Polycrystalline silicon reduction furnace control method based on multiple control variables of forward mechanism model

Abstract

The invention relates to a polycrystalline silicon reduction furnace control method based on multiple control variables of a forward mechanism model, and the method comprises the following steps: S1, collecting and calculating a reaction state, namely a reflected real-time deposition rate and a reaction heat loss rate; the control parameters in the PID control process are determined based on the difference value of the real-time deposition rate curve and a target deposition rate curve, a solid line deposition curve is generated based on the real-time deposition rate of the reaction process sequence, and the target deposition rate curve is a preset curve; s3, executing a PID control process: performing PID control on the hydrogen input amount and the reaction current of the reaction by using the control parameters in the step S2; according to the method, the optimal deposition rate curve is determined through data analysis, and the deposition rate is calculated in real time in the control process. And the ratio of current heat in the furnace is calculated by utilizing a thermodynamic formula, and hydrogen and current are accurately regulated and controlled by combining the ratio with a deposition rate difference value, so that the stability and the accuracy of the production process are remarkably improved.

Description

Technical Field

[0001] The invention relates to the technical field of polysilicon reduction control, and in particular to a polysilicon reduction furnace control method based on a forward mechanism model with multiple control variables. Background Art

[0002] Polysilicon is the core raw material for the manufacture of solar photovoltaic devices. The optimization and efficiency improvement of its production process have a profound impact on the development of the entire renewable energy industry. In the context of the current global energy transformation and low-carbon economy, the photovoltaic industry is expanding rapidly, and the demand for polysilicon has also surged. Therefore, improving the production efficiency and quality of polysilicon while reducing its production cost is a key issue in the industry.

[0003] The modified Siemens process is currently the most commonly used method for producing polysilicon. It uses chemical vapor deposition (CVD) technology to react hydrogen and trichlorosilane (SiHCl3) under high temperature conditions to deposit silicon on silicon rods to form polysilicon. This process needs to be carried out in a high-temperature reduction furnace, and precise control of the furnace temperature is the key to ensuring the quality and production efficiency of polysilicon. However, due to technical and equipment limitations, it is unrealistic to install sensors in the furnace, and installing sensors outside the furnace cannot accurately measure, which poses a great challenge to temperature control.

[0004] Traditional control methods usually rely on the operator's experience and observation, judging the reaction in the furnace by observing the power curve, exhaust gas temperature, voltage curve, etc., and then adjusting the hydrogen, current, etc. based on manual experience. However, these curves are only the apparent reaction curves of the reduction furnace, not the core parameter indicators. In addition, manual observation is subjective and has errors, which may lead to unstable and inaccurate control, thus affecting the yield and quality of polysilicon. Summary of the invention

[0005] In view of the fact that the existing polysilicon reduction control process relies on apparent curves such as voltage, current, and electric power, its control process is unstable and it is difficult to reliably control the polysilicon output, quality, and unit power consumption of the reduction process. The present invention provides a polysilicon reduction furnace control method based on a forward mechanism model with multiple control variables.

[0006] The technical solution of the present invention provides a polysilicon reduction furnace control method based on a forward mechanism model with multiple control variables, comprising the following steps.

[0007] S1. Collect and calculate the reaction state: the reaction state includes the real-time deposition rate ΔM of the reaction and the heat loss rate ΔQ of the reaction; the real-time deposition rate ΔM is the rate of change of the mass M of the polysilicon rods in the reduction furnace, and the heat loss rate ΔQ is the ratio of the heat loss Q of the reaction to the total energy E input; S2. Inferring control parameters: Determining control parameters in the PID control process based on the difference between the real-time deposition rate curve and the target deposition rate curve, the control parameters include a proportional adjustment parameter P, an integral adjustment parameter I and a differential adjustment parameter D; wherein the solid line deposition curve is generated based on the real-time deposition rate ΔM of the reaction process sequence, and the target deposition rate curve is a preset curve; S3, execute PID control process: use the control parameters in step S2 to perform PID control on the amount of hydrogen input and the reaction current; adjust the amount of hydrogen input ; Reaction current adjustment ;k p1 , k i1 , k d1 , k p2 , k i2 , k d2 is the control coefficient; where k p2 , k i2 , k d2 At least one of has a preset value set, and its value is related to the heat loss ΔQ of the reaction.

[0008] Preferably, in the step S1 of collecting and calculating the reaction state, the mass M of the polycrystalline silicon rod is:

[0009] Where, m is the number of current loops in the polysilicon reduction furnace, M i is the mass of the polysilicon rod of the i-th phase loop, calculated by the following formula:

[0010] Where ρ2 is the resistivity of the silicon rod, L is the total length of the silicon rod, and n i I is the number of silicon rod pairs in the i-th phase loop calculated in the reduction furnace; i is the i-th phase loop current; U i is the i-th phase loop voltage; ρ is the density of polysilicon rods.

[0011] Preferably, in the step of collecting and calculating the reaction state in S1, Gaussian filtering is performed on the mass M or the deposition rate ΔM of the polycrystalline silicon rod.

[0012] Preferably, in the step of collecting and calculating the reaction state in S1, the heat loss Q of the reaction is the heat of the tail gas, and the total energy E is the input electrical energy:

[0013] Where P is the input electrical power and T is the time difference.

[0014] Preferably, in the step of collecting and calculating the reaction state in S1, the heat loss Q of the reaction is calculated as follows: According to the chemical reaction equation in the above production process, the heat formula in each chemical reaction during the reaction process can be obtained as follows:

[0015]

[0016]

[0017]

[0018]

[0019] in: is the specific heat capacity of TCS, is the average mass flow rate of TCS in the corresponding time, is the mass of silicon rod at time t2, is the mass of silicon rod at time t1, is the relative atomic mass of elemental silicon, is the relative molecular mass of silicon tetrachloride, is the relative molecular mass of trichlorosilane, It is the tail gas temperature minus the feed temperature in the corresponding time; is the specific heat capacity of gaseous hydrogen chloride, is the specific heat capacity of gaseous silicon tetrachloride, is the specific heat capacity of hydrogen, is the average mass flow rate of hydrogen in the corresponding time, is the relative molecular mass of hydrogen chloride.

[0020] Preferably, in the step S2 of inferring control parameters, the target deposition curve is synthesized based on a historical deposition rate curve that meets the single power consumption standard and meets preset requirements.

[0021] Preferably, in the S2 control parameter inference step, the polysilicon reduction process is divided into several stages, and in each stage, a segment of the historical deposition rate curve with the best local single power consumption in that stage is selected, and then the segments obtained in each stage are spliced ​​together to synthesize the target deposition rate curve.

[0022] Preferably, in the step S2 of inferring control parameters, the optimal deposition rate curve at each moment is obtained by sequentially connecting the segments to obtain the final target deposition rate curve.

[0023] Preferably, in the step S2 of inferring control parameters, the target deposition rate curve is , the real-time deposition rate is recorded as , the current time is , the historical data of one time step forward is , proportional adjustment parameters , integral adjustment parameter And the differential adjustment parameters .

[0024] Preferably, in the step S3 of executing the PID control process, the hydrogen input adjustment amount ; Reaction current adjustment ; Among them, for the hydrogen input adjustment amount, the proportional adjustment coefficient 、 Integral adjustment coefficient , differential adjustment coefficient ; For the reaction current adjustment, the proportional adjustment coefficient , integral adjustment coefficient ,when Between 40% and 50%, ,when Less than 40%, ,when Greater than 50%, .

[0025] The present invention provides a new control strategy that uses the deposition rate calculated by current and voltage as the control target. The method determines the optimal deposition rate curve through data analysis and calculates the deposition rate in real time during the control process. The current heat percentage in the furnace is calculated using a thermodynamic formula, and this percentage is combined with the deposition rate difference and input into the PID control module to accurately control hydrogen and current respectively. This automated control strategy adjusts both current and hydrogen at the same time, rather than adjusting a single variable. The stability and accuracy of the production process are significantly improved, thereby ultimately increasing the yield and quality of polysilicon and ensuring low unit electricity consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A schematic flow chart of a polysilicon reduction furnace control method based on a forward mechanism model with multiple control variables according to the present application; Figure 2 This is the average value curve data of the deposition rate obtained by grouping historical data in this application. DETAILED DESCRIPTION

[0027] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. In this specification, the size ratios in the drawings do not represent the actual size ratios, but are only used to reflect the relative position relationship and connection relationship between the components. Components with the same name or the same number represent similar or identical structures and are only for illustrative purposes.

[0028] In the reduction control process of polysilicon, the process parameters are controlled by positive feedback through the variable combination of deposition rate and reaction heat to achieve optimized control of polysilicon quality and power consumption. Figure 1 As shown, based on the preset process plan, the general control process in one control cycle includes the following steps.

[0029] S1. Collect and calculate the reaction state. The reaction state includes the real-time deposition rate ΔM of the reaction and the heat loss rate ΔQ of the reaction. The real-time deposition rate ΔM is the rate of change of the mass M of the polysilicon rods in the reduction furnace. The heat loss rate ΔQ is the ratio of the heat loss Q of the reaction to the total energy E input. The heat loss mainly refers to the heat carried away by the tail gas.

[0030] S2. Infer control parameters. Determine the control parameters in the PID control process based on the difference between the real-time deposition rate curve and the target deposition rate curve, and the control parameters include a proportional adjustment parameter P, an integral adjustment parameter I, and a differential adjustment parameter D. The solid line deposition curve is generated based on the real-time deposition rate ΔM of the reaction process sequence, and the target deposition rate curve is a preset curve.

[0031] S3, execute PID control process. Use the control parameters in step S2 to perform PID control on the amount of hydrogen input and the reaction current. Hydrogen input adjustment amount ; Reaction current adjustment . k p1 , k i1 , k d1 , k p2 , k i2 , k d2 is the control coefficient. p2 , k i2 , k d2 At least one of has a preset value set, and its value is related to the heat loss ΔQ of the reaction.

[0032] The following is one specific embodiment of the above control process.

[0033] S1. Collect and calculate the reaction status.

[0034] During the operation of the reduction furnace, the feed amount, voltage, current, etc. of the reaction are generally controlled by the preset process plan. In addition, the temperature, flow rate and power consumption related parameters are usually collected and recorded in real time. On this basis, the deposition rate and heat loss of the reaction can be inferred.

[0035] Specifically, the deposition rate ΔM is the change in the mass M of the polysilicon rod over time, that is, The mass M of polysilicon in the reduction furnace cannot be calculated directly, but it can be estimated more accurately based on the collected parameters.

[0036] As the cross-sectional area of ​​polysilicon increases and the resistance decreases during the deposition and growth process, the electrical characteristics of the reduction furnace change accordingly. However, based on experience, it can be determined that the resistivity of polysilicon does not change much at a certain temperature. For polysilicon reduction, in order to ensure the quality of the discharge, the polysilicon in the furnace is generally controlled at around 1050°C. Therefore, during the polysilicon reduction process, it can be basically assumed that the resistivity of the polysilicon rod is basically unchanged, so it can be preliminarily inferred that the volume V of the polysilicon rod in the single-phase circuit in the reduction furnace is i .

[0037]

[0038] Among them, ρ2 is the resistivity of the silicon rod, which can be considered as a constant value. When heated to 1050℃ (1050℃ is the optimal temperature for the reduction reaction of polysilicon), the resistivity is , the temperature does not change much, so the resistivity is assumed to be constant. The 1000 in the formula is the dimension difference, and when based on the International System of Units, the factor 1000 degenerates to 1. L is the total length of the silicon rod, including the length of the crossbeam at the top of the silicon rod, which can be determined before the reaction starts.

[0039] For a certain reduction furnace, in order to make full use of the reduction furnace space and improve the utilization rate, multiple current loops can be set in the same reduction furnace, each current loop is a phase loop below. In each phase loop, multiple polysilicon rods can also be connected in series to achieve simultaneous deposition.

[0040] n i I is the number of silicon rod pairs of the phase loop calculated in the reduction furnace; i is the i-th phase loop current (A); U i is the i-th phase circuit voltage (V); In order to obtain the total mass of the polysilicon rods in the reduction furnace, it is necessary to sum up the mass of the silicon rods in all phase loops. At the same time, we determine the mass M of the polysilicon rods in each phase. i , which obviously satisfies Where ρ is the density of the polysilicon rod, which is generally considered to be a constant value. We ignore the density difference caused by the quality of the polysilicon material. Depending on the specific design of the reduction furnace, there is a certain number of loops m, for example, m=6 in the embodiment of the present application. Then the mass M of the polysilicon rod is:

[0041] The first-order derivative of the polysilicon mass with respect to time calculated based on the collected data is the required deposition rate, that is, .

[0042] In the actual calculation process, the deposition rate can be discretized as the ratio of the deposition rate within a certain time step to the time step. For example, if the time step is 20 minutes, then:

[0043] Among them, M 0min is the quality of the current polysilicon rod, M -20min is the mass of the silicon rod 20 minutes ago.

[0044] Preferably, the quality of the polysilicon rod is mainly calculated by the collected values ​​of the voltage and current of the reduction furnace. However, due to the noise interference in the collection process, the fluctuation range of the relevant collection amount is large, and there is measurement noise. Therefore, in the process of deriving and calculating the deposition rate, it is necessary to filter the polysilicon quality or deposition rate to ensure the smoothness of the result and the stability and rapid convergence of the subsequent PID control process. The denoising process can apply Gaussian filtering to process the polysilicon quality curve or the deposition rate curve.

[0045]

[0046] in is the distance between the data point and the center of the Gaussian filter, is the standard deviation of the Gaussian distribution, which determines the width of the filter. Higher values ​​result in a wider filter, which has a smoothing effect on the data. The selection of needs to obtain their autocorrelation coefficients from historical data and use the autocorrelation coefficients to determine.

[0047] In addition, the heat loss rate ΔQ is crucial to determine the heat loss Q and the total energy E of the reaction.

[0048] The total energy input of the reaction is relatively simple, mainly in the form of electrical energy, so

[0049] Where P is power and T is time difference. The specific formula is , the unit is J For the reduction reaction Time to The average power supply during the time, in watts (W), .

[0050] The heat loss Q mainly comes from the exhaust gas. In this reaction process, the heat of the exhaust gas temperature, that is, the heat taken away by the output, can be calculated through the chemical reaction equation and thermodynamic formula. The chemical reaction equation to be used is:

[0051]

[0052] According to the chemical reaction equation in the above production process, the heat formula in each chemical reaction during the reaction process can be obtained as follows:

[0053]

[0054]

[0055]

[0056] in: is the specific heat capacity of TCS (trichlorosilane), which is ; is the average mass flow rate of TCS in the corresponding time, in units of ; is the mass of the silicon rod at time t2, the calculation method refers to the previous steps is the mass of the silicon rod at time t1, and the calculation method refers to the above steps; is the relative atomic mass of elemental silicon, which is ; is the relative molecular mass of silicon tetrachloride, which is ; is the relative molecular mass of trichlorosilane, which is l; It is the tail gas temperature minus the feed temperature in the corresponding time; is the specific heat capacity of gaseous hydrogen chloride, which is ; is the specific heat capacity of gaseous silicon tetrachloride, which is ; is the specific heat capacity of hydrogen, which is ; is the average mass flow rate of hydrogen in the corresponding time, in units of ; is the relative molecular mass of hydrogen chloride, which is ; After adding all the above heat formulas, we get the total heat taken away by the exhaust gas:

[0057] The heat loss rate at this moment is:

[0058] S2. Infer control parameters.

[0059] The control process is the feedback control parameters required in the PID control process, including the proportional adjustment parameter P, the integral adjustment parameter I and the differential adjustment parameter D. The control parameters are derived from the process difference, which is usually determined by the real-time deposition rate and the ideal deposition rate curve determined in advance. To this end, it is necessary to determine the ideal target deposition rate curve through certain means.

[0060] In simple terms, we screen out the deposition rate curves that meet the requirements from a large amount of historical data. In the screening process of this embodiment, the deposition rate curves corresponding to abnormal reaction process, material density rate that does not meet the requirements, reaction process timeout, etc. are generally eliminated, and then the target deposition rate curve is selected or synthesized based on the remaining deposition rate curves. The purpose is to retain the characteristics of the historical deposition rate curves with excellent single power consumption in the target deposition rate curve as much as possible.

[0061] The target deposition rate curve can be determined by taking the globally optimal deposition rate curve in the complete reduction process. However, considering the staged nature of the polysilicon reduction process, the global optimality does not guarantee the optimality in each stage. In actual data, there is generally an intersection between deposition rate curves with better power consumption, which means that the stages with higher local deposition rates corresponding to different reduction processes may be different from each other, so there is room for optimization.

[0062] First of all, the polysilicon reduction process is generally divided into five different stages, namely, the feeding period, the stabilizing period, the reducing period, the peak period, and the leveling period. The reaction states in the reduction furnace are different in the five processes. Therefore, the complete reduction process can be divided into several stages according to the corresponding stages of the polysilicon reduction process. In each stage, the segmentation of the locally optimal historical deposition rate curve in the stage is selected. The standard is that the average deposition rate in the stage is the largest. Then, the segments obtained in each stage are spliced ​​to synthesize the target deposition rate curve. When splicing, the adjacent segments can be appropriately smoothed to avoid the violent process at the connection affecting the stability of the control process.

[0063] In addition, it is also possible to preferably not consider the stages of the specific reduction process, but to screen and form a target deposition rate curve based on data characteristics. In short, considering the intersection of historical deposition rates, after obtaining the historical target curve that meets the conditions, the optimal deposition rate curve at each moment is taken and connected in segments in sequence to obtain the final target deposition rate curve, which is equivalent to obtaining the maximum deposition rate envelope of all historical deposition rate curves that meet the conditions.

[0064] Based on the target deposition rate curve, the control parameters can be determined as follows.

[0065] Assume that the target deposition rate curve is , the real-time deposition rate is recorded as , assuming the current time is , the historical data of one time step forward is In this embodiment, the time step is 10 minutes, and the target difference is obtained. , , then define the proportional adjustment parameter , integral adjustment parameter And the differential adjustment parameters .

[0066] The reason why deposition rate is chosen as the control target is that after analyzing historical data, it can be found that the correlation between power consumption and deposition rate is better. Therefore, compared with other control quantities, the effect of energy-saving control achieved by controlling deposition rate should be better. The general control target curve uses data such as power or voltage that can be directly obtained through sensors. However, these two curves are very affected by the furnace itself and the operator's control, and the convergence is not good. In order to achieve the purpose of process convergence, the applicant even needs to shorten the global length of the control process to prevent global divergence. An empirical evidence of a similar approach is that in the applicant's individual historical schemes, there is a technical solution that divides the process into different stages in the entire global reduction process and then adopts different control strategies. Its main purpose is to avoid the problem of divergence of characterization quantity control.

[0067] In order to determine whether the deposition rate can be used as a control target curve, the historical production data obtained was first screened and anomalies were removed, and then divided into three groups according to the single power consumption. During the exploration process, the three groups were taken from the first 30%, the middle 40%, and the last 30% in descending order. The mean of the voltage, power, and deposition rate of all the data in the three groups of high, medium, and low is calculated, and the change curve of the mean during the reaction process is shown in the following figure: Figure 2As shown, it can be found that the red line with high power consumption has a lower average deposition rate, and the blue line with low power consumption has a higher deposition rate. In other words, the lower the power consumption, the higher the average deposition rate. Further research is done to find the trace of the covariance of the three curves of voltage, power, and deposition rate for all data (the trace of the covariance ( ), the larger the value, the higher the correlation and the better the convergence. The original data is relatively sensitive and is not provided here. It is found that the trace of the covariance between the deposition rate and the single power consumption is the largest. On the one hand, it shows that the correlation between the deposition rate and the single power consumption is good, which is consistent with the initial judgment. On the other hand, due to the large absolute value of the deposition rate, the adjustment space provided is more abundant and more suitable for application in the PID control process. Therefore, the deposition rate is more suitable as the control target curve of this project.

[0068]

[0069] S3. Execute PID control process.

[0070] Use the control parameters in step S2 to perform PID control on the amount of hydrogen input and the reaction current. Hydrogen input adjustment amount ; Reaction current adjustment .

[0071] Among them, for the hydrogen input adjustment amount, the proportional adjustment coefficient can be set 、 Integral adjustment coefficient , differential adjustment coefficient For the reaction current adjustment, you can set , The value of is preferably classified and adjusted according to the heat loss rate. Between 40% and 50%, ,when Less than 40%, ,when Greater than 50%, .

[0072] This note describes a new control strategy that uses the deposition rate calculated from current and voltage as the control target. The method first determines the optimal deposition rate curve through data analysis, and calculates the deposition rate in real time during the control process. Then, the current heat percentage in the furnace is calculated using thermodynamic formulas, and this percentage is combined with the deposition rate difference and input into the PID control module to accurately control the hydrogen and current respectively. This automated control strategy adjusts both current and hydrogen at the same time, rather than adjusting a single variable. The stability and accuracy of the production process are significantly improved, which ultimately increases the yield and quality of polysilicon.

[0073] This control method has shown significant technical advantages in polysilicon production: 1. Precise control: This method can accurately measure the growth weight of silicon rods through real-time calculation, thereby achieving precise control. In addition, the control cycle is shortened to once every 10 minutes, which significantly improves the control frequency and density. 2. High adaptability: The deposition rate is calculated based on real-time current and voltage data, so it can more accurately reflect the growth status of silicon rods in the furnace. Compared with traditional control methods, the latter usually relies on initial observations of power and later observations of voltage, and cannot track the same indicator throughout the process, so the adaptability is poor. 3. Improved automation level: This method controls current and hydrogen at the same time, forming a double closed loop. The current regulation ratio can be better obtained through thermodynamic formulas, thereby achieving precise control of current. This method reduces the reliance on human experience and makes control more automated. 4. The selection of control targets is more reasonable: The mathematical and statistical analysis method is used to quantify the correlation between target curves by calculating the trace of covariance, providing a scientific basis for the selection of control targets. This method is objective and quantitative, avoiding the failure of control due to unreasonable selection of control targets.

[0074] The above content only describes the preferred implementation mode of the present invention, and does not limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solution of the present invention by ordinary technicians in this field should fall within the protection scope determined by the claims of the present invention.

Claims

1. A polysilicon reduction furnace control method based on a forward mechanism model with multiple control variables, characterized in that: The steps include: S1. Collect and calculate the reaction state: the reaction state includes the real-time deposition rate ΔM of the reaction and the heat loss rate ΔQ of the reaction; the real-time deposition rate ΔM is the rate of change of the mass M of the polysilicon rods in the reduction furnace, and the heat loss rate ΔQ is the ratio of the heat loss Q of the reaction to the total energy E input; S2. Inferring control parameters: Determining control parameters in the PID control process based on the difference between the real-time deposition rate curve and the target deposition rate curve, the control parameters include a proportional adjustment parameter P, an integral adjustment parameter I and a differential adjustment parameter D; wherein the solid line deposition curve is generated based on the real-time deposition rate ΔM of the reaction process sequence, and the target deposition rate curve is a preset curve; S3, execute PID control process: use the control parameters in step S2 to perform PID control on the amount of hydrogen input and the reaction current; adjust the amount of hydrogen input ; Reaction current adjustment ;k p1 , k i1 , k d1 , k p2 , k i2 , k d2 is the control coefficient; where k p2 , k i2 , k d2 At least one of has a preset value set, and its value is related to the heat loss ΔQ of the reaction.

2. The polysilicon reduction furnace control method based on the forward mechanism model with multiple control variables as claimed in claim 1, characterized in that: In the step S1 of collecting and calculating the reaction state, the mass M of the polycrystalline silicon rod is: Where, m is the number of current loops in the polysilicon reduction furnace, M i is the mass of the polysilicon rod of the i-th phase loop, calculated by the following formula: Where ρ2 is the resistivity of the silicon rod, L is the total length of the silicon rod, and n i I is the number of silicon rod pairs in the i-th phase loop calculated in the reduction furnace; i is the i-th phase loop current; U i is the i-th phase loop voltage; ρ is the density of polysilicon rods.

3. The polysilicon reduction furnace control method based on the forward mechanism model with multiple control variables as claimed in claim 2, characterized in that: In the step of collecting and calculating the reaction state, S1 performs Gaussian filtering on the mass M or the deposition rate ΔM of the polycrystalline silicon rod.

4. The polysilicon reduction furnace control method based on the forward mechanism model with multiple control variables as claimed in claim 1, characterized in that: In the step S1 of collecting and calculating the reaction state, the heat loss Q of the reaction is the heat of the tail gas, and the total energy E is the input electrical energy: Where P is the input electrical power and T is the time difference.

5. The polysilicon reduction furnace control method based on forward mechanism model and multiple control variables as claimed in claim 4, characterized in that: In the step of collecting and calculating the reaction state in S1, the heat loss Q of the reaction is calculated as follows: According to the chemical reaction equation in the above production process, the heat formula in each chemical reaction during the reaction process can be obtained as follows: in: is the specific heat capacity of TCS, is the average mass flow rate of TCS in the corresponding time, is the mass of silicon rod at time t2, is the mass of silicon rod at time t1, is the relative atomic mass of elemental silicon, is the relative molecular mass of silicon tetrachloride, is the relative molecular mass of trichlorosilane, It is the tail gas temperature minus the feed temperature in the corresponding time; is the specific heat capacity of gaseous hydrogen chloride, is the specific heat capacity of gaseous silicon tetrachloride, is the specific heat capacity of hydrogen, is the average mass flow rate of hydrogen in the corresponding time, is the relative molecular mass of hydrogen chloride.

6. The polysilicon reduction furnace control method based on forward mechanism model and multiple control variables as claimed in claim 1, characterized in that: In the step S2 of inferring control parameters, the target deposition curve is synthesized based on a historical deposition rate curve that meets the single power consumption standard and meets the preset requirements.

7. The polysilicon reduction furnace control method based on forward mechanism model and multiple control variables as claimed in claim 6, characterized in that: In the S2 control parameter inference step, the polysilicon reduction process is divided into several stages, and in each stage, the segment of the historical deposition rate curve with the best local single power consumption in that stage is selected, and then the segments obtained in each stage are spliced ​​together to synthesize the target deposition rate curve.

8. The polysilicon reduction furnace control method based on forward mechanism model and multiple control variables as claimed in claim 6, characterized in that: In the step S2 of inferring control parameters, the optimal deposition rate curve at each moment is obtained by sequentially connecting the segments to obtain the final target deposition rate curve.

9. The polysilicon reduction furnace control method based on forward mechanism model and multiple control variables as claimed in any one of claims 6 to 8, characterized in that: In the S2 control parameter inference step, the target deposition rate curve is: , the real-time deposition rate is recorded as , the current time is , the historical data of one time step forward is , proportional adjustment parameters , integral adjustment parameter And the differential adjustment parameters .

10. The polysilicon reduction furnace control method based on forward mechanism model and multiple control variables as claimed in claim 9, characterized in that: In the PID control process step of S3, the hydrogen input adjustment amount ; Reaction current adjustment ; Among them, for the hydrogen input adjustment amount, the proportional adjustment coefficient 、 Integral adjustment coefficient , differential adjustment coefficient ; For the reaction current adjustment, the proportional adjustment coefficient , integral adjustment coefficient ,when Between 40% and 50%, ,when Less than 40%, ,when Greater than 50%, .

Images