannot be introduced. A method for predicting the oxygen consumption of silicon liquid oxidation refining based on historical data is proposed. The method first collects the historical data of silicon liquid oxidation refining, and screens the historical data to obtain the data source; then the difference between the refining start temperature and the refining end temperature in the data source is recorded as the temperature drop value, all the temperature drop values ​​are pre-processed to obtain the error coefficient; and the refining oxygen consumption prediction value is calculated according to the given calculation model. The refining oxygen consumption prediction value obtained by the calculation model of the present invention is accurate and can be actually applied to the silicon liquid oxidation refining process. The same amount of oxygen as the refining oxygen consumption prediction value can be introduced into the silicon bag, which will not cause too much or too little oxygen to be blown into the silicon bag, and can meet the requirements of on-site use.

Description

Technical Field

[0001] The invention belongs to the technical field of industrial silicon smelting, and in particular relates to a method for predicting oxygen consumption in silicon liquid oxidation refining based on historical data. Background Art

[0002] The electrothermal smelting method of industrial silicon is to add a certain proportion of silica and carbonaceous reducing agent as raw materials into the submerged arc furnace, input electric energy into the submerged arc furnace through electrodes for heating, promote the melting of raw materials and chemical reactions, and the reaction ends after a certain period of time, and silicon liquid flows out from the furnace nozzle. Since the raw materials contain a large amount of impurities, some of the impurities in the raw materials and reducing agents will inevitably be brought into the silicon liquid under high temperature and strong reduction conditions. Among them, the main impurities are Al, Ca, etc., which seriously affect the performance and use of industrial silicon products. In order to reduce the Al and Ca content in industrial silicon products, the commonly used method to remove impurities is oxidation refining. A certain amount of oxygen is blown into and out of the bottom of the silicon bag. Oxygen enters the silicon liquid. At a certain temperature, Al and Ca in the silicon liquid are oxidized by oxygen, which accelerates slag formation and removes impurities as soon as possible, thereby realizing slag-silicon separation.

[0003] However, during the refining process, due to the large temperature difference of the silicon liquid, it is difficult to determine the amount of oxygen that should be blown in. It can only be determined by manual observation during the oxygen blowing process. If the smoke concentration above the silicon liquid becomes clear, a large amount of tofu-like aggregation is formed around the edge of the silicon liquid when it rolls, and the silicon slag at the edge of the silicon bag is slightly sticky, it means that the refining is completed and the oxygen injection can be stopped.

[0004] Due to the high temperature and dusty environment, employees are unable to judge the state of the silicon liquid normally, often causing too much or too little oxygen to be blown in. If too much oxygen is blown in, the remaining elements in the silicon liquid will be oxidized, including the oxidation of silicon, which will directly cause the loss of silicon liquid; and in the oxidation refining process, the impurity elements react with oxygen to form oxides. Since the oxidation reaction is an exothermic reaction, it will release a high amount of heat, causing the temperature of the silicon liquid to rise, and the silicon bag refractory will be corroded at high temperature for a long time, which will reduce the service life of the silicon bag. If too little oxygen is blown in, the oxidation reaction will not be sufficient, and the heat released will be low, which will cause the temperature of the silicon liquid to drop, making it difficult to separate the silicon liquid and silicon slag, and the silicon liquid will stick to the wall of the silicon bag, thus affecting the output and quality of industrial silicon products.

[0005] Therefore, it is necessary to design a method to predict the oxygen consumption during oxidative refining of silicon liquid. Summary of the invention

[0006] The purpose of the present invention is to address the above-mentioned problems and to provide a method for predicting oxygen consumption in silicon liquid oxidation refining based on historical data. The method can predict the amount of oxygen required for the silicon liquid oxidation refining process, thereby facilitating the introduction of an appropriate amount of oxygen into the silicon bag in actual production.

[0007] The purpose of the present invention is implemented by the following technical solutions:

[0008] A method for predicting oxygen consumption of silicon liquid oxidation refining based on historical data comprises the following steps:

[0009] A: Collect historical data of silicon liquid oxidation refining, and screen the historical data to obtain a data source;

[0010] B: The difference between the refining start temperature and the refining end temperature in the data source is recorded as the temperature drop value, and all the temperature drop values ​​are preprocessed to obtain the error coefficient; in the actual refining process, the reaction between the blown oxygen and Al and Ca is an exothermic reaction. According to different refining start temperatures and in order to achieve different refining end temperatures, the oxygen blowing amount will be adjusted accordingly to ensure the casting temperature. Therefore, the temperature change, i.e., the temperature drop value, is used as an influencing factor to calculate the error coefficient;

[0011] C: The predicted value of refining oxygen consumption calculated according to formula (1);

[0012] YL=(0.88*(Al a -Al b )+0.4*(Ca a -Ca b ))*M*(RE+1) / ρ 空 (1);

[0013] Among them, YL is the predicted value of refining oxygen consumption, Al a Al content of the sample out of the furnace, Al b To end the refining, Al content and Ca a is the Ca content of the furnace sample, Ca b is the Ca content at the end of refining, M is the weight of silicon liquid, RE is the error coefficient, ρ 空 is the air density. According to the formula (1), the predicted value of refining oxygen consumption can be calculated, and the prediction is accurate and can meet the needs of on-site use.

[0014] Furthermore, the historical data includes Al content of the out-of-furnace sample, Ca content of the out-of-furnace sample, refining start temperature, refining end temperature, actual refining oxygen consumption of this furnace, silicon liquid weight, Al content at the end of refining, and Ca content at the end of refining.

[0015] Furthermore, in step A, the screening includes: screening out data in the historical data that simultaneously satisfies the conditions that the Al content at the end of refining is ≤0.2%, the Ca content at the end of refining is ≤0.1%, and the refining end temperature is 1500~1560°C as the data source, so as to remove abnormal data and interference in the historical data set.

[0016] Furthermore, in step B, the pretreatment comprises the following steps:

[0017] B1: Substitute the Al content of the sample out of the furnace, the Al content at the end of refining, the Ca content of the sample out of the furnace, the Ca content at the end of refining, the weight of the silicon liquid, and the air density in the data source into formula (2) to calculate the theoretical oxygen consumption of each silicon liquid oxidation refining;

[0018] LL=(0.88*(Al a -Al b )+0.4*(Ca a -Ca b ))*M* / ρ 空 (2);

[0019] Among them, LL is the theoretical oxygen consumption;

[0020] B2: Substituting the actual refining oxygen consumption of the current furnace in the data source and the theoretical oxygen consumption into formula (3) to calculate the error coefficient;

[0021] RE = (SL-LL) / SL (3);

[0022] Among them, SL is the actual refining oxygen consumption of this furnace;

[0023] B3: Perform linear fitting on the temperature drop value and the error coefficient to obtain formula (4);

[0024] RE = f * CZ + b (4);

[0025] Among them, f is the regression coefficient, CZ is the temperature drop value, and b is the error value.

[0026] Beneficial effect: The present invention is a method for predicting oxygen consumption in silicon liquid oxidation refining, which provides influencing factors and calculation models that affect oxygen consumption in silicon liquid oxidation refining, and can calculate the predicted value of refining oxygen consumption based on the field data of refining. The predicted value of refining oxygen consumption obtained by the calculation model of the present invention is accurate, and the amount of oxygen required to be introduced in the silicon liquid oxidation refining process can be obtained through prediction, which is convenient for introducing a suitable amount of oxygen into the silicon bag in actual production, and will not cause too much or too little oxygen to be blown into the silicon bag. It can be actually applied to the silicon liquid oxidation refining process to meet field use. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] Example 1

[0029] A method for predicting oxygen consumption in silicon liquid oxidation refining based on historical data comprises step A, step B and step C.

[0030] Step A: Collect historical data of silicon liquid oxidation refining, and screen the historical data to obtain the data source; the historical data includes Al content of the furnace sample, Ca content of the furnace sample, refining start temperature, refining end temperature, actual refining oxygen consumption of this furnace, silicon liquid weight, Al content at the end of refining, and Ca content at the end of refining; the data that simultaneously meets the conditions of Al content at the end of refining ≤0.2%, Ca content at the end of refining ≤0.1%, and refining end temperature of 1500~1560℃ in the historical data are screened out as the data source to remove abnormal data and interference in the historical data set.

[0031] Step B: Record the difference between the refining start temperature and the refining end temperature in the data source as the temperature drop value, pre-process all the temperature drop values ​​to obtain the error coefficient;

[0032] The preprocessing includes step B1, step B2, and step B3:

[0033] Step B1: Substitute the Al content of the sample out of the furnace, the Al content at the end of refining, the Ca content of the sample out of the furnace, the Ca content at the end of refining, the weight of the silicon liquid, and the air density in the data source into formula (2) to calculate the theoretical oxygen consumption of each silicon liquid oxidation refining;

[0034] LL=(0.88*(Al a -Al b )+0.4*(Ca a -Ca b ))*M* / ρ 空 (2);

[0035] Among them, LL is the theoretical oxygen consumption, Al a Al content of the sample out of the furnace, Al b To end the refining, Al content and Ca a is the Ca content of the furnace sample, Ca b is the Ca content at the end of refining, M is the weight of silicon liquid, ρ 空 is the air density;

[0036] Step B2: Substitute the actual refining oxygen consumption and theoretical oxygen consumption of the current furnace in the data source into formula (3) to calculate the error coefficient;

[0037] RE = (SL-LL) / SL (3);

[0038] Among them, RE is the error coefficient, SL is the actual refining oxygen consumption of this furnace;

[0039] Step B3: Perform linear fitting on the temperature drop value and the error coefficient to obtain equation (4);

[0040] RE = f * CZ + b (4);

[0041] Among them, f is the regression coefficient, CZ is the temperature drop value, and b is the error value.

[0042] Step C: Calculate the predicted value of refining oxygen consumption according to formula (1);

[0043] YL=(0.88*(Al a -Al b )+0.4*(Ca a -Ca b ))*M*(RE+1) / ρ 空 (1);

[0044] Among them, YL is the predicted value of refining oxygen consumption.

[0045] Example 2

[0046] A method for predicting oxygen consumption in silicon liquid oxidation refining based on historical data comprises step A, step B and step C.

[0047] Step A: Collect historical data of silicon liquid oxidation refining from the ore-fired furnace and silicon bag in a silicon production workshop, and screen the historical data to obtain the data source; the historical data include Al content of the furnace sample, Ca content of the furnace sample, refining start temperature, refining end temperature, actual refining oxygen consumption of this furnace, silicon liquid weight, Al content at the end of refining, and Ca content at the end of refining; screen out the data in the historical data that simultaneously meet the conditions of Al content at the end of refining ≤0.2%, Ca content at the end of refining ≤0.1%, and refining end temperature of 1500~1560℃ as the data source to remove abnormal data and interference in the historical data set.

[0048] Step B: Record the difference between the refining start temperature and the refining end temperature in the data source as the temperature drop value, pre-process all the temperature drop values ​​to obtain the error coefficient;

[0049] The preprocessing includes step B1, step B2, and step B3:

[0050] Step B1: Substitute the Al content of the sample out of the furnace, the Al content at the end of refining, the Ca content of the sample out of the furnace, the Ca content at the end of refining, the weight of the silicon liquid, and the air density in the data source into formula (2) to calculate the theoretical oxygen consumption of each silicon liquid oxidation refining;

[0051] LL=(0.88*(Al a -Al b )+0.4*(Ca a -Cab ))*M* / ρ 空 (2);

[0052] Where LL is the theoretical oxygen consumption (m 3 )、Al a Al content (%), Al b The Al content (%) and Ca content at the end of refining a Ca content (%), Ca b is the Ca content at the end of refining (%), M is the weight of silicon liquid (kg), ρ 空 is the air density (kg / m 3 );

[0053] Step B2: Substitute the actual refining oxygen consumption and theoretical oxygen consumption of the current furnace in the data source into formula (3) to calculate the error coefficient;

[0054] RE = (SL-LL) / SL (3);

[0055] Among them, RE is the error coefficient, SL is the actual refining oxygen consumption of this furnace (m 3 );

[0056] Step B3: Perform linear fitting on the temperature drop value and the error coefficient to obtain equation (4), i.e., the linear regression equation of the temperature drop value and the error coefficient; the fitting results are shown in Table 1-3;

[0057] RE = 1.724 - 0.006 * CZ (4);

[0058] Among them, CZ is the temperature drop value (℃).

[0059] Table 1 Analysis of variance

[0060]

[0061] Table 2 Model summary

[0062]

[0063] Table 3 Regression coefficients

[0064]

[0065] As can be seen from Table 1-3, the Sig shows a significant level of <0.005, the hypothesis test is valid, and the regression equation is significant; the adjusted R 2 =0.739, indicating that the independent variable, namely the temperature drop value, can explain 73.9% of the variation of the error coefficient, and the result of the Debin-Watson test is around 2, indicating that the sample meets the independence; the error coefficient and the temperature drop value are significantly related.

[0066] Step C: Calculate the predicted value of refining oxygen consumption according to formula (1);

[0067] YL=(0.88*(Al a -Al b )+0.4*(Ca a -Ca b ))*M*(2.724-0.006*CZ) / ρ 空 (1);

[0068] Among them, YL is the predicted value of refining oxygen consumption (m 3 ).

[0069] Example 3

[0070] In an actual silicon liquid oxidation refining process in a silicon production workshop, the Al content of the furnace sample was 2.1%, the Ca content of the furnace sample was 1.9%, the refining start temperature was 1650℃, the weight of the silicon liquid was 5000kg, the target refining end temperature was 1550℃, the target refining end Al content was 0.2%, and the target refining end Ca content was 0.1%. 空 =1.429kg / m 3 According to the refining start temperature of 1650℃ - refining end temperature of 1550℃, the temperature drop value is calculated to be 100℃.

[0071] According to formula (1) of Example 2, YL=(0.88*(2.1%-0.2%)+0.4*(1.9%-0.1%))*5000kg*(2.724-0.006*100℃) / 1.429kg / m 3 =178m 3 According to the on-site experience, the actual refining oxygen consumption of this furnace was measured to be 175m 3 , the error between the predicted value of refining oxygen consumption and the actual refining oxygen consumption of this furnace is 0.016. The error is small, indicating that the predicted value of refining oxygen consumption predicted by the method for predicting oxygen consumption of silicon liquid oxidation refining of the present invention is accurate, and can be applied to silicon liquid oxidation refining process, through the Al content of the furnace sample, the Ca content of the furnace sample, the refining start temperature, the weight of silicon liquid, the target refining end temperature, the target refining end Al content, the target refining end Ca content, ρ 空 The predicted value of refining oxygen consumption is calculated using the same parameters, and the same amount of oxygen as the predicted value of refining oxygen consumption is introduced into the silicon bag.

[0072] Example 4

[0073] In an actual silicon liquid oxidation refining process in a silicon production workshop, the Al content of the furnace sample was 1.4%, the Ca content of the furnace sample was 1.5%, the refining start temperature was 1670℃, the weight of the silicon liquid was 5500kg, the target refining end temperature was 1550℃, the target refining end Al content was 0.2%, and the target refining end Ca content was 0.1%. 空 =1.429kg / m 3 According to the refining start temperature of 1670℃ - refining end temperature of 1550℃, the temperature drop value is calculated to be 120℃.

[0074] According to formula (1) of Example 2, YL=(0.88*(1.4%-0.2%)+0.4*(1.5%-0.1%))*5500kg*(2.724-0.006*120℃) / 1.429kg / m 3 =125m 3 According to the on-site experience, the actual refining oxygen consumption of this furnace was measured to be 127m 3 , the error between the predicted value of refining oxygen consumption and the actual refining oxygen consumption of this batch is 0.018. The error is small, indicating that the predicted value of refining oxygen consumption obtained by the method for predicting oxygen consumption of silicon liquid oxidation refining of the present invention is accurate, and can be applied to silicon liquid oxidation refining process, through the Al content of the furnace sample, the Ca content of the furnace sample, the refining start temperature, the weight of silicon liquid, the target refining end temperature, the target refining end Al content, the target refining end Ca content, ρ 空 The predicted value of refining oxygen consumption is calculated using the same parameters, and the same amount of oxygen as the predicted value of refining oxygen consumption is introduced into the silicon bag.

[0075] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting oxygen consumption of silicon liquid oxidation refining based on historical data, characterized in that: It includes the following steps: A: Collect historical data of silicon liquid oxidation refining, and screen the historical data to obtain a data source; B: Record the difference between the refining start temperature and the refining end temperature in the data source as the temperature drop value, and pre-process all the temperature drop values ​​to obtain an error coefficient; The pretreatment The following steps are included: B1: Substitute the Al content of the sample out of the furnace, the Al content at the end of refining, the Ca content of the sample out of the furnace, the Ca content at the end of refining, the weight of the silicon liquid, and the air density in the data source into formula (2) to calculate the theoretical oxygen consumption of each silicon liquid oxidation refining; LL=(0.88*(Al a -Al b )+0.4*(Ca a -Ca b ))*M / ρ 空 (2); Among them, LL is the theoretical oxygen consumption, Al a Al content of the sample out of the furnace, Al b To end the refining, Al content and Ca a is the Ca content of the furnace sample, Ca b is the Ca content at the end of refining, M is the weight of silicon liquid, ρ 空 is the air density; B2: Substituting the actual refining oxygen consumption of the current furnace in the data source and the theoretical oxygen consumption into formula (3) to calculate the error coefficient; RE = (SL-LL) / SL (3); Among them, RE is the error coefficient, SL is the actual refining oxygen consumption of this furnace; B3: Perform linear fitting on the temperature drop value and the error coefficient to obtain formula (4); RE = f * CZ + b (4); Among them, f is the regression coefficient, CZ is the temperature drop value, and b is the error value; C: The predicted value of refining oxygen consumption calculated according to formula (1); YL=(0.88*(Al a -Al b )+0.4*(Ca a -Approx b ))*M*(RE+1) / ρ 空 (1) Among them, YL is the predicted value of refining oxygen consumption.

2. The method for predicting oxygen consumption of silicon liquid oxidation refining based on historical data according to claim 1 is characterized in that: The historical data include Al content of the out-of-furnace sample, Ca content of the out-of-furnace sample, refining start temperature, refining end temperature, actual refining oxygen consumption of this furnace, silicon liquid weight, Al content at the end of refining, and Ca content at the end of refining.

3. The method for predicting oxygen consumption of silicon liquid oxidation refining based on historical data according to claim 1 is characterized in that: In step A, the screening includes: screening out data in the historical data that simultaneously satisfies the conditions that the Al content at the end of refining is ≤0.2%, the Ca content at the end of refining is ≤0.1%, and the refining end temperature is 1500-1560°C as the data source, so as to remove abnormal data and interference in the historical data set.

Images