Constant flow and constant pressure control method and system for magnetic suspension centrifugal compressor

Abstract

The invention belongs to the technical field of magnetic suspension centrifugal compressor control, and particularly relates to a magnetic suspension centrifugal compressor constant flow and constant pressure control method and system.The method comprises the steps that multi-parameter data of a compressor are obtained, and a working condition analysis model is constructed through the multi-parameter data and historical operation data; training and updating the constructed working condition analysis model; adjusting a working condition demand weight according to a judgment result of the working condition analysis model, introducing a prediction control algorithm according to historical data and working condition characteristics, and performing loss compensation through a dynamic response model established by predicting a current change trend; dynamic adjustment and optimization are carried out based on the real-time working condition and the change rate of pressure and flow; and accurately controlling the output power of the compressor through a dynamic optimization result. Dynamic adjustment and optimization are carried out based on the real-time working condition and the change rate of the pressure and the flow, the precision of pressure and flow control is further improved, and adaptability is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of magnetic suspension centrifugal compressor control, and in particular relates to a constant flow and constant pressure control method and system for a magnetic suspension centrifugal compressor. Background Art

[0002] Magnetic levitation centrifugal compressors are increasingly used in many industrial fields due to their significant advantages such as high efficiency, energy saving, low carbon and environmental protection, and intelligent design. With the diversification and refinement of industrial production, the application processes of different users have increasingly complex requirements for the operation of compressors. In actual working conditions, although the fixed speed mode can meet basic needs in most cases, in certain specific process scenarios, such as precise aeration, combustion support, cooling, textile air shuttle air supply and gas transportation in chemical production, there are strict and dynamically changing requirements for gas flow and pressure.

[0003] The constant pressure and constant flow operation mode of the existing magnetic levitation centrifugal compressor mainly adjusts the speed (PID adjustment) through PLC to adapt to the changes in external pressure and flow. However, this traditional control method gradually exposes many problems when facing complex and changeable working conditions. On the one hand, when the working conditions change frequently or there are instantaneous large fluctuations, its response speed is relatively lagging, and it is difficult to quickly and accurately adjust the pressure and flow to the target value, which easily leads to an increase in unstable factors in the production process. For example, in the chemical reaction process, a small deviation in flow or pressure may affect the reaction rate and product quality. On the other hand, the traditional control method is not smooth enough during the mode switching process, which is easy to cause system fluctuations. At the same time, due to the lack of comprehensive consideration of various working condition parameters, over-adjustment or under-adjustment may occur, which not only affects production efficiency, but also increases equipment wear and energy consumption, and shortens the service life of the equipment. In addition, the existing control scheme has poor adaptability to normal pressure and flow fluctuations, and it is easy to cause excessive temperature rise of the inverter and the motor due to frequent speed adjustment, which further affects the stability and reliability of the equipment. At this stage, a constant flow and constant pressure control method and system for magnetic levitation centrifugal compressors are needed. Summary of the invention

[0004] In order to solve the problems of insufficient control accuracy and poor adaptability to working conditions of traditional compressors, the present invention provides a constant flow and constant pressure control method and system for a magnetically suspended centrifugal compressor.

[0005] In the first aspect, the present invention provides a method for controlling a constant flow and constant pressure of a magnetically suspended centrifugal compressor, which adopts the following technical solution: A constant flow and constant pressure control method for a magnetic suspension centrifugal compressor, comprising: Obtain multi-parameter data of the compressor, including compressor gas pressure, temperature, humidity and gas composition data; Use multi-parameter data and historical operation data to build a working condition analysis model, and train and update the built working condition analysis model; Adjust the weight of the working condition demand according to the judgment results of the working condition analysis model, including adjusting the weight of pressure and flow using an exponential decay curve; Introducing a predictive control algorithm based on historical data and operating characteristics, and compensating for losses by predicting the current trend of change and establishing a dynamic response model; Dynamic adjustment and optimization based on real-time working conditions and the rate of change of pressure and flow; The output power of the compressor is precisely controlled through dynamic optimization results.

[0006] Furthermore, the use of multi-parameter data and historical operating data to construct an operating condition analysis model includes cleaning the acquired multi-parameter data to remove abnormal data, and using the cleaned data as input to construct an operating condition analysis model with an input layer, a hidden layer and an output layer as a framework, wherein the number of nodes in the input layer is the same as the number of parameters, and the output layer has two nodes, which are respectively used to represent the weights of the current operating condition for the pressure control demand and the flow control demand, and the nodes of the hidden layer are calculated according to the number of nodes in the input layer and the output layer.

[0007] Furthermore, the constructed operating condition analysis model is trained and updated, including inputting historical operating data as a training set into the operating condition analysis model, using a mean square error function to calculate the error between the output pressure and flow control weights and the actual operating condition demand weights, and then back-propagating the error from the output layer to the input layer through a back-propagation algorithm, and adjusting the connection weights and bias items of each layer in turn.

[0008] Furthermore, the training and updating of the constructed working condition analysis model also includes using an incremental learning method to add new data generated by the compressor during operation to the training set, calculating the mean square error function after adding the new data, and updating the weights and bias terms based on the gradient descent algorithm and using the mean square error function. The calculation formula of the mean square error function is: , in, is the actual working condition demand weight of the kth sample in the original training set, is the model’s prediction weight for the kth sample in the original training set, and They are the actual working condition demand weight and model prediction weight of the new data respectively.

[0009] Furthermore, the predictive control algorithm is introduced according to the historical data and operating conditions, including fitting the transfer function model of pressure and flow using the least squares method, obtaining the dynamic response model based on the current fan speed, inverter output frequency and the established transfer function model, and calculating the expected change of pressure and flow during the switching process according to the dynamic response model, setting the allowable pressure fluctuation range and the allowable flow fluctuation range, and comparing them with the expected change.

[0010] Furthermore, the dynamic response model established by predicting the current change trend performs loss compensation, including loss compensation when the expected change is greater than the allowable fluctuation range, and determines the expected pressure compensation value by reversely solving the pressure dynamic response model, and calculates the required output frequency change based on the pressure compensation value in combination with the pressure-frequency relationship model and the dynamic response model.

[0011] Furthermore, the dynamic adjustment optimization based on real-time working conditions and the rate of change of pressure and flow includes adding temperature, humidity and gas composition data on the basis of the dynamic response model, redefining the difference between pressure and flow through a compensation algorithm, and performing secondary adjustment on the output frequency change according to the difference between pressure and flow.

[0012] In a second aspect, a constant flow and constant pressure control system for a magnetically suspended centrifugal compressor comprises: The data acquisition module is configured to: acquire multi-parameter data of the compressor, including acquiring compressor gas pressure, temperature, humidity and gas composition data; The model module is configured to: construct a working condition analysis model using multi-parameter data and historical operation data, and train and update the constructed working condition analysis model; The conversion module is configured to: adjust the weight of the working condition demand according to the judgment result of the working condition analysis model, including adjusting the weight of the pressure and the flow rate by using an exponential decay curve; The compensation module is configured to: introduce a predictive control algorithm based on historical data and operating condition characteristics, and perform loss compensation by predicting a dynamic response model established by the current change trend; The optimization module is configured to: perform dynamic adjustment optimization based on real-time working conditions and the change rates of pressure and flow; The execution module is configured to: accurately control the output power of the compressor through dynamic optimization results.

[0013] In a third aspect, the present invention provides a computer-readable storage medium storing a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device, for example, a method for controlling a constant flow and constant pressure of a magnetically levitated centrifugal compressor.

[0014] In a fourth aspect, the present invention provides a terminal device comprising a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, wherein the instructions are suitable for being loaded and executed by the processor as a method for controlling a constant flow and constant pressure of a magnetically levitation centrifugal compressor.

[0015] In summary, the present invention has the following beneficial technical effects: 1. The present invention acquires multi-parameter data including gas pressure, temperature, humidity and gas composition, and builds a working condition analysis model in combination with historical operation data, which can more comprehensively and accurately reflect the actual working condition of the compressor. The input layer, hidden layer and output layer architecture of the model are reasonably designed, taking into account the number of parameters and output requirements, making the judgment of the weight of pressure control requirements and flow control requirements more accurate, which helps to achieve fine distinction and targeted control of different working conditions, and improves the adaptability of the entire system to complex working conditions.

[0016] 2. The present invention performs dynamic adjustment and optimization based on real-time working conditions and the rate of change of pressure and flow. Whether it is a constant pressure mode or a constant flow mode, various environmental factors and working condition changes are taken into account. The pressure and flow differences are redefined by a compensation algorithm, and secondary adjustment is performed. It can effectively reduce errors caused by factors such as temperature, humidity and gas composition changes, further improve the accuracy of pressure and flow control, and increase adaptability.

[0017] 3. The present invention utilizes normal distribution for data cleaning, which can remove outliers in multi-parameter data, improve data quality, and provide more reliable input for the construction of the working condition analysis model. The mean square error function and back propagation algorithm are used to train the model, which can accurately measure the error between the model prediction weight and the actual working condition demand weight, and effectively adjust the connection weights and bias items of each layer to continuously optimize the model, thereby improving the accuracy of the model in working condition analysis.

[0018] 4. The present invention adopts an incremental learning method to add newly generated data to the training set, and updates the weights and bias terms based on the new mean square error function and gradient descent algorithm, so that the model can timely learn the new operating conditions of the compressor during operation, avoiding the resource consumption of frequent retraining of the model. At the same time, it ensures that the model can always adapt to the dynamic changes of the operating conditions and maintain the ability to accurately judge different operating conditions, effectively improving the timeliness and accuracy of the model.

[0019] 5. The present invention introduces a predictive control algorithm, establishes a dynamic response model by fitting the transfer function model of pressure and flow, and calculates the expected change. By comparing with the allowable fluctuation range, loss compensation is performed when the expected change exceeds the range. This prediction and compensation mechanism can deal with the pressure and flow fluctuation problems that may occur during the working condition switching process in advance, effectively avoid system instability caused by working condition changes, ensure the stable operation of the compressor during various working condition conversion processes, and increase working condition adaptability.

[0020] 6. The present invention cooperates with each other in a series of processes from multi-parameter data collection, working condition analysis model construction and training update, predictive control and loss compensation to dynamic adjustment optimization to form a closed-loop control system. This comprehensive optimization method can maintain the stable operation of the system under complex and changeable working conditions for a long time, reduce equipment wear and failure risks caused by changes in working conditions, and improve the reliability and service life of the compressor system. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a schematic diagram of the overall flow of a constant flow and constant pressure control method for a magnetically suspended centrifugal compressor according to Example 1 of the present invention. DETAILED DESCRIPTION

[0022] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0023] Example 1 Reference Figure 1 , a constant flow and constant pressure control method for a magnetic suspension centrifugal compressor of this embodiment includes: Obtain multi-parameter data of the compressor, including compressor gas pressure, temperature, humidity and gas composition data; Use multi-parameter data and historical operation data to build a working condition analysis model, and train and update the built working condition analysis model; Adjust the weight of the working condition demand according to the judgment results of the working condition analysis model, including adjusting the weight of pressure and flow using an exponential decay curve; Introducing a predictive control algorithm based on historical data and operating characteristics, and compensating for losses by predicting the current trend of change and establishing a dynamic response model; Dynamic adjustment and optimization based on real-time working conditions and the rate of change of pressure and flow; The output power of the compressor is precisely controlled through dynamic optimization results.

[0024] Specifically, the following steps are included: S1. Obtaining multi-parameter data of the compressor, including obtaining compressor gas pressure, temperature, humidity and gas composition data; The inlet and outlet of the compressor are respectively equipped with a temperature sensor, a humidity sensor and a gas composition sensor. The temperature sensor is a platinum resistance temperature sensor, which has the characteristics of high precision and good stability. The humidity sensor is a capacitive humidity sensor, which can accurately measure the humidity content in the gas. The gas composition sensor uses a gas sensor based on the infrared absorption principle or the electrochemical principle. This embodiment not only installs a pressure transmitter and flow sensor commonly used in compressors to monitor the outlet pressure of the fan and calculate the flow rate using existing parameters, but also adds temperature, humidity, and gas composition sensors to comprehensively monitor the gas state parameters at the inlet and outlet of the fan.

[0025] S2. Use multi-parameter data and historical operation data to build a working condition analysis model, and train and update the built working condition analysis model; The 3σ principle is used to clean the acquired multi-parameter data, and the cleaned pressure, temperature, humidity, and gas composition multi-parameter data are used as the input layer nodes of the working condition analysis model. The working condition analysis model adopts a neural network architecture, in which the number of input layer nodes corresponds to the number of parameters collected by the sensor. For example, if there are 4 sensors that collect pressure, temperature, humidity, and gas composition data respectively, then 4 nodes are set in the input layer, and each node receives the real-time data value of the corresponding parameter and passes it to the hidden layer for processing.

[0026] Hidden layer setting: The number of hidden layers and the number of nodes in each layer need to be determined according to the complexity of the actual data and the accuracy requirements of the working condition analysis. In this embodiment, the structure of the hidden layer is preliminarily determined by an empirical formula, which is: , where m is the number of hidden layer nodes, n is the number of input layer nodes, and l is the number of output layer nodes to calculate the number of hidden layer nodes. The hidden layer nodes are connected to the input layer nodes through connection weights, and the ReLU activation function is used to perform nonlinear transformation on the input data to extract feature information from the data.

[0027] Output layer definition: The output layer sets two nodes, which are used to represent the weights of the current working conditions for pressure control demand and flow control demand. The value range of these two weights is between 0 and 1, and the sum of the two is 1. Through the calculation of the output layer nodes and the normalization processing of the softmax function, the final pressure control weight and flow control weight are obtained to guide the subsequent intelligent mode switching and control strategy adjustment.

[0028] The sorted historical operation data is divided into training set and test set according to a certain ratio. The working condition analysis model is trained using the training set data. The back propagation algorithm is used to calculate the error gradient, and the parameters are updated according to the gradient descent method. In each training iteration, the mean square error function is used to calculate the error between the output pressure and flow control weights and the actual working condition demand weights. The mean square error function is expressed as: , in, is the number of training samples, is the actual working condition demand weight, is the weight predicted by the model, and then the error is back-propagated from the output layer to the input layer through the back-propagation algorithm, and the connection weights and bias items of each layer are adjusted in turn, so that the model gradually approaches the optimal solution.

[0029] Then, the incremental learning method is used to add the data generated during the operation of the compressor to the training set and retrain the model as new data. ,in For new input parameters, such as new pressure, temperature, humidity, gas composition, etc. measurements; The corresponding actual working condition demand data, such as new pressure control weight and flow control weight, are added to the existing training set. Suppose the original training set is , the training set after adding new data becomes: , in the working condition analysis model, the model parameters ( Connection weights and The update of the bias term is based on the gradient descent algorithm. In incremental learning, the parameter update formula needs to consider the impact of new data. Taking the connection weight as an example, its update formula is: , in, is the learning rate, which determines the step size of parameter update. is the old value of the connection weight between the ith input neuron and the jth hidden layer neuron, is the new value of the connection weight between the i-th input neuron and the j-th hidden layer neuron after updating, is the mean square error function after including the new data, calculated as: , in, is the actual working condition demand weight of the kth sample in the original training set, is the model’s prediction weight for the kth sample in the original training set, and They are the actual working condition demand weight and model prediction weight of the new data, Indicates the number of training set samples after adding new data, bias term The update formula is similar to the weight update formula: , in, is the new value of the bias term after update, It is represented as the old value of the bias term, thereby completing the update of the weight and bias term.

[0030] S3. Adjust the weight of the working condition demand according to the judgment result of the working condition analysis model, including adjusting the weight of pressure and flow by using an exponential decay curve; When the operating condition analysis model determines that the operating condition changes from focusing on flow demand to pressure demand, or from pressure demand to flow demand, an exponential decay curve is used to adjust the control weight. For example, the control weight formula for the constant flow mode is: , in, , is the switching time, is the attenuation coefficient, In order to control the weight of the traffic pattern at time t, the characteristics of the exponential function make the weight drop faster in the initial stage and gradually slow down as time goes by, which meets the requirements of rapid response to working condition changes in the early stage of switching and smooth transition in the later stage. For example, if T=10 seconds, , at t=0, ; When t=5 seconds, , when t=10 seconds, , we can see that the flow control weight gradually decreases. The fixed pressure mode control weight , then it gradually increases from 0 to 1 accordingly, realizing a smooth transition of the control weights of the two modes, where is the pressure mode control weight at time t.

[0031] S4. Introduce a predictive control algorithm based on historical data and operating characteristics, and compensate for losses by predicting the current change trend and establishing a dynamic response model; 1. When the compressor is in stable operation, a step change is applied to the output frequency of its inverter. At the same time, a high-precision pressure transmitter and flow sensor are used to quickly collect the pressure and flow change data at the fan outlet over time to obtain the pressure response data P(t) and flow response data Q(t). Then, the transfer function model is fitted. For the pressure response, it is assumed that the transfer function model is: , in, is the pressure gain, which represents the steady-state gain of the system, It is a pure lag link. is the pure lag time, is the time constant, Typical expression of the first-order inertia link, s is the Laplace transform variable, and the collected pressure step response data P(t) is used to estimate the transfer function model parameters , the curve is fitted by the least square method, the collected pressure data P(t) is Laplace transformed to obtain P(S), and then the assumed transfer function model is With the input step signal (whose Laplace transform is ) to obtain the theoretical output response , adjust the model parameters by the least squares method , and , so that the theoretical output response Inverse Laplace transform The sum of squares of the errors between the actual measured pressure response data P(t) is minimized.

[0032] For the flow response, the transfer function model is also assumed to be: , in, is the pressure gain, which represents the steady-state gain of the system, It is a pure lag link. is the pure lag time, is the time constant, Typical expression of the first-order inertia link, using the same method as the pressure response above to estimate the model parameters using the flow step response data Q(t) , and ,Through such system identification process, a transfer function model is obtained that can ,accurately describe the dynamic response characteristics of the fan system pressure and ,flow, which provides a basis for the subsequent dynamic response prediction and ,control strategy optimization.

[0033] When the working condition changes (such as from flow demand to pressure demand), the current inverter output frequency is known. , fan speed And the system dynamic response model including the pressure response model , and flow response model First, the target inverter output frequency is calculated according to the control strategy. , According to the change of the constant pressure mode control weight and the pressure setting value, it is calculated by the pre-established pressure-frequency relationship model. Then, the change of the inverter output frequency is calculated: , Will As input signals, they are substituted into the pressure and flow dynamic response models respectively to predict the changing trends of pressure and flow in the future after switching. For pressure prediction, by solving the pressure dynamic response model, the pressure The differential equation after the switch or the Laplace transform method is used to calculate the pressure change curve over time. Similarly, for flow prediction, the flow change curve over time is calculated. These prediction curves can show the approximate changes in pressure and flow after switching in advance, providing a basis for early adjustment.

[0034] Allowable fluctuation range of set pressure And flow rate allowable fluctuation range , compare the expected change with the allowable fluctuation range to determine whether it is possible to exceed the allowable range. According to the predicted pressure and flow change curves, if it is found that the pressure will drop instantly after switching (for example, it is predicted that the pressure will be lower than the set value in the first few seconds after switching), loss compensation measures need to be taken, and the inverter output frequency adjustment required for advance adjustment is calculated. The expected pressure compensation value is determined by reversely solving the pressure dynamic response model. According to the pressure compensation value, combined with the pressure-frequency relationship and the dynamic response model, the required output frequency change is calculated.

[0035] S5. Dynamic adjustment and optimization based on real-time working conditions and the change rate of pressure and flow, including dynamic optimization in constant pressure mode and dynamic optimization in constant flow mode; Temperature, humidity and gas composition data are added to the dynamic response model, and the difference between pressure and flow is redefined through the compensation algorithm: , in, and are the temperature and humidity compensation coefficients, and It is the real-time temperature and humidity value. and are the temperature and humidity parameters under standard conditions. is the pressure setting value, The actual pressure value is calculated in the same way. The difference between the actually measured flow value and the preset value is calculated. The output frequency change is adjusted secondary according to the pressure and flow difference. The adjustment effect can be further optimized by adjusting the output frequency change secondary according to the redefined pressure and flow difference. When the pressure and flow difference changes due to temperature, humidity or gas composition changes, the inverter output frequency can be adjusted timely and accurately through secondary adjustment, so that the pressure and flow return to the set value faster.

[0036] S6. Accurately control the output power of the compressor through dynamic optimization results; Based on the methods described in S1 to S5, the motor speed is controlled by adjusting the inverter output frequency through PLC automatic control, thereby realizing constant pressure and constant flow control. The motor input voltage and frequency are adjusted by the inverter to realize precise power control. The calculated power target value is compared with the actual power measurement value. If there is a deviation between the actual power and the target power, the inverter output frequency is adjusted to make the actual power approach the target power. The inverter output frequency is gradually reduced by continuously adjusting it until the power control accuracy requirement is met.

[0037] Example 2 The difference between this embodiment and embodiment 1 is that this embodiment provides a constant flow and constant pressure control system for a magnetic suspension centrifugal compressor, including: The data acquisition module is configured to: acquire multi-parameter data of the compressor, including acquiring compressor gas pressure, temperature, humidity and gas composition data; The model module is configured to: construct a working condition analysis model using multi-parameter data and historical operation data, and train and update the constructed working condition analysis model; The conversion module is configured to: adjust the weight of the working condition demand according to the judgment result of the working condition analysis model, including adjusting the weight of the pressure and the flow rate by using an exponential decay curve; The compensation module is configured to: introduce a predictive control algorithm based on historical data and operating condition characteristics, and perform loss compensation by predicting a dynamic response model established by the current change trend; The optimization module is configured to: perform dynamic adjustment optimization based on real-time working conditions and the change rates of pressure and flow; The execution module is configured to: accurately control the output power of the compressor through dynamic optimization results.

[0038] A computer-readable storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device, for a method for controlling a constant flow and constant pressure of a magnetically suspended centrifugal compressor.

[0039] A terminal device includes a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, wherein the instructions are suitable for being loaded and executed by the processor, a method for controlling a constant flow and constant pressure of a magnetically suspended centrifugal compressor.

[0040] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for controlling constant flow and constant pressure of a magnetically suspended centrifugal compressor, characterized in that: include: Obtain multi-parameter data of the compressor, including compressor gas pressure, temperature, humidity and gas composition data; Use multi-parameter data and historical operation data to build a working condition analysis model, and train and update the built working condition analysis model; Adjust the weight of the working condition demand according to the judgment results of the working condition analysis model, including adjusting the weight of pressure and flow using an exponential decay curve; Introducing a predictive control algorithm based on historical data and operating characteristics, and compensating for losses by predicting the current trend of change and establishing a dynamic response model; Dynamic adjustment and optimization based on real-time working conditions and the rate of change of pressure and flow; The output power of the compressor is precisely controlled through dynamic optimization results.

2. A method for controlling constant flow and constant pressure of a magnetic suspension centrifugal compressor according to claim 1, characterized in that: The method uses multi-parameter data and historical operation data to construct an operating condition analysis model, including cleaning the acquired multi-parameter data to remove abnormal data, and using the cleaned data as input to construct an operating condition analysis model with an input layer, a hidden layer and an output layer as a framework, wherein the number of nodes in the input layer is the same as the number of parameters, the output layer has two nodes, which are respectively used to represent the weights of the current operating condition for the pressure control demand and the flow control demand, and the nodes of the hidden layer are calculated according to the number of nodes in the input layer and the output layer.

3. A method for controlling constant flow and constant pressure of a magnetic suspension centrifugal compressor according to claim 2, characterized in that: The constructed operating condition analysis model is trained and updated, including inputting historical operating data as a training set into the operating condition analysis model, using a mean square error function to calculate the error between the output pressure and flow control weights and the actual operating condition demand weights, and then back-propagating the error from the output layer to the input layer through a back-propagation algorithm, and adjusting the connection weights and bias items of each layer in turn.

4. A method for controlling constant flow and constant pressure of a magnetic suspension centrifugal compressor according to claim 3, characterized in that: The training and updating of the constructed working condition analysis model also includes using an incremental learning method to add new data generated by the compressor during operation to the training set, calculating the mean square error function after adding the new data, and updating the weights and bias items based on the gradient descent algorithm and using the mean square error function. The calculation formula of the mean square error function is: ， in, is the actual working condition demand weight of the kth sample in the original training set, is the model’s prediction weight for the kth sample in the original training set, and They are the actual working condition demand weight and model prediction weight of the new data respectively.

5. A method for controlling constant flow and constant pressure of a magnetic suspension centrifugal compressor according to claim 4, characterized in that: The predictive control algorithm is introduced according to historical data and operating conditions, including using the least squares method to fit the transfer function model of pressure and flow, obtaining the dynamic response model based on the current fan speed, inverter output frequency and the established transfer function model, and calculating the expected changes in pressure and flow during the switching process according to the dynamic response model, setting the allowable pressure fluctuation range and the allowable flow fluctuation range, and comparing them with the expected changes.

6. A method for controlling constant flow and constant pressure of a magnetic suspension centrifugal compressor according to claim 5, characterized in that: The loss compensation is performed by using a dynamic response model established by predicting the current change trend, including performing loss compensation when the expected change is greater than the allowable fluctuation range, determining the expected pressure compensation value by reversely solving the pressure dynamic response model, and calculating the required output frequency change based on the pressure compensation value in combination with the pressure-frequency relationship model and the dynamic response model.

7. A method for controlling constant flow and constant pressure of a magnetic suspension centrifugal compressor according to claim 6, characterized in that: The dynamic adjustment optimization based on real-time working conditions and the rate of change of pressure and flow includes adding temperature, humidity and gas composition data on the basis of the dynamic response model, redefining the difference between pressure and flow through a compensation algorithm, and performing secondary adjustment on the output frequency change according to the difference between pressure and flow.

8. A constant flow and constant pressure control system for a magnetically suspended centrifugal compressor, executing the method according to claim 1, characterized in that: include: The data acquisition module is configured to: acquire multi-parameter data of the compressor, including acquiring compressor gas pressure, temperature, humidity and gas composition data; The model module is configured to: construct a working condition analysis model using multi-parameter data and historical operation data, and train and update the constructed working condition analysis model; The conversion module is configured to: adjust the weight of the working condition demand according to the judgment result of the working condition analysis model, including adjusting the weight of the pressure and the flow rate by using an exponential decay curve; The compensation module is configured to: introduce a predictive control algorithm based on historical data and operating condition characteristics, and perform loss compensation by predicting a dynamic response model established by the current change trend; The optimization module is configured to: perform dynamic adjustment optimization based on real-time working conditions and the change rates of pressure and flow; The execution module is configured to: accurately control the output power of the compressor through dynamic optimization results.

9. A computer-readable storage medium storing a plurality of instructions, characterized in that: The instructions are suitable for being loaded and executed by a processor of a terminal device according to a method for controlling a constant flow and constant pressure of a magnetically suspended centrifugal compressor as claimed in claim 1.

10. A terminal device, comprising a processor and a computer-readable storage medium, wherein the processor is used to implement each instruction; and the computer-readable storage medium is used to store multiple instructions, characterized in that: The instructions are suitable for being loaded by a processor and executed by a constant flow and constant pressure control method for a magnetically suspended centrifugal compressor as claimed in claim 1.

Images