Improved tokamak magnet power supply capacity optimization method and system thereof

Abstract

This invention relates to the field of tokamak magnet power supply technology, and provides an improved gray prediction method for optimizing tokamak magnet power supply capacity. The method includes: transforming the weights of the original sequence of the gray prediction model; constructing background values ​​using the composite Simpson formula and simplifying the composite Simpson formula; verifying the predicted displacement using known plasma displacement information, and accurately predicting the next cycle displacement signal by verifying and training the gray prediction model; and obtaining the magnet power supply reference signal in advance by predicting the plasma displacement. When the plasma experiences a small displacement shift, the magnet power supply only needs to output a small current to achieve plasma displacement excitation balance control, thereby reducing the capacity and optimizing the magnet power supply capacity. This solves the problem of the continuously increasing demand for tokamak magnet power supply capacity affecting the practical application of fusion devices.

Description

Technical Field

[0001] This invention relates to the field of tokamak magnet power supply technology, and more specifically to an improved gray prediction method for optimizing tokamak magnet power supply capacity. Background Technology

[0002] The tokamak device is crucial for achieving controlled nuclear fusion, and the displacement control of the plasma directly impacts the success of this process. The tokamak magnet power supply can rapidly establish a corresponding magnetic field based on the real-time displacement signal of the plasma, achieving magnetic confinement control of the plasma displacement and realizing plasma displacement feedback balance control. The design of the tokamak magnet power supply is based on the capacity required for the maximum displacement balance control of the plasma. Plasma control experts estimate the required capacity of the magnet power supply, and the magnet power supply designers then design a magnet power supply with a corresponding capacity to excite the load coil, achieving closed-loop magnetic confinement control of the plasma.

[0003] Tokamak devices involve a wide range of disciplines. The operational performance of magnet power supplies, such as poloidal field power supplies and fast-control power supplies, depends on their response speed and maximum current output capability. To achieve plasma displacement balance control under large displacement offsets, high demands are placed on the capacity and response performance of the magnet power supply. In his May 2015 doctoral dissertation, "Simulation and Experimental Study of Passive Stability and Active Control of Vertical Displacement in EAST," Liu Lei of the University of Science and Technology of China described how the tokamak magnet power supply project team estimated the magnet power supply capacity by establishing a power supply model based on the power supply's response delay and current ramp-up delay, setting the maximum acceptable initial plasma displacement to achieve robust control, and thus estimating the relevant magnet power supply capacity. With the development of controlled nuclear fusion engineering, the demand for increased tokamak magnet power supply capacity will continue to rise. Considering that controlled fusion devices will ultimately benefit humanity through grid-connected power generation, the current demand for continuously and significantly increasing magnet power supply capacity will inevitably seriously affect the practical application of fusion devices. Optimizing the plasma control system and improving its performance will help reduce the ever-increasing demand for magnet power capacity, maximizing the commercial value of controlled nuclear fusion.

[0004] The control objective of some magnet power supplies, such as fast-control power supplies, is to rapidly track the reference signal given by the plasma control system (PCS), output a corresponding current to excite the load coil, and achieve magnetic confinement control of the plasma's vertical displacement. The shorter the time for output current and voltage establishment on the load coil, the faster plasma displacement balance control can be achieved. Furthermore, the faster the output current or the smaller the plasma displacement deviation, the smaller the target current required to control the plasma to the equilibrium position. Therefore, rapid control of the magnet power supply's output current, or predictive control of plasma displacement, is an important way to reduce the required current for plasma displacement balance control and a crucial method for optimizing the capacity of the tokamak magnet power supply. Employing a digital control structure in the magnet power supply, improving the dynamic response speed of the output current, and reducing digital control delay and inherent system delay are all important methods to improve power supply control performance. The inherent system delay can be addressed by predicting the plasma displacement in advance to obtain the reference signal, allowing the magnet power supply to output a small current quickly in advance. By predicting plasma displacement and obtaining a reference current in advance, a small current can be output to pull the plasma back to its equilibrium position when a small displacement deviation occurs. This is an important way to reduce the magnet power supply capacity, optimize the magnet power supply capacity, and achieve stable control of plasma displacement balance.

[0005] Grey GM(1,1) prediction can predict target quantities without requiring a precise mathematical model of the target, and the prediction process is simple, fast, and computationally convenient. Existing technology includes Chinese invention patent CN113190535A, "A Gas Concentration Prediction Method Based on an Improved Grey-Long Short-Term Memory Neural Network Combined Model," which discloses an improved grey GM(1,1) model prediction method. This method utilizes ensemble learning to combine an improved grey prediction model with a long short-term memory neural network to establish an improved grey-long short-term memory neural network combined prediction model, achieving accurate gas concentration prediction. The paper "Application of Improved Grey Model with Initial Value Correction and Function Transformation in Pipeline Corrosion Depth Prediction," published in the *Journal of Safety and Environment* in June 2023, improves the initial value of grey prediction by increasing the weight of new information, achieving accurate prediction of pipeline corrosion depth. The paper "Research on Grey Settlement Prediction Model Based on Background Value Optimization," published in the *Journal of Surveying and Mapping and Spatial Geographic Information* in December 2022, utilizes the Cortes trapezoidal formula to reconstruct the background value, optimizing the grey GM(1,1) model and improving the prediction accuracy of underground settlement.

[0006] The aforementioned patent applications and papers modified the gray GM(1,1) prediction model, improving the gray GM(1,1) model or reconstructing the original sequence and background values. However, the reconstruction function or improved algorithm used is too complex, which is not conducive to the implementation of digital control of the tokamak magnet power supply and puts a heavy burden on the digital processor. Furthermore, in the process of gray prediction predicting the target quantity, the original data is not used for repeated prediction verification and correction, and the accuracy of the target quantity prediction needs to be further improved. Consequently, the optimization of the magnet power supply capacity and the plasma balance displacement control effect are poor. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to reduce the power supply capacity of the tokamak magnet by accurately predicting plasma displacement and obtaining displacement deviation in advance.

[0008] The present invention solves the above-mentioned technical problems through the following technical means:

[0009] This invention provides an improved gray prediction method for optimizing the power supply capacity of a tokamak magnet, comprising the following steps:

[0010] S1. Collect plasma displacement data from the last four switching cycles provided by PCS to form the original sequence X of the gray GM(1,1) prediction model. (0) The original sequence expression is

[0011] X (0) =[X (0) (1),X (0) (2),X (0) (3),X (0) (4)] (1)

[0012] And the weights of each piece of information in the original sequence are transformed;

[0013] S2, based on the original sequence X (0) Generate the cumulative sequence X (1) (k), the cumulative sequence expression is:

[0014]

[0015] Then, the background value Z is reconstructed using the simplified Simpson formula. (1) (k);

[0016] S3, Based on background value Z (1) (k) and the original displacement sampling data are used to calculate the gray development coefficient a and gray quantity b; based on the gray development coefficient a, gray quantity b, and the original sequence X... (0) The prediction formula for obtaining the original displacement signal is obtained.

[0017] S4. Predict based on the prediction formula. The value is calculated and the prediction error is checked; the displacement output value of the next cycle is predicted using a prediction formula that meets the error check.

[0018] S5. Predicted displacement The plasma displacement deviation is obtained by comparing it with the reference displacement. The transfer function is calculated based on the displacement deviation and the reference signal, and the reference signal is obtained by calculation.

[0019] Furthermore, the weights of each piece of information in the original sequence are transformed in step S1, and the specific transformation expression is as follows:

[0020]

[0021] Furthermore, the reconstruction of the background value using the simplified Simpson formula in step S2 specifically involves:

[0022] In the composite Simpson formula, the sampling period from k to k+1 is divided into m equal intervals, and the cumulative sequence curve expression is set as f(x). i ), including

[0023]

[0024] The expression for solving the background value in the approximation integral process using the composite Simpson formula is as follows:

[0025]

[0026] The accumulated sequence X within the sampling period k to k+1 (1) (k) to X (1) If the change value of (k+1) is divided into m equal parts, then the cumulative change value of the sequence in the m intervals is...

[0027]

[0028] X in the cumulative sequence (1) (k) and X (1) The value of (k+1) is known, therefore let

[0029] f(x i ) = X (1) (k)+iΔ,i=0,1,…,m-1 (14)

[0030]

[0031] The composite Simpson formula is simplified by combining formulas (12) to (15), and the reconstructed background value is

[0032]

[0033] Further, step S3, based on the background value Z (1) (k) and the original displacement sampling data are used to calculate the gray development coefficient a and gray quantity b. The specific calculation method is as follows:

[0034]

[0035] X n =[X (0) (2)X (0) (3)X (0) (4)] T (7)

[0036] Furthermore, the prediction formula for the original displacement signal mentioned in step S3 is as follows:

[0037]

[0038] Furthermore, step S4 involves using a prediction formula that satisfies the error check to predict the displacement output value for the next cycle. Specifically:

[0039] Let the acceptable prediction error be e. The error check formula is used to determine whether the predicted value meets the prediction error requirement. The error check formula is:

[0040]

[0041] If the error check is satisfied, then the next displacement output value is predicted according to the prediction formula.

[0042] Otherwise, update the weights of the new information in the original sequence, recalculate the gray development coefficient 'a' and the gray quantity 'b', and obtain the new predicted value according to the prediction formula. And re-verify until the error check is satisfied.

[0043] Furthermore, the reference signal transfer function calculated in step S5 is as follows:

[0044]

[0045] Among them, K p K i K d The coefficients of the transfer function are calculated for the reference signal, where s is the Laplace operator; the input to the transfer function is the obtained plasma displacement deviation, and the output is the reference signal supplied to the tokamak magnet power supply.

[0046] This invention also provides an improved gray prediction tokamak magnet power supply capacity optimization system, which employs the above-described method during system operation and includes the following modules:

[0047] The acquisition module is used to acquire plasma displacement data from the most recent four switching cycles provided by the PCS to form the original sequence X of the gray GM(1,1) prediction model. (0) The original sequence expression is

[0048] X (0) =[X (0) (1),X (0) (2),X (0) (3),X (0) (4)] (1)

[0049] And the weights of each piece of information in the original sequence are transformed;

[0050] Background value calculation module, used to calculate the background value based on the original sequence X (0) Generate the cumulative sequence X (1) (k), the cumulative sequence expression is:

[0051]

[0052] Then, the background value Z is reconstructed using the simplified Simpson formula. (1) (k);

[0053] The gray prediction module is used to predict based on the background value Z. (1) (k) and the original displacement sampling data are used to calculate the gray development coefficient a and gray quantity b; based on the gray development coefficient a, gray quantity b, and the original sequence X... (0) The prediction formula for obtaining the original displacement signal is obtained.

[0054] The error verification module is used to predict based on the prediction formula. The value is calculated and the prediction error is checked; the displacement output value of the next cycle is predicted using a prediction formula that meets the error check.

[0055] Reference signal output module, used to predict the obtained displacement The plasma displacement deviation is obtained by comparing it with the reference displacement. The transfer function is calculated based on the displacement deviation and the reference signal, and the reference signal is obtained by calculation.

[0056] Furthermore, the acquisition module includes an information weight transformation unit, and the transformation expression used during its operation is specifically as follows:

[0057]

[0058] Furthermore, the background value calculation module includes a Simpson background value reconstruction unit, which operates as follows:

[0059] In the composite Simpson formula, the sampling period from k to k+1 is divided into m equal intervals, and the cumulative sequence curve expression is set as f(x). i ), including

[0060]

[0061] The expression for solving the background value in the approximation integral process using the composite Simpson formula is as follows:

[0062]

[0063] The accumulated sequence X within the sampling period k to k+1 (1) (k) to X (1) If the change value of (k+1) is divided into m equal parts, then the cumulative change value of the sequence in the m intervals is...

[0064]

[0065] X in the cumulative sequence (1) (k) and X (1) The value of (k+1) is known, therefore let

[0066] f(x i ) = X (1) (k)+iΔ,i=0,1,…,m-1 (14)

[0067]

[0068] The composite Simpson formula is simplified by combining formulas (12) to (15), and the reconstructed background value is

[0069]

[0070] The advantages of this invention are:

[0071] (1) This invention improves the influence of old information on prediction accuracy in traditional gray prediction by transforming the weight of each piece of information in the original sequence to increase the influence weight of new information, thereby further improving the prediction accuracy of the gray GM(1,1) prediction model.

[0072] (2) This invention uses the Simpson formula to replace the median theorem to construct the background value, which reduces the background value construction error in the traditional gray GM(1,1) prediction model. It also simplifies the composite Simpson formula, ensuring the accuracy of gray prediction without increasing the amount of computation, and ensuring that the improved gray GM(1,1) prediction model can quickly and accurately predict plasma displacement.

[0073] (3) This invention uses known displacement information to perform error verification on the predicted displacement until a gray GM(1,1) prediction model that meets the error requirements is trained, thus realizing the closed-loop feedback verification function of gray prediction. Using the gray GM(1,1) prediction model that meets the error requirements trained to perform displacement prediction greatly improves the accuracy of plasma displacement prediction.

[0074] (4) By combining the improved gray GM(1,1) prediction model to predict the plasma displacement trajectory in advance, the reference signal can be obtained in advance. When the plasma displacement is slightly deviated, only a small current needs to be output from the tokamak magnet power supply to excite the load coil. The plasma displacement balance control can be achieved through the magnetic field, which reduces the requirements for the power capacity of the magnet power supply and realizes the optimization of the tokamak magnet power supply capacity. Attached Figure Description

[0075] Figure 1 This is a schematic diagram of the plasma displacement balance control method for optimizing the power supply capacity of a tokamak magnet using the improved gray GM(1,1) model in an example of the present invention.

[0076] Figure 2 This is a schematic diagram showing the comparison of background values ​​before and after the improvement in an example of the present invention;

[0077] Figure 3 The figure shows the simulation results of plasma balance control in an example of the present invention. Detailed Implementation

[0078] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0079] Example 1

[0080] This invention provides an improved gray prediction method for optimizing the power supply capacity of tokamak magnets, such as... Figure 1 As shown, it includes the following steps:

[0081] S1. Collect plasma displacement data from the last four switching cycles provided by PCS to form the original sequence X of the gray GM(1,1) prediction model. (0) The original sequence expression is

[0082] X (0) =[X (0) (1),X (0) (2),X(0) (3),X (0) (4)] (1)

[0083] New information often better aligns with the predicted trend. To reduce the influence of old information and increase the weight of new information, the weights of each piece of information in the original sequence are transformed to amplify the impact of the new information. The transformation expressions for each piece of information are as follows:

[0084]

[0085] The transformed δ(1)~δ(4) form a new original sequence, and the oldest information δ(1) in the new original sequence also contains some new information X. (0) (4) This increases the weight of new information.

[0086] S2, based on the original sequence X (0) Generate the cumulative sequence X (1) (k), the cumulative sequence expression is:

[0087]

[0088] Then, based on the concept of integration, the accumulated sequence X is... (1) (k) in the (k-1)th sampling period time t k-1 Time t up to the kth sampling period k The background value Z is obtained by integrating the integral on the surface. (1) (k), the background value expression is:

[0089]

[0090] In grey prediction, to simplify calculations, the median theorem is used to approximate the background value. The approximate expression for the background value is as follows:

[0091]

[0092] However, the background values ​​constructed using the median theorem have certain errors. Therefore, this embodiment uses a simplified Simpson formula to reconstruct the background values, ensuring the accuracy of gray prediction without increasing the computational load.

[0093] like Figure 2 As shown, the background value Z before improvement (1) (k) is X (1) (k) and X (1) The area of ​​the right trapezoid formed by the (k+1) line and the time axis from k to k+1, and the improved background value Z. (1) (k) is X (1) (k) and X (1)The area of ​​the curvilinear trapezoid enclosed by the (k+1) curve and the k~k+1 time axis. The improved background value construction process uses a simplified Simpson formula to approximate the curve integral process, making the background value construction more accurate and ensuring the accuracy of gray prediction without increasing the computational load. In the composite Simpson formula, the sampling period time of k~k+1 is divided into m equal intervals, and the cumulative sequence curve expression is set as f(x i Furthermore, the specific expression requires complex calculations or fitting to obtain, among which...

[0094]

[0095] The expression for solving the background value in the approximation integral process using the composite Simpson formula is as follows:

[0096]

[0097] The cumulative sequence curve expression f(x) i The specific expression is difficult to obtain, therefore the accumulated sequence X within the sampling period k to k+1 is used. (1) (k) to X (1) If the change value of (k+1) is divided into m equal parts, then the cumulative change value of the sequence in the m intervals is...

[0098]

[0099] X in the cumulative sequence (1) (k) and X (1) The value of (k+1) is known, therefore let

[0100] f(x i ) = X (1) (k)+iΔ,i=0,1,…,m-1 (14)

[0101]

[0102] The composite Simpson formula is simplified by combining formulas (12) to (15). The simplified Simpson formula is used to approximate the integral process and replace the mean value theorem to reconstruct the background value. The background value reconstructed by the simplified Simpson formula is

[0103]

[0104] Using the simplified Simpson formula described in formula (16) to approximate the integration process instead of the mean value theorem to reconstruct the background value not only ensures higher accuracy of the reconstructed background value but also simplifies the computation of the integration process.

[0105] S3, Based on background value Z (1)(k) and the original displacement sampling data are used to calculate the gray development coefficient a and gray quantity b. The formulas for calculating a and b are as follows:

[0106]

[0107] X n =[X (0) (2)X (0) (3)X (0) (4)] T (7)

[0108] Based on the gray development coefficient a, gray quantity b, and the original sequence X (0) The original displacement signal is predicted using the following formula:

[0109]

[0110] S4. Predict based on the prediction formula. The value is calculated and the prediction error is checked. Then, the displacement output value for the next cycle is predicted using a prediction formula that meets the error check. Let the acceptable prediction error be e. Use the error check formula to determine whether the predicted value meets the prediction error requirement. The error check formula is:

[0111]

[0112] If the error verification formula is satisfied, then the next displacement output value is predicted according to the prediction formula. Otherwise, update the weights of the new information in the original sequence, and then continue to update the values ​​of a and b to make predictions. The next displacement output value is predicted only when the error check formula is satisfied.

[0113] In steps S3 and S4 above, it is first necessary to predict the plasma displacement signal for the fourth switching cycle based on the plasma displacement data from the last four switching cycles. Compare the predicted displacement signal with the actual displacement signal X (0) (4) Compare the predicted and actual displacement signals to achieve error verification until the predicted displacement signal and the actual displacement signal meet the prediction error, thus achieving the purpose of gray GM(1,1) prediction closed-loop verification. Then predict and output the plasma displacement signal for the next switching cycle. This ensures the accuracy of displacement prediction.

[0114] S5. Predicted displacement The plasma displacement deviation is obtained by comparing it with a reference displacement. The transfer function is then calculated based on the displacement deviation and the reference signal to obtain the reference signal. The tokamak magnet power supply tracks the reference signal and outputs current to excite the load coil, achieving magnetic balance control of the plasma displacement.

[0115] The reference signal is obtained by obtaining the plasma displacement deviation in advance. When the plasma displacement has a small deviation, only a small reference signal is needed to supply the tokamak magnet power supply. The magnet power supply only needs to output a small current to excite the load coil. In other words, only a magnet power supply with a small capacity is needed to achieve magnetic balance control of plasma displacement.

[0116] The transfer function of the reference signal is calculated as follows

[0117]

[0118] Among them, K p K i K d The coefficients of the transfer function are calculated for the reference signal, where s is the Laplace operator; in the traditional function, the input is the plasma displacement deviation obtained in advance, and the output is the reference signal given to the tokamak magnet power supply.

[0119] In this embodiment, a simulation analysis is performed on the tokamak magnet power supply capacity optimization method based on improved grey prediction. The simulation parameters are: reference signal amplitude ±10V, load inductance 160μH, inductor internal resistance 25mΩ, magnet power supply feedback current sensor ratio 50000:1, plasma discharge current set to 1mA, and reference signal transfer function parameter K. p =100, K i =24, K d =3, the initial plasma displacement deviation is 10cm. An improved grey prediction method is used to predict the plasma displacement in advance, so as to obtain the reference signal ahead of time. For example... Figure 3 As shown in the simulation waveform diagram, the reference signal is limited to within ±10V, and the plasma displacement can be quickly controlled to a stable zero position. The peak current required by the tokamak magnet power supply is 215.7kA, and the peak voltage required is 5.45kV. In comparison, when the plasma discharge current in the CFETR nuclear fusion device is set to 1mA, the designers specified a required magnet power supply capacity of ±7.46kV / ±309.74kA. Through grey prediction to predict the plasma displacement in advance, the reference signal can be output earlier when the plasma deviates slightly. This significantly reduces the maximum current required by the magnet power supply and the maximum voltage required to be established at the output terminal. This demonstrates that the present invention can significantly reduce the power capacity of the tokamak magnet power supply, optimizing the magnet power supply capacity while also ensuring good plasma displacement balance control performance.

[0120] Example 2

[0121] It should be further explained that, based on the same inventive concept, this embodiment provides an improved gray prediction tokamak magnet power capacity optimization system. The system executes the method described in Embodiment 1 during operation and includes the following modules:

[0122] The acquisition module is used to acquire plasma displacement data from the most recent four switching cycles provided by the PCS to form the original sequence X of the gray GM(1,1) prediction model. (0) The original sequence expression is

[0123] X (0) =[X (0) (1),X (0) (2),X (0) (3),X (0) (4)] (1)

[0124] And the weights of each piece of information in the original sequence are transformed;

[0125] Background value calculation module, used to calculate the background value based on the original sequence X (0) Generate the cumulative sequence X (1) (k), the cumulative sequence expression is:

[0126]

[0127] Then, the background value Z is reconstructed using the simplified Simpson formula. (1) (k);

[0128] The gray prediction module is used to predict based on the background value Z. (1) (k) and the original displacement sampling data are used to calculate the gray development coefficient a and gray quantity b; based on the gray development coefficient a, gray quantity b, and the original sequence X... (0) The prediction formula for obtaining the original displacement signal is obtained.

[0129] The error verification module is used to predict based on the prediction formula. The value is calculated and the prediction error is checked; the displacement output value of the next cycle is predicted using a prediction formula that meets the error check.

[0130] Reference signal output module, used to predict the obtained displacement The plasma displacement deviation is obtained by comparing it with the reference displacement, and the transfer function is calculated based on the displacement deviation and the reference signal to obtain the reference signal.

[0131] The acquisition module includes an information weight transformation unit, and the transformation expression used during its operation is as follows:

[0132]

[0133] The background value calculation module includes a Simpson background value reconstruction unit, which operates as follows:

[0134] In the composite Simpson formula, the sampling period from k to k+1 is divided into m equal intervals, and the cumulative sequence curve expression is set as f(x). i ), including

[0135]

[0136] The expression for solving the background value in the approximation integral process using the composite Simpson formula is as follows:

[0137]

[0138] The accumulated sequence X within the sampling period k to k+1 (1) (k) to X (1) If the change value of (k+1) is divided into m equal parts, then the cumulative change value of the sequence in the m intervals is...

[0139]

[0140] X in the cumulative sequence (1) (k) and X (1) The value of (k+1) is known, therefore let

[0141] f(x i ) = X (1) (k)+iΔ,i=0,1,…,m-1 (14)

[0142]

[0143] The composite Simpson formula is simplified by combining formulas (12) to (15), and the reconstructed background value is

[0144]

[0145] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An improved gray prediction method for optimizing the power supply capacity of tokamak magnets, characterized in that, Includes the following steps: S1. Collect plasma displacement data from the last four switching cycles provided by PCS to form the original sequence of the gray GM(1,1) prediction model. X (0) The original sequence expression is （1） The weights of each piece of information in the original sequence are then transformed as follows: （10） Transformed (1)~ (4) formed a new original sequence X (0)’ ; S2, based on the new original sequence X (0)’ Generate an accumulated sequence X (1)’ ( k The cumulative sequence expression is: （2） Then, the background value is reconstructed using the simplified Simpson formula. Z (1) ( k ); S3, Based on background value Z (1) ( k ) and the new original sequence X (0)’ Calculate the grey development coefficient a and gray quantity b According to the grey development coefficient a、 Gray quantity b and the new original sequence X (0)’ The prediction formula for obtaining the original displacement signal is obtained. S4. Predict based on the prediction formula. The value is calculated and the prediction error is checked; the displacement output value of the next cycle is predicted using a prediction formula that meets the error check. ; S5. Predicted displacement The plasma displacement deviation is obtained by comparing it with the reference displacement. The transfer function is calculated based on the displacement deviation and the reference signal, and the reference signal is obtained by calculation.

2. The improved grey prediction method for optimizing the power supply capacity of a tokamak magnet according to claim 1, characterized in that, Step S2, which involves reconstructing the background value using the simplified Simpson formula, specifically includes: In the composite Simpson formula, k ~ k +1 sampling period time is divided into m For each interval, the cumulative sequence curve expression is set as follows: f ( x i ), including （11） The expression for solving the background value in the approximation integral process using the composite Simpson formula is as follows: （12） Will k ~ k Accumulated sequence within +1 sampling period X (1)’ ( k )arrive X (1)’ ( k +1) Change value m Divide into equal parts, then m The cumulative sequence change value in each interval is （13） In the cumulative sequence X (1)’ ( k )and X (1)’ ( k The value of +1 is known, therefore let （14） （15） The composite Simpson formula is simplified by combining formulas (12) to (15), and the reconstructed background value is （16）。 3. The improved grey prediction method for optimizing the power supply capacity of a tokamak magnet according to claim 2, characterized in that, Step S3, based on background value Z (1) ( k ) and the new original sequence X (0)’ Calculate the grey development coefficient a and gray quantity b The specific calculation method is as follows: （5） （6） （7）。 4. The improved grey prediction method for optimizing the power supply capacity of a tokamak magnet according to claim 3, characterized in that, The prediction formula for the original displacement signal mentioned in step S3 is as follows: （8）。 5. The improved grey prediction method for optimizing the power supply capacity of a tokamak magnet according to claim 4, characterized in that, Step S4 describes using a prediction formula that satisfies the error check to predict the displacement output value for the next cycle. Specifically: Let the acceptable prediction error be... e The error check formula is used to determine whether the predicted value meets the prediction error. The error check formula is as follows: （9） If the error check is satisfied, then the next displacement output value is predicted according to the prediction formula. ; Otherwise, update the weights of the new information in the original sequence and recalculate the gray development coefficient. a and gray quantity b The new predicted value is obtained based on the prediction formula. And re-verify until the error check is satisfied.

6. The improved grey prediction method for optimizing the power supply capacity of a tokamak magnet according to claim 1, characterized in that, The reference signal transfer function calculated in step S5 is as follows: （17） in, K p , K i , K d Calculate the coefficients of the transfer function for the reference signal. s The Laplace operator is used; the input to the transfer function is the obtained plasma displacement deviation, and the output is the reference signal given to the tokamak magnet power supply.

7. An improved grey prediction-based tokamak magnet power capacity optimization system, characterized in that, The system operates using the method described in any one of claims 1-6, comprising the following modules: The acquisition module is used to acquire plasma displacement data from the most recent four switching cycles provided by the PCS to form the original sequence of the gray GM(1,1) prediction model. X (0) The original sequence expression is （1） The weights of each piece of information in the original sequence are then transformed as follows: （10） Transformed (1)~ (4) formed a new original sequence X (0)’ ; Background value calculation module, used based on the new original sequence X (0)’ Generate an accumulated sequence X (1)’ ( k The cumulative sequence expression is: （2） Then, the background value is reconstructed using the simplified Simpson formula. Z (1) ( k ); The gray prediction module is used to predict based on background values. Z (1) ( k ) and the new original sequence X (0)’ Calculate the grey development coefficient a and gray quantity b According to the grey development coefficient a、 Gray quantity b and the new original sequence X (0)’ The prediction formula for obtaining the original displacement signal is obtained. The error verification module is used to predict based on the prediction formula. The value is calculated and the prediction error is checked; the displacement output value of the next cycle is predicted using a prediction formula that meets the error check. ; Reference signal output module, used to predict the obtained displacement The plasma displacement deviation is obtained by comparing it with the reference displacement, and the transfer function is calculated based on the displacement deviation and the reference signal to obtain the reference signal.

8. The improved grey prediction tokamak magnet power capacity optimization system according to claim 7, characterized in that, The background value calculation module includes a Simpson background value reconstruction unit, which operates as follows: In the composite Simpson formula, k ~ k +1 sampling period time is divided into m For each interval, the cumulative sequence curve expression is set as follows: f ( x i ), including （11） The expression for solving the background value in the approximation integral process using the composite Simpson formula is as follows: （12） Will k ~ k Accumulated sequence within +1 sampling period X (1)’ ( k )arrive X (1)’ ( k +1) Change value m Divide into equal parts, then m The cumulative sequence change value in each interval is （13） In the cumulative sequence X (1)’ ( k )and X (1)’ ( k The value of +1 is known, therefore let （14） （15） The composite Simpson formula is simplified by combining formulas (12) to (15), and the reconstructed background value is （16）。

Images