Polysilicon reduction furnace control method, apparatus, and readable storage medium

Abstract

The application provides a polysilicon reduction furnace control method, equipment and a readable storage medium. The polysilicon reduction furnace control method comprises a control stage of hydrogen flow PID control based on total power of the polysilicon reduction furnace, in which a first time period is used as a control cycle of a PID control algorithm to cyclically perform PID control; and in one control cycle, a proportional error term and a differential error term of the total power of the polysilicon reduction furnace are determined based on a target power curve, and an adjustment amount of hydrogen flow in the PID control is determined accordingly. The application finely adjusts the reaction process of the polysilicon reduction furnace based on the target power curve, the control algorithm combines the current operating conditions and the expected changes, calculates the power target change value in real time according to the furnace condition, and gives the most reasonable hydrogen adjustment amount, so that the reduction furnace can be stably operated according to the optimal curve. This control strategy not only optimizes the energy use and reduces the energy consumption, but also improves the production quality of the polysilicon through accurate control.

Description

Technical Field

[0001] This invention relates to the field of polysilicon reduction furnace control technology, specifically to a polysilicon reduction furnace control method, equipment, and readable storage medium. Background Technology

[0002] Polysilicon is a crucial raw material for the photovoltaic industry. With the continuous development and utilization of renewable energy, the supply and demand of polysilicon are constantly increasing. Optimistically, global polysilicon demand is expected to reach 1.13 million tons by 2030. Currently, the mainstream method for producing polysilicon is the modified Siemens process. This method utilizes the principle of chemical vapor deposition, using hydrogen to reduce trichlorosilane at a high temperature of around 1100 degrees Celsius in a reduction furnace to obtain polysilicon deposited on a silicon core.

[0003] In traditional industrial manufacturing processes, especially in the operation of critical equipment such as reduction furnaces, the control of polysilicon reduction furnaces often relies on human experience and periodic monitoring. Traditional manual control methods are primarily based on the operator's intuitive judgment of the process equipment and intermittent manual adjustments. For example, when a reduction reaction occurs inside the polysilicon reduction furnace, the operator needs to rely on observing the furnace interior through a viewing window (such as the clarity and color intensity) to judge the internal conditions and make targeted adjustments. This production method heavily relies on experience and has significant limitations. First, manual control depends on the skills and experience of individual operators, making it difficult to ensure the consistency of the production process, and the quality and yield of polysilicon cannot be effectively guaranteed. Second, novice operators may not be able to accurately judge the equipment status and make targeted adjustments, increasing production risks and uncertainties. Finally, manual adjustments often have a delay in response; operators may not be able to quickly respond to sudden changes in the production process, especially in continuous production, where even slight carelessness can lead to production accidents or quality problems. Furthermore, manual observation can hardly achieve the precision that machines can, especially when dealing with chemical reactions and heat treatment processes that require fine-tuning. Summary of the Invention

[0004] To achieve precise control over the polysilicon reduction process, this invention provides a polysilicon reduction furnace control method, equipment, and readable storage medium.

[0005] The technical solution of this invention provides a control method for a polysilicon reduction furnace, including a control stage for PID control of hydrogen flow rate based on the total power of the polysilicon reduction furnace. In this control stage, PID control is performed cyclically with a first time period T1 as the control cycle of the PID control algorithm. Within one control cycle, the proportional error term E of the total power of the polysilicon reduction furnace is determined based on the target power curve. p and the differential error term E d Based on this, the adjustment amount ΔH of hydrogen flow rate in PID control is determined.

[0006] Preferably, the control formula for determining the adjustment amount ΔH of hydrogen flow rate in PID control is as follows:

[0007] ,

[0008] Where k p k d These are the preset control parameters.

[0009] Preferably, the control formula for determining the adjustment amount ΔH of hydrogen flow rate in PID control is as follows:

[0010] ,

[0011] Where k p k d k w For the preset control parameters, the correction term E w Determination based on the state parameters of the polysilicon reduction furnace.

[0012] Preferably, the modification term E w The correction term E is determined based on the tail gas temperature change rate ΔT of the polycrystalline silicon reduction furnace. w The formula is:

[0013]

[0014] Among them, T h This is the preset threshold for the rate of change of exhaust gas temperature.

[0015] Preferably, the proportional error term E for determining the total power of the polysilicon reduction furnace based on the target power curve within one control cycle is... p and the differential error term E d The steps are as follows:

[0016] The first-order difference ΔP of the target power curve at time point t is calculated using the calculation period T2. goal,t :

[0017] ,

[0018] Among them, P goal,t P represents the power value at time point t under the target power curve. goal,(t-T2) The historical power value at the corresponding historical time point after calculating the period T2 before the target power curve's time point t is calculated.

[0019] Based on the power history curves collected during the current process, the first-order difference ΔP of the electrical power of the polysilicon reduction furnace at time point t is determined. real,t :

[0020] ,

[0021] Among them, P real,t P represents the actual power value collected at time point t. real,(t-T2) The actual power value collected at the historical time point corresponding to the calculation period T2 before time point t;

[0022] Taking the current time point t1 as time point t, calculate the difference P between the actual power change and the ideal power change at the current time point t1. diff,t1 :

[0023] ,

[0024] Take an interval period T3, and push the current time point t1 forward by the interval period T3 to obtain a historical time point t2;

[0025] Using historical time point t2 as time point t, the difference P between the actual power change and the ideal power change at historical time point t2 is calculated. diff,t2 :

[0026] ,

[0027] but,

[0028] proportional error term E p With differential error term E d for:

[0029] ,

[0030] Among them, E i This is the integral error term for PID control.

[0031] Preferably, a target power consumption and a target density are set, and the target power curve is obtained by averaging the corresponding time points of historical production cycles that meet the conditions that the power consumption does not exceed the target power consumption and the density of the obtained polysilicon is not less than the target density.

[0032] Preferably, it further includes a second control stage for PID control of hydrogen flow rate based on the voltage drop of the polysilicon reduction furnace, the second control stage following the control stage for PID control of hydrogen flow rate based on the total power of the polysilicon reduction furnace.

[0033] Preferably, the second control stage includes the following steps:

[0034] Determine the voltage representative value V of the polysilicon reduction furnace, which is a function of the voltage of several polysilicon groups in the polysilicon reduction furnace;

[0035] Based on the voltage representative value V, the following is calculated:

[0036] The target voltage drop V is obtained by subtracting the time period T4 from time point t and using a time interval T5 as the calculation period. goal :

[0037] ,

[0038] Starting from time point t, advance the time period T6, and use a time interval T5 as the calculation period to obtain the voltage drop at time T6 as the historical target voltage drop V. past,goal :

[0039] ,

[0040] c is a preset correction term, and c can take the value of zero;

[0041] Using the current time t as the base point, and pre-setting a time span k, the difference in voltage drop at time t is obtained as follows:

[0042] ,

[0043] Using the current time t as the base point, and a preset time span k, after pushing forward a preset time length L, the difference in historical voltage drop at time t is:

[0044] ,

[0045] Determine the proportional error term E of PID control p and the differential error term E d :

[0046] ,

[0047] Among them, E i This refers to the integral error of PID control.

[0048] PID control is performed based on the PID control formula, which is:

[0049] ,

[0050] Where k p k d These are the preset control parameters.

[0051] The technical solution of the present invention also provides a control device for a polysilicon reduction furnace, including...

[0052] Memory, which stores computer programs;

[0053] A processor is configured to execute the computer program, which, when executed, implements the polysilicon reduction furnace control method described in any of the preceding claims.

[0054] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the polysilicon reduction furnace control method described in any of the above claims.

[0055] This invention uses a target power curve as a benchmark to finely adjust the reaction process of a polysilicon reduction furnace. The control algorithm combines current operating conditions and expected changes, calculates the target power change value in real time based on the furnace condition, and provides the most reasonable hydrogen adjustment amount to ensure that the reduction furnace operates stably according to the optimal curve. Overall, this control strategy not only optimizes energy use and reduces energy consumption, but also improves the production quality of polysilicon through precise control. Attached Figure Description

[0056] Figure 1 This is an example of the resistance variation curve of a reduction furnace in historical production. Detailed Implementation

[0057] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0058] The polysilicon reduction furnace control method of the present invention achieves relatively stable and reliable operation control by applying power target-based PID control throughout the complete or partial production cycle of the polysilicon reduction furnace. An embodiment of this control scheme is given below.

[0059] In this embodiment, the total power of the polysilicon reduction furnace is used as the control target, and the adjustment of hydrogen flow rate is used as the control means to control the state of the polysilicon reduction furnace. This control process can be implemented based on a PID control algorithm.

[0060] Specifically, in the first stage of control, PID control is performed iteratively using the first time period T1 as the control cycle. Within one control cycle, the proportional error term E of the power of the polysilicon reduction furnace is determined. p and the differential error term E d Based on this, the adjustment amount ΔH of the hydrogen flow rate is determined:

[0061]

[0062] Where k p k d These are control parameters determined based on experience or initial testing.

[0063] Generally, in the control process described above, a target power curve is preset using other methods. The proportional error term of the PID control can be determined based on the absolute difference between the current power reading and the power value at the same time point on the target power curve, thus achieving PID regulation in a general sense. However, in the control process of a polysilicon reduction furnace, relying solely on the error value at a fixed sampling time point cannot avoid local disturbances in the system, leading to reduced control efficiency and fluctuations in the actual power curve. Even after trying multiple sets of different control parameters, the fluctuation problem cannot be effectively resolved. Therefore, this embodiment provides a first-order differential PID control method based on the target power curve to improve the accuracy and stability of power control.

[0064] To obtain the error term based on the first-order difference, this implementation scheme first determines the calculation period T2. The first-order difference ΔP of the target power curve at time point t is then calculated using the calculation period T2. goal :

[0065] .

[0066] Among them, P goal,t P represents the power value at time point t under the target power curve. goal,(t-T2) The historical power value is calculated by pushing forward the calculation period T2 from the time point t under the target power curve to the corresponding historical time point.

[0067] Simultaneously, based on the power history curve collected during the current process, the first-order difference ΔP of the electrical power of the polysilicon reduction furnace at time point t is determined. real :

[0068] .

[0069] Among them, P real,t P represents the actual power value collected at time point t. real,(t-T2) The actual power value collected at the historical time point corresponding to the calculation period T2 before time point t.

[0070] Therefore, the difference P between the actual power change and the ideal power change is obtained. diff :

[0071] .

[0072] Based on the above first-order differential logic, if the current time point t1 is defined as time point t in the formula, the difference P between the actual power change and the ideal power change at the current time point t1 is obtained. diff,t1 :

[0073] .

[0074] Taking an interval period T3, which is usually equal to the first control period T1, we push the current time point t1 forward by the interval period T3 to obtain a historical time point t2. Based on the above first-order differential logic, if we define the historical time point t2 as the time point t in the formula, we obtain the difference P between the actual power change and the ideal power change at the historical time point t2. diff,t2 :

[0075] .

[0076] The difference P between the actual power change and the ideal power change at the current time point t1 is calculated. diff,t1 E, the proportional error term in the PID control process p The difference P between the actual power change and the ideal power change at the aforementioned historical time point t2 is calculated. diff,t2 E, the integral error term of the control process i ,Right now:

[0077] .

[0078] Therefore, the differential error of the PID control process is determined as follows:

[0079] .

[0080] The obtained proportional error E p and differential error E d By inputting the control equation, the corresponding hydrogen flow adjustment ΔH can be obtained.

[0081] In specific control cases, the appropriate first time period T1, calculation period T, and interval period T3 can be selected based on the specific circumstances of the polycrystalline silicon reactor. For example, the first time period T1=T3=20min and the calculation period T2=60min can be selected.

[0082] Returning to the PID control scheme, k p k d As a system control parameter in the PID control process, its specific value needs to be determined based on the specific conditions of the polysilicon reduction furnace. Generally, it is determined through trial operation; in this application, k is selected. p =0.5,k d =0.3.

[0083] In the aforementioned control process, the error term of the PID control process is obtained based on first-order differential calculation. The selected calculation period for the differential calculation determines that the PID control process has relatively good stability and reduces target fluctuations in the power following process. However, there are also disadvantages, namely, the response speed of the PID control process may be slow, and it cannot effectively handle sudden instantaneous changes. Therefore, a further technical solution can introduce a correction term E into the PID control process. w This is used to compensate for the shortcomings of the original scheme in terms of system transient response. Specifically, the governing equation for determining the adjustment amount ΔH of the hydrogen flow rate is:

[0084]

[0085] Among them, the correction term E w The specific state changes of the polysilicon reduction furnace are determined, such as exhaust gas temperature, current, voltage, or cooling water temperature. In this embodiment, exhaust gas temperature is preferred because changes in current and voltage are too sensitive and easily affected by irrelevant factors, making it difficult to establish a stable correspondence with the polysilicon growth rate within the furnace. Using current or voltage can easily cause control fluctuations, or even, in severe cases, system overload. Cooling water temperature and flow rate are not considered for the sake of replicability. For any different polysilicon reduction furnace, the cooling water system is affected not only by the reaction conditions within the furnace but also by the water circulation system. This means that for each polysilicon reduction furnace, the phenomena reflected in the cooling water system by changes in the furnace state are different, leading to inconsistent relationships between cooling water changes and furnace states in different furnace systems. Therefore, when migrating the control system to different polysilicon reduction furnaces for production, the PID control parameters need to be recalibrated, making it impossible to replicate the control scheme of this invention on a large scale at low cost. The solution using exhaust gas temperature monitoring for correction can effectively address these issues. The exhaust gas temperature is directly output from the internal environment of the polysilicon reduction furnace, thus accurately reflecting the internal reaction conditions. Furthermore, the exhaust gas from the polysilicon reduction furnace is collected directly through an exhaust pipe. The exhaust gas temperature not only responds promptly to changes in reaction conditions within the furnace but also exhibits good consistency across different reduction furnace production processes, making it an ideal state parameter for error correction. Based on test data from the trial operation phase, k... w The value is around 0.5.

[0086] The correction term E can be determined by the exhaust gas temperature using the following method. wIn the corresponding PID control loop, the rate of change ΔT of the exhaust gas temperature is calculated using the collected exhaust gas temperature values. If the exhaust gas temperature changes too rapidly, hydrogen is added to bring the internal reaction process back to normal. If the exhaust gas temperature change is normal, no additional influence is needed on the aforementioned control process. Therefore, this control logic is implemented by pre-setting a threshold Th for the rate of change of the exhaust gas temperature per unit time. w The following is determined:

[0087] This means that the correction term E w The system is triggered only when the rate of change in exhaust gas temperature exceeds a certain level, thus preventing the system from changing too rapidly in a short period of time.

[0088] As mentioned above, in the first stage, the total power is used as the control target to control the hydrogen flow rate to achieve PID control. The basic logic is that electrical power indirectly reflects the silicon deposition rate in the polysilicon reduction furnace. The amount of raw material input can affect the chemical vapor deposition process of trichlorosilane in the reduction furnace, thus affecting the silicon deposition rate. We can certainly consider controlling the current to regulate the reaction's electrical power, but adjusting the current essentially acts on the circuit containing the silicon core, not on the raw material for chemical vapor deposition. Therefore, adjusting the current might significantly adjust the electrical power, but it's merely a change in the circuit itself and cannot directly affect the material input and control system. Furthermore, in the modified Siemens method, to ensure the conversion rate of trichlorosilane and improve the silicon conversion rate, hydrogen is used in excess during the actual feeding process. Therefore, the deposition rate of polysilicon inside is generally regulated by controlling the amount of hydrogen input.

[0089] In the first stage, the above technical solution chose power as the control target, rather than the resistance of the polycrystalline silicon core, mainly due to considerations of control accuracy. For example... Figure 1The figure shows the resistance variation curve of a polycrystalline silicon reduction furnace during one production cycle. Basically, throughout the complete production cycle, the resistance of the silicon core varies between 0 and 10 ohms. Especially after a period of deposition, the resistance of the silicon core quickly drops below 1 ohm, and for most of the subsequent production time, the resistance of the silicon core slowly decreases only between 0 and 1 ohm. This shows that although the resistance of the silicon core does change as the deposition process progresses, the absolute value of the resistance change is very small, fluctuating only between 0 and 10 ohms, and even more so, for most of the time, the resistance variation is limited to 0-1 ohms. Because of this, if the silicon core resistance is used as the control target, on the one hand, the small difference in the control target's change is limited by the PID control algorithm itself, making effective control at this level of precision difficult to achieve. Therefore, using resistance as the control target is not effective in practice. On the other hand, even if resistance control at this level of precision can be achieved, another problem that must be faced is that because the resistance value is too small, any disturbance in the production process will cause a significant change in the resistance value. Even errors in the data acquisition process will cause a significant change in the test value. Therefore, if the resistance value is used as the control target, the noise signal will have a significant impact on the measured value because the resistance value is too small. Even if effective PID control can be achieved, it will be difficult to achieve the expected stable control target.

[0090] In practice, the production process of a polysilicon reduction furnace typically allows for the regulation of the system current, while the voltage is passively changed. This means that the reaction process can be controlled when necessary by comprehensively managing the hydrogen flow rate, trichlorosilane flow rate, and current magnitude. In this case, parameters such as voltage, exhaust gas temperature, and cooling water temperature are passively altered dependent variables. Therefore, any scheme that uses current or voltage as a single control variable or target has certain limitations, as the two are related rather than independent. Based on the above discussion, using the reduction furnace's electrical power as the control target is necessary in this scheme and has significant implications for stable and reliable process control.

[0091] In the above scheme, the target power curve is predetermined and is an integral part of the production process. To achieve high-quality, low-energy-consumption polysilicon reduction production, the target power curve can be determined as follows: Monitoring data is collected from the same polysilicon reduction furnace over multiple complete production cycles (one production cycle refers to the complete production process of producing one batch of polysilicon rods from furnace start-up to shutdown). These multiple complete production cycles are then screened, retaining those in which the density of the polysilicon rods is not lower than a first threshold ρ. h And the electricity consumption per unit does not exceed the second threshold W hSeveral production cycles are selected as the chosen production cycles. To obtain the target power curve, the following operation is performed: The average power at the same moment within the selected production cycle is taken as the target power value at that moment. The target power curve is then fitted based on the target power values ​​at multiple moments within a single production cycle. In the process of averaging the power values ​​at the same moment, there may inevitably be some missing data in certain production cycles. In this case, the production cycles with missing data are not included, and the remaining valid data are averaged.

[0092] The preset target power curve can be determined based on several complete production cycles of the same reduction furnace under the same material specifications and other conditions. Specifically, it involves selecting the best-performing cycles from several historical complete production cycles that meet certain conditions, and taking their average as the target power curve. For example, a target energy consumption per unit area and a target density can be set. When the energy consumption per unit area in certain production cycles does not exceed the target energy consumption and the density of the obtained polysilicon is not less than the target density, these production cycles are considered excellent cycles. Then, the target power curve is obtained by averaging the recorded power curves at the corresponding times. In the example above, the target energy consumption per unit area can be 42 kWh / kg, and the target density can be 65%.

[0093] It should be noted that for the production process control of polysilicon reduction furnaces, the material input, including the flow rates of trichlorosilane and hydrogen, can obviously be actively controlled. Generally, the current of the reduction furnace can also be controlled by a controller. It can be understood that the resistance of the silicon core is only related to the internal reaction conditions of the reduction furnace. While the current is controllable, the voltage cannot be actively controlled; it is generally monitored and acquired. Correspondingly, other system outputs, such as cooling water temperature and exhaust gas temperature, are also passively changed. In this case, the same polysilicon reduction furnace mentioned above should have the same production process strategy across multiple complete production cycles, that is, only controlling the hydrogen flow rate should be used as a control method, while the same control rules are typically applied to the trichlorosilane flow rate and current. For example, a pre-defined trichlorosilane feeding schedule determines the trichlorosilane flow rate at each moment within the production cycle, and the current is given by a preset current variation curve.

[0094] In actual polysilicon production, the above embodiments still have certain problems. The main problem is that after the polysilicon reduction furnace system has been running for a period of time, the resistance of the silicon core decreases and the rate of change is not significant, causing the change in system power to slow down. Since the above PID adjustment strictly relies on the power difference component for regulation, when the power slows down, the difference component will gradually degenerate to zero, at which point the PID adjustment process based on the power curve cannot be applied. To solve this problem, the technical solution of the present invention provides a second embodiment with staged control.

[0095] The second embodiment divides the production process of the polysilicon reduction furnace into two consecutive stages: the first stage is the target power control stage, and the second stage is the voltage difference control stage.

[0096] Based on production practice and actual production data, it was found that the voltage drop curve is relatively flat in the later stages, exhibiting good linearity. Furthermore, different reduction furnaces operate under different conditions, resulting in varying voltage drop rates, making it difficult to achieve a uniform target curve for the same power output. Utilizing the linear voltage drop characteristic to achieve downstream control ensures both control accuracy and robustness while also accommodating the application of the control method across different reduction furnaces.

[0097] Voltage difference control can also be considered as voltage drop control, aiming to make the subsequent control process as smooth as possible with a linear decrease, thereby ensuring stable and effective polysilicon deposition. The specific implementation of voltage difference control in this stage first requires determining a standard voltage calculation method, referred to here as the voltage representative value V. Its function is to hide the voltage differences from different sources, facilitating accurate and stable calculation of voltage differences. The voltage representative value is a quantity related to the voltages at the ends of multiple silicon rods within the reduction furnace, such as their average value. The differential PID control process based on the voltage representative value is roughly as follows.

[0098] Since voltage difference control does not use historical curves as control targets, but only controls the smoothness of the voltage drop, the representative voltage value V only needs to be determined from historical data of the current process. Taking the current time point t as the time base, advancing the time period T4, and using a time interval T5 as the calculation period, the voltage drop at time T4 can be obtained as the target voltage drop V. goal .

[0099]

[0100] Using the same forward time period T6, and with a time interval T5 as the calculation period, the voltage drop at time T6 can be obtained as the historical target voltage drop V. past,goal .

[0101]

[0102] In the above formula, c is a preset correction term, which can be zero, i.e., no correction.

[0103] The measured voltage value used as a reference is calculated by differential calculation based on the current time t. With a preset time span k, the voltage drop difference at time t can be obtained as follows:

[0104] .

[0105] Meanwhile, the measured historical voltage value, used as a reference, is calculated using differential calculation based on the current time t. With a preset time span k, the difference in historical voltage drop at time t after shifting forward by a preset time length L can be obtained as follows:

[0106] .

[0107] Therefore, the PID tuning parameters at this stage can be determined as follows:

[0108] .

[0109] Returning to the PID control scheme, k p k d These are the preset PID tuning parameters, so the hydrogen flow rate can be controlled based on the PID control formula to achieve smooth and linear control of the voltage drop. In practice, to optimize the system response, the PID control parameters are piecewise; if |E p If |<2, then use a smaller gain value, such as k. p =0.1, k d =0.03. And if |E p If |≥2, then use a larger gain value, such as choosing K. p =0.2, K d =0.01.

[0110] Returning to the starting point of the voltage drop control scheme described above, the specific formula for expressing the voltage representative value V is not unique. However, to comprehensively characterize the system's state, it generally needs to be a function of the voltages of each group of silicon rods in the system. Here is an example. For large polycrystalline silicon reduction furnaces, there may be groups consisting of multiple inner and outer rings of polycrystalline silicon rods, and the number of polycrystalline silicon rods in each group is not unique. Normally, the voltage difference between the ends of polycrystalline silicon rods in different groups is not particularly significant, but the voltage between the inner and outer rings is quite noticeable. Considering only the voltages of a few polycrystalline silicon rods or several groups may not fully reflect the true state of the polycrystalline silicon furnace. Therefore, the voltage representative value V can usually be obtained by summing the voltages of all groups of polycrystalline silicon rods and dividing by the number of polycrystalline silicon rods, thus obtaining the average voltage value across a single polycrystalline silicon rod. Taking a polycrystalline silicon furnace with only inner and outer rings of polycrystalline silicon rods as an example, the outer ring voltage V can usually be measured. out and inner ring voltage V in Furthermore, the number N of polycrystalline silicon rods is known at the time of feeding.

[0111]

[0112] In this embodiment, the timing of the two-stage switching can be more precisely determined to ensure a smooth switching process in the polysilicon reduction furnace, conforming to the actual internal reaction conditions. The two-stage switching method, determined through long-term practice, is based on the actual feed table design and power curve. The switching criterion is that when the amount of trichlorosilane (TCS) begins to decrease over time and the power decrease per hour is less than a certain value, it indicates that the power change is not significant and effective control based on the power curve is difficult. In practical applications, the threshold for the power decrease per hour is generally taken as 20kW.

[0113] In actual control processes, the data collected by the system often fluctuates significantly due to various disturbances, including process noise. To ensure that the collected data better reflects the actual reaction process, Gaussian filtering is required on the data used in the control process. The filtering method is as follows. First, we define two objective functions, one in one-dimensional space and one in two-dimensional space. In one-dimensional space, the objective function is: ,in It is the distance from the center point. It is the standard deviation; in two-dimensional space, the objective function becomes: , here and These represent the horizontal and vertical distances from the midpoint to the center point in two-dimensional space, respectively. The value of σ (standard deviation) has a decisive impact on the filtering effect. It controls the width of the filtering window, thus affecting the balance between smoothness and edge preservation. Choosing an appropriate σ value is crucial for achieving the desired image or signal processing results. In this implementation example, a σ value of 100 can be selected. However, different σ values ​​need to be chosen based on specific filtering requirements in different situations.

[0114] Compared with traditional manual control responses and traditional adaptive control systems, this invention has the following advantages:

[0115] The first-order difference of the target power curve is used as a reference. Compared to using the absolute value of power, the difference can more sensitively capture instantaneous changes in the system state. This method allows the control system to react more quickly to sudden changes, thereby optimizing adjustment speed and efficiency. Furthermore, the differential data provides a direct perspective on how the system changes over time, which helps in adjusting and optimizing dynamic behavior.

[0116] The application of a control strategy allows for precise adjustments to the reaction process within the furnace. The control algorithm combines current operating conditions and anticipated changes, calculating the target power change in real time based on the furnace condition and providing the most reasonable hydrogen adjustment amount to ensure the reduction furnace operates stably according to the optimal curve. Overall, this control strategy not only optimizes energy use and reduces energy consumption but also improves the production quality of polysilicon through precise control.

[0117] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0118] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0119] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0120] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0121] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for controlling a polycrystalline silicon reduction furnace, characterized in that, This includes a control stage for PID control of hydrogen flow rate based on the total power of the polysilicon reduction furnace. In this stage, PID control is performed iteratively with a first time period T1 as the control cycle. Within one control cycle, the proportional error term E of the total power of the polysilicon reduction furnace is determined based on the target power curve. p and the differential error term E d Based on this, the adjustment amount ΔH of the hydrogen flow rate in PID control is determined. The proportional error term E for determining the total power of the polysilicon reduction furnace based on the target power curve within a control cycle is... p and the differential error term E d The steps are as follows: The first-order difference ΔP of the target power curve at time point t is calculated using the calculation period T2. goal,t : ， Among them, P goal,t P represents the power value at time point t under the target power curve. goal,(t-T2) The historical power value at the corresponding historical time point after calculating the period T2 before the target power curve's time point t is calculated. Based on the power history curves collected during the current process, the first-order difference ΔP of the electrical power of the polysilicon reduction furnace at time point t is determined. real,t : ， Among them, P real,t P represents the actual power value collected at time point t. real,(t-T2) The actual power value collected at the historical time point corresponding to the calculation period T2 before time point t; Taking the current time point t1 as time point t, calculate the difference P between the actual power change and the ideal power change at the current time point t1. diff,t1 : ， Take an interval period T3, and push the current time point t1 forward by the interval period T3 to obtain a historical time point t2; Using historical time point t2 as time point t, the difference P between the actual power change and the ideal power change at historical time point t2 is calculated. diff,t2 : ， but, proportional error term E p With differential error term E d for: ， Among them, E i This is the integral error term for PID control.

2. The polycrystalline silicon reduction furnace control method as described in claim 1, characterized in that, The control formula for determining the adjustment amount ΔH of hydrogen flow rate in PID control is as follows: ， Where k p k d These are the preset control parameters.

3. The polycrystalline silicon reduction furnace control method as described in claim 1, characterized in that, The control formula for determining the adjustment amount ΔH of hydrogen flow rate in PID control is as follows: ， Where k p k d k w For the preset control parameters, the correction term E w Determination based on the state parameters of the polysilicon reduction furnace.

4. The polycrystalline silicon reduction furnace control method as described in claim 3, characterized in that, Correction Item E w The correction term E is determined based on the tail gas temperature change rate ΔT of the polycrystalline silicon reduction furnace. w The formula is: Among them, T h This is the preset threshold for the rate of change of exhaust gas temperature.

5. The polycrystalline silicon reduction furnace control method as described in claim 1, characterized in that, Set target power consumption and target density, and take the average of the corresponding time points of the historical production cycle that meets the conditions that power consumption does not exceed the target power consumption and the density of the obtained polysilicon is not less than the target density to obtain the target power curve.

6. The polycrystalline silicon reduction furnace control method according to any one of claims 1-5, characterized in that, It also includes a second control stage for PID control of hydrogen flow rate based on the voltage drop of the polysilicon reduction furnace, which follows the control stage for PID control of hydrogen flow rate based on the total power of the polysilicon reduction furnace.

7. The polycrystalline silicon reduction furnace control method as described in claim 6, characterized in that, The second control phase includes the following steps: Determine the voltage representative value V of the polysilicon reduction furnace, which is a function of the voltage of several polysilicon groups in the polysilicon reduction furnace; Based on the voltage representative value V, the following is calculated: The target voltage drop V is obtained by subtracting the time period T4 from time point t and using a time interval T5 as the calculation period. goal : ， Starting from time point t, advance the time period T6, and use a time interval T5 as the calculation period to obtain the voltage drop at time T6 as the historical target voltage drop V. past,goal : ， c is a preset correction term, and c can take the value of zero; Using the current time t as the base point, and pre-setting a time span k, the difference in voltage drop at time t is obtained as follows: ， Using the current time t as the base point, and a preset time span k, after pushing forward a preset time length L, the difference in historical voltage drop at time t is: ， Determine the proportional error term E of PID control p and the differential error term E d : ， Among them, E i This refers to the integral error of PID control. PID control is performed based on the PID control formula, which is: ， Where k p k d These are the preset control parameters.

8. A control device for a polycrystalline silicon reduction furnace, characterized in that, include Memory, which stores computer programs; A processor for executing the computer program, which, when executed, implements the polysilicon reduction furnace control method as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a processor, causes the processor to perform the steps of the polysilicon reduction furnace control method as described in any one of claims 1-7.

Images