Evaporator energy storage process control method and system based on fractional order MIMO nonlinearity

Abstract

The invention discloses an evaporator energy storage process control method and system based on fractional order MIMO nonlinearity. Historical operation data of the evaporator are obtained, and corresponding fractional order MIMO nonlinear system models are constructed according to different dryness intervals; operating data containing dryness is collected in real time, the fractional order MIMO model of the current dryness is input, and a prediction result is generated; comparing the prediction result with the real-time data, and if the deviation exceeds a threshold value, adjusting the parameters and re-executing; otherwise, calculating a dynamic weight to optimize the model in the next stage; real-time data are analyzed through fractional order MIMO models under different dryness degrees, abnormal stages are identified, and early warning is carried out; and comparing real-time analysis results of different dryness interval models, identifying dryness fluctuation amplitude and pressure oscillation frequency data, and then judging an anomaly type. According to the method, reliable control over the energy storage process of the evaporator is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of evaporators, and specifically to an evaporator energy storage process control method and system based on fractional-order MIMO nonlinearity. Background Technique

[0002] As one of the core components of a refrigeration system, the evaporator exchanges heat with the external environment through a low-temperature condensed working fluid to achieve phase change endothermic heat absorption, thereby generating a cooling effect. Its structure includes a heat exchange unit and a separation chamber: the former is responsible for transferring the heat energy required for vaporization, and the latter ensures the complete separation of the medium state. In the heat management process, the evaporator has the characteristics of absorbing and temporarily storing heat energy, which can not only efficiently recover the waste heat of the system to improve the energy conversion efficiency, but also coordinate the internal heat flow distribution to ensure the stable operation of the equipment. However, the environmental temperature and pressure conditions will restrict its heat storage performance. Therefore, it is necessary to optimize the working parameters through machine learning intelligent control means to ensure that the system maintains the best working conditions.

[0003] For this reason, an evaporator energy storage process control method and system based on fractional-order MIMO nonlinearity are proposed. Summary of the Invention

[0004] The purpose of the present invention is to provide an evaporator energy storage process control method and system based on fractional-order MIMO nonlinearity, which realizes reliable control of the evaporator energy storage process to solve the technical problems mentioned in the background technique.

[0005] To solve the above technical problems, the specific technical solutions of the present invention are as follows:

[0006] An evaporator energy storage process control method based on fractional-order MIMO nonlinearity includes:

[0007] Obtain the historical operation data of the evaporator, and construct a corresponding fractional-order MIMO nonlinear system model according to different dryness intervals;

[0008] Real-time collect the operation data including dryness, input it into the fractional-order MIMO nonlinear system model of the current dryness, and generate a prediction result;

[0009] Compare the prediction result with the real-time data. If the deviation exceeds the threshold, adjust the parameters and re-execute; otherwise, calculate the dynamic weight to optimize the model in the next stage;

[0010] Analyze the real-time data through the fractional-order MIMO nonlinear system model at different dryness levels, identify the abnormal stage and give an early warning;

[0011] After the early warning is triggered, compare the real-time analysis results of the models in different dryness intervals, identify the data of the dryness fluctuation amplitude and the pressure oscillation frequency, and then determine the type of abnormality.

[0012] Preferably, compare the temperature, pressure, flow rate, dryness, and energy parameters in the real-time data and the predicted data, and calculate the comprehensive deviation value;

[0013] For different dryness intervals, map the comprehensive deviation to the dynamic weight DW through the arctangent function:

[0014]

[0015] where, arctan() represents the arctangent function; ΔD represents the allowable reaction deviation value; C th represents the comprehensive deviation threshold; C represents the real-time comprehensive deviation value.

[0016] Preferably, the calculation formula of the real-time comprehensive deviation value C:

[0017]

[0018] where, β k is the weight factor corresponding to temperature, pressure, flow rate, dryness, and energy; Vreal is the real-time parameter value; Vpre is the predicted parameter value.

[0019] Preferably, the state space representation of the fractional-order MIMO nonlinear system model is:

[0020]

[0021] where, D α is the fractional derivative, 0 < α < 1; X(t) is the state variable; S is the state matrix, I is the input matrix, O is the output matrix; T is the transfer matrix; DW is the dynamic weight, U(t) is the input variable, including temperature, pressure, flow rate, and dryness parameters; d(t) is the external disturbance; y(t) is the output variable.

[0022] Preferably, for heat exchange tubes with a pipe diameter less than 8 mm: According to the historical data of the two-phase evaporation stage divided by different dryness, train different prediction sub-models respectively;

[0023] When the dryness ≤ 0.75, the historical data also includes the liquid film thickness change rate and the bubble growth time. Perform nonlinear fitting on the historical data, use the nucleate boiling coefficient as the regularization term, and capture the lag of the liquid film dynamics and bubble behavior by introducing a lag variable;

[0024] When the dryness > 0.75, the historical data also includes the turbulence intensity and the droplet entrainment rate; perform nonlinear fitting on the historical data, use the Reynolds number as the regularization term; capture the correlation between the stability of the flow and the energy output by introducing the Lyapunov exponent.

[0025] Preferably, the abnormal types are divided into recoverable abnormalities and structural abnormalities, and the determination method is:

[0026] If the amplitude of fluctuations in several degrees is less than 20% of the rated value and the duration is less than 5 seconds, or the correlation coefficient between the pressure oscillation frequency and the spectrum of the preset normal operating condition is greater than 0.8, it is determined as a recoverable abnormality;

[0027] If the several degrees continuously deviate from the safety threshold for more than 30 seconds, or the Mahalanobis distance between the pressure-flow coupling relationship and the historical degradation data is less than the preset threshold, it is determined as a structural abnormality.

[0028] Preferably, after the abnormal warning module identifies the abnormal reaction stage, it further includes an automatic adjustment mechanism, which triggers corresponding countermeasures according to the type and severity of the abnormality;

[0029] For recoverable abnormalities, the fractional-order MIMO nonlinear system model automatically adjusts the operating parameters, and the operating parameters include input power and fluid flow rate, and the energy storage process of the evaporator is restored to the normal state by changing the operating parameters;

[0030] For structural abnormalities, the fractional-order MIMO nonlinear system model automatically switches to the standby control mode, and at the same time issues an alarm to notify the maintenance personnel to conduct inspections and repairs;

[0031] The automatic adjustment mechanism includes a preset adjustment strategy library, which is formulated based on historical data and expert experience, and covers the parameter adjustment range and control logic under different abnormal conditions.

[0032] In addition, on the other hand of the present invention, an evaporator energy storage process control system based on fractional-order MIMO nonlinearity includes: a system control module, a data acquisition module, a data processing module, a data control module, a data judgment module, an abnormal warning module, an abnormal type determination module, and an adaptive regulation module;

[0033] Among them, the system control module is used to control the start, pause, and stop of system equipment;

[0034] The data acquisition module is used to obtain real-time data collected by sensors and historical reaction data in the historical database;

[0035] The data processing module is used to input the historical reaction data into the fractional-order MIMO nonlinear system model of the current dryness to obtain dryness prediction data for each dryness stage of the energy storage process;

[0036] The data control module is used to control the reaction variables of each stage by using the fractional-order MIMO nonlinear system model of each stage to obtain reaction result data for each stage;

[0037] The data judgment module is used to compare the reaction result data of each stage with the real-time data numerically; analyze the reaction result data and the reaction prediction data to obtain the dynamic weight; then, apply the dynamic weight of the current stage to the fractional-order MIMO nonlinear system model of the next stage to constrain the reaction variables in the next reaction stage.

[0038] The abnormal warning module is used to identify the abnormal reaction stage according to the real-time data; if an abnormal reaction stage is identified, a warning is issued.

[0039] The abnormal type determination module is used to determine the abnormal type according to the dryness fluctuation amplitude and the pressure oscillation frequency data.

[0040] The evaporator energy storage process control method and system based on fractional-order MIMO nonlinear of the present invention have the following advantages:

[0041] 1. The present invention proposes a reaction stage prediction function for predicting the data characteristics of each stage of the evaporator energy storage process; this function includes two steps: stage identification and stage prediction; the process of stage identification uses the historical reaction data with marked dryness to train the reaction stage identification model, which is to distinguish the dryness data of different stages; then, the process of stage prediction uses multiple stage data sets to train and obtain the dryness stage prediction model of each reaction stage, and obtain the reaction prediction data of each reaction stage of the energy storage process; this function can obtain accurate prediction results for each stage, provide reliable data support for subsequent reaction data control, and further improve the effectiveness and reliability of the evaporator energy storage process control.

[0042] 2. The present invention proposes a fractional-order MIMO nonlinear system model based on dynamic weight to suppress the deviation problem of reaction parameters caused by stage conversion and environmental changes during the energy storage process; the process of obtaining the dynamic weight is calculated by comparing the allowed reaction deviation value, the comprehensive deviation threshold and the real-time comprehensive deviation value of each dryness stage; then, the dynamic weight calculated in the current reaction stage is embedded into the state space model of the next stage to adaptively adjust the reaction parameters during the energy storage process, so as to more effectively control the deviation of reaction variables, thereby improving the effectiveness and reliability of the evaporator energy storage process control.

[0043] 3. The present invention proposes an abnormal warning function for identifying the abnormal reaction stage and providing warning prompt information. This function can identify abnormal situations including stage reverse jumps and stage crossings through detailed dryness analysis of real-time data. By defining the expected stage transition sequence and encoding the stages, this function can effectively identify and predict abnormal situations during the reaction process and issue warnings in a timely manner when abnormalities are predicted. By capturing the dynamic changes and trends before and after the stage and efficient abnormal control, the stability and efficiency of the system control process are effectively improved, thereby enhancing the effectiveness and reliability of the evaporator energy storage process control.

[0044] 4. The present invention proposes a determination method for abnormal types. By integrating multi-dimensional characteristic parameters such as dryness mutation gradient and energy efficiency deviation, it can accurately distinguish transient disturbances (such as medium flow fluctuations) from structural degradations (such as tube wall scaling or equipment aging), and the misjudgment rate is significantly reduced compared with the traditional threshold method. At the same time, based on the sliding mode variable structure adaptive control strategy and the multi-model set and rolling optimization strategy, the abnormal positioning accuracy is significantly improved, avoiding redundant control actions caused by mis-triggering of a single parameter, reducing unplanned downtime and operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic diagram of the control process of the evaporator energy storage process of the present invention;

[0046] Figure 2 It is a schematic diagram of the structure of the reaction stage identification model of the present invention;

[0047] Figure 3 It is a schematic diagram of the structure of the reaction stage prediction model of the present invention;

[0048] Figure 4 It is a schematic diagram of the structure of the evaporator energy storage process control system based on fractional-order MIMO nonlinearity of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] In order to better understand the purpose, structure and function of the present invention, the control method and system for the evaporator energy storage process based on fractional-order MIMO nonlinearity proposed by the present invention will be further described in detail below with reference to the drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0050] Embodiment 1

[0051] The dryness adopts the general definition in the field of evaporators, which is the weight ratio of the working medium steam in wet saturated steam. That is, it refers to the mass percentage of dry steam contained in each kilogram of wet steam.

[0052] The dryness X = (h x - h f ) / (h s - h f ). Wherein, h x is the enthalpy of wet steam, h f is the enthalpy of saturated water, and h s is the enthalpy of saturated steam.

[0053] In the embodiments of the present application, the specific implementation process will be realized by the evaporator energy storage process control method based on fractional - order MIMO nonlinearity of the present invention; the evaporator energy storage process control method based on fractional - order MIMO nonlinearity includes:

[0054] Obtain the historical operation data of the evaporator, and construct a corresponding fractional - order MIMO nonlinear system model according to different dryness intervals;

[0055] Collect the operation data including dryness in real - time, input it into the fractional - order MIMO nonlinear system model of the current dryness, and generate a prediction result;

[0056] Compare the prediction result with the real - time data. If the deviation exceeds the threshold, adjust the parameters and re - execute; otherwise, calculate the dynamic weight to optimize the model for the next stage;

[0057] Analyze the real - time data through the fractional - order MIMO nonlinear system models at different dryness levels, identify the abnormal stage and give an early warning;

[0058] After the early warning is triggered, compare the real - time analysis results of the models in different dryness intervals, identify the data of the dryness fluctuation amplitude and the pressure oscillation frequency, and then determine the type of abnormality.

[0059] The control flow of the evaporator energy storage process can refer to Figure 1 , specifically as follows:

[0060] Collect the real - time data including dryness information;

[0061] Input the real - time data into the fractional - order MIMO nonlinear system model of the current stage to obtain the reaction result data of the current stage;

[0062] Compare the reaction result data of the current stage with the reaction prediction data. If the difference between the two is greater than the preset threshold, adjust the adjustable parameters of the current stage and re - input them into the system model of the current stage for processing; otherwise, analyze the reaction result data and the reaction prediction data to obtain the dynamic weight;

[0063] Operate on the next stage according to the fractional - order MIMO nonlinear system model based on the dynamic weight;

[0064] Identify the abnormal reaction stage based on the real-time data; if the abnormal reaction stage is identified, issue a warning.

[0065] In this embodiment, a control method for the evaporator energy storage process based on fractional-order MIMO nonlinearity is proposed. Obtain and record the historical reaction data of the evaporator energy storage process; for each reaction stage, construct a corresponding fractional-order MIMO nonlinear system model; use the reaction stage prediction model to obtain the reaction prediction data for each reaction stage; repeat the evaporator energy storage process with the initial reaction parameters until the energy storage process ends. This method can effectively improve the effectiveness and reliability of the control of the evaporator energy storage process.

[0066] For specific illustration, the present invention is described in conjunction with the following embodiments as follows:

[0067] Obtain the historical reaction data of the evaporator energy storage process; the historical reaction data is recorded according to the dryness.

[0068] Furthermore, the specific implementation process of training the corresponding reaction stage prediction model for each reaction stage according to the reaction stage identification model to obtain the reaction prediction data for each reaction stage of the energy storage process includes:

[0069] Perform annotation processing on the historical reaction data, use the reaction feature data without reaction stage annotation as the independent variable data, and use the reaction stage annotation as the dependent variable data. After training, obtain the reaction stage identification model.

[0070] The structure of the reaction stage identification model in this embodiment is as Figure 2 shown, including: an input layer, a multi-scale feature extraction layer, an attention enhancement layer, a multi-scale feature recognition layer, and a fusion output layer; the input layer is used to convert the annotated (dryness) historical reaction data into the feature domain for subsequent processing; the multi-scale feature extraction layer uses convolution operations of sizes 3×3, 5×5, and 7×7 to extract multi-scale convolution features; the attention enhancement layer uses spatial attention and channel attention mechanisms to enhance the multi-scale convolution features; the multi-scale feature recognition layer is used to recognize the enhanced multi-scale convolution features to obtain multi-scale recognition features; finally, the output layer first performs channel fusion on the multi-scale recognition features, and then uses a convolutional layer to convert the fused recognition features into the recognition stage data result for output.

[0071] Training the corresponding reaction stage prediction model for each reaction stage includes:

[0072] Divide the reaction characteristic data according to the reaction stage annotation to obtain multiple dryness stage data sets; for each of the stage data sets, use the input reaction characteristic data of each stage as independent variable data and the output reaction characteristic data as dependent variable data, and perform training respectively to obtain the reaction stage prediction model corresponding to each of the reaction stages;

[0073] In this embodiment, the reaction stage prediction model uses a hybrid network combining CNN and LSTM. The structure of this model is as Figure 3 shown; the hybrid network adopts a dual-branch structure, inputs the input data into two sub-networks of CNN and LSTM respectively for feature processing, and finally fuses the output features of different networks to output the prediction result; CNN uses 3 convolutional layers to extract spatial convolutional features; the LSTM network uses 3 LSTM layers to extract long-term dependence features that change over time.

[0074] In this embodiment, a reaction stage prediction function is proposed to predict the data characteristics of each stage of the evaporator energy storage process; this function includes two steps: stage identification and stage prediction; the process of stage identification uses the labeled historical reaction data to train the reaction stage identification model, which is to distinguish the reaction data of different dryness stages; then, the process of stage prediction uses multiple stage data sets to train and obtain the reaction stage prediction model of each dryness stage, and obtains the reaction prediction data of each energy storage process reaction stage; this function can obtain accurate prediction results for each stage, provide reliable data support for subsequent reaction data control, and further improve the effectiveness and reliability of the evaporator energy storage process control.

[0075] Construct a corresponding fractional-order MIMO nonlinear system model according to each energy storage process reaction stage;

[0076] Repeat the evaporator energy storage process with the initial reaction parameters until the energy storage process ends;

[0077] Further, in the evaporator energy storage control process, compare the reaction temperature information, reaction pressure information, reaction dryness information, reaction flow information, and reaction output energy information in the reaction result data with the reaction temperature prediction information, reaction pressure prediction information, reaction dryness prediction information, reaction flow prediction information, and reaction output energy prediction information in the reaction prediction data respectively to obtain a real-time comprehensive deviation value;

[0078] Further, compare the temperature, pressure, flow, dryness, and energy parameters in the real-time data and the prediction data, and calculate the comprehensive deviation value;

[0079] Since traditional modeling has always used time series for analysis, introducing the parameter of dryness in the evaporator field for modeling and division greatly increases the scientific nature of the MIMO nonlinear system model. The advancement of this method can be verified through subsequent experiments. The innovative introduction of the variable of dryness can greatly improve the prediction accuracy of different models at different dryness levels.

[0080] For example, the dryness can be divided into three intervals: (0, 0.75), [0.75, 0.8], [0.8, 1) for data division and prediction, and three different fractional-order MIMO models are established.

[0081] For different dryness intervals, the comprehensive deviation is mapped to the dynamic weight DW through the arctangent function:

[0082]

[0083] Among them, arctan() represents the arctangent function; ΔD represents the allowable reaction deviation value; C th represents the comprehensive deviation threshold; C represents the real-time comprehensive deviation value.

[0084] The energy storage process of the evaporator involves the nonlinear dynamic characteristics of multivariable coupling. Especially in the two-phase evaporation stage, the change of dryness will significantly affect key parameters such as liquid film thickness and turbulence intensity. Traditional control methods (such as PID control or static weight model) are difficult to adapt to such complex working conditions, and often cause control lag or overshoot due to fixed thresholds or weights. By introducing the dynamic weight (DW) calculation mechanism, the comprehensive deviation is mapped to the weight value through the arctangent function to achieve the adaptive adjustment of the weight.

[0085] Traditional methods often control based on the deviation of a single parameter (such as temperature or pressure), ignoring the coupling effect between multivariables. By comprehensively considering the real-time deviation values of temperature, pressure, flow rate, dryness and energy parameters. Through dynamic weight mapping and multi-parameter comprehensive evaluation, the problem of insufficient adaptability of traditional methods in the nonlinear and multivariable control of evaporators is solved. Its innovation lies in: improving the model self-adaptability, enhancing the control accuracy of multiple working conditions, optimizing the complex dynamic response, and ensuring the engineering reliability through mathematical stability design. Experimental data and comparison results (as shown in Table 2) further prove that this method improves the reasonable range ratio of energy storage heat loss from 80.7% of traditional methods to 95.2%, which is significantly better than the existing technology.

[0086] Furthermore, the calculation formula of the real-time comprehensive deviation value C:

[0087]

[0088] Among them, β kare weight factors corresponding to temperature, pressure, flow, dryness and energy; Vreal is the real-time parameter value; Vpre is the predicted parameter value.

[0089] Evaluate the system state from a global perspective. For example, in the nucleate boiling stage with dryness ≤ 0.75, the nonlinear coupling effect between the liquid film thickness and the bubble growth time is significant. The comprehensive deviation calculation can more accurately capture such dynamic characteristics and avoid the one-sidedness of single parameter control. k ) is flexibly configured, and the system can adjust parameter priorities according to different working conditions (such as high / low dryness range) to further optimize the control strategy.

[0090] The heat and mass transfer process in the two-phase evaporation stage of the evaporator has strong nonlinear and time-varying characteristics, and it is difficult for traditional linear models to accurately describe the liquid film-turbulence coupling behavior. Through the closed-loop feedback of the fractional-order MIMO model and real-time data under the dynamic weight mechanism, the adaptability of the model to stage transitions is significantly improved. For example, when the dryness transitions from the low range (≤0.75) to the high range (>0.75), the system quickly adjusts the gains of the input matrix and the transfer matrix through dynamic weights, effectively suppressing the output deviation caused by sudden changes in turbulence intensity or fluctuations in droplet entrainment rate.

[0091] The introduction of the inverse tangent function not only provides nonlinear mapping capabilities, but its monotonic bounded characteristics also ensure the mathematical stability of weight adjustment and avoid the gradient explosion problem that may be caused by the traditional exponential function. The data in Tables 1 and 2 show the advancement of this scheme compared to the traditional PID control using the integral differential control strategy.

[0092] In this embodiment, the reaction temperature weight factor, reaction pressure weight factor, reaction dryness weight factor, reaction flow weight factor and reaction output energy weight factor are all set to 0.2. Of course, the weight factors can also be adjusted according to actual conditions.

[0093] The setting of the comprehensive prediction and evaluation threshold in this embodiment will be affected by factors such as the type of evaporator and the reaction environment. Therefore, the comprehensive prediction and evaluation threshold needs to be set by relevant personnel in this field based on actual conditions and experience summary, and is not unique.

[0094] In this embodiment, a dynamic weight is proposed to improve the fractional-order MIMO nonlinear control model in the next reaction stage; the weight first numerically compares the comprehensive prediction evaluation difference and the allowable reaction deviation value, and then uses the arctangent function to scale the numerical comparison result to between [-1, 1]; wherein, the comprehensive prediction evaluation difference represents the degree of difference between the comprehensive prediction evaluation threshold and the comprehensive prediction evaluation value; the comprehensive prediction evaluation value is obtained by comprehensively evaluating the reaction result data and the reaction prediction data; the dynamic weight can adaptively adjust the reaction parameters in the system model, avoiding the decline of the system model accuracy caused by huge deviations due to external disturbances, thereby improving the effectiveness and reliability of the evaporator energy storage process control.

[0095] Further, operate on the next stage according to the fractional-order MIMO nonlinear system model based on the dynamic weight;

[0096] The fractional-order MIMO nonlinear system model based on the dynamic weight is a multi-dimensional state space mathematical model, represented by a state space and state variables;

[0097] Among them, the state space of the fractional-order MIMO nonlinear system based on the dynamic weight is expressed as:

[0098]

[0099] Among them, D α is the fractional derivative, 0 < α < 1; X(t) is the state variable; S is the state matrix, I is the input matrix, O is the output matrix; T is the transfer matrix; DW is the reaction prediction weight, U(t) is the input variable, including temperature, pressure, flow rate and dryness parameters; d(t) is the external disturbance; y(t) is the output variable.

[0100] The state variable X(t) can be expressed as:

[0101]

[0102] Among them, the output variable y(t) can be expressed as [y1(x), y2(x), y3(x), y4(x)] T , T represents the transpose operation, and different components respectively represent the output medium flow rate, output medium temperature, output pressure and output energy.

[0103] In this embodiment, α is set to 0.5; the state matrix, input matrix, output matrix and transfer matrix in the fractional-order MIMO nonlinear system are all 4×4 in size, and the internal parameters are constantly changing with time, and the initial parameters can be set according to the actual situation.

[0104] In this embodiment, a fractional-order MIMO nonlinear system model based on dynamic weights is proposed to suppress the deviation problem of reaction parameters caused by stage conversion and environmental changes during the energy storage process; the acquisition process of the dynamic weights is calculated by comparing the difference between the reaction result data and the corresponding reaction prediction data of each stage with the reaction deviation value; then, the dynamic weights calculated in the current reaction stage are embedded into the state space model of the next stage to adaptively adjust the reaction parameters during the energy storage process, so as to more effectively control the deviation of reaction variables, thereby improving the effectiveness and reliability of the evaporator energy storage process control.

[0105] To verify the control robustness of the liquid film-turbulent coupling behavior during the dryness interval switching of 6mm / 7mm / 8mm pipe diameters, 6mm, 7mm, and 8mm copper-nickel alloy heat exchange pipes treated with surface nano-coatings are used, and R32 refrigerant is used.

[0106] Dynamically adjust the evaporation pressure to 0.4MPa ± 0.05MPa; stepwise adjust the inlet flow rate to gradually reach: 0.8m / s; continuously and gradually control the dryness range: 0.2 - 0.95.

[0107] The two-phase evaporation stage during the evaporator energy storage process is one of the hot research directions in the existing research and development. Due to the strong coupling and sensitivity to external interference and other nonlinear characteristics in the two-phase evaporation stage, it is difficult for traditional PID control to accurately model and respond. And the two-phase evaporation stage involves the dynamic characteristics of multiphase flow and the coupling mechanism of heat transfer and mass transfer, so it has always been one of the important core reasons for the low prediction accuracy of the evaporator energy storage process control method. Therefore, corresponding optimization designs are focused on the two-phase evaporation stage.

[0108] Based on the arctangent function dynamic weight DW calculation model parameter offset ΔW:

[0109]

[0110] Among them, N is the number of calculated data; DW pre is the predicted data of the dynamic weight; DW real is the real-time data of the dynamic weight;

[0111] Based on the ratio of the liquid film oscillation frequency f film and the main turbulent frequency f turb calculate the coupling degree η:

[0112]

[0113] The mean square error calculation formula before and after the correction of the dynamic weight is as follows:

[0114]

[0115] The partial experimental results are shown in Table 1 below. After adopting the dynamic weight correction, the root mean square error (RMSE) of the 6-mm pipe diameter decreases from 0.45 to 0.21, verifying the robustness of this method under complex working conditions.

[0116] The change rate of liquid film thickness is the change rate of liquid film thickness with time (dδ / dt), which directly affects the heat transfer resistance between the evaporator wall and the working fluid. By monitoring this parameter in real time, the non-linear influence of the dynamic shrinkage or expansion of the liquid film on the heat flux can be quantified. The bubble growth time characterizes the time period from the nucleation of a single bubble to its detachment from the wall, reflecting the activity of nucleate boiling. A shorter growth time means a higher bubble generation frequency, which can improve the heat transfer efficiency. The average bubble growth time of a single bubble is obtained through statistics of a large amount of data. The nucleate boiling coefficient is introduced into the model as a regularization term to constrain the non-linear relationship between the liquid film thickness and the bubble growth time and prevent overfitting. Its physical meaning lies in balancing the correlation between the evaporation rate and the bubble detachment frequency. The lag variable captures the phase difference between the liquid film dynamics and the bubble behavior by introducing a time delay term (such as the liquid film response lag time τ). For example, the local overheating caused by the liquid film contraction will trigger the nucleation of new bubbles after a certain delay.

[0117] When the dryness is relatively low (dryness ≤ 0.75), a thin film is easily formed inside the evaporator, and a large number of bubbles will be generated at the same time. Existing technologies often ignore the dynamic changes of these details, resulting in inaccurate control. By introducing the "change rate of liquid film thickness", since the thinner the liquid film, the higher the heat transfer efficiency, but if the liquid film is too thin, it is easily broken. By monitoring the change rate of liquid film thickness in real time, the operating state of the evaporator can be adjusted in time to avoid the influence of the too-thin or too-thick liquid film on the heat transfer effect. By introducing the "bubble growth time", since the faster the bubbles are generated (i.e., the shorter the generation time), the higher the heat transfer efficiency. By controlling the bubble growth time, the heat transfer speed of the evaporator can be optimized. By introducing the "nucleate boiling coefficient" as a balancing parameter, the relationship between the liquid film thickness and the bubble generation can be coordinated. For example, when the liquid film shrinks rapidly, the system will automatically adjust the bubble generation frequency to prevent the "incoordination" of the two from causing unstable heat transfer. In actual operation, there is a small time difference between the liquid film change and the bubble behavior (for example, after the liquid film shrinks, the bubbles may be generated with a delay). After introducing the "lag variable", the model can more realistically reflect this dynamic delay and avoid the control command from missing the time point.

[0118] The turbulence intensity is the ratio of the turbulent pulsation velocity to the average velocity, quantifying the intensity of the vortices in the flow. A high turbulence intensity will increase the risk of liquid film rupture and reduce the heat transfer uniformity. The droplet entrainment rate is the mass fraction of the entrained droplets in the vapor flow, which directly affects the energy transfer efficiency of the two-phase flow. The Reynolds number Among them, ρ is the density of the fluid; v is the flow velocity of the fluid; d is the characteristic length; μ is the dynamic viscosity of the fluid. The Reynolds number is used to divide the flow state and optimize the fitting of turbulent parameters. By dynamically adjusting the Reynolds number weight, the mass transfer coefficient under high quality can be adaptively corrected. The early warning of flow instability is realized by calculating the maximum Lyapunov exponent λ of the flow velocity field, and it is judged whether the system is in a chaotic state (λ > 0) or a stable state (λ ≤ 0).

[0119] When the quality is relatively high (quality > 0.75), the liquid basically turns into steam, and the flow becomes violent and chaotic (turbulent). The calculation methods of the existing technologies are difficult to accurately calculate such complex situations. By introducing the "turbulence intensity" to measure the "degree of chaos" of the steam flow. The larger the value of the turbulence intensity, the more unstable the flow, and it may cause the liquid film to break. By monitoring the turbulence intensity in real time, measures can be taken in advance (such as reducing the flow velocity) to prevent the system from getting out of control. By introducing the "droplet entrainment rate" to measure the proportion of droplets mixed in the steam. Too many droplets will reduce the energy transfer efficiency and may also damage the equipment. Controlling this parameter can ensure that the steam is "purer" and improve the energy utilization rate. By introducing the "Reynolds number" as a balance parameter to judge whether the flow state is stable or turbulent. By dynamically adjusting the weight of the Reynolds number, the model can automatically adapt to different working conditions, such as optimizing the heat transfer coefficient under high quality. After introducing the "Lyapunov exponent", the model predicts whether the system will "get out of control". If the exponent is positive, it means that the flow is about to enter a chaotic state; a negative exponent represents stability. By calculating this exponent in real time, the system can give an early warning and adjust the parameters.

[0120] Table 1. Comparison of experimental data for different pipe diameters

[0121]

[0122] Furthermore, an abnormal reaction stage is identified according to the real-time data; if the abnormal reaction stage is identified, a warning is given;

[0123] Among them, the specific implementation process of identifying an abnormal reaction stage according to the real-time data and giving a warning if the abnormal reaction stage is identified includes:

[0124] Define the expected stage transition sequence according to the reaction stage;

[0125] Furthermore, each of the reaction stages is assigned a unique increasing numerical code according to the expected stage transition sequence to obtain a coding rule;

[0126] Furthermore, the reaction stages in the real-time data are coded according to the coding rule to generate a corresponding stage label vector;

[0127] Furthermore, perform a dryness analysis on the real-time data containing the phase label vectors to obtain the number of phase reverse jumps and the number of phase crossings;

[0128] Among them, the calculation formulas for obtaining the number of phase reverse jumps N and the number of phase crossings M are expressed as:

[0129]

[0130] Among them, K represents the total number of dryness data of the real-time data; represents the phase reverse jump indicator function of the i-th dryness data; represents the phase crossing indicator function of the i-th dryness data, i represents the index of the dryness data, V i represents the phase label vector of the i-th dryness data, V i+1 represents the phase label vector of the (i + 1)-th dryness data, and δ is a preset threshold when performing the incremental numerical coding;

[0131] In this embodiment, the preset threshold δ is set to 0.95; the preset threshold will vary depending on the reaction environment, reaction equipment, and reaction conditions, and relevant personnel in this field need to set it flexibly according to the actual situation.

[0132] Furthermore, if the number of phase reverse jumps is greater than the first threshold and / or the number of phase crossings is greater than the second threshold, it is identified as the abnormal reaction phase and a warning is issued.

[0133] In this embodiment, an abnormal warning function is proposed to identify the abnormal reaction phase and provide a warning prompt message; this function can identify abnormal situations including phase reverse jumps and phase crossings through a detailed dryness analysis of real-time data; this function can effectively identify and predict abnormal situations during the reaction process and issue a warning in a timely manner when an abnormality is predicted by defining the expected phase transition order and encoding the phases; this function effectively improves the stability and efficiency of the system control process through capturing the dynamic changes and trends before and after the phase and efficient abnormal control, thereby improving the effectiveness and reliability of the evaporator energy storage process control.

[0134] To verify the actual effect of the evaporator energy storage process control method based on fractional-order MIMO nonlinearity proposed by the present invention, multiple groups of comparative experiments were designed. Among them, Method 1 applied the evaporator energy storage process control method based on fractional-order MIMO nonlinearity proposed by the present invention; Method 2 only used the fractional-order MIMO nonlinear system model for the evaporator energy storage process control, omitting the reaction stage prediction process and dynamic weight calculation; Method 3 applied the reaction stage prediction model to predict the parameters of the evaporator energy storage reaction process for control, omitting the fractional-order MIMO nonlinear system model based on dynamic weight; Method 4 adopted the traditional manual operation of reaction parameters to control the evaporator energy storage process.

[0135] Table 2. Comparison of the effects of different control schemes

[0136]

[0137] As shown in Table 2, the comprehensive method (Method 1) combining the reaction stage prediction and the fractional-order MIMO nonlinear system model based on dynamic weight performs the best, indicating that the method proposed by the present invention is the most effective in the control of the evaporator energy storage process.

[0138] Embodiment 2

[0139] As an implementation manner of the present invention, referring to Figure 4 , the evaporator energy storage process control system based on fractional-order MIMO nonlinearity includes: a system control module, a data acquisition module, a data processing module, a data control module, a data judgment module, and an abnormal warning module;

[0140] Among them, the system control module is used to control the start, pause, and stop of the system equipment;

[0141] The data acquisition module is used to obtain the real-time data collected by the sensor and the historical reaction data in the historical database;

[0142] The data processing module is used to input the historical reaction data into the reaction stage prediction model to obtain the reaction prediction data of each energy storage process reaction stage;

[0143] Furthermore, the specific implementation process of the data processing module for inputting the historical reaction data into the reaction stage prediction model to obtain the reaction prediction data of each energy storage process reaction stage includes:

[0144] Perform annotation processing on the historical reaction data, use the reaction feature data after removing the reaction stage annotation as the independent variable data, and use the reaction stage annotation as the dependent variable data, and obtain the reaction stage recognition model after training;

[0145] Training a corresponding reaction stage prediction model for each of the said reaction stages, including:

[0146] Dividing the said reaction feature data according to the said reaction stage annotation to obtain multiple stage data sets; for each of the said stage data sets, using the input reaction feature data of each stage as independent variable data and the output reaction feature data as dependent variable data, and training respectively to obtain the said reaction stage prediction model corresponding to each of the said reaction stages;

[0147] Based on the said reaction stage identification model and the said reaction stage prediction model trained for each of the said reaction stages, obtaining reaction prediction data for each energy storage process reaction stage.

[0148] The said data control module is used to control the reaction variables of each of the said stages by using the fractional order MIMO nonlinear system model of each stage to obtain the reaction result data of each of the said stages;

[0149] The said data judgment module is used to numerically compare the said reaction result data of each of the said stages with the said reaction prediction data; and analyze the said reaction result data and the said reaction prediction data to obtain a dynamic weight; then, applying the said dynamic weight of the current stage to the fractional order MIMO nonlinear system model of the next stage to constrain the said reaction variables in the next reaction stage;

[0150] Further, the specific implementation process of the said data judgment module for numerically comparing the said reaction result data of each of the said stages with the said reaction prediction data and analyzing the said reaction result data and the said reaction prediction data to obtain a dynamic weight includes:

[0151] Obtaining the reaction result data and reaction prediction data of each stage;

[0152] Comparing the reaction temperature information, reaction pressure information, reaction flow information, and reaction output energy information in the said reaction result data with the reaction temperature prediction information, reaction pressure prediction information, reaction flow prediction information, and reaction output energy prediction information in the said reaction prediction data respectively to obtain a comprehensive deviation value;

[0153] The calculation formula of the real-time comprehensive deviation value C:

[0154]

[0155] where β k is the weight factor corresponding to temperature, pressure, flow rate, dryness, and energy; Vreal is the real-time parameter value; Vpre is the predicted parameter value.

[0156] Compare the temperature, pressure, flow rate, dryness, and energy parameters in the real-time data with the predicted data, and calculate the comprehensive deviation value;

[0157] For different dryness intervals, map the comprehensive deviation to the dynamic weight DW through the arctangent function:

[0158]

[0159] where, arctan() represents the arctangent function; ΔD represents the allowable reaction deviation value; C th represents the comprehensive deviation threshold; C represents the real-time comprehensive deviation value.

[0160] Furthermore, the data judgment module inputs the dynamic weight of the current stage into the fractional-order MIMO nonlinear system model of the next stage in the data control module, and controls the reaction variable according to the state space model of the fractional-order MIMO nonlinear system model; where, the state space model is expressed as:

[0161]

[0162] where, D α is the fractional derivative, 0 < α < 1; X(t) is the state variable; S is the state matrix, I is the input matrix, O is the output matrix; T is the transfer matrix; DW is the reaction prediction weight, U(t) is the input variable, including temperature, pressure, flow rate, and dryness parameters; d(t) is the external disturbance; y(t) is the output variable.

[0163] The abnormal warning module is used to identify the abnormal reaction stage according to the real-time data; if the abnormal reaction stage is identified, a warning is issued.

[0164] Furthermore, the abnormal warning module is used to identify the abnormal reaction stage according to the real-time data; the specific implementation process of issuing a warning if the abnormal reaction stage is identified includes:

[0165] Define the expected stage transition order according to the reaction stage information;

[0166] Furthermore, assign a unique increasing numerical code to each of the reaction stages according to the expected stage transition order to obtain the coding rule;

[0167] Furthermore, encode the reaction stage in the real-time data according to the coding rule to generate the corresponding stage label vector;

[0168] Furthermore, perform dryness analysis on the real-time data including the stage label vector to obtain the number of stage reverse jumps and the number of stage crossings;

[0169] Further, if the number of reverse jumps in the stage is greater than the first threshold and / or the number of stage crossings is greater than the second threshold, it is identified as the abnormal reaction stage and a warning is issued;

[0170] Further, the abnormal types are divided into recoverable abnormalities and structural abnormalities, and the abnormal type is determined according to the dryness fluctuation amplitude and the pressure oscillation frequency data; the determination method is as follows:

[0171] If the dryness fluctuation amplitude is less than 20% of the rated value and the duration is less than 5 seconds, or the correlation coefficient between the pressure oscillation frequency and the preset normal operating condition spectrum is greater than 0.8, it is determined as a recoverable abnormality;

[0172] If the dryness continuously deviates from the safety threshold for more than 30 seconds, or the Mahalanobis distance between the pressure-flow coupling relationship and the historical degradation data is less than the preset threshold, it is determined as a structural abnormality;

[0173] Further, after the abnormal warning module identifies the abnormal reaction stage, it further includes an automatic adjustment mechanism, which triggers corresponding countermeasures according to the abnormal type and severity;

[0174] For recoverable abnormalities, the system automatically adjusts the operating parameters, such as adjusting the input power, changing the fluid flow rate, etc., to restore the energy storage process of the evaporator to the normal state;

[0175] For structural abnormalities, the system automatically switches to the standby control mode and simultaneously issues an alarm to notify the maintenance personnel to conduct inspections and repairs;

[0176] The automatic adjustment mechanism includes a preset adjustment strategy library, which is formulated based on historical data and expert experience and covers the parameter adjustment range and control logic under different abnormal conditions.

[0177] It can be understood that the present invention is described through some embodiments. Those skilled in the art know that without departing from the spirit and scope of the present invention, various changes or equivalent replacements can be made to these features and embodiments. Additionally, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application belong to the scope protected by the present invention.

Claims

1. The evaporator energy storage process control method based on fractional-order MIMO nonlinearity is characterized by: include: Obtain the historical operating data of the evaporator and construct the corresponding fractional-order MIMO nonlinear system model according to different dryness ranges; Collect operating data including dryness in real time, input the fractional-order MIMO nonlinear system model of the current dryness, and generate prediction results; Compare the prediction results with the real-time data. If the deviation exceeds the threshold, adjust the parameters and re-execute. Otherwise, the dynamic weights are calculated to optimize the next stage model; Analyze real-time data through fractional-order MIMO nonlinear system models at different dryness levels, identify abnormal stages and issue warnings; After the early warning is triggered, the real-time analysis results of the models in different dryness intervals are compared, the dryness fluctuation amplitude and pressure oscillation frequency data are identified, and then the abnormality type is determined.

2. The evaporator energy storage process control method based on fractional-order MIMO nonlinearity according to claim 1 is characterized in that: Compare the temperature, pressure, flow, dryness and energy parameters in real-time data and predicted data, and calculate the comprehensive deviation value; For different dryness ranges, the comprehensive deviation is mapped to the dynamic weight DW through the inverse tangent function: Wherein, arctan() represents the inverse tangent function; ΔD represents the allowable reaction deviation value; C th It is represented as the comprehensive deviation threshold; C is represented as the real-time comprehensive deviation value.

3. The evaporator energy storage process control method based on fractional-order MIMO nonlinearity according to claim 2 is characterized in that: The calculation formula of the real-time comprehensive deviation value C is: Among them, β k are weight factors corresponding to temperature, pressure, flow, dryness and energy; Vreal is the real-time parameter value; Vpre is the predicted parameter value.

4. The evaporator energy storage process control method based on fractional-order MIMO nonlinearity according to claim 1 is characterized in that: The state space representation of the fractional-order MIMO nonlinear system model is: Among them, D α is the fractional derivative, 0<α<1; X(t) is the state variable; S is the state matrix, I is the input matrix, O is the output matrix; T is the transfer matrix; DW is the dynamic weight, U(t) is the input variable, including temperature, pressure, flow and dryness parameters; d(t) is the external disturbance; y(t) is the output variable.

5. The evaporator energy storage process control method based on fractional-order MIMO nonlinearity according to claim 2 is characterized in that: For heat exchange tubes with a diameter of less than 8mm: divide the historical data of the two-phase evaporation stage according to different dryness, and train different prediction sub-models respectively; When the dryness is ≤0.75, the historical data also includes the liquid film thickness change rate and the bubble growth time. Nonlinear fitting is performed on the historical data, and the nucleate boiling coefficient is used as a regularization term. The hysteresis of the liquid film dynamics and bubble behavior is captured by introducing a hysteresis variable. When the dryness is greater than 0.75, the historical data also includes turbulence intensity and droplet entrainment rate; nonlinear fitting is performed on the historical data, and the Reynolds number is used as a regularization term; and the relationship between flow stability and energy output is captured by introducing the Lyapunov exponent.

6. The evaporator energy storage process control method based on fractional-order MIMO nonlinearity according to claim 1, characterized in that: The abnormality types are divided into recoverable abnormality and structural abnormality, and the determination method is as follows: If the fluctuation amplitude is less than 20% of the rated value and the duration is less than 5 seconds, or the correlation coefficient between the pressure oscillation frequency and the preset normal operating condition spectrum is greater than 0.8, it is determined to be a recoverable abnormality; If a certain degree of deviation from the safety threshold continues for more than 30 seconds, or the Mahalanobis distance between the pressure-flow coupling relationship and the historical degradation data is less than the preset threshold, it is judged as a structural anomaly.

7. The evaporator energy storage process control method based on fractional-order MIMO nonlinearity according to claim 6 is characterized in that: After the abnormal stage is identified, it includes an automatic adjustment mechanism to trigger corresponding response measures according to the type and severity of the abnormality; For recoverable anomalies, the fractional-order MIMO nonlinear system model automatically adjusts operating parameters, including input power and fluid flow, and restores the evaporator energy storage process to a normal state by changing the operating parameters; In case of structural anomalies, the fractional-order MIMO nonlinear system model automatically switches to the backup control mode and issues an alarm to notify maintenance personnel to conduct inspection and repair; The automatic adjustment mechanism includes a preset adjustment strategy library, which is formulated based on historical data and expert experience and covers parameter adjustment ranges and control logic under different abnormal situations.

8. An evaporator energy storage process control system based on fractional-order MIMO nonlinearity, using the evaporator energy storage process control method based on fractional-order MIMO nonlinearity as claimed in any one of claims 1 to 7, characterized in that: include: System control module, data acquisition module, data processing module, data control module, data judgment module, abnormal warning module, abnormal type judgment module and adaptive control module; Wherein, the system control module is used to control the start, pause and stop of the system equipment; The data acquisition module is used to obtain real-time data collected by sensors and historical response data in the historical database; The data processing module is used to input the historical response data into the fractional-order MIMO nonlinear system model of the current dryness to obtain dryness prediction data of each dryness stage of the energy storage process; The data control module is used to control the reaction variables of each stage by using the fractional-order MIMO nonlinear system model of each stage to obtain the reaction result data of each stage; The data judgment module is used to compare the reaction result data of each stage with the real-time data; and analyze the reaction result data and the reaction prediction data to obtain dynamic weights; then, the dynamic weights of the current stage are applied to the fractional-order MIMO nonlinear system model of the next stage to constrain the reaction variables in the next reaction stage; The abnormal warning module is used to identify the abnormal reaction stage according to the real-time data; if the abnormal reaction stage is identified, a warning is issued; The abnormality type determination module is used to determine the abnormality type according to the dryness fluctuation amplitude and the pressure oscillation frequency data.

Images