Centrifugal machine temperature control parameter optimization method and system

Abstract

The invention relates to the technical field of centrifugal machine control, in particular to a centrifugal machine temperature control parameter optimization method and system, and the method comprises the steps: collecting a plurality of parameters of a centrifugal machine, and constructing a sample data set; an XGBoost stage classification model is trained based on the sample data set, the working condition probability of the centrifugal machine is output, the chamber temperature and the refrigeration power of the centrifugal machine at the next moment are preset, and a temperature precision penalty coefficient, a temperature change rate risk coefficient and an energy consumption penalty coefficient are calculated; the weights of the temperature precision penalty coefficient, the temperature change rate risk coefficient and the energy consumption penalty coefficient are dynamically adjusted based on the proportions of the steady speed, the accelerated speed and the decelerated speed in the real-time working condition, an objective function is constructed, and a comprehensive control error is calculated; and multiple groups of historical data most similar to the working condition at the current moment are selected as reference data, all the reference data are traversed, the compressor rotating speed corresponding to the reference data with the minimum comprehensive control error is selected as the optimal rotating speed, the optimal rotating speed is issued to the compressor for execution, and the temperature control precision of the centrifugal machine is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of centrifuge control. In particular, it relates to a centrifuge temperature control parameter optimization method and system. BACKGROUND

[0002] A centrifuge is a key device for separating samples in the fields of biochemistry, pharmacy, medical examination, etc. Its working principle is to drive a rotor to rotate at high speed by a motor to generate centrifugal force and realize sample separation. In this process, a large amount of heat is generated by the motor coil dissipation of Joule heat and the rotor and air friction, resulting in a sharp change in the internal temperature of the centrifuge cavity. Many biological samples are highly sensitive to temperature, and the centrifugation process must be carried out in a constant low-temperature environment. Temperature overshoot or subcooling will cause the sample to denature and inactivate, seriously affecting the reliability and repeatability of experimental results.

[0003] At present, the temperature control of a centrifuge widely adopts a traditional PID control algorithm. However, this method has inherent defects in the variable working condition operation of the centrifuge, including: a fixed set of PID parameters cannot simultaneously meet the different requirements of the controller in different operating stages such as acceleration, steady speed and deceleration. In the acceleration stage, the thermal disturbance is severe, and the controller needs to respond quickly to prevent overshoot. In the steady speed stage, the control needs to be smooth and low in energy consumption. In the deceleration stage, the heat generation is reduced, and it is necessary to prevent subcooling. The PID controller is difficult to achieve the best performance in all stages. There is no priority switching, and only the minimum temperature deviation is used as a single target, which easily leads to problems such as high energy consumption in the steady speed stage and subcooling in the deceleration stage. SUMMARY

[0004] To solve the problem that the PID control algorithm cannot meet the requirements of different working conditions when the existing centrifuge is operated in different working conditions, resulting in temperature subcooling or overshoot of the centrifuge and affecting the reliability of experimental results, the present application provides solutions in the following aspects.

[0005] In a first aspect, a centrifuge temperature control parameter optimization method comprises: training an XGBoost stage classification model based on a sample data set, selecting a target XGBoost model based on recall rate, and outputting probabilities of the centrifuge being in three working conditions of acceleration, steady speed, and deceleration based on the target XGBoost model; presetting a chamber temperature and a compressor refrigeration power of the centrifuge at a next time, calculating a temperature accuracy penalty coefficient based on a deviation of the chamber temperature at the next time and a preset temperature, calculating a temperature change rate risk coefficient based on the chamber temperature at the next time and a current time, calculating an energy consumption penalty coefficient based on the refrigeration power at the next time and a maximum value of the refrigeration power in historical data, dynamically adjusting weights of the temperature accuracy penalty coefficient, the temperature change rate risk coefficient, and the energy consumption penalty coefficient based on proportions of the steady speed, the acceleration, and the deceleration in a real-time working condition, and constructing a target function, and calculating a comprehensive control error of the centrifuge at the next time based on the target function; selecting a plurality of groups of historical data most similar to a current working condition as reference data, selecting a compressor speed corresponding to reference data making the comprehensive control error minimum as an optimal speed in traversing each reference data, and delivering the optimal speed to the compressor for execution.

[0006] By collecting a plurality of parameters of the centrifuge, the running state of the centrifuge can be comprehensively reflected, the XGBoost model is used to realize real-time and accurate probability identification of the three working conditions of acceleration, steady speed, and deceleration of the centrifuge, the weights of temperature accuracy, energy consumption optimization, and temperature change suppression are dynamically adjusted based on proportions of the steady speed, the acceleration, and the deceleration in a real-time working condition, for example, more attention is paid to temperature accuracy and energy consumption in the steady speed stage, and more attention is paid to suppression of temperature change in the acceleration / deceleration stage, thereby realizing adaptive optimization; by constructing a comprehensive target function including temperature accuracy penalty, temperature change rate risk, and energy consumption penalty, multiple control targets that may conflict with each other, such as rapid cooling, small temperature fluctuation, and energy saving, are converted into a single quantifiable optimization target, the weights are used to dynamically balance each sub-target, thereby realizing multi-objective collaborative optimization rather than single-objective optimization, by presetting the chamber temperature and the refrigeration power at the next time and calculating corresponding penalty coefficients, temperature overshoot, severe fluctuation, or energy consumption surge can be suppressed in advance rather than being passively responsive to errors, thereby realizing intelligentization, high precision, energy saving, and adaptivity of centrifuge temperature control, and significantly improving sample safety, equipment life, and operation economy.

[0007] Preferably, the proportion of the steady speed in the real-time working condition is calculated by multiplying the steady speed probability in the real-time working condition by 2 and adding the acceleration probability and the deceleration probability as a denominator, and taking the steady speed probability as a numerator, to obtain the temperature accuracy weight and the energy consumption optimization weight, and the temperature accuracy weight and the energy consumption optimization weight are used as weights of the temperature accuracy penalty coefficient and the energy consumption penalty coefficient.

[0008] Preferably, the probability of steady speed in the real-time working condition of the centrifuge is multiplied by 2, and then added to the probability of speed-up and the probability of speed-down to serve as the denominator, and the sum of the probability of speed-up and the probability of speed-down serves as the numerator to calculate the proportion of the sum of the probability of speed-up and the probability of speed-down in the real-time working condition, so as to obtain the temperature change inhibition weight and serve as the weight of the temperature change rate risk coefficient.

[0009] By multiplying the probability of steady speed by 2, the proportion of the probability of steady speed in the denominator is improved, and thus the weight of temperature precision is increased, so that the temperature control precision is greatly improved, and the safety of sample centrifugation operation is ensured. In the speed-up / speed-down stage, by strengthening the temperature change inhibition weight, the temperature change rate is significantly reduced, and sample damage caused by sudden temperature change is avoided. In the steady speed or low load working condition, by increasing the energy consumption optimization weight, the compressor speed is actively reduced to the minimum requirement to meet the temperature precision, and energy waste caused by traditional "overcooling" or "excessive refrigeration" is avoided. Long-term operation can significantly reduce the energy consumption of the centrifuge.

[0010] Preferably, the calculation method of the temperature precision penalty coefficient is as follows: the square of the difference between the temperature at the next time point in the centrifugal cavity and the set temperature is calculated as the temperature deviation degree, and the ratio of the temperature deviation degree to the maximum value of the temperature deviation degree in the historical data is calculated as the temperature precision penalty coefficient.

[0011] Preferably, the calculation method of the temperature change rate risk coefficient is as follows: the change rate of the temperature at the next time point in the centrifugal cavity compared with the current temperature in one sampling period is calculated as the temperature change rate, and the ratio of the temperature change rate to the maximum value of the temperature change rate in the historical data is calculated as the temperature change rate risk coefficient.

[0012] Preferably, the calculation method of the energy consumption penalty coefficient is as follows: the ratio of the refrigeration power of the compressor at the next time point to the maximum value of the refrigeration power in the historical data is calculated as the energy consumption penalty coefficient.

[0013] Preferably, the reference data is selected in the following manner: based on the probability of each working condition at the current time point, the working condition with the maximum probability is selected as the dominant working condition, the data in all historical data that is the same as the current dominant working condition is selected, the Euclidean distance between each selected historical data and the current time data is calculated, and multiple historical data is selected as the reference data based on the Euclidean distance.

[0014] The historical data is first divided into three parts of "dominant working condition" of speed-up / steady / speed-down to avoid cross-stage interference, and only multiple samples that are most similar to the current time point are left by using the Euclidean distance to exclude extreme or abnormal working conditions. Since the reference data is highly similar to the current working condition, the mechanism simulation interpolation error is smaller, the accuracy of the comprehensive control error score is effectively improved, and the optimal speed prediction accuracy is significantly improved.

[0015] In a second aspect, a centrifuge temperature control parameter optimization system comprises a processor and a memory, the memory storing computer program instructions which, when executed by the processor, implement any of the centrifuge temperature control parameter optimization methods.

[0016] The present application has the following effects: 1. By collecting multiple parameters of the centrifuge, the running state of the centrifuge can be comprehensively reflected. The XGBoost model is used to perform real-time and accurate probability identification on the three working conditions of the centrifuge, namely, the acceleration, the steady speed and the deceleration. The weights of the temperature precision, the energy consumption optimization and the temperature change suppression are dynamically adjusted according to the proportion of the steady speed, the acceleration and the deceleration in the real-time working condition. For example, more attention is paid to the temperature precision and the energy consumption in the steady speed stage, and more attention is paid to the suppression of temperature change in the acceleration / deceleration stage, so that adaptive optimization is realized. By constructing a comprehensive objective function including temperature precision penalty, temperature change rate risk and energy consumption penalty, multiple control objectives that may conflict with each other, such as rapid cooling, small temperature fluctuation and energy saving, are converted into a single quantifiable optimization objective. The weights are dynamically balanced to realize multi-objective collaborative optimization rather than single-objective optimization. By presetting the chamber temperature and the refrigeration power at the next moment and calculating the corresponding penalty coefficient, temperature overshoot, severe fluctuation or energy consumption surge can be suppressed in advance rather than being passively responsive to errors. The intelligentization, high precision, energy saving and adaptability of the centrifuge temperature control are realized, and the sample safety, equipment life and operation economy are significantly improved.

[0017] 2. In the acceleration / deceleration stage, the temperature change rate is significantly reduced by strengthening the temperature change suppression weight, and sample damage caused by sudden temperature change is avoided. In the steady speed or low load working condition, the compressor speed is actively reduced to the minimum requirement to meet the temperature precision by increasing the energy consumption optimization weight, and energy waste caused by traditional "overcooling" or "excessive refrigeration" is avoided. Long-term operation can significantly reduce the energy consumption of the centrifuge. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 is a method flowchart of steps S1-S4 in the centrifuge temperature control parameter optimization method of the embodiment of the present application.

[0019] Figure 2 is a structure flowchart of the centrifuge temperature control parameter optimization system of the embodiment of the present application. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments of the present application.

[0021] The specific embodiments of the present application will be described in detail below with reference to the drawings.

[0022] Referring to Figure 1 A centrifuge temperature control parameter optimization method includes steps S1-S4, as follows: S1: Collecting multiple parameters of the centrifuge, constructing sample data containing each parameter of the centrifuge based on time sequence characteristics, and constructing a sample data set.

[0023] The rotor speed, cavity temperature, compressor speed and compressor refrigeration power of the centrifuge are synchronously collected by high-precision online sensors and collection cards. The parameters to be collected also include the preset temperature of the cavity, which is set artificially. The collection frequency of each parameter can be once per second, and 1000 complete operation cycles of 50 centrifuges are accumulated.

[0024] The collected data is preprocessed, and abnormal values are removed using the 3σ criterion. When the continuous missing is less than 5 sampling points, linear interpolation is used to complete the missing. When the continuous missing is greater than 5 points, it is marked as an invalid cycle. According to the change rate of the compressor speed, it is marked as three working conditions of speed-up, steady speed and speed-down.

[0025] Based on the time sequence characteristics, the rotor speed, cavity temperature, set temperature, compressor speed and compressor refrigeration power of the centrifuge are constructed as sample data of the centrifuge, and a sample data set is constructed based on each sample data.

[0026] S2: Training an XGBoost phase classification model based on the sample data set, selecting a target XGBoost model based on the recall rate, and outputting the probability of the centrifuge being in the three working conditions of speed-up, steady speed and speed-down based on the target XGBoost model.

[0027] Based on the preprocessed sample data set, a sliding window method is used for feature construction. The step length of the sliding window can be 10s. Based on the five-dimensional feature vector of the rotor speed, cavity temperature, set temperature, compressor speed and compressor refrigeration power, the five-dimensional feature vector specifically includes the average speed of the centrifuge rotor for 10s, the instantaneous change rate of the centrifuge rotor speed, the deviation degree of the actual temperature and the set temperature, the temperature change trend, and the average refrigeration power of the compressor for 10s.

[0028] For each sampling time, a unified five-dimensional feature vector can be generated according to the above definition. The feature vector is strictly aligned with the running phase label corresponding to the time, ensuring the complete consistency of the feature and the label in the time dimension, and laying a reliable data foundation for the accurate training and online identification of the subsequent model. In other embodiments, more statistical features (variance, slope, spectral features, etc.) can be included according to actual needs to further improve the adaptability to complex working conditions.

[0029] The feature vector data set is divided into training set and test set in time sequence with a ratio of 8:2, that is, the first 80% is selected for training and the last 20% is selected for testing along the time axis. The optimal hyperparameter combination of XGBoost (eXtreme Gradient Boosting) is determined by grid search and five-fold cross-validation. During the model training and parameter optimization process, the recall rate of each operating condition (acceleration, steady speed, deceleration) is taken as the core evaluation index, and the model with the best performance in acceleration, steady speed and deceleration recall rate is preferentially selected as the target XGBoost model.

[0030] Real-time acquisition of centrifuge parameters, based on the target XGBoost model to output the probability of the centrifuge at the current time in each operating condition.

[0031] S3: The chamber temperature of the centrifuge at the next time and the refrigeration power of the compressor are preset, the temperature accuracy penalty coefficient is calculated based on the deviation of the chamber temperature at the next time and the preset temperature, the temperature change rate risk coefficient is calculated based on the chamber temperature at the next time and the current time, the energy consumption penalty coefficient is calculated based on the refrigeration power at the next time and the maximum value of the refrigeration power in the historical data, the weights of the temperature accuracy penalty coefficient, the temperature change rate risk coefficient and the energy consumption penalty coefficient are dynamically adjusted based on the proportion of steady speed, acceleration and deceleration in the real-time operating condition, and the target function is constructed, the comprehensive control error of the centrifuge at the next time is calculated based on the target function.

[0032] The proportion of steady speed in the real-time operating condition is calculated by multiplying the steady speed probability by 2 and adding the acceleration probability and the deceleration probability as the denominator, and taking the steady speed probability as the numerator, to obtain the temperature accuracy weight and the energy consumption optimization weight, and as the weights of the temperature accuracy penalty coefficient and the energy consumption penalty coefficient. The proportion of acceleration and deceleration in the real-time operating condition is calculated by multiplying the steady speed probability by 2 and adding the acceleration probability and the deceleration probability as the denominator, and taking the sum of the acceleration probability and the deceleration probability as the numerator, to obtain the temperature change inhibition weight and as the weight of the temperature change rate risk coefficient.

[0033] The square of the difference between the temperature at the next time in the centrifugal cavity and the set temperature is calculated as the temperature deviation degree, and the ratio of the temperature deviation degree to the maximum value of the temperature deviation degree in the historical data is calculated and normalized as the temperature accuracy penalty coefficient.

[0034] The temperature change rate of the temperature at the next time in the centrifugal cavity compared with the current time in a sampling period is calculated as the temperature change rate, and the ratio of the temperature change rate to the maximum value of the temperature change rate in the historical data is calculated and normalized as the temperature change rate risk coefficient.

[0035] The ratio of the refrigeration power of the compressor at the next time to the maximum value of the refrigeration power in the historical data is calculated and normalized as the energy consumption penalty coefficient.

[0036] The temperature precision weight is taken as the weight of the temperature precision penalty coefficient, the temperature change inhibition weight is taken as the weight of the temperature change rate risk coefficient, and the energy consumption optimization weight is taken as the weight of the energy consumption penalty coefficient, to construct a target function and calculate the comprehensive control error of the centrifuge at the next moment.

[0037] The formula of the target function is: Among them, represents the comprehensive control error of the centrifuge, The smaller the value of is, the smaller the comprehensive control error, that is, the higher the temperature precision, the lower the risk of overshoot or overcooling, and the more optimal the energy consumption, and the purpose of constructing the target function is to select the compressor speed corresponding to the minimum value of the comprehensive control error at the next moment, represents the temperature precision weight, represents the temperature precision penalty coefficient, represents the temperature change inhibition weight, represents the temperature change rate risk coefficient, represents the energy consumption optimization weight, represents the energy consumption penalty coefficient, , , , The value of is output in real time based on the target XGBoost model.

[0038] S4: Select a plurality of groups of historical data most similar to the current working condition as reference data, traverse each reference data, select the compressor speed corresponding to the reference data that minimizes the comprehensive control error as the optimal speed, and issue it to the compressor for execution.

[0039] When selecting similar historical data, the working condition with the highest probability is taken as the dominant working condition based on the probability of each working condition at the current moment, the data in all historical data that is the same as the dominant working condition at the current moment is selected, and the Euclidean distance between each selected historical data and the data at the current moment is calculated. The smaller the Euclidean distance, the higher the similarity. Based on the Euclidean distance from small to large, the first 20% of historical moments are selected as reference data. If the number of the first 20% moments is less than 50, the selection ratio is gradually increased until the number of reference data is greater than 50. Ensure that the number of reference data is greater than 50 to ensure statistical effectiveness. The 50 historical data selected are most similar to the current working condition. In other embodiments, the number of selected reference data can be flexibly adjusted as needed.

[0040] ​The compressor speed data of the next moment corresponding to each reference data is acquired as a candidate speed, the 95% confidence interval of the candidate speed is calculated by using a normal distribution method, and the effective candidate speed of the current dominant working condition is obtained. The range can cover more than 92% of the effective working conditions in the historical data. Each effective working condition is substituted into the centrifuge and refrigeration system mechanism simulation model one by one, and the cavity temperature and compressor refrigeration power data of the next moment corresponding to each reference data are output. Based on the cavity temperature, the temperature precision penalty coefficient and the temperature change rate risk coefficient are calculated, based on the compressor refrigeration power, the energy consumption penalty coefficient is calculated, based on the real-time data, the working condition probability is output through the target XGBoost model, and the temperature precision weight, the temperature change suppression weight and the energy consumption optimization weight are calculated. Each reference data corresponding to the comprehensive control error is calculated by traversing each reference data through the target function.

[0041] The system comprises a processor and a memory, and the memory stores computer program instructions which, when executed by the processor, implement the centrifuge temperature control parameter optimization method according to the first aspect of the application.

[0042] The system further comprises a communication bus and a communication interface and other components familiar to those skilled in the art, the settings and functions of which are known in the art, and thus will not be described here.

[0043] The centrifuge temperature control parameter optimization method and system provided by the application realize the intelligentization, high precision, energy saving and self-adaptation of the centrifuge temperature control, and significantly improve the sample safety, equipment life and operation economy.

[0044] It should be noted that, for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of the present application. Therefore, the protection scope of the present application patent should be subject to the appended claims.

Claims

1. A method of optimizing a centrifuge temperature control parameter, the method comprising: The method comprises the following steps: Collecting a plurality of parameters of the centrifuge, constructing sample data containing each parameter of the centrifuge based on time sequence characteristics, and constructing a sample data set; Training an XGBoost stage classification model based on the sample data set, selecting a target XGBoost model based on recall rate, and outputting probabilities of the centrifuge being in three working conditions of acceleration, constant speed and deceleration based on the target XGBoost model; Pre-setting the chamber temperature of the centrifuge at the next moment and the refrigeration power of the compressor, calculating a temperature accuracy penalty coefficient based on the deviation of the chamber temperature at the next moment and the preset temperature, calculating a temperature change rate risk coefficient based on the chamber temperature at the next moment and the current moment, calculating an energy consumption penalty coefficient based on the refrigeration power at the next moment and the maximum value of the refrigeration power in the historical data, dynamically adjusting the weights of the temperature accuracy penalty coefficient, the temperature change rate risk coefficient and the energy consumption penalty coefficient based on the proportions of the constant speed, the acceleration and the deceleration in the real-time working condition, and constructing a target function, calculating the comprehensive control error of the centrifuge at the next moment based on the target function; Selecting a plurality of historical data similar to the current working condition as reference data, selecting the compressor speed corresponding to the reference data with the minimum comprehensive control error as the optimal speed, and issuing the optimal speed to the compressor for execution.

2. The method of claim 1, wherein, The proportion of constant speed in the real-time working condition is calculated by taking the constant speed probability in the real-time working condition multiplied by 2 and added to the acceleration probability and the deceleration probability as the denominator, and taking the constant speed probability as the numerator, to obtain the temperature accuracy weight and the energy consumption optimization weight, which are used as the weights of the temperature accuracy penalty coefficient and the energy consumption penalty coefficient.

3. The method of claim 1, wherein, The proportion of the sum of acceleration and deceleration in the real-time working condition is calculated by taking the constant speed probability in the real-time working condition multiplied by 2 and added to the acceleration probability and the deceleration probability as the denominator, and taking the sum of the acceleration probability and the deceleration probability as the numerator, to obtain the temperature change inhibition weight and use it as the weight of the temperature change rate risk coefficient.

4. The method of claim 1, wherein, The calculation method of the temperature accuracy penalty coefficient is: The square of the difference between the next moment temperature of the centrifugal cavity and the set temperature is calculated as the temperature deviation degree, and the ratio of the temperature deviation degree to the maximum value of the temperature deviation degree in the historical data is calculated as the temperature accuracy penalty coefficient.

5. The method of claim 1, wherein, The calculation method of the temperature change rate risk coefficient is: The change rate of the next moment temperature of the centrifugal cavity compared with the current moment temperature in a sampling period is calculated as the temperature change rate, and the ratio of the temperature change rate to the maximum value of the temperature change rate in the historical data is calculated as the temperature change rate risk coefficient.

6. The method of claim 1, wherein, The calculation method of the energy consumption penalty coefficient is: The ratio of the next moment refrigeration power of the compressor to the maximum value of the refrigeration power in the historical data is calculated as the energy consumption penalty coefficient.

7. The method of claim 1, wherein, The selection method of the reference data is: Based on the probabilities of each working condition at the current moment, the working condition with the maximum probability is selected as the dominant working condition, the data in all historical data which is the same as the current dominant working condition is selected, the Euclidean distance between each selected historical data and the current moment data is calculated, and a plurality of historical data is selected as reference data based on the Euclidean distance.

8. A centrifuge temperature control parameter optimization system, comprising: The method comprises the following steps: A processor and a memory, the memory stores computer program instructions, when the computer program instructions are executed by the processor, the centrifuge temperature control parameter optimization method according to any one of claims 1-7 is realized.

Images