An optimization method for semiconductor refrigeration control system

By using an optimized bidirectional long and short-term memory network and energy efficiency control algorithm in semiconductor refrigeration control systems, combining time gating and adaptive attention mechanisms, the stability and energy efficiency problems of the system in the face of complex environment changes are solved, and the system is realized. Accurate temperature regulation and energy efficiency management improve the environmental performance and sustainability of the system.

CN118654423BActive Publication Date: 2024-12-03WUXI GUANYA REFRIGERATION TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411110495.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-14
Publication Date
2024-12-03
Estimated Expiration
2044-08-14

AI Technical Summary

Technical Problem

Existing semiconductor refrigeration control systems are difficult to effectively process the time correlation and dynamic changes in data, resulting in the inability to capture the true laws of system state changes over time, and are unstable in the face of complex environmental changes and equipment operation characteristics, and it is difficult to achieve dynamics. Adjustment and optimization lead to inefficiency and energy waste.

Method used

The optimized two-way long and short-term memory network is used to process the feature vectors, combine the regression model and periodic trend components to predict temperature, and calculate the total energy efficiency through energy efficiency control algorithm, optimize energy efficiency through frequency update, and introduce time gating and adaptive attention mechanism enhancement Time sensitivity, improved Fourier series analog periodic trends, and calculated energy efficiency using squared voltage, exponential decay, normalization of load, and logarithmic and sinusoidal functions of frequency.

Benefits of technology

Improves the accuracy and reliability of temperature prediction, reduces energy waste, and improves the operating efficiency of semiconductor refrigeration control systems, ensuring optimal performance under various load conditions, reducing operating costs, and improving environmental performance and sustainability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118654423B_ABST
    Figure CN118654423B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of semiconductor refrigeration control systems, and in particular to an optimization method for semiconductor refrigeration control systems. The content includes: acquiring and standardizing the original data points, generating and processing feature vectors based on the standardized data points, and obtaining weighted hidden states; based on the weighted hidden states, using regression model fitting to obtain the output of the regression model; calculating the predicted temperature of the semiconductor refrigeration control system, generating a control signal based on the difference between the predicted temperature and the actual temperature and mapping it to the actual operating parameters, and regulating the temperature of the semiconductor refrigeration control system; calculating the total energy efficiency based on the predicted temperature, and optimizing it through frequency updates. It solves the problem that the real law of the system state changing over time cannot be captured in the predictive control process; the system is unstable in the face of complex environmental changes and equipment operating characteristics, and it is difficult to achieve effective dynamic adjustment and optimization, resulting in low system energy efficiency and energy waste.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of semiconductor refrigeration control systems, and in particular to an optimization method for semiconductor refrigeration control systems. Background Art

[0002] In the industrial scenarios of new energy battery testing and semiconductor production, there are extremely high requirements for the accuracy and reliability of temperature control, because temperature fluctuations directly affect the test results of battery performance and the production quality of semiconductor products. Optimizing semiconductor refrigeration control systems has become a technical problem that needs to be solved urgently. An optimization method for semiconductor refrigeration control systems has demonstrated unique technical advantages, which can achieve precise temperature control, improve the reliability and stability of semiconductor refrigeration control systems, and reduce energy efficiency.

[0003] my country's patent application number: CN201610052667.5, publication date: 2016.06.22, discloses a constant temperature control system for lithium battery performance testing, including a temperature detection module, a temperature control module, a temperature adjustment module, and a lithium battery performance testing module. The temperature detection module includes a temperature sensor, the temperature control module includes a single-chip microcomputer, the temperature adjustment module includes a cooling drive power supply circuit and a heating drive power supply circuit, and the lithium battery performance testing module connects the lithium battery and tests various performance parameters of the lithium battery. The constant temperature control system for lithium battery performance testing described in the invention has the advantages of using semiconductor cooling plates and heating plates to adjust the temperature of the constant temperature box, which is green and environmentally friendly, the cooling plates can work continuously, have a long life, are easy to use, and can stably control the ambient temperature of the lithium battery performance testing module, which can effectively avoid the influence of external temperature on the lithium battery performance test and improve the test accuracy of lithium battery performance.

[0004] However, the above technology has at least the following technical problems: it is difficult to effectively process the time correlation and dynamic change characteristics in the data during the prediction and control process, resulting in the inability to capture the true law of system state changes over time; it is unstable when facing complex environmental changes and equipment operating characteristics, and lacks comprehensive consideration of periodic trends and nonlinear influencing factors; it is difficult to achieve effective dynamic adjustment and optimization, and cannot flexibly respond to the energy efficiency impact brought about by equipment load and environmental changes, resulting in low system energy efficiency and energy waste. Summary of the invention

[0005] The present invention provides an optimization method for a semiconductor refrigeration control system to solve the problems that it is difficult to effectively process the time correlation and dynamic change characteristics in the data during the prediction and control process, resulting in the inability to capture the true law of the system state changing over time; the unstable performance when facing complex environmental changes and equipment operating characteristics, and the lack of comprehensive consideration of periodic trends and nonlinear influencing factors; and the difficulty in achieving effective dynamic adjustment and optimization, and the inability to flexibly respond to the energy efficiency impact brought about by equipment load and environmental changes, resulting in low system energy efficiency and energy waste.

[0006] The present invention provides a method for optimizing a semiconductor refrigeration control system, which specifically includes the following technical solutions:

[0007] A method for optimizing a semiconductor refrigeration control system comprises the following steps:

[0008] S1. Obtain and standardize the original data points, generate feature vectors based on the standardized data points; use an optimized bidirectional long short-term memory network to process the feature vectors to obtain weighted hidden states; based on the weighted hidden states, use the regression model to fit the output of the regression model;

[0009] S2, obtaining the predicted temperature of the semiconductor refrigeration control system through a refrigeration temperature prediction algorithm, generating a control signal according to the difference between the predicted temperature and the actual temperature of the semiconductor refrigeration control system, mapping the control signal to an actual operating parameter, and regulating the temperature of the semiconductor refrigeration control system by controlling the actual operating parameter;

[0010] S3, based on the predicted temperature of the semiconductor refrigeration control system, the total energy efficiency is calculated using the energy efficiency control algorithm, and the total energy efficiency is optimized through frequency update;

[0011] Among them, the specific calculation formula of total energy efficiency is:

[0012] ,

[0013] in, Indicates total energy efficiency; Indicates the voltage of semiconductor refrigeration control system; Indicates the semiconductor refrigeration control system at time point The predicted temperature; represents the standardized temperature parameter; represents the production load; represents the standardized load parameter; Indicates the operating frequency of the semiconductor refrigeration control system; Indicates the maximum frequency parameter.

[0014] Preferably, the S1 specifically includes:

[0015] In the optimized bidirectional long short-term memory network, the temporal gating mechanism and the adaptive attention mechanism are introduced.

[0016] Preferably, the S1 specifically includes:

[0017] The time gating mechanism controls the influence of the feature vector at the current time point and the hidden state at the previous time point on the current state by introducing a time gating unit into the structure of the bidirectional long short-term memory network.

[0018] Preferably, the S1 specifically includes:

[0019] The attention weights of the hidden states are calculated through an adaptive attention mechanism.

[0020] Preferably, the S2 specifically includes:

[0021] The refrigeration temperature prediction algorithm predicts the temperature of the semiconductor refrigeration control system by combining the output of the regression model and the periodic trend component.

[0022] Preferably, the S2 specifically includes:

[0023] The periodic trend component is simulated by a modified Fourier series.

[0024] Preferably, the S3 specifically includes:

[0025] The energy efficiency control algorithm is based on the predicted temperature of the semiconductor refrigeration control system. The total energy efficiency is calculated by introducing the square of the voltage, the exponential decay of the predicted temperature, the normalization of the load, the logarithm of the frequency and the sine function.

[0026] Preferably, the S3 specifically includes:

[0027] The total energy efficiency is adjusted by frequency update to obtain a new frequency value; the specific formula for frequency update is:

[0028] ,

[0029] in, Indicates the new frequency value; Indicates the operating frequency of the semiconductor refrigeration control system; represents the learning rate; represents the momentum factor; represents the momentum term; represents the second derivative weighting factor; Total energy efficiency Frequency The first derivative of ; Total energy efficiency Frequency The second derivative of

[0030] Based on the new frequency value, the energy efficiency control algorithm is applied again to calculate the new total energy efficiency.

[0031] The beneficial effects of the technical solution of the present invention are:

[0032] 1. By introducing the time gating mechanism and adaptive attention mechanism, the optimized bidirectional long short-term memory network can capture the time points The enhanced time sensitivity enables the optimized bidirectional long short-term memory network to accurately identify time points. patterns and trends in the feature vectors, thereby improving the accuracy and reliability of the analysis;

[0033] 2. The temperature is predicted by combining the output of the regression model with the refrigeration temperature prediction algorithm of the periodic trend component, which effectively captures the periodic fluctuations caused by the equipment operation cycle or environmental factors, adds key periodic information to the temperature prediction, and improves the accuracy and reliability of the temperature prediction in combination with the output of the regression model, which is conducive to the precise regulation of the semiconductor refrigeration control system;

[0034] 3. Based on the semiconductor refrigeration control system at the time point The predicted temperature is calculated through the energy efficiency control algorithm to obtain the total energy efficiency, and the total energy efficiency is adjusted through frequency update. This not only reduces energy waste and improves the operating efficiency of the semiconductor refrigeration control system, but also ensures that the semiconductor refrigeration control system can maintain optimal performance under various load conditions. Effective energy efficiency management greatly reduces operating costs, while improving the environmental performance and sustainability of the semiconductor refrigeration control system. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flow chart of an optimization method for a semiconductor refrigeration control system described in the present invention. DETAILED DESCRIPTION

[0036] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0038] The specific scheme of the optimization method of a semiconductor refrigeration control system provided by the present invention is described in detail below with reference to the accompanying drawings.

[0039] See attached Figure 1 , which shows a flow chart of an optimization method for a semiconductor refrigeration control system provided by an embodiment of the present invention, the method comprising the following steps:

[0040] S1. Obtain and standardize the original data points, generate feature vectors based on the standardized data points; use an optimized bidirectional long short-term memory network to process the feature vectors to obtain weighted hidden states; based on the weighted hidden states, use the regression model to fit the output of the regression model;

[0041] Obtain raw data points through the sensor network and integrate the acquired raw data points into a data set , Represents the number of original data points in the data set, Indicates The original data points, , the raw data points include the internal temperature, ambient temperature, voltage, current, system load, operating frequency, humidity, and timestamp of the semiconductor refrigeration control system. The sensor network can be specifically set according to the specific implementation scenario and is not limited here;

[0042] The original data points in the data set are standardized to obtain standardized data points, so that the original data points are processed on a unified scale, the influence of the dimension is eliminated, and the accuracy of the analysis is improved. The specific formula is:

[0043] ,

[0044] in, represents the normalized data points; Indicates original data points; Representation dataset the median; Representation dataset The interquartile range;

[0045] The normalized data points are used to generate time points The eigenvector of , so as to be input into the optimized bidirectional long short-term memory network for further processing. The specific method of generating the feature vector is a technical means well known to those skilled in the art and will not be described in detail here;

[0046] Use an optimized bidirectional long short-term memory network to classify time points The eigenvector of Process and get the time point The optimized bidirectional long short-term memory network enhances the processing capability of the bidirectional long short-term memory network for time sensitivity by introducing a time gating mechanism and an adaptive attention mechanism;

[0047] Use an optimized bidirectional long short-term memory network to classify time points The eigenvector of The specific steps for processing are as follows:

[0048] The first step is to calculate the time gate;

[0049] By introducing a time gating mechanism, the processing capability of the bidirectional long short-term memory network for time sensitivity is enhanced; the time gating mechanism controls the time point by introducing an additional time gating unit into the basic structure of the bidirectional long short-term memory network The eigenvector of and the impact of the hidden state at the previous time point on the current state;

[0050] The specific formula for calculating the time-gated output is:

[0051] ,

[0052] in, Indicates at a point in time Time-gated output for the flow of feature vectors and control of time points The eigenvector of and the hidden state at the previous time point Impact on current status; express An activation function compresses the time-gated output to between 0 and 1 to facilitate nonlinear transformation; Represents the weight matrix of time gating, used for time points The eigenvector of ; Represents the time-gated recurrent weight matrix for the hidden state at the previous time point , It can be a hidden state initialized to zero; represents the bias term of time gating;

[0053] The second step is to apply temporal gating to the bidirectional LSTM network;

[0054] For the forward LSTM network:

[0055] ,

[0056] in, Represents the forward long short-term memory network at time point The hidden state of Represents element-by-element multiplication; represents the long short-term memory network; Indicates the positive long short-term memory network at the previous time point The hidden state of

[0057] For the reverse LSTM network:

[0058] ,

[0059] in, Represents the reverse long short-term memory network at time point The hidden state of Represents element-by-element multiplication; represents the long short-term memory network; Represents the reverse long short-term memory network at the next time point The hidden state of

[0060] The third step is to merge the forward and reverse hidden states;

[0061] The hidden states of the forward and reverse LSTM networks are merged to get the time point The hidden state of is:

[0062] ,

[0063] in, Represents the time point obtained by merging the hidden states of the forward and reverse LSTM networks The hidden state of

[0064] The fourth step is to introduce the adaptive attention mechanism;

[0065] An adaptive attention mechanism is introduced to calculate the attention weight. The adaptive attention mechanism is a technology that enhances the ability of the bidirectional long short-term memory network to recognize the importance of a specific time point. The adaptive attention mechanism weights the hidden state by calculating the attention weight of the time point.

[0066] The specific formula for calculating the attention weight is:

[0067] ,

[0068] in, Indicates time point The attention weight of represents the weight matrix of the third attention layer, represents the weight matrix of the second attention layer, Represents the weight matrix of the first attention layer, used for time points The eigenvector of Calculation of weights; Represents the rectified linear unit activation function, which is used to add nonlinear transformation; represents the hyperbolic tangent activation function, which is used for nonlinear transformation; Represents the hidden state weight matrix of the first attention layer, which is used for the hidden state at the previous time point Calculation of weights; represents the bias term of the first attention layer; Represents the bias term of the second attention layer; represents the normalized exponential function, which normalizes the weights to ensure that the sum is 1;

[0069] Step 5: Calculate the weighted hidden state;

[0070] Based on the calculated time point The attention weight , by optimizing the bidirectional long short-term memory network, different time points are given different importance. Weighted hidden state The calculation is as follows:

[0071] ,

[0072] in, Indicates time point The weighted hidden state of Indicates time point The attention weight of Represents the time point obtained by merging the hidden states of the forward and reverse LSTM networks The hidden state of

[0073] Based on time point The weighted hidden state of , using the regression model To fit the output of the regression model The specific regression model is a technical means well known to those skilled in the art and will not be described in detail here.

[0074] S2, obtaining the predicted temperature of the semiconductor refrigeration control system through a refrigeration temperature prediction algorithm, generating a control signal according to the difference between the predicted temperature and the actual temperature of the semiconductor refrigeration control system, mapping the control signal to an actual operating parameter, and regulating the temperature of the semiconductor refrigeration control system by controlling the actual operating parameter;

[0075] The temperature of the semiconductor refrigeration control system is predicted by a refrigeration temperature prediction algorithm to achieve more accurate regulation; the refrigeration temperature prediction algorithm predicts the temperature by combining the output of the regression model and the periodic trend component;

[0076] By simulating the periodic trend component through the improved Fourier series, the periodic fluctuations caused by the equipment operation cycle or environmental factors can be accurately described, and key periodic information can be added to the temperature prediction. The specific formula is:

[0077] ,

[0078] in, Represents the periodic trend component, that is, the semiconductor refrigeration control system at the time point Periodic temperature fluctuations; Represents the index of the number of terms in the Fourier series, indicating the frequency components; Represents the total number of terms in the Fourier series, that is, the number of frequency components; Indicates The amplitude adjustment coefficient of the cosine component in the term is used to adjust the amplitude of the cosine wave; Indicates The amplitude adjustment coefficient of the sine component in the term is used to adjust the amplitude of the sine wave; represents the exponential decay factor, which is used to control the decay rate of the cosine component. is the attenuation constant; represents the exponential enhancement factor, which is used to control the enhancement rate of the sinusoidal component. is the enhancement constant; and Represent the cosine function and sine function, respectively, which are used to describe periodic fluctuations; Represents the angular frequency of the cosine function, which determines the periodic changes of the cosine function and can simulate and capture high-frequency periodic changes in the refrigeration system; Represents the angular frequency of the sine function, which determines the periodic changes of the sine function and is used to simulate and capture low-frequency periodic changes in the refrigeration system; and are the phase shifts of the cosine and sine functions, respectively;

[0079] Combining the output of the regression model and the cyclical trend component To predict the temperature, the specific formula is:

[0080] ,

[0081] in, Indicates the semiconductor refrigeration control system at time point The predicted temperature; represents a logistic function; is a parameter that controls the steepness of the logistic function curve; represents the output of the regression model; Represents the adjustment of the sine function output and the periodic trend component the coefficient of the magnitude of the product; Represents the sine function, which is used to simulate periodic changes; Represents the angular frequency of the sine function, determines the frequency of periodic changes, can accurately simulate the complex periodic temperature changes in the semiconductor refrigeration control system, and improve the accuracy of temperature prediction; Represents the phase of the sine function and adjusts the position of the periodic trend on the time axis;

[0082] According to the semiconductor refrigeration control system at the time point The predicted temperature and semiconductor refrigeration control system at the time point The control signal is calculated based on the difference in actual temperature. The specific formula is:

[0083] ,

[0084] in, Indicates at a point in time The control signal is used to adjust the parameters of the semiconductor refrigeration control system, such as voltage and current, to achieve precise control of the temperature of the semiconductor refrigeration control system; Indicates the temperature difference, which is the temperature difference of the semiconductor refrigeration control system at a certain point in time. Predicted temperature With actual temperature The difference between ; Indicates the proportional coefficient, which adjusts the control signal and the temperature difference The proportional relationship between them;

[0085] Map the control signal to actual operating parameters, such as voltage and current; control the temperature of the semiconductor refrigeration control system by controlling the actual operating parameters. The specific formula is:

[0086] ,

[0087] in, Indicates actual operating parameters; Indicates the minimum practical operating parameter value; Indicates the maximum actual operating parameter value; Represents normalization at time point The control signal , the control signal Adapted to the actual operating parameter range, Indicates at a point in time The minimum value of the control signal, Indicates at a point in time The maximum value of the control signal.

[0088] S3. Based on the predicted temperature of the semiconductor refrigeration control system, the total energy efficiency is calculated using the energy efficiency control algorithm, and the total energy efficiency is optimized through frequency update. Under different working points, the energy efficiency of the semiconductor refrigeration control system is difficult to effectively manage, which will lead to energy waste and reduced efficiency.

[0089] The total energy efficiency is calculated by the energy efficiency control algorithm; the energy efficiency control algorithm is based on the semiconductor refrigeration control system at the time point The predicted temperature is adjusted by introducing the square of the voltage, the exponential decay of the predicted temperature, the normalization of the load, and the logarithmic and sine functions of the frequency, and the total energy efficiency can be accurately calculated.

[0090] The specific calculation formula for total energy efficiency is:

[0091] ,

[0092] in, Indicates total energy efficiency; Indicates voltage, specifically the voltage of the semiconductor refrigeration control system, the square of the voltage It is used to increase the nonlinearity of the formula to reflect the important influence of voltage change on the energy efficiency of semiconductor refrigeration control system; Indicates the semiconductor refrigeration control system at time point The predicted temperature directly affects the working efficiency and energy consumption of the semiconductor refrigeration control system; Represents the standardized temperature parameter, used for standardization , in order to adjust the impact of energy efficiency at different temperatures; represents an exponential decay function, which is used to simulate the effect of temperature on energy efficiency. The higher the temperature, the lower the energy efficiency. Indicates production load, the size of the load directly affects energy efficiency; Represents standardized load parameters, used to standardize production loads , so that the formula can be flexible to deal with different load conditions; Indicates frequency, specifically the frequency at which the semiconductor refrigeration control system operates. Changes in frequency will affect the performance and energy efficiency of the semiconductor refrigeration control system. represents a logarithmic function, which is used to increase the nonlinear effect of frequency on energy efficiency and ensure that the change of energy efficiency remains smooth; Represents a sinusoidal function, so that the response of the formula to frequency varies periodically, reflecting the performance differences at different frequency operating points; Represents the maximum frequency parameter, used to normalize the frequency , ensuring that the sine function varies within the appropriate frequency range;

[0093] In order to further optimize the overall energy efficiency, the overall energy efficiency is adjusted by frequency update. The specific formula for frequency update is:

[0094] ,

[0095] in, Indicates the new frequency value; Indicates frequency, specifically the frequency at which the semiconductor refrigeration control system operates; represents the learning rate, which is a factor that controls the frequency adjustment pace; Total energy efficiency Frequency The first derivative of is the main directional indicator of the optimized frequency; represents the momentum factor, which is a factor that helps overcome local extreme values ​​and accelerates convergence, making the adjustment process more stable and smooth; represents the momentum term, initially is 0; represents the second derivative weighting factor, which controls the influence of the second derivative on the frequency adjustment; Energy efficiency Frequency The second derivative of

[0096] Based on the new frequency value, the energy efficiency control algorithm is applied again to calculate the new total energy efficiency , set a threshold for energy efficiency change , threshold It can be set according to the specific implementation scenario and is not limited here. When the energy efficiency change calculated after the frequency adjustment is less than the threshold, that is, , stop adjusting to ensure that the energy efficiency reaches the optimal state, otherwise continue to adjust until the energy efficiency change is less than the threshold;

[0097] Through the above process, dynamic adjustment and optimization of the energy efficiency of the semiconductor refrigeration control system at different working points are realized. Through the energy efficiency control algorithm and frequency update mechanism, the energy efficiency management capability of the semiconductor refrigeration control system is improved, energy waste is reduced, and the efficiency of the semiconductor refrigeration control system is improved.

[0098] In summary, an optimization method for a semiconductor refrigeration control system is completed.

[0099] The order of the embodiments of the invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0100] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0101] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A method for optimizing a semiconductor refrigeration control system, characterized in that: The following steps are involved: S1. Obtain and standardize the original data points, generate feature vectors based on the standardized data points; use an optimized bidirectional long short-term memory network to process the feature vectors to obtain weighted hidden states; based on the weighted hidden states, use the regression model to fit the output of the regression model; The specific implementation steps of the optimized bidirectional long short-term memory network are as follows: The first step is to introduce the time gating mechanism and calculate the time gating; The second step is to apply time gating to the bidirectional LSTM network to obtain the hidden states of the forward and reverse LSTM networks. The third step is to merge the forward and reverse hidden states; The fourth step is to introduce the adaptive attention mechanism and calculate the attention weight; The fifth step is to calculate the weighted hidden state to obtain the weighted hidden state; S2, obtaining the predicted temperature of the semiconductor refrigeration control system through a refrigeration temperature prediction algorithm, generating a control signal according to the difference between the predicted temperature and the actual temperature of the semiconductor refrigeration control system, mapping the control signal to an actual operating parameter, and regulating the temperature of the semiconductor refrigeration control system by controlling the actual operating parameter; The refrigeration temperature prediction algorithm simulates the periodic trend component through an improved Fourier series, and the specific formula of the improved Fourier series is: , in, Indicates at a point in time The cyclical trend component of Represents the index of the number of terms in the Fourier series; represents the total number of terms in the Fourier series; Indicates The amplitude adjustment coefficient of the cosine component in the term; Indicates The amplitude adjustment coefficient of the sinusoidal component in the term; represents the exponential decay factor; is the attenuation constant; represents the exponential enhancement factor; is the enhancement constant; represents the angular frequency of the cosine function; represents the angular frequency of the sine function; and are the phase shifts of the cosine and sine functions, respectively; Combining the output of the regression model with the periodic trend component, the predicted temperature is obtained as follows: , in, Indicates the semiconductor refrigeration control system at time point The predicted temperature; represents a logistic function; is a parameter that controls the steepness of the logistic function curve; represents the output of the regression model; Represents the adjustment of the sine function output and the periodic trend component the coefficient of the magnitude of the product; represents the angular frequency of the sine function; Represents the phase of the sine function; S3. Based on the predicted temperature of the semiconductor refrigeration control system, the total energy efficiency is calculated using the energy efficiency control algorithm. The specific calculation formula for the total energy efficiency is: , in, Indicates total energy efficiency; Indicates the voltage of semiconductor refrigeration control system; Indicates the semiconductor refrigeration control system at time point The predicted temperature; represents the standardized temperature parameter; represents the production load; represents the standardized load parameter; Indicates the operating frequency of the semiconductor refrigeration control system; Indicates the maximum frequency parameter; The total energy efficiency is adjusted by frequency update to obtain a new frequency value; the specific formula for frequency update is: , in, Indicates the new frequency value; represents the learning rate; represents the momentum factor; represents the momentum term; represents the second derivative weighting factor; Total energy efficiency Frequency The first derivative of ; Total energy efficiency Frequency The second derivative of Based on the new frequency value, the energy efficiency control algorithm is applied again to calculate the new total energy efficiency.

2. The optimization method of a semiconductor refrigeration control system according to claim 1, characterized in that: The S1 specifically includes: The time gating mechanism controls the influence of the feature vector at the current time point and the hidden state at the previous time point on the current state by introducing a time gating unit into the structure of the bidirectional long short-term memory network.

3. The optimization method of a semiconductor refrigeration control system according to claim 1, characterized in that: The S3 specifically includes: The energy efficiency control algorithm is based on the predicted temperature of the semiconductor refrigeration control system. The total energy efficiency is calculated by introducing the square of the voltage, the exponential decay of the predicted temperature, the normalization of the load, the logarithm of the frequency and the sine function.