Cryogenic nitrogen making machine cluster control method, system and device and storage medium

Abstract

The present application relates to the technical field of control of deep cooling nitrogen making machine, in particular to a kind of deep cooling nitrogen making machine cluster control method, system, device and storage medium.The steps of control method include: the real-time parameters of air of the air compressor outlet of each deep cooling nitrogen making machine in cluster, the current state parameters of molecular sieve adsorber and pre-cooling machine operating parameter are collected, and the regeneration demand of each molecular sieve adsorber in future preset time period is predicted.The present application predicts future regeneration demand by collecting the operating parameter of each unit in cluster and based on time series model, provides "time window" for the dynamic adjustment of waste heat utilization;On this basis, the dynamic heat contribution value is calculated based on air compressor outlet parameter, and the branch valve opening is adjusted in advance accordingly, the "time matching" and "quantity matching" of waste heat supply and regeneration demand are realized, and the imbalance of heat supply and demand caused by adjustment delay is avoided.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of control of deep cooling nitrogen making machines, in particular to a deep cooling nitrogen making machine cluster control method, system, device and storage medium. BACKGROUND

[0002] As a technical support in the field of industrial gas preparation, the cluster operation of deep cooling nitrogen making machines has become the mainstream direction to meet the demand for large-scale and high-stability nitrogen supply. The deep cooling nitrogen making machines in the cluster realize oxygen-nitrogen separation of air through a molecular sieve adsorber, and the periodic regeneration of the molecular sieve is the key to maintaining the adsorption performance. The regeneration process requires stable high-temperature hot gas flow, and the existing technology generally adopts a combination of air compressor outlet waste heat recovery and electric heater to meet the heat demand, that is, the form of main heat source and supplementary heat source, and the air compressor outlet temperature is regulated by a pre-cooling machine to ensure the adsorption efficiency of the molecular sieve. However, the existing cluster control scheme still has the limitation of superimposing single unit logic, and its defects are the lack of dynamic cooperation at the system level:

[0003] On the one hand, there is a time difference between the prediction of regeneration demand and the real-time adjustment of waste heat utilization, that is, sensor data processing, waste heat calculation and valve adjustment depend on serial execution of the software layer, resulting in delay of adjustment instructions, which cannot accurately match the waste heat supply and demand at the start of regeneration, and often causes problems such as insufficient heat requiring electric heating to supplement energy or excessive heat increasing pre-cooling load.

[0004] On the other hand, the lag response characteristics of the pre-cooling machine are not linked with the valve adjustment, and the misalignment of the control timing of the two will cause fluctuations in the air compressor outlet temperature, which not only reduces the regulation efficiency of the pre-cooling machine, but also affects the regeneration effect of the molecular sieve, resulting in increased energy consumption and unstable adsorption efficiency.

[0005] On the other hand, the existing technology does not optimize the waste heat recovery efficiency, pre-cooling energy consumption and regeneration effect as a system-level target, and often leads to imbalance of overall performance due to optimization of local links. Therefore, the technical problem to be solved by the present application is: how to build a dynamic cooperative control mechanism in the deep cooling nitrogen making machine cluster to achieve the simultaneous achievement of energy consumption reduction and regeneration efficiency improvement. SUMMARY

[0006] The present disclosure proposes a deep cooling nitrogen making machine cluster control method, system, device and storage medium, which aims to overcome at least one defect in the prior art.

[0007] To achieve the above-mentioned purpose, the technical scheme disclosed by the present application is as follows:

[0008] According to one aspect of the present disclosure, a deep cooling nitrogen making machine cluster control method is provided, comprising the steps of:

[0009] Collecting real-time parameters of air at the outlet of the air compressor of each cryogenic nitrogen generator in the cluster, current state parameters of the molecular sieve adsorber, and operating parameters of the pre-cooler, predicting the regeneration requirement of each molecular sieve adsorber in a future preset time period;

[0010] According to the real-time parameters of air at the outlet of the air compressor and the regeneration requirement, calculating a dynamic heat contribution value of waste heat at the outlet of the air compressor, and adjusting the opening degree of the branch valve from the outlet of the air compressor to the molecular sieve electric heater based on the dynamic heat contribution value;

[0011] According to the adjustment amount of the opening degree of the branch valve and the lag response difference of the pre-cooler, adjusting the frequency of the frequency converter of the pre-cooler in linkage, and introducing a lag time compensation factor to offset the lag response difference;

[0012] Based on the real-time operating state and regeneration effect feedback of each cryogenic nitrogen generator in the cluster, dynamically adjusting the linkage relationship between the opening degree of the branch valve and the frequency of the frequency converter of the pre-cooler through a multi-objective optimization model, so as to cooperatively optimize the power consumption of the molecular sieve electric heater and the pre-cooler and the regeneration efficiency of the molecular sieve adsorber.

[0013] Further, the prediction process of the regeneration requirement is as follows: an autoregressive integral moving average model is used to fit the time series of the adsorption saturation degree of each molecular sieve adsorber, and the trend of the adsorption saturation degree in the future T hours is predicted; when the predicted adsorption saturation degree reaches a regeneration threshold, a regeneration requirement signal of the corresponding molecular sieve adsorber is generated, and the regeneration requirement signal includes a regeneration start time, a target temperature, and a required heat for regeneration.

[0014] Further, the dynamic heat contribution value quantifies the coupling time-varying characteristics of the temperature thermal lag effect and the dirt accumulation by changing the upper limit, and the expression is as follows:

[0015] wherein Q contrib (t) is the dynamic heat contribution value at time t, λ f is the equivalent thermal conductivity of the dirt layer, A eff is the effective heat exchange area of the waste heat recovery device, k δ is the dirt deposition reference coefficient, T out (τ) is the air outlet temperature of the air compressor at historical time τ, α is the temperature influence index, and β is the thermal lag attenuation coefficient.

[0016] Further, the calculation method of the lag time compensation factor is as follows:

[0017] The time difference between the frequency converter frequency adjustment time and the pre-cooler outlet temperature stabilization time in the historical operating data of the pre-cooler is collected, and the lag time of the pre-cooler is fitted;

[0018] The compensation factor k is calculated according to the lag timeγ :

[0019] wherein T sample is the parameter acquisition period, and γ is the lag time;

[0020] The compensation factor k γ is multiplied by the frequency adjustment amount of the precooler frequency converter to obtain an advanced adjustment frequency value to offset the adjustment lag of the precooler.

[0021] Further, the multi-objective optimization model adopts a non-dominated sorting genetic algorithm II, and the steps include:

[0022] An optimization target is set, including minimizing the total power consumption P heater of the molecular sieve electric heater, minimizing the total power consumption P cooler of the precooler, and maximizing the adsorption efficiency η adsorb after regeneration of the molecular sieve;

[0023] The branch valve opening θ i of each unit and the frequency f i of the precooler frequency converter are taken as decision variables, and a Pareto frontier solution set is generated through non-dominated sorting and crowded distance calculation;

[0024] The optimal decision variable is selected from the Pareto frontier solution set according to the priority of the cluster, and the operating parameters of each unit are dynamically adjusted.

[0025] Further, it further includes a cluster load transfer control step:

[0026] When the precooler load of any of the cryogenic nitrogen generators is too high, the load amount ΔP cool that needs to be transferred by the cryogenic nitrogen generator is calculated.

[0027] According to the waste heat recovery margin of other cryogenic nitrogen generators in the cluster, the load amount ΔP cool is distributed to N cryogenic nitrogen generators with the largest margin, N≥1.

[0028] The branch valve opening of the distributed cryogenic nitrogen generator is adjusted to increase the waste heat recovery amount, and the frequency of the precooler frequency converter is simultaneously reduced to share the precooler pressure of the high-load unit.

[0029] Further, it further includes a cluster redundancy control step in a fault state:

[0030] When the outlet branch valve of the air compressor of any of the cryogenic nitrogen generators is detected to be faulty, the waste heat recovery branch of the cryogenic nitrogen generator is cut off, and the pure electric heating regeneration mode is switched to;

[0031] calculating the heat required for the regeneration of the cryogenic nitrogen generator, and broadcasting the heat required for the regeneration to other cryogenic nitrogen generators in the cluster;

[0032] According to the waste heat margin of other cryogenic nitrogen generators, the cryogenic nitrogen generator with the largest waste heat margin is selected as the waste heat supply end, and the waste heat of the waste heat supply end is transported to the molecular sieve electric heater of the fault cryogenic nitrogen generator through the cluster waste heat delivery pipe network;

[0033] Adjust the branch valve opening of the waste heat supply end cryogenic nitrogen generator to the maximum, and reduce the frequency converter frequency of the pre-cooler;

[0034] Adjust the electric heater power of the fault unit, and only supplement the heat demand not covered by the waste heat supply end.

[0035] Further, when adjusting the frequency converter frequency of the pre-cooler, an adaptive correction of the ambient temperature is introduced, and the steps include:

[0036] Collect the current ambient temperature, and linearly increase the pre-cooler frequency converter frequency adjustment amount when the current ambient temperature is greater than the preset high temperature environment value;

[0037] When the current ambient temperature is greater than the preset low temperature environment value and less than or equal to the preset high temperature environment value, the original adjustment amount is maintained;

[0038] When the current ambient temperature is less than the preset low temperature environment value, linearly reduce the pre-cooler frequency converter frequency adjustment amount.

[0039] Further, it also includes a correction step of the molecular sieve regeneration effect:

[0040] After the molecular sieve adsorber completes the regeneration, the dew point temperature of the outlet air is collected to reflect the adsorption efficiency;

[0041] If the dew point temperature is higher than the target dew point temperature, calculate the regeneration heat deficiency:

[0042] , Q target is the target heat of regeneration, Q contrib is the dynamic heat contribution value, Q heater is the heat provided by the electric heater; according to the regeneration heat deficiency ΔQ lack , increase the branch valve opening and the electric heater power of the next regeneration until the dew point temperature reaches the target value.

[0043] Further, the step of adjusting the branch valve opening from the air compressor outlet to the molecular sieve electric heater includes:

[0044] According to the predicted regeneration demand start time, advance Δt=2γ+T sampleAdjusting the branch valve opening; linearly adjusting the valve opening from the current value to the target value to ensure that the amount of waste heat recovery is equal to the amount of regeneration demand when the regeneration demand starts.

[0045] Further, the multi-objective optimization model also updates the model parameters using a forgetting factor recursive least squares algorithm, and the steps include:

[0046] Collecting the latest operating data of the cluster every preset period, including the power consumption of the electric heater, the power consumption of the precooler, and the adsorption efficiency of the molecular sieve, and correcting the target weight coefficients in the model using a forgetting factor recursive least squares algorithm to adapt to changes in the operating state of the cluster to stabilize the optimization effect.

[0047] Further, it further includes segmented linkage control of the precooler frequency converter, divides the branch valve opening into multiple intervals, and sets different frequency adjustment strategies for each interval.

[0048] Further, the multiple intervals include a low interval, an intermediate interval, and a high interval; in the low interval, the frequency of the precooler frequency converter is adjusted in a linear proportion, and the frequency change amount is proportional to the valve opening change amount; in the intermediate interval, the frequency of the precooler frequency converter is adjusted in a nonlinear exponential manner, and the frequency change amount decreases at an accelerated rate as the valve opening increases; in the high interval, the frequency of the precooler frequency converter is adjusted in a saturation limiting manner, and the frequency is not lower than a preset percentage of the rated frequency to avoid unit shutdown.

[0049] According to another aspect of the present disclosure, a cryogenic nitrogen generator cluster control system is provided for implementing the cryogenic nitrogen generator cluster control method as described above, comprising:

[0050] A data acquisition module for acquiring air compressor outlet parameters, molecular sieve state parameters, and precooler operating parameters of each unit in the cluster;

[0051] A demand prediction module for predicting future regeneration demand based on a time series model;

[0052] A dynamic waste heat calculation module for calculating the dynamic heat contribution value of the air compressor outlet waste heat;

[0053] A valve adjustment module for adjusting the branch valve opening in advance according to the dynamic waste heat calculation result;

[0054] A precooler lag compensation module and a linkage module for calculating the frequency regulator adjustment amount with compensation according to the branch valve opening adjustment amount and the precooler lag characteristics, and linkage controlling the precooler;

[0055] A multi-objective optimization module for optimizing the power consumption of the electric heater, the power consumption of the precooler, and the regeneration efficiency by a non-dominated sorting genetic algorithm II to dynamically adjust the operating parameters of each unit;

[0056] Cluster load transfer and redundancy module, when the load of a certain machine is too high or fails, it is used to distribute the load to other machines to keep the cluster stable.

[0057] Regeneration effect correction module, for correcting the operating parameters of the next regeneration according to the adsorption efficiency of the regenerated molecular sieve.

[0058] According to another aspect of the present disclosure, a cryogenic nitrogen generator cluster control device is provided, comprising a cryogenic nitrogen generator cluster control system as described above, and further comprising:

[0059] Edge side sensor preprocessing unit, integrated with the air compressor outlet parameter sensor of the data acquisition module, built-in hardware level filtering and temperature-flow correlation feature extraction circuit, pre-processes the high-frequency sampling data of the air compressor outlet and outputs the dynamic coupling feature value to the dynamic waste heat calculation module;

[0060] Time-sensitive network bus interaction unit, connected to the data acquisition module, dynamic waste heat calculation module and valve adjustment module, used to control the transmission delay of the dynamic waste heat result to the valve adjustment instruction;

[0061] Pre-cooler variable frequency drive compensation circuit, connected to the pre-cooler lag compensation module and the frequency converter, built-in hardware PID compensation algorithm, adjusts the frequency of the frequency converter according to the heat dissipation lag time of the pre-cooler to perform hardware level lag compensation;

[0062] Cluster redundancy switching unit, connected to the cluster load transfer module and the unit load sensor, realizes load-redundancy matrix through FPGA hardware logic, and triggers load transfer when the unit load is too high or fails;

[0063] Molecular sieve adsorption efficiency calibration unit, connected to the regeneration effect correction module and the molecular sieve outlet infrared sensor, built-in hardware level gas concentration signal processing circuit, real-time calculates the nitrogen purity-adsorption efficiency correlation value, and outputs the calibrated regeneration parameter threshold to the dynamic waste heat calculation and multi-objective optimization module.

[0064] According to another aspect of the present disclosure, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to realize the steps of the cryogenic nitrogen generator cluster control method as described above.

[0065] The beneficial effects of the present application are:

[0066] The present application solves the technical problem of insufficient dynamic collaboration of the cryogenic nitrogen generator cluster by constructing a full-link collaborative control logic of regeneration demand prediction-dynamic waste heat calculation-advance valve regulation-pre-cooling lag compensation-multi-objective optimization.

[0067] Specifically, the application provides a "time window" for dynamic adjustment of waste heat utilization by collecting operation parameters of each unit in the cluster and predicting future regeneration demand based on a time series model; on this basis, a dynamic heat contribution value is calculated based on outlet parameters of the air compressor, and the branch valve opening degree is adjusted in advance based on the heat contribution value, so that "time matching" and "quantity matching" of waste heat supply and regeneration demand are realized, and imbalance between heat supply and demand caused by adjustment delay is avoided.

[0068] Further, in view of the hysteresis response characteristic of the pre-cooling machine, a hysteresis time compensation factor is introduced to adjust the frequency of the pre-cooling machine frequency converter, so that the time sequence misalignment of valve adjustment and pre-cooling control is eliminated, the outlet temperature of the air compressor is stabilized, and the pre-cooling machine adjustment efficiency is improved.

[0069] Further, by taking the power consumption of the electric heater, the power consumption of the pre-cooling machine and the regeneration efficiency of the molecular sieve as the overall target through a multi-objective optimization model, the operation parameters of each unit are dynamically adjusted, and energy consumption optimization and regeneration effect improvement at the cluster level are realized.

[0070] The control method of the application is not a simple combination of each link, but a demand-supply-regulation collaborative adjustment mechanism is constructed through deep coupling of each step, which not only guarantees the stability of molecular sieve regeneration and continuously meets the adsorption efficiency, but also significantly reduces the energy consumption of the cluster, and optimizes the energy consumption of the electric heater and the pre-cooling machine, thereby providing a systematic technical solution for efficient and stable operation of the deep cold nitrogen generator cluster.

[0071] The above description is only a summary of the technical scheme of the application, in order to more clearly understand the technical means of the application, and the content of the specification can be implemented. The following describes a preferred embodiment of the application with reference to the drawings. BRIEF DESCRIPTION OF DRAWINGS

[0072] Figure 1 The flow chart of the deep cold nitrogen generator cluster control method in an embodiment of the application is shown in the figure.

[0073] Figure 2 The ARIMA prediction contrast graph of the adsorption saturation in an embodiment of the application is shown in the figure.

[0074] Figure 3 The three-dimensional surface graph of the dynamic heat contribution value in an embodiment of the application is shown in the figure.

[0075] Figure 4 The hysteresis response and compensation effect contrast graph of the pre-cooling machine in an embodiment of the application is shown in the figure.

[0076] Figure 5 The thermal map of the dynamic heat contribution value changing with the dirt coefficient in an embodiment of the application is shown in the figure.

[0077] Figure 6Pareto front and decision space diagram for multi-objective optimization in an embodiment of the present application;

[0078] Figure 7 Three-dimensional surface diagram for energy consumption-efficiency collaborative optimization in an embodiment of the present application;

[0079] Figure 8 Precooling lag time and compensation factor relationship diagram in an embodiment of the present application;

[0080] Figure 9 Fault redundancy load transfer thermal diagram in an embodiment of the present application;

[0081] Figure 10 Dew point temperature iteration curve diagram for regeneration effect correction in an embodiment of the present application;

[0082] Figure 11 RLS weight update curve diagram for forgetting factor in an embodiment of the present application. DETAILED DESCRIPTION

[0083] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0084] The term "comprising" and any variation thereof in the specification and claims of the present application is intended to cover not exclusively including, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. In addition, the use of "and / or" in the specification and claims means at least one of the connected objects, for example, A and / or B, means including single A, single B, and three cases of A and B.

[0085] In the embodiments of the present application, the words such as "exemplary" or "for example" are used to mean an example, illustration or description. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the use of "exemplary" or "like", "such as", "for example" and the like approximate words aims to present the relevant concept in a specific way.

[0086] The present application provides the following preferred embodiments:

[0087] Embodiment one: to solve the problem of lack of system-level dynamic coordination in the operation of cryogenic nitrogen generator cluster, the embodiment provides a cryogenic nitrogen generator cluster control method. By building a coupling mechanism of demand prediction, waste heat dynamic calculation, lag compensation and multi-objective optimization, the synergistic optimization of energy consumption and regeneration efficiency is realized, as shown in Figure 1 The flow of the control method is as follows:

[0088] S100: Collecting real-time parameters of air at the outlet of the air compressor of each cryogenic nitrogen generator in the cluster, current state parameters of the molecular sieve adsorber and operating parameters of the pre-cooler, predicting the regeneration demand of each molecular sieve adsorber in a future preset time period.

[0089] S200: According to the real-time parameters of air at the outlet of the air compressor and the regeneration demand, calculating the dynamic heat contribution value of the waste heat at the outlet of the air compressor, and adjusting the opening degree of the branch valve from the outlet of the air compressor to the molecular sieve electric heater based on the dynamic heat contribution value.

[0090] S300: According to the adjustment amount of the branch valve opening degree and the lag response difference of the pre-cooler, the frequency of the pre-cooler frequency converter is adjusted, and the lag time compensation factor is introduced to offset the lag response difference.

[0091] S400: Based on the real-time operating state and regeneration effect feedback of each cryogenic nitrogen generator in the cluster, the linkage relationship between the branch valve opening degree and the pre-cooler frequency converter frequency is dynamically adjusted through a multi-objective optimization model, so as to synergistically optimize the power consumption of the molecular sieve electric heater and the pre-cooler and the regeneration efficiency of the molecular sieve adsorber.

[0092] Specifically, the real-time parameters of air temperature, pressure and flow rate at the outlet of the air compressor of each cryogenic nitrogen generator in the cluster are collected, which directly reflects the recoverable potential of the waste heat of the air compressor; the current adsorption saturation, bed temperature and outlet dew point temperature of the molecular sieve adsorber are collected, which are used to judge the attenuation degree of the adsorption performance of the molecular sieve; the frequency converter frequency, inlet temperature and outlet temperature of the pre-cooler are collected, which reflect the adjustment state of the pre-cooler. Based on the collected time series data of the adsorption saturation of the molecular sieve adsorber, an autoregressive integrated moving average model is used for fitting. The model eliminates the non-stationarity of the data through difference processing, captures random fluctuations through moving average terms, and outputs the future adsorption saturation trend in a preset time period. When the predicted adsorption saturation reaches the preset regeneration threshold, the regeneration demand signal of the corresponding molecular sieve adsorber is generated, which includes the regeneration start time, the target bed temperature and the heat required for regeneration, providing a time node and quantitative demand reference for the subsequent dynamic adjustment of waste heat supply. As shown in Figure 2 Through the fitting of the autoregressive integrated moving average model to the time series of adsorption saturation, the future trend of adsorption saturation can be accurately predicted, ensuring the timeliness and accuracy of the regeneration demand signal.

[0093] Further, according to the real-time parameters of the air compressor outlet air and the regeneration demand, the dynamic heat contribution value of the air compressor outlet waste heat is calculated. The calculation of the dynamic heat contribution value needs to consider the time-varying thermal resistance characteristics of the waste heat recovery device, that is, the coupling effect of dirt accumulation and temperature hysteresis: the equivalent thermal conductivity of the dirt layer gradually decreases with the increase of the running time, the effective heat exchange area decreases with the attachment of dirt, which directly affects the heat transfer efficiency; the temperature influence index reflects the nonlinear enhancement effect of the outlet temperature of the air compressor on heat transfer, the higher the outlet temperature, the more the heat transferred per unit time increases exponentially; the thermal hysteresis decay coefficient reflects the time delay of heat transfer from the air compressor outlet to the molecular sieve electric heater, and the influence of the outlet temperature at the historical moment on the current heat contribution decreases over time. Based on these time-varying parameters, combined with the historical data of the outlet temperature of the air compressor, the dynamic heat contribution value is calculated by variable upper limit integration, and the integral term includes the outlet temperature at the historical moment, the dirt deposition correction coefficient and the thermal hysteresis decay term. The integral result is the available waste heat at the current moment. As shown in Figure 3 , the dynamic heat contribution value presents a three-dimensional surface feature with the change of time and the outlet temperature of the air compressor. The higher the outlet temperature and the shorter the running time, the less the dirt accumulation, and the greater the dynamic heat contribution value. As shown in Figure 5 , the change of the dirt coefficient will significantly affect the dynamic heat contribution value. The longer the running time, the more the dirt accumulation, the lower the equivalent thermal conductivity, the smaller the effective heat exchange area, and the lower the heat transfer efficiency. Therefore, the dirt deposition correction coefficient needs to be introduced in real time in the calculation to ensure the accuracy of the result.

[0094] Further, based on the dynamic heat contribution value, the branch valve opening degree from the air compressor outlet to the molecular sieve electric heater is adjusted in advance: according to the predicted start time of the regeneration demand, the valve opening degree is linearly adjusted from the current value to the target value in advance for a certain period of time, which is determined by combining the pre-cooler lag time and the parameter acquisition period, to ensure that the waste heat recovery amount matches the heat required for regeneration at the moment of the start of the regeneration demand. If the dynamic heat contribution value is less than the heat required for regeneration, the valve opening degree is increased to increase the waste heat recovery amount; if the dynamic heat contribution value is greater than the heat required for regeneration, the valve opening degree is reduced to reduce the waste heat recovery amount, to avoid excess heat increasing the pre-cooling load.

[0095] Furthermore, based on the adjustment amount of the branch valve opening and the hysteresis response difference of the precooler, the frequency of the precooler's inverter is adjusted in a coordinated manner. The hysteresis response difference of the precooler stems from its mechanical and control characteristics. When adjusting the inverter frequency, the compressor speed needs to change gradually, and the flow rate and temperature of the cooling medium also need time to stabilize. Therefore, there is a time difference between the adjustment command and the actual effect. To quantify this time difference, the inverter frequency adjustment time and the precooler outlet temperature stabilization time are collected from the precooler's historical operating data. The time difference between the two is calculated and fitted to obtain the precooler's hysteresis time. A hysteresis time compensation factor is introduced to offset the hysteresis response difference. The compensation factor is 1 plus the ratio of the hysteresis time to the parameter acquisition period. Its physical meaning is to advance the precooler's adjustment command to counteract the hysteresis effect. Figure 4 As shown, the initial response of the precooler exhibits significant lag; after the adjustment command is issued, a lag time constant τ is required to achieve the target effect. However, by introducing a lag time compensation factor, the response time can be significantly shortened. Figure 8 As shown, the compensation factor increases linearly with the increase of the precooling lag time. The longer the lag time, the larger the compensation factor, in order to offset the more significant lag effect. Specifically, the adjustment amount of the branch valve opening is converted into the basic adjustment amount of the precooler inverter frequency. Then, this basic adjustment amount is multiplied by the lag time compensation factor to obtain the pre-adjusted frequency value. When the branch valve opening increases, resulting in an increase in waste heat recovery, the precooler frequency is increased in advance to reduce the air compressor outlet temperature and prevent excessive temperature from affecting the molecular sieve adsorption efficiency. When the branch valve opening decreases, resulting in a decrease in waste heat recovery, the precooler frequency is decreased in advance to avoid overcooling, reduce precooling energy consumption, and ensure that the temperature regulation effect of the precooler reaches the target synchronously when the valve opening adjustment is completed, avoiding air compressor outlet temperature fluctuations caused by misalignment of the control timing.

[0096] Furthermore, based on the real-time operating status and regeneration effect feedback of each cryogenic nitrogen generator within the cluster, an adaptive multi-objective optimization model dynamically adjusts the linkage between the branch valve opening and the precooler inverter frequency. The real-time operating status of the cluster includes the power consumption of the electric heaters of each unit, the power consumption of the precooler, the air compressor outlet temperature, and the molecular sieve bed temperature; the regeneration effect feedback includes the adsorption efficiency of the molecular sieve after regeneration and the uniformity of the bed temperature, where the adsorption efficiency is indirectly reflected by the outlet dew point temperature. The objectives of the multi-objective optimization model are to minimize the total power consumption of the molecular sieve electric heaters, minimize the total power consumption of the precooler, and maximize the adsorption efficiency of the molecular sieve after regeneration. Using the branch valve opening and the precooler inverter frequency of each unit as decision variables, a non-dominated sorting genetic algorithm II is used to solve the problem. This algorithm divides the solutions into different levels through non-dominated sorting, preserving the optimal solution at the Pareto front; it maintains the diversity of solutions through congestion distance calculation, avoiding getting trapped in local optima. Figure 6As shown, the multi-objective optimization Pareto front demonstrates the trade-off relationship between the pre-cooler power consumption, electric heater power consumption, and adsorption efficiency. Each point on the front is the optimal solution under different objectives. Every preset period, the latest cluster operation data is collected, and the forgetting factor recursive least squares algorithm is used to correct the target weight coefficients in the model. The forgetting factor gives higher weight to new data, and the weight of old data decays over time, ensuring that the model adapts to changes in the cluster operating state, such as changes in environmental temperature and efficiency decline due to unit wear. For example Figure 11 As shown, through the updating of the target weight coefficients by the forgetting factor recursive least squares algorithm, the changes in the cluster operating state can be tracked in real time, ensuring the adaptive ability of the model. According to the priority of the cluster, such as energy consumption priority or efficiency priority, the optimal decision variable is selected from the Pareto front solution set: if the current priority of the cluster is energy consumption priority, the solution with the minimum total power consumption of the electric heater and the pre-cooler is selected; if the priority is efficiency priority, the solution with the highest adsorption efficiency is selected, and finally the branch valve opening degree and the pre-cooler frequency of each unit are dynamically adjusted to realize adaptive optimization of the linkage relationship.

[0097] The benefits of this embodiment are that the time and heat demand of regeneration are determined through data collection and regeneration demand prediction, the available waste heat is quantified through dynamic heat contribution value calculation, the precise matching of heat supply in time and quantity is achieved through advanced valve adjustment, the timing synchronization of valve adjustment and pre-cooling adjustment is achieved through pre-cooler lag compensation, and the collaborative optimization of system-level objectives is achieved through adaptive multi-objective optimization. Each step forms a closed loop of demand prediction, heat calculation, valve adjustment, pre-cooling compensation, and optimization adjustment, and the output of each step is the input of the next step, finally building a dynamic collaborative mechanism at the system level to support the efficient operation of the deep cold nitrogen generator cluster.

[0098] Embodiment Two: To solve the problem of not timely triggering of the regeneration demand of the molecular sieve adsorber in the deep cold nitrogen generator cluster and the inaccurate quantification of the regeneration parameters, this embodiment further optimizes the regeneration demand prediction mechanism, and realizes accurate prediction of the adsorption saturation degree and quantitative output of the regeneration demand through a time series model.

[0099] It can be understood that the adsorption saturation is an index reflecting the attenuation of the adsorption performance of the molecular sieve, which presents a trend of rising over time, and is affected by random fluctuations of variables such as air compressor outlet temperature and air humidity. In this embodiment, the ARIMA model is used to fit the time series of the adsorption saturation of each molecular sieve adsorber. The model eliminates the non-stationarity of the time series through difference processing, captures the dependence of the adsorption saturation on the historical value through the autoregressive term, and processes the random error through the moving average term, so as to completely depict the variation law of the adsorption saturation. Based on the fitted model, the trend of the adsorption saturation in the next T hours is output; it should be understood that the value of T hours should be determined in combination with the molecular sieve adsorption period and the cluster regulation response time to ensure the practicality of the prediction results.

[0100] Further, when the predicted adsorption saturation reaches the preset regeneration threshold, the system automatically generates a regeneration demand signal for the corresponding molecular sieve adsorber. The regeneration start time is the time when the predicted curve intersects with the threshold, the target temperature is determined based on the optimal regeneration temperature interval of the molecular sieve adsorbent, and the heat required for regeneration is calculated by the cumulative increment of the adsorption saturation and the heat capacity of the molecular sieve bed. As shown in Figure 2 The predicted curve of the ARIMA model has high consistency with the actual adsorption saturation curve, can accurately capture the rising trend and fluctuation characteristics of the adsorption saturation, and ensures the timeliness and accuracy of the regeneration demand signal.

[0101] The benefit of this embodiment is that the ARIMA model realizes accurate prediction of the adsorption saturation, avoiding the hysteresis of relying on real-time values to trigger regeneration; the quantitative regeneration demand signal provides a clear time node and parameter reference for the dynamic adjustment of subsequent waste heat supply, ensuring the synergy of the regeneration process and the waste heat supply, and supporting the stable operation of the cryogenic nitrogen generator cluster.

[0102] Embodiment three: To solve the problem of quantitative deviation of heat contribution under the coupling effect of temperature hysteresis and fouling accumulation in the waste heat recovery of the cryogenic nitrogen generator cluster, this embodiment further refines the calculation method of the dynamic heat contribution value, and realizes accurate quantification of the heat contribution by using the variable upper limit integral model to depict the time-varying characteristics of the two, which is expressed as follows:

[0103] , wherein Q contrib (t) is the dynamic heat contribution value at time t, λ f is the equivalent thermal conductivity of the fouling layer, A eff is the effective heat exchange area of the waste heat recovery device, k δ is the fouling deposition reference coefficient, T out (τ) is the air compressor outlet temperature at historical time τ, α is the temperature influence index, and β is the thermal hysteresis decay coefficient.

[0104] The waste heat of the cryogenic nitrogen generator cluster mainly comes from the high-temperature air at the outlet of the air compressor. The heat transfer to the molecular sieve regeneration system has a significant thermal hysteresis effect. The heat contribution at the current time not only depends on the current temperature, but also is affected by the temperature at the historical time, and the influence gradually decays over time. At the same time, the heat exchange surface of the waste heat recovery device will produce dirt deposition due to long-term operation, which will hinder heat conduction, resulting in changes in effective heat exchange area and heat conduction efficiency over time. The traditional heat calculation method ignores the coupling time-varying characteristics of the two, and only calculates based on real-time temperature and fixed heat exchange parameters, which is difficult to accurately reflect the actual heat contribution.

[0105] In this embodiment, the variable upper limit integral formula is used to calculate the dynamic heat contribution value, and the integral interval covers all historical τ points from the initial time to the current time t, which means that the dynamic heat contribution value is the cumulative result of the historical temperature influence. The negative β of the exponential term e multiplied by t minus τ power describes the time decay characteristics of the historical temperature influence through the decay coefficient β. The closer the historical temperature to the current time, the greater the influence on the current heat contribution, and vice versa, which is exponentially decaying, consistent with the physical law of thermal hysteresis. The equivalent thermal conductivity λ f of the dirt layer in the formula directly reflects the degree of hindering of heat conduction by dirt. As the running time increases, the accumulation of dirt leads to the decrease of λ f , and thus the heat transfer efficiency; the dirt deposition reference coefficient k δ is related to the rate of dirt accumulation, and the larger k δ , the faster the thickness of the dirt layer increases in the same time, and the faster the decline rate of λ f . The two work together to quantify the time-varying characteristics of dirt accumulation. The effective heat exchange area A eff is adjusted in real time in combination with the dirt deposition situation to ensure that the model responds to the actual state of the heat exchange system. The temperature influence index α is used to describe the nonlinear influence of the outlet temperature of the air compressor on the heat contribution. When the outlet temperature is higher than a certain threshold, the heat contribution increment brought by the temperature rise will gradually slow down, which is consistent with the nonlinear law of heat transfer.

[0106] As shown in Figure 3 , the dynamic heat contribution value presents a continuous surface feature with the change of the historical sequence of the outlet temperature of the air compressor and the dirt coefficient, reflecting the coupling influence of temperature thermal hysteresis and dirt accumulation; as shown in Figure 5 , when the dirt coefficient increases, i.e. the dirt layer becomes thicker, the dynamic heat contribution value shows a regional downward trend, which is consistent with the action logic of λ f in the model. It should be understood that the parameters in the model are not fixed values, but are dynamically updated with the change of running time and working conditions: λ f is corrected by regularly detecting the thickness of the dirt layer, β is adjusted based on the thermal resistance characteristics of the waste heat transfer path, and α is calibrated through the correlation analysis of historical temperature and heat contribution.

[0107] The benefit of the embodiment is that by incorporating the time-varying attenuation of temperature hysteresis and the time-varying obstruction of dirt accumulation into the same quantitative framework through the upper limit integral model, the precise calculation of dynamic heat contribution value is realized; the model not only covers the cumulative impact of historical temperature, but also adjusts the heat transfer efficiency in real time through the dirt-related parameters, providing accurate quantitative basis for the subsequent heat supply and demand matching of the regeneration system, supporting the dynamic optimization of the waste heat utilization of the cryogenic nitrogen machine cluster.

[0108] Embodiment four: To solve the problem of lag between frequency adjustment instruction and outlet temperature stability in the pre-cooler adjustment process of the cryogenic nitrogen machine, the embodiment further optimizes the lag time compensation strategy of the pre-cooler, realizes advance adjustment by quantifying the lag time and calculating the compensation factor, and offsets the impact of adjustment lag on temperature control.

[0109] Specifically, the pre-cooler of the cryogenic nitrogen machine is responsible for cooling the high-temperature air at the outlet of the air compressor, and the frequency regulation of its frequency converter is directly related to the refrigeration output, which in turn affects the outlet temperature; however, due to the heat capacity of the heat exchanger, the flow inertia of the medium and other physical properties, the outlet temperature cannot be stabilized immediately after frequency regulation, and it needs to go through a certain lag time to reach the target value. In the conventional adjustment mode, the system issues a frequency instruction based on the current temperature deviation, and when the temperature stabilizes, the deviation has often been enlarged, and the stability of subsequent molecular sieve adsorption is thus affected. This embodiment fits the lag time based on historical data, and then calculates the compensation factor based on the correlation between the lag time and the collection period, turning the frequency regulation from passive response to advance prediction.

[0110] Specifically, the frequency regulation time of the frequency converter and the outlet temperature stabilization time in the historical operation data of the pre-cooler are collected, the time difference between the two times is calculated, and the lag time γ of the pre-cooler is obtained by fitting these time differences; the fitting process needs to reflect the trend of the lag time changing with the working condition, for example, using linear regression or exponential model to process multiple sets of historical time differences to obtain the γ value that meets the current operating state. Further, combined with the parameter collection period T sample Calculate the compensation factor k γ , the calculation method is:

[0111] ; wherein the parameter collection period T sample is the sampling interval of the system to the pre-cooler state (such as frequency, temperature), which is directly related to the sensing frequency of the system to the state change, that is, T sample is smaller, the system tracks the pre-cooler state more frequently, and a larger compensation factor is needed to realize timely adjustment under the same lag time. Then, multiply the original frequency converter frequency regulation amount by the compensation factor k γ , get the frequency value of advance adjustment, and send this value as the actual adjustment instruction to the pre-cooler frequency converter.

[0112] It's important to understand that the compensation factor works by amplifying the frequency adjustment, thus accelerating the rate of change in the precooler outlet temperature and ensuring the temperature accurately reaches the target value within the lag time. For example, if conventional adjustment requires a frequency change Δf to adjust the temperature from T1 to T2, the temperature needs a time interval γ to stabilize; after compensation, the adjustment frequency becomes k. γ Δf, the rate of temperature change accelerates, and it can stabilize earlier within time γ, thus offsetting the hysteresis effect. For example... Figure 4 As shown, the time it takes for the original response curve to reach 90% of the regulation effect is significantly longer than that of the compensated response curve. The time difference between the two directly reflects the effect of the compensation strategy on offsetting the hysteresis. Figure 8 The three-dimensional surface plot shows the compensation factor k γ The relationship between the lag time γ and the acquisition period Tsample is as follows: when γ increases or Tsample decreases, k γ The trend is increasing, which is consistent with the compensation logic that "the more obvious the lag and the more frequent the sampling, the greater the lead time is required".

[0113] Furthermore, the lag time γ needs to be dynamically updated because the thermal resistance of the precooler changes with operating time. For example, fouling on the heat exchanger surface can increase thermal resistance, thus extending the lag time γ. The system needs to periodically collect the latest adjustment and stability data to refit γ and ensure the compensation factor k is correct. γ Matches the current operating conditions. Parameter acquisition period T sample The settings are based on the system's control accuracy requirements, and a balance needs to be struck between control accuracy and data processing load. sample Too small a value will increase the system's computational load, while too large a value may lead to untimely compensation. Adjustments need to be made based on actual operational needs.

[0114] The advantage of this embodiment is that, by quantifying the correlation between lag time and compensation factor, it achieves the correction of the lead time of the precooler adjustment, enabling the outlet temperature to track the target value more timely; the compensation strategy only corrects the frequency adjustment amount through software algorithm, without changing the hardware structure of the precooler, thus adapting to the upgrade needs of existing cryogenic nitrogen generator clusters; combined with Figure 4 Response comparison and Figure 8 The relationship between the factors changes can be intuitively verified, ensuring the effectiveness and feasibility of the strategy.

[0115] Example 5: To address the synergistic optimization problem among the power consumption of the molecular sieve electric heater, the power consumption of the precooler, and the adsorption efficiency after regeneration in the operation of a cryogenic nitrogen generator cluster, this example further refines the multi-objective optimization strategy based on the non-dominated sorting genetic algorithm II. By setting clear optimization objectives and decision variables, a Pareto front solution set is generated and the optimal operating parameters are selected by matching the cluster priority.

[0116] The stable operation of the cryogenic nitrogen generator cluster depends on the balance of the temperature control of the pre-cooling system, the energy supply of the regeneration system, and the adsorption efficiency of the molecular sieve. The reduction of the frequency of the pre-cooling machine frequency converter can reduce power consumption, but may cause the outlet temperature to rise, increasing the adsorption load of the molecular sieve; the increase of the power consumption of the electric heater can strengthen the regeneration effect, but will increase the operating cost; and the adsorption efficiency of the molecular sieve is directly related to the purity and yield of the product nitrogen. Conventional single-objective optimization often has to sacrifice one thing for another, for example, pursuing low energy consumption may lead to insufficient regeneration and decreased adsorption efficiency, and vice versa. This embodiment incorporates the three conflicting objectives into the same model through a multi-objective optimization framework, and finds a set of solutions that cannot improve one objective without sacrificing another by using the global search capability of the non-dominated sorting genetic algorithm II.

[0117] Specifically, first, three optimization objectives are set: minimizing the total power consumption P heater of the molecular sieve electric heater cooler , minimizing the total power consumption P adsorb of the pre-cooling machine, and maximizing the adsorption efficiency η heater of the molecular sieve after regeneration. These objectives cover the core cost and performance indicators of the cluster operation, and their conflict is reflected in the fact that the increase of the electric heater power consumption can increase the regeneration temperature and strengthen the adsorption capacity of the molecular sieve, but will directly increase the energy consumption; the increase of the pre-cooling machine frequency can reduce the outlet temperature and optimize the state of the air before adsorption, but will increase the power consumption of the pre-cooling system. To quantify these objectives, a correlation model needs to be established based on the historical operation data of the cluster, such as the linear relationship between P cooler and the regeneration temperature, the quadratic function relationship between P adsorb and the frequency of the frequency converter, and the nonlinear relationship between η i and the pre-cooling outlet temperature and the regeneration temperature. These relationships can be intuitively displayed through the energy-efficiency synergy optimization three-dimensional surface graph: the height of the surface represents the adsorption efficiency, the two dimensions of the bottom surface are the power consumptions of the pre-cooling machine and the electric heater, and the inclination direction of the surface reflects the trade-off trend between the objectives, i.e., when the power consumption of the pre-cooling machine increases from 20 kW to 40 kW, the adsorption efficiency increases from 85% to 92%, but the power consumption of the electric heater needs to decrease from 15 kW to 10 kW to maintain the balance of the total energy consumption.

[0118] Further, the decision variables are determined: the branch valve opening degree θ i of each unit and the frequency f i of the pre-cooling machine frequency converterThe branch valve opening controls the air flow distribution into each nitrogen generator, which directly affects the processing load of the single unit; the precooler frequency converter regulates the refrigeration capacity, which determines the outlet air temperature, both of which jointly affect the energy consumption and adsorption efficiency. The decision variables need to satisfy the actual operation constraints: the branch valve opening range is 0 to 100% to avoid unbalanced flow distribution, and the precooler frequency converter frequency range is 20 to 50 Hz to ensure stable operation of the compressor, which ensures that the generated solution has engineering feasibility.

[0119] Further, the non-dominated sorting genetic algorithm II is used to generate the Pareto front solution set. The algorithm first generates an initial population, usually containing 100 to 200 individuals, each corresponding to a set of decision variables (θ i , f i ). Then, non-dominated sorting is performed: the solutions in the population are classified according to the dominance relationship, and the solutions in the same non-dominated front are not dominated by each other, i.e., they cannot improve the other objective value without reducing one objective value. For example, if solution A has P heater = 12 kW, P cooler = 25 kW, and η adsorb = 90%, and solution B has P heater = 15 kW, P cooler = 22 kW, and η adsorb = 88%, then solution A and solution B are not dominated by each other and belong to the same front. Then, the crowding distance is calculated: for each solution in the non-dominated front, the density of the surrounding solutions is calculated, and the larger the crowding distance, the better the diversity of the solution, avoiding the solution set from concentrating in a certain area, such as excessive bias towards low energy consumption or high efficiency. The next generation population is generated by operations such as single-point crossover and mutation such as randomly perturbing 10% of the variable values, and the solutions with large crowding distances in the non-dominated front are retained to ensure that the solution set converges to the Pareto front and maintains diversity.

[0120] Further, after the iteration is terminated, usually for 50 to 100 generations, the final non-dominated front obtained is the Pareto front solution set, which corresponds to the multi-objective optimization Pareto front and decision space diagram of Figure 6 , where the blue points in the diagram constitute the Pareto front, the red ideal point represents the theoretical point with the optimal three objectives, P heater = 10 kW, P cooler = 20 kW, and η adsorb = 95%, and the black best compromise point is the actual optimal solution selected according to the cluster priority.

[0121] Further, the optimal decision variables are selected from the Pareto front solution set according to the operating priority of the cluster. If the cluster is in a working condition with high nitrogen purity requirement, i.e., regeneration efficiency priority, the solution with the maximum η adsorb is selected, such as Figure 6 in which η adsorbFor the point with 93% probability, the corresponding decision variables are increasing the electric heater power consumption to 18kW (raising the regeneration temperature to 200℃) and increasing the precooler frequency to 45Hz (lowering the outlet temperature to 10℃). If it is during peak electricity consumption periods, then energy consumption is prioritized, and option P is chosen. heater With P cooler The solution with the minimum sum, for example, at a point with a total power consumption of 30kW, corresponds to decision variables such as reducing the electric heater power consumption to 10kW and the precooler frequency to 25Hz; if a balance between energy consumption and efficiency is required, then the solution with the maximum congestion distance is chosen, for example... Figure 6 The optimal compromise point in the middle, its P heater =14kW, P cooler =28kW、η adsorb =91%, to ensure that all targets are at a relatively good level.

[0122] It's important to understand that the value of the Pareto front solution set lies in providing all possible trade-offs, while priority matching translates the algorithm output into practically executable runtime parameters. For example, when a cluster needs to improve regeneration efficiency, the system will select η from the Pareto front. adsorb The highest solution corresponds to adjusting the branch valve openings to 80%, 75%, 85%, 70%, and 90% for units 1 to 5 respectively, thus increasing the flow rate of the high-load units. The precooler frequency is adjusted to 42Hz, which reduces the outlet temperature. These parameter adjustments are reflected in… Figure 7 In the three-dimensional surface plot, that is, from the inefficient region (P) of the surface. heater =10kW, P cooler =20kW、η adsorb =85%) towards high-efficiency areas (P heater =18kW, P cooler =45kW、η adsorb =93%) moved.

[0123] The advantage of this embodiment is that it achieves multi-objective collaborative optimization through the non-dominated sorting genetic algorithm II, avoiding the limitations of single-objective optimization; the Pareto front solution set covers all possible trade-off points, providing a flexible parameter selection space for the cluster; combined with Figure 6 and Figure 7 The visualization allows maintenance personnel to intuitively understand the relationships between various objectives and quickly match priorities to select the optimal parameters. Through this embodiment, the adjustment of operating parameters for the cryogenic nitrogen generator cluster has shifted from experience-driven to data-driven, ensuring a balance between energy consumption and efficiency under different operating conditions.

[0124] Embodiment six: To solve the problem of pressure distribution of the pre-cooling machine of a single unit in the cluster of deep cold nitrogen making machines, the embodiment further refines the cluster load transfer control logic, and realizes the pre-cooling pressure dispersion of the high-load unit through the linkage adjustment of the waste heat recovery margin matching and the operating parameters.

[0125] Specifically, when the frequency of the pre-cooling machine frequency converter of a certain deep cold nitrogen making machine exceeds 80% of the rated frequency, the system first calculates the pre-cooling load ΔP that needs to be transferred by the unit cool , which corresponds to the load size that needs to be dispersed to reduce the pre-cooling load of the unit to the safe range. Then the system obtains the waste heat recovery margin of other deep cold nitrogen making machines in the cluster, that is, the difference between the actual recovered heat and the maximum recoverable heat of each unit. The larger the margin, the stronger the ability of the unit to replace pre-cooling with waste heat. Based on this margin distribution, the system distributes ΔP cool to the N units with the largest margin, N≥1, to ensure that the load distribution matches the waste heat utilization capacity of the units.

[0126] Further, for the units allocated with the load, the system first adjusts the branch valve opening degree to increase the air flow into the unit, thereby increasing the waste heat recovery of the unit; and simultaneously reduces the frequency of the pre-cooling machine frequency converter. Because the increase in waste heat recovery can partially replace the refrigeration demand of the pre-cooling machine, the operating load of the pre-cooling machine can be reduced without affecting the air pre-cooling effect. As shown in the load transfer thermodynamic diagram Figure 9 , the system quickly locates the units that can receive the load based on the margin distribution of each unit to realize the linkage of dynamic load distribution and parameter adjustment, wherein the color depth corresponds to the margin size.

[0127] The benefit of the embodiment is that the precise dispersion of the high-load pressure of the pre-cooling machine is realized through the linkage of the waste heat recovery margin and the load transfer, avoiding the equipment wear caused by the long-term high-load operation of a single unit; and the simultaneous adjustment of the branch valve opening degree and the pre-cooling machine frequency ensures the stability of the overall pre-cooling effect of the cluster during the load transfer process, maintaining the consistency of the molecular sieve adsorption efficiency.

[0128] Embodiment seven: To solve the problem of regenerative heat supply of a single air compressor outlet branch valve in the cluster of deep cold nitrogen making machines, the embodiment further refines the redundancy control process under the fault state, and maintains the stability of the regeneration effect of the fault unit through the linkage adjustment of the cluster waste heat sharing and the operating parameters.

[0129] In this fact example, when a valve fault of the air compressor outlet branch of a deep cooling nitrogen generator is detected, the system first cuts off the waste heat recovery branch of the unit, switches to the pure electric heating regeneration mode, and the valve fault may cause abnormal air flow in the waste heat recovery branch. Cutting off the branch can avoid the disturbance of abnormal waste heat input to the stability of the regeneration process. Then the system calculates the heat required for the molecular sieve regeneration of the fault unit, which corresponds to the minimum heat threshold required to maintain the adsorption performance of the molecular sieve, and broadcasts this requirement to all other deep cooling nitrogen generators in the cluster to ensure the targeting of subsequent waste heat supply.

[0130] Further, the system obtains the waste heat margin of other units in the cluster, i.e. the difference between the actual heat recovered by each unit and the maximum recoverable heat, and the larger the margin, the more waste heat the unit can output. Based on this margin distribution, the system selects the unit with the largest margin as the waste heat supply end, as shown in the fault redundancy load transfer thermodynamic diagram. Figure 9 As shown in the fault redundancy load transfer thermodynamic diagram, the color depth directly reflects the margin level of each unit, and the supply end with sufficient waste heat output capacity can be quickly located. After selecting the supply end, the system transfers the waste heat of the supply end to the molecular sieve electric heater of the fault unit through the cluster waste heat delivery pipe network, realizes the cross-unit transfer of waste heat, and fills the heat gap of the fault unit using cluster resources.

[0131] It should be understood that in order to improve the waste heat output capacity of the supply end, the system adjusts the branch valve opening degree to the maximum to increase the air flow into the supply end, i.e. the increase of air flow will increase the waste heat recovery of the unit, thereby providing more waste heat for the fault unit; simultaneously, the frequency converter frequency of the pre-cooler of the supply end is reduced, because the increase of waste heat recovery can partially replace the refrigeration demand of the pre-cooler, ensuring that the air pre-cooling effect of the supply end is not affected while outputting waste heat. For the fault unit, the system adjusts the power of the electric heater to only supplement the heat demand not covered by the waste heat of the supply end, which not only reduces the electric heating energy consumption by using cluster waste heat, but also supplements the difference through electric heating to ensure that the total heat required for regeneration meets the requirements of molecular sieve regeneration.

[0132] The benefits of this embodiment are that through the operation mode switching of the fault unit and the cluster waste heat sharing, the heat supply required for molecular sieve regeneration is maintained in the valve fault state; at the same time, the parameter linkage adjustment of the supply end and the fault unit ensures the efficiency of waste heat utilization and the stability of regeneration effect, avoiding the influence of the fault on the overall nitrogen production performance of the cluster.

[0133] Embodiment Eight: To solve the problem of adjustment deviation caused by environmental temperature fluctuation in the frequency adjustment process of the pre-cooler frequency converter, this embodiment further optimizes the linkage adjustment logic of the pre-cooler frequency, and introduces an adaptive correction mechanism for the environmental temperature.

[0134] Specifically, the system first collects the current ambient temperature, which directly reflects the initial enthalpy state of the air and is a key factor affecting the pre-cooling refrigeration demand. When the current ambient temperature is greater than the preset high-temperature ambient value, it indicates that the initial temperature of the air is high, and the pre-cooling machine needs a larger refrigeration capacity to reduce the air to the target temperature range. At this time, the system linearly increases the pre-cooling machine frequency converter frequency adjustment amount, that is, the linear adjustment continuity matches the gradual influence of the ambient temperature on the refrigeration demand, avoiding the sudden change of the adjustment amount causing the pre-cooling effect to fluctuate. Figure 4 As shown in the pre-cooling machine hysteresis response and compensation effect comparison diagram, the pre-cooling machine response curve after adaptive correction is closer to the ideal trajectory, which verifies the compensation effect of the increase of the adjustment amount on the increase of the refrigeration demand in a high-temperature environment.

[0135] Further, when the current ambient temperature is greater than the preset low-temperature ambient value and less than or equal to the preset high-temperature ambient value, it indicates that the ambient temperature is in a normal range, and the influence of the initial enthalpy of the air on the pre-cooling demand is within the coverage range of the inherent adjustment capability of the pre-cooling machine. The system maintains the original adjustment amount to ensure that the pre-cooling machine operates stably according to the original logic and maintains the consistency of the pre-cooling effect.

[0136] Further, when the current ambient temperature is less than the preset low-temperature ambient value, it indicates that the initial temperature of the air is low, and the pre-cooling machine needs less refrigeration capacity. The system linearly reduces the pre-cooling machine frequency converter frequency adjustment amount, that is, reduces the adjustment amount to avoid energy waste caused by excessive operation of the pre-cooling machine, and prevent excessive refrigeration from causing the air temperature to be lower than the target value, avoiding the efficiency fluctuation of the subsequent molecular sieve adsorption caused by the air temperature being too low; as shown in the pre-cooling hysteresis compensation factor surface diagram, the adjustment trend of the compensation factor decreases with the temperature in a low-temperature environment, which is consistent with the logic of the linear reduction of the adjustment amount in this embodiment, and together ensures the adjustment accuracy of the pre-cooling machine in a low-temperature environment. Figure 8

[0137] It should be understood that the adaptive correction of the ambient temperature is not independent of the original linkage adjustment logic, but is embedded as a supplementary module in the whole process, that is, the system synchronously completes the correction judgment every time the ambient temperature is collected, ensuring that the adjustment amount matches the current environmental conditions in real time. This mechanism cooperates with the original load linkage and hysteresis compensation logic to form a more perfect pre-cooling machine control strategy: the ambient temperature affects the initial enthalpy of the air, the enthalpy determines the pre-cooling demand, and the demand is converted into adaptive adjustment of the adjustment amount, which ultimately points to the stability of the air temperature after pre-cooling.

[0138] The benefit of this embodiment is that through the adaptive correction of the ambient temperature, the pre-cooling machine frequency converter frequency adjustment amount dynamically matches the ambient temperature change, avoiding the effect deviation of the fixed adjustment amount in an extreme temperature; the smoothness of the linear adjustment maintains the stability of the air temperature after pre-cooling, providing stable inlet conditions for the subsequent molecular sieve adsorption, and ensuring the continuity of the cluster nitrogen production process.

[0139] ​Embodiment Nine: To solve the problem of reduced adsorption efficiency caused by insufficient regeneration in the molecular sieve regeneration process, this embodiment further refines the correction mechanism of the molecular sieve regeneration effect, and realizes dynamic adjustment of the regeneration parameters through dew point temperature feedback.

[0140] Specifically, when the molecular sieve adsorber completes the regeneration process, the dew point temperature of the outlet air is immediately collected. The dew point temperature directly reflects the water content of the air, and the core goal of regeneration is to desorb the water adsorbed by the molecular sieve to restore its adsorption activity, so the dew point temperature is a direct representation of the regeneration effect. When the collected dew point temperature is higher than the target dew point temperature, it means that there is still water that has not been completely desorbed in the molecular sieve, and the regeneration is insufficient. At this time, the system calculates the heat deficiency ΔQ lack , the formula is:

[0141] , Q target is the target heat required for regeneration, which is determined based on the molecular sieve adsorption capacity and the current environmental conditions; Q heater is the direct heat provided by the electric heater for the regeneration process; Q contrib is the dynamic heat contribution value, which is the indirect heat input to the regeneration process by air pretreatment through auxiliary modules such as pre-coolers, as shown in the surface of the dynamic heat contribution value with time and outlet temperature in Figure 3 , which clearly presents the real-time calculation logic of Q contrib , ensuring the accuracy of the heat deficiency evaluation.

[0142] Further, according to the calculated ΔQ lack , the system adjusts the parameters of the next regeneration process: on the one hand, it increases the valve opening of the regeneration branch to increase the flow of the regeneration gas, that is, a larger flow can more fully carry away the water desorbed from the surface of the molecular sieve, strengthening the heat and mass transfer efficiency; on the other hand, it increases the power of the electric heater to directly supplement the heat deficiency required for regeneration. It needs to be understood that this adjustment is achieved through iteration: after each regeneration, the parameters of the next time are adjusted according to the dew point temperature feedback until the dew point temperature drops to within the target value. As shown in the regeneration effect correction curve in Figure 10 , the initial dew point temperature is higher than the target, and after two parameter adjustments, the dew point temperature gradually approaches and stabilizes within the target range, clearly presenting the closed-loop optimization process of the correction mechanism, that is, the dew point temperature drop after each adjustment in the curve is directly related to the calculated value of ΔQ lack , which confirms the pertinence of the correction logic.

[0143] It should be noted that the correction mechanism of the regeneration effect is connected with the control logic of other modules in the cluster: the dynamic calculation of Q contrib is based on the heat contribution model of Figure 3 , and the adjustment of the electric heater power is related to theFigure 11 The RLS weight update shown, it is to be noted that the electric heater weight is one of the target weights, and this linkage ensures the coordination of the regeneration parameter adjustment and the modules such as the pre-cooler and the electric heater, and avoids the local adjustment from causing the overall system fluctuation. For example, when the electric heater power is increased, Figure 11 The electric heater weight in the formula is dynamically updated according to the regeneration effect feedback, and the power adjustment is balanced with the overall energy consumption control of the electric heater. Specifically, the RLS algorithm dynamically updates the weight coefficients of the electric heater, the pre-cooler and the regeneration efficiency through the following calculation formula:

[0144] wherein W(n) is the weight vector of the nth iteration, W(n-1) is the weight vector of the (n-1)th iteration, i.e. the weight value of the last iteration; K(n) is the gain matrix of the nth iteration, which is used to reflect the influence degree of the current input variable on the weight adjustment, and the greater the value, the more sensitive the weight is to the error; σ is a forgetting factor, which controls the weight decay of the historical data; ϕ(n) T is the transpose of the input vector of the nth iteration, which contains the dew point temperature deviation, the regeneration heat shortage ΔQ lack and other state variables; ε(n) is the error signal of the nth iteration, i.e. the difference between the actual dew point temperature and the target dew point temperature, which directly represents the deviation between the regeneration effect and the expectation. This weight update logic ensures that the electric heater power adjustment not only responds to the deviation of the regeneration effect, but also avoids the energy consumption fluctuation caused by frequent adjustment, and realizes the coordinated optimization of the regeneration effect and the energy consumption.

[0145] The benefits of the embodiment are that the quantitative evaluation of the regeneration effect is realized through the direct feedback of the dew point temperature, the targeted adjustment of the regeneration parameters is realized in combination with the calculation of the heat shortage, and it is ensured that the molecular sieve can recover sufficient adsorption activity after each regeneration; the iterative logic of dynamic adjustment adapts to the decay of the molecular sieve performance with the use time and the change of the environmental conditions, maintains the long-term stability of the regeneration effect, provides reliable inlet conditions for the subsequent adsorption process, and guarantees the continuity and reliability of the cluster nitrogen production.

[0146] The benefits of the embodiment are that the quantitative evaluation of the regeneration effect is realized through the direct feedback of the dew point temperature, the targeted adjustment of the regeneration parameters is realized in combination with the calculation of the heat shortage, and it is ensured that the molecular sieve can recover sufficient adsorption activity after each regeneration; the iterative logic of dynamic adjustment adapts to the decay of the molecular sieve performance with the use time and the change of the environmental conditions, maintains the long-term stability of the regeneration effect, provides reliable inlet conditions for the subsequent adsorption process, and guarantees the continuity and reliability of the cluster nitrogen production.

[0147] Embodiment ten: To solve the problem that the amount of waste heat recovery and the amount of regeneration demand cannot be matched in real time due to system lag when starting regeneration, this embodiment further refines the lead adjustment logic of the valve at the branch from the air compressor outlet to the electric heater of the molecular sieve, realizes the pre-adjustment of the valve opening degree by determining a reasonable lead time Δt, and ensures the accurate matching of heat supply and demand when starting regeneration.

[0148] Specifically, the calculation logic of the lead time Δt is based on the system lag characteristics and the valve adjustment demand, and the calculation formula is Δt = 2γ + T sample . Wherein γ is the thermal lag time of the precooler, as shown in the precooler lag compensation factor surface Figure 8 , γ represents the response delay of the precooler from receiving the adjustment instruction to outputting the heat change, which is caused by the internal thermal inertia of the precooler and the time loss of fluid transmission; and the value of 2γ covers the double lag of the waste heat recovery system: on the one hand, the response of the precooler itself needs γ time, and on the other hand, the waste heat needs γ time from the air compressor outlet to the electric heater of the molecular sieve and is effectively absorbed by the regeneration system, and the superposition of the double lag determines that 2γ needs to be reserved as a buffer time. sample T is the time for the valve to adjust from the current opening to the target opening linearly, which is based on the mechanical response characteristics of the valve, that is, linear adjustment can avoid the system pressure fluctuation caused by sudden change of opening, which is consistent with the RLS weight update logic shown in Figure 11 , to ensure the stability of the system operation.

[0149] Further, the adjustment process takes the adsorption saturation prediction result shown in Figure 2 as the trigger starting point: when the adsorption saturation prediction value reaches the regeneration threshold, the system outputs the regeneration demand start time, and then starts the valve opening degree adjustment in advance by Δt. During the adjustment process, the valve opening degree linearly transitions from the current value to the target value - the target opening degree is determined by the matching relationship between the regeneration demand and the amount of waste heat recovery, and the regeneration demand is calculated based on the dynamic heat contribution model of Figure 3 . It needs to be explained that Q target is the target heat of regeneration, and the amount of waste heat recovery is one of the components. It needs to be understood that the design of this linear adjustment is to disperse the change of the valve opening degree in the T sample period, to avoid the interference of local parameter mutation on the collaborative operation of the precooler, electric heater and other modules, for example, when the valve opening degree linearly increases, the precooler weight in Figure 11 will be updated smoothly at the same time, to ensure that the precooler output matches the valve opening degree adjustment.

[0150] Further, when the regeneration demand is officially started, the valve has completed the linear adjustment and stabilized at the target opening, at which time the 2γ hysteresis time is just exhausted: the heat output response of the pre-cooler is completed, the residual heat also completes the transfer from the air compressor outlet to the regeneration system, and the residual heat recovery amount just matches the regeneration demand amount. This logic is highly consistent with the pre-cooling hysteresis compensation factor surface of Figure 8 , that is, γ, as the core hysteresis parameter of the pre-cooler, directly participates in the lead time calculation, ensuring that the adjustment logic is targeted to the pre-cooler characteristics; and Figure 2 the adsorption saturation prediction of the pre-cooling hysteresis compensation factor surface of

[0151] The benefit of the embodiment is that by combining the pre-cooler hysteresis characteristics and the valve adjustment time to determine the lead time, the accurate matching of the residual heat recovery amount and the demand amount at the regeneration start is realized, avoiding the insufficient regeneration of the molecular sieve or the waste of energy caused by the imbalance between heat supply and demand; at the same time, the linear adjustment mode ensures the stability of the system operation, unifies the cooperative control logic of the pre-coolers, electric heaters and other modules in the cluster, and maintains the stable and efficient operation of the overall system.

[0152] Embodiment Eleven: To solve the problem that the multi-objective optimization model cannot adapt to the dynamic changes of the cluster operation state due to fixed parameters, the embodiment further refines the parameter update logic of the multi-objective optimization model, and periodically corrects the target weight coefficients by using the forgetting factor recursive least squares algorithm to maintain the stability of the optimization effect.

[0153] Specifically, the core of the multi-objective optimization model is to balance the weight coefficients of the electric heater power consumption, the pre-cooler power consumption and the molecular sieve adsorption efficiency, which directly determine the best compromise point position of the Pareto frontier shown in Figure 6 . When the cluster operation state changes, for example, the power consumption of the pre-cooler of a certain nitrogen generator increases due to the increase of the ambient temperature, or the adsorption efficiency of the molecular sieve slowly decays due to the use time, the fixed weight coefficients will make the decision of the optimization model deviate from the current optimal state. Therefore, the embodiment sets to collect the latest operation data of the cluster every preset period, including the real-time power consumption of the electric heater, the real-time power consumption of the pre-cooler and the current adsorption efficiency of the molecular sieve, and inputs these data into the forgetting factor recursive least squares algorithm to correct the target weight coefficients.

[0154] Further, the design logic of the forgetting factor recursive least squares algorithm is to apply weight decay to the historical data: over time, the influence of old data on model parameters gradually decreases, and newly collected operation data is given a higher weight, so that the model can dynamically track the latest state of the cluster. This logic is highly consistent with the pre-cooling hysteresis compensation factor surface of Figure 11This is intuitively reflected in the three weight curves: the weight of the electric heater, the weight of the precooler, and the weight of the regeneration efficiency. These curves exhibit smooth dynamic changes during the iteration process, with a trend conforming to the exponential decay law. This avoids the interference of sudden weight changes on the system while ensuring the model's rapid response to new states. For example, when the power consumption of the electric heater continuously increases due to grid voltage fluctuations, the electric heater weight coefficient will be adjusted through the algorithm, causing the multi-objective optimization model to focus more on reducing the power consumption of the electric heater in subsequent optimizations, while maintaining a balance with the power consumption of the precooler and the molecular sieve adsorption efficiency. Conversely, when the molecular sieve adsorption efficiency decreases due to insufficient regeneration, the regeneration efficiency weight will be adjusted accordingly, pushing the model optimization direction towards improving regeneration efficiency.

[0155] It is important to understand that this parameter update logic is related to... Figure 7 The energy consumption-efficiency co-optimization surface is highly synergistic: the correction of the weighting coefficients synchronously adjusts the optimal decision space of the surface, ensuring that the optimal compromise point always falls within the balance region of the electric heater power consumption, precooler power consumption, and molecular sieve adsorption efficiency under the current state. For example, when the precooler power consumption weight increases, the optimal compromise point of the co-optimization surface will shift towards a lower precooler power consumption, but always keep the adsorption efficiency within an acceptable range, ensuring the stability of the optimization effect.

[0156] The advantage of this embodiment is that by periodically collecting the latest operating data and using the forgetting factor recursive least squares algorithm to correct the target weight coefficients, the multi-objective optimization model can dynamically track changes in the cluster's operating status, maintain the optimal balance between the power consumption of the electric heater, the power consumption of the precooler, and the molecular sieve adsorption efficiency, ensure the stability of the optimization effect, and avoid deviations in optimization decisions caused by changes in the state.

[0157] Example 12: To address the issue that a single adjustment strategy in the linkage control of the precooler frequency converter and branch valve opening cannot adapt to the dynamic characteristics of the valve across its entire range, this example further optimizes the control logic of the precooler by adopting a segmented linkage control method for the branch valve opening. The valve opening is divided into multiple intervals, and different frequency adjustment strategies are set for each interval.

[0158] The cooling effect of the precooler is directly related to the opening degree of the branch valves. Changes in valve opening alter the airflow entering the precooler, thus affecting its heat load. If a single frequency regulation strategy, such as linear regulation, is used throughout, the rapid frequency change at small valve openings will exacerbate the precooler's thermal hysteresis effect. Figure 4The pre-cooling machine hysteresis response and compensation effect comparison chart shows that linear regulation at small opening degree easily leads to response overshoot, and when the valve is fully opened, the low frequency affects the stable operation of the pre-cooling machine. Therefore, in this embodiment, the valve opening degree is divided into multiple intervals, and each interval corresponds to a frequency regulation logic that adapts to its characteristics: when the valve opening degree is in different intervals, the frequency change law of the frequency converter matches the dynamic characteristics of the valve opening degree, that is, the small opening degree interval focuses on response speed, the medium opening degree interval focuses on non-linear adaptation, and the large opening degree interval focuses on operation protection. This segmented logic and Figure 8 The characteristics of the pre-cooling hysteresis compensation factor surface are coordinated: different opening degree intervals correspond to different regions of the surface, and the frequency regulation strategy and the compensation factor work together to optimize the pre-cooling machine's thermal hysteresis compensation effect.

[0159] The benefit of this embodiment is that by segmenting the branch valve opening degree control, the frequency converter frequency change better adapts to the dynamic characteristics of the valve full range, avoiding the regulation hysteresis or overshoot problem caused by a single strategy, improving the linkage control accuracy of the pre-cooling machine and the valve opening degree, and maintaining the stable operation of the pre-cooling system.

[0160] Embodiment Thirteen: To solve the problem of insufficient adaptability of the segmented linkage control in different intervals in Embodiment Twelve, this embodiment further refines the specific regulation strategy of the three intervals, divides the branch valve opening degree into low, medium and high intervals, and sets different frequency regulation logics for each interval.

[0161] Specifically, the low interval corresponds to the small opening degree range of the valve, at which time the air flow entering the pre-cooling machine is small, and the pre-cooling machine's refrigeration capacity is linearly related to the frequency of the frequency converter. This embodiment uses linear proportional regulation in this interval: the frequency change is proportional to the valve opening degree change, and the frequency increases by a fixed proportion when the valve opening degree increases, ensuring linear matching of the refrigeration capacity and the air flow, that is, this regulation method matches Figure 8 The linear region of the pre-cooling hysteresis compensation factor surface, the linear change of the compensation factor cooperates with the frequency regulation, and the response speed of the pre-cooling under small load is optimized.

[0162] Further, the medium interval corresponds to the medium opening degree range of the valve, at which time the air flow increases, leading to a decrease in the marginal effect of the pre-cooling heat exchange, and linear regulation cannot adapt to the non-linear demand of the thermal load. This embodiment uses non-linear exponential regulation in this interval: the frequency change decreases at an accelerating rate as the valve opening degree increases, and when the valve opening degree increases from the lower limit to the upper limit of the medium interval, the frequency increase rate gradually slows down, so that the refrigeration capacity increase rate matches the thermal load demand. This strategy corresponds to Figure 8 The non-linear region of the surface, the non-linear change of the compensation factor cooperates with the frequency regulation, avoiding energy waste or insufficient pre-cooling under medium load.

[0163] Further, the high interval corresponds to a large valve opening range, at which the air flow is close to full load, and the pre-cooling machine refrigeration load peaks. In this interval, the saturation limit regulation is used in this embodiment: the frequency is not lower than the preset percentage of the rated frequency, and the valve opening exceeds the lower limit of the high interval, the frequency is no longer increased, and is maintained above a certain percentage of the rated frequency. This strategy corresponds to Figure 8 the saturation region of the curved surface, the upper limit of the compensation factor cooperates with the frequency limit to avoid the unit from being unable to maintain operation due to excessively low frequency, and to ensure stable operation under high load.

[0164] The benefits of this embodiment are that, through the differentiated regulation strategies of the three intervals, the frequency regulation is more accurately adapted to the system characteristics under different openings, the low interval ensures the response speed under small load, the middle interval adapts to the nonlinear demand under medium load, and the high interval protects the unit safety, thereby improving the adaptability of the segmented linkage control.

[0165] Embodiment fourteen: To solve the problems of insufficient coordination of each unit, insufficient waste heat utilization, and dynamic load response lag in the operation of a cryogenic nitrogen generator cluster, this embodiment provides a cryogenic nitrogen generator cluster control system, which integrates data acquisition, demand prediction, dynamic waste heat calculation, and multi-module cooperative control through modular design, to realize efficient and stable operation of the cluster.

[0166] Specifically, the data acquisition module acquires, through sensors deployed on each unit of the cluster, the temperature, pressure, and flow parameters of the air compressor outlet, the adsorption saturation of the molecular sieve, the regeneration inlet and outlet temperature and water content, and the frequency converter operating frequency, inlet and outlet medium temperature, and branch valve opening of the pre-cooling machine. All collected data are transmitted to the system core controller through an industrial bus, providing real-time and accurate basic data support for the operation of subsequent modules. The demand prediction module analyzes the collected molecular sieve adsorption saturation historical data based on a time series model, and predicts the regeneration demand time and regeneration load size of each unit within the next 24 hours, combined with the nitrogen demand in the future production plan, such as Figure 2 The adsorption saturation ARIMA prediction comparison chart as shown in the figure can give a regeneration demand warning 30 minutes to 1 hour in advance, providing a time window for subsequent dynamic waste heat utilization and valve regulation.

[0167] Further, the dynamic waste heat calculation module takes the air compressor outlet parameters as input, calculates the dynamic heat contribution value of the air compressor outlet waste heat to the regeneration system based on heat exchange principles and real-time operating conditions, which reflects the proportion of waste heat that can replace the electric heater under the current working condition. Combined with the regeneration demand prediction result, the waste heat utilization strategy is determined, such as Figure 3a three-dimensional surface plot of the dynamic heat contribution value of the air compressor, which is dynamically adjusted with the changes of the outlet temperature, flow rate and molecular sieve regeneration stage, and provides a quantitative basis for the valve adjustment module to utilize the waste heat. The valve adjustment module adjusts the opening degree of the branch valve connected to the air compressor outlet and the regeneration system in advance according to the heat contribution value output by the dynamic waste heat calculation module. When the waste heat contribution value is higher than the threshold value, the valve opening degree is increased in advance to introduce more waste heat. When the waste heat contribution value decreases, the valve is appropriately closed to avoid a shortage of heat in the regeneration system when the waste heat is insufficient. This advance adjustment strategy can shorten the heat response lag time of the regeneration system to less than 50% of the original scheme.

[0168] Further, the pre-cooler lag compensation module and the linkage module receive the opening degree adjustment amount of the valve adjustment module, and simultaneously collect the inlet and outlet temperatures, operating frequency and lag time parameters of the pre-cooler. Based on the pre-cooler lag compensation factor calculation diagram, the pre-cooler lag response and compensation effect comparison diagram and the pre-cooling lag time and compensation factor relationship diagram, the lag compensation factor of the pre-cooler is calculated, and then the frequency adjustment amount of the frequency converter with compensation is obtained, so as to ensure that the refrigerating capacity of the pre-cooler changes synchronously with the adjustment of the branch valve opening degree, and avoid fluctuations in the molecular sieve inlet temperature caused by pre-cooling lag. Figure 4 the pre-cooler lag response and compensation effect comparison diagram and the pre-cooling lag time and compensation factor relationship diagram, the lag compensation factor of the pre-cooler is calculated, and then the frequency adjustment amount of the frequency converter with compensation is obtained, so as to ensure that the refrigerating capacity of the pre-cooler changes synchronously with the adjustment of the branch valve opening degree, and avoid fluctuations in the molecular sieve inlet temperature caused by pre-cooling lag. Figure 8 the pre-cooler lag response and compensation effect comparison diagram and the pre-cooling lag time and compensation factor relationship diagram, the lag compensation factor of the pre-cooler is calculated, and then the frequency adjustment amount of the frequency converter with compensation is obtained, so as to ensure that the refrigerating capacity of the pre-cooler changes synchronously with the adjustment of the branch valve opening degree, and avoid fluctuations in the molecular sieve inlet temperature caused by pre-cooling lag. Figure 6 the pre-cooler lag response and compensation effect comparison diagram and the pre-cooling lag time and compensation factor relationship diagram, the lag compensation factor of the pre-cooler is calculated, and then the frequency adjustment amount of the frequency converter with compensation is obtained, so as to ensure that the refrigerating capacity of the pre-cooler changes synchronously with the adjustment of the branch valve opening degree, and avoid fluctuations in the molecular sieve inlet temperature caused by pre-cooling lag. Figure 7 the pre-cooler lag response and compensation effect comparison diagram and the pre-cooling lag time and compensation factor relationship diagram, the lag compensation factor of the pre-cooler is calculated, and then the frequency adjustment amount of the frequency converter with compensation is obtained, so as to ensure that the refrigerating capacity of the pre-cooler changes synchronously with the adjustment of the branch valve opening degree, and avoid fluctuations in the molecular sieve inlet temperature caused by pre-cooling lag.

[0169] Further, the cluster load transfer and redundancy module monitors the load rate and operating state of each nitrogen generator in real time. When the load rate of a nitrogen generator exceeds 90% or the nitrogen generator is out of order, the module distributes the load of the faulty generator to other normal generators according to the load balancing principle based on the residual capacity and operating efficiency of each unit, so as to Figure 9 the fault redundancy load transfer thermodynamic diagram, the load distribution process is completed within 10 minutes, ensuring that the total nitrogen production of the cluster remains stable and meets the user's demand for nitrogen. The regeneration effect correction module collects the adsorption saturation and dew point temperature parameters after regeneration after each molecular sieve regeneration is completed, evaluates the regeneration effect, and if the adsorption efficiency is lower than the target value, adjusts the electric heater power or regeneration time of the next regeneration according to the regeneration effect correction dew point temperature iterative curve diagram, and gradually improves the regeneration effect through iterative correction to ensure that the molecular sieve always maintains a high adsorption capacity. Figure 10 the fault redundancy load transfer thermodynamic diagram, the load distribution process is completed within 10 minutes, ensuring that the total nitrogen production of the cluster remains stable and meets the user's demand for nitrogen. The regeneration effect correction module collects the adsorption saturation and dew point temperature parameters after regeneration after each molecular sieve regeneration is completed, evaluates the regeneration effect, and if the adsorption efficiency is lower than the target value, adjusts the electric heater power or regeneration time of the next regeneration according to the regeneration effect correction dew point temperature iterative curve diagram, and gradually improves the regeneration effect through iterative correction to ensure that the molecular sieve always maintains a high adsorption capacity.

[0170] The benefits of the embodiment are that the functions of data acquisition, demand prediction, dynamic waste heat utilization, pre-cooling lag compensation, multi-objective optimization, load redundancy and regeneration correction are integrated through modular design, closed-loop cooperation is formed between the modules, the efficiency of cluster operation is realized, the stability under dynamic load is guaranteed, the system energy consumption is reduced through waste heat utilization and multi-objective optimization, and the regeneration efficiency is improved.

[0171] Embodiment Fifteen: To solve the problems of hardware layer data processing delay, insufficient compensation response speed, poor timeliness of redundancy switching and limited accuracy of adsorption efficiency calibration in the control of the cryogenic nitrogen generator cluster, the embodiment provides a control device. On the basis of the cluster control system, hardware and software modules are deeply cooperated through the addition of hardware level functional units, and the real-time performance and control accuracy of the device are improved.

[0172] Specifically, the edge side sensor preprocessing unit is integrated with the air compressor outlet parameter sensor of the data acquisition module, and a hardware level low pass filter circuit and a temperature flow dynamic coupling feature extraction circuit are built-in. The high frequency sampling data of the air compressor outlet temperature and flow are filtered in real time, the noise introduced by electromagnetic interference is removed, and the temperature and flow parameters are converted into dynamic coupling characteristic values reflecting the waste heat capacity through hardware logic, and are directly output to the dynamic waste heat calculation module. This unit avoids the delay of software layer data processing, makes the calculation of dynamic heat contribution value more consistent with the real-time state of the air compressor outlet waste heat, and provides a more reliable input basis for the accurate generation of the dynamic heat contribution value three-dimensional surface graph as shown in the figure. Figure 3

[0173] Further, the time-sensitive network bus interaction unit connects the data acquisition module, the dynamic waste heat calculation module and the valve adjustment module, and uses the time synchronization mechanism and flow scheduling function of the time-sensitive network to control the transmission delay of the dynamic waste heat calculation result to the valve adjustment instruction within milliseconds. This low-delay transmission ensures that the valve adjustment module can execute the advance adjustment strategy based on the latest waste heat contribution value, avoiding the problem that the valve opening adjustment is out of sync with the waste heat change due to data transmission lag.

[0174] Further, the pre-cooling machine variable frequency drive compensation circuit connects the pre-cooling machine lag compensation module and the pre-cooling machine frequency converter, and has a built-in hardware PID compensation algorithm. When the pre-cooling machine lag compensation module outputs the compensation factor, this circuit directly corrects the frequency of the frequency converter according to the pre-cooling machine heat dissipation lag time at the hardware layer, realizing hardware level lag compensation. Compared with software compensation, the parallel processing characteristics of hardware PID greatly improve the compensation response speed, so that the refrigerating capacity change of the pre-cooling machine more accurately matches the branch valve opening adjustment, further optimizing the synchronization of the pre-cooling machine lag response and compensation effect as shown in the figure. Figure 4

[0175] ​​Further, the FPGA-based cluster redundancy switching unit connects the cluster load transfer and redundancy module and the unit load sensor, and realizes the load redundancy matrix through the parallel hardware logic of the FPGA. The unit receives the load rate data of each unit in real time, and when the load of a nitrogen generator is too high or fails, the load distribution instruction is triggered directly through the hardware logic to quickly transfer the load of the failed unit to other normal units. This hardware-level redundancy switching mechanism greatly shortens the load transfer time, ensures Figure 9 The load distribution process in the fault redundancy load transfer thermodynamic diagram shown in the figure can be completed in a shorter time, maintaining the stability of the cluster nitrogen production.

[0176] Further, the molecular sieve adsorption efficiency calibration unit connects the regeneration effect correction module and the molecular sieve outlet infrared sensor, and is provided with a hardware-level gas concentration signal processing circuit. The unit collects the concentration signal of the nitrogen gas at the outlet of the molecular sieve in real time, converts the concentration value into the purity of the nitrogen gas through hardware logic, calculates the correlation value between the purity of the nitrogen gas and the adsorption efficiency, and outputs the calibrated regeneration parameter threshold to the dynamic waste heat calculation module and the multi-objective optimization module. The calibrated threshold enables the dynamic waste heat calculation module to more accurately match the waste heat utilization and regeneration demand, and also enables the parameter adjustment of the multi-objective optimization module to be more in line with the actual adsorption capacity of the molecular sieve, providing Figure 10 The dew point temperature iteration curve for the regeneration effect correction shown in the figure provides a more accurate correction basis.

[0177] The benefits of the present embodiment are that through the cooperative design of hardware units and software modules, the underlying performance of the device is optimized from data preprocessing, transmission delay, compensation response, redundancy switching and efficiency calibration, etc. The real-time, accuracy and stability of the cluster control are further improved, and the linkage effect of the hardware layer and the software layer is strengthened, providing a more solid hardware support for the efficient operation of the cluster of cryogenic nitrogen generators.

[0178] Embodiment 16: To solve the problem of executable and cross-device deployment of the cluster control method of the cryogenic nitrogen generator, the present embodiment provides a computer readable storage medium, which stores a computer program to realize the solidification and reuse of control logic, ensuring that the cluster control method can be consistently executed on different hardware platforms.

[0179] Specifically, the computer program carried by the computer readable storage medium contains the complete logic of the cluster control of the cryogenic nitrogen generator, which, when loaded and executed by the processor, first initializes the communication link with the data acquisition module, the valve adjustment module, the pre-cooler frequency converter and the unit load sensor, and establishes a real-time data transmission channel. Subsequently, the program performs the following steps in turn: receiving the air compressor outlet temperature, pressure, flow, molecular sieve adsorption saturation and regeneration temperature, pre-cooler frequency converter frequency and other parameters transmitted by the data acquisition module, these data constitute the basis for subsequent calculation; running the time series model to analyze the historical data of the molecular sieve adsorption saturation, predicting the regeneration demand time and load size of each unit within the next 24 hours, corresponding to the adsorption saturation prediction result shown in FIG. 6; calculating the dynamic waste heat contribution value based on the air compressor outlet parameters, which directly guides the valve adjustment module to adjust the branch valve opening degree to match the waste heat utilization and the regeneration heat demand; calculating the compensation factor according to the pre-cooler lag time to drive the pre-cooler frequency drive compensation circuit to adjust the frequency converter frequency, corresponding to the pre-cooling lag time and compensation factor relationship shown in FIG. 7; performing multi-objective optimization on the power consumption of the electric heater, the power consumption of the pre-cooler and the regeneration efficiency through the non-dominated sorting genetic algorithm II, outputting the optimal operating parameters, corresponding to the multi-objective optimization Pareto frontier solution set shown in FIG. 8; when the unit load is too high or a fault is detected, triggering the load distribution logic of the cluster load transfer and redundancy module, corresponding to the fault redundancy load transfer thermodynamic diagram shown in FIG. 9; after each regeneration is completed, the electric heater power or time for the next regeneration is corrected according to the correlation value of the molecular sieve outlet nitrogen purity and the adsorption efficiency, corresponding to the dew point temperature iteration curve of the regeneration effect correction shown in FIG. 10. During the entire process, the program realizes instruction issuance and data return through the calling of the hardware interface, ensuring the consistency of the control logic and the hardware execution. Figure 2 Figure 8 Figure 6 Figure 9 Figure 10

[0180] The benefit of the present embodiment is that the cluster control method of the cryogenic nitrogen generator is solidified into an executable program through the computer readable storage medium, avoiding control effect deviation caused by hardware differences, while providing a convenient carrier for updating the control logic, which only needs to replace the program in the storage medium to realize the upgrade of the control strategy, improving the maintainability and expandability of the cluster control system, and making the control method better adapt to the dynamic operation demand of the cryogenic nitrogen generator cluster.

[0181] Although the present application has been described above with reference to the preferred embodiments thereof, it is to be understood that the application is not limited to the embodiments described above but that there are a variety of modifications and variations which can be made thereto without departing from the spirit of the present application. Such modifications and variations are to be considered as falling within the purview of the present application as defined by the appended claims and equivalents thereof.​​​​​

Claims

1. A method for controlling a cluster of cryogenic nitrogen generators, characterized in that the steps include... include: The system collects real-time air compressor outlet parameters, current status parameters of molecular sieve adsorbers, and precooler operating parameters of each cryogenic nitrogen generator in the cluster, and predicts the regeneration requirements of each molecular sieve adsorber within a preset time period. Based on the real-time parameters of the air outlet of the air compressor and the regeneration requirements, the dynamic heat contribution value of the waste heat at the air compressor outlet is calculated, and the opening of the branch valve from the air compressor outlet to the molecular sieve electric heater is adjusted in advance based on the dynamic heat contribution value. Based on the adjustment amount of the branch valve opening and the hysteresis response difference of the precooler, the frequency of the precooler's inverter is adjusted in linkage, and a hysteresis time compensation factor is introduced to offset the hysteresis response difference. Based on the real-time operating status and regeneration effect feedback of each cryogenic nitrogen generator in the cluster, the linkage between the opening degree of the branch valve and the frequency of the precooler inverter is dynamically adjusted through a multi-objective optimization model to synergistically optimize the power consumption of the molecular sieve electric heater and the precooler, as well as the regeneration efficiency of the molecular sieve adsorber.

2. The cryogenic nitrogen generator cluster control method as described in claim 1, characterized in that, The process of predicting the regeneration demand is as follows: the time series of adsorption saturation of each molecular sieve adsorber is fitted using an autoregressive integral moving average model to predict the trend of adsorption saturation change in the next T hours; when the predicted adsorption saturation reaches the regeneration threshold, a regeneration demand signal for the corresponding molecular sieve adsorber is generated, which includes the regeneration start time, target temperature and heat required for regeneration.

3. The cryogenic nitrogen generator cluster control method as described in claim 1, characterized in that, The dynamic heat contribution value is expressed as follows: The time-varying characteristics of the coupling between the temperature thermal hysteresis effect and fouling accumulation are quantified by variable upper limit integral quantification. , where Q contrib (t) represents the dynamic heat contribution value at time t, λ f A is the equivalent thermal conductivity of the fouling layer. eff k represents the effective heat exchange area of ​​the waste heat recovery device. δ T is the baseline coefficient for dirt deposition. out (τ) represents the air compressor outlet temperature at historical time τ, α represents the temperature influence index, β represents the thermal hysteresis decay coefficient, and t represents the current time.

4. The cryogenic nitrogen generator cluster control method as described in claim 1, characterized in that, The lag time compensation factor is calculated as follows: The time difference between the inverter frequency adjustment moment and the precooler outlet temperature stabilization moment in the historical operating data of the precooler is collected, and the lag time of the precooler is obtained by fitting. According to the hysteresis time a compensation factor k is calculated γ : T sample γ is the parameter acquisition period, and γ is the lag time. The compensation factor k γ The compensation factor k is multiplied by the frequency adjustment amount of the pre-cooler frequency converter to obtain the frequency value of the advance adjustment to offset the adjustment lag of the pre-cooler.

5. The cryogenic nitrogen generator cluster control method as described in claim 1, characterized in that, The multi-objective optimization model employs a non-dominated sorting genetic algorithm II, the steps of which include: Setting optimization goals, including minimizing the total power consumption P of the molecular sieve electric heater heater , minimizing the total power consumption P of the pre-cooling machine cooler , and maximizing the adsorption efficiency η after molecular sieve regeneration adsorb ; The branch valve opening θ of each unit i The frequency f of the pre-cooling machine frequency converter i The decision variables, through non-dominated sorting and crowded distance calculation, generate the Pareto frontier solution set; Based on the cluster priority, the optimal decision variables are selected from the Pareto front solution set, and the operating parameters of each unit are dynamically adjusted.

6. The cryogenic nitrogen generator cluster control method as described in claim 1, characterized in that, It also includes cluster load balancing control steps: When the precooler load of any of the aforementioned cryogenic nitrogen generators is too high, calculate the load ΔP that needs to be transferred from the cryogenic nitrogen generator. cool ; Based on the waste heat recovery margin of the other cryogenic nitrogen generators in the cluster, the load ΔP cool Allocate the N cryogenic nitrogen generators with the largest margin, where N≥1; Adjust the branch valve opening of the assigned cryogenic nitrogen generator to increase waste heat recovery, and simultaneously reduce the frequency of its precooler inverter to share the precooling pressure of the high-load unit.

7. The cryogenic nitrogen generator cluster control method as described in claim 1, characterized in that, It also includes cluster redundancy control steps in fault conditions: When a fault is detected in the air compressor outlet branch valve of any of the cryogenic nitrogen generators, the waste heat recovery branch of the cryogenic nitrogen generator is cut off and switched to pure electric heating regeneration mode. Calculate the heat required for regeneration of the cryogenic nitrogen generator and broadcast the heat requirement for regeneration to other cryogenic nitrogen generators in the cluster; Based on the waste heat margin of other cryogenic nitrogen generators, the cryogenic nitrogen generator with the largest waste heat margin is selected as the waste heat supply end, and the waste heat of the waste heat supply end is transported to the molecular sieve electric heater of the faulty cryogenic nitrogen generator through the cluster waste heat transmission pipeline network. Adjust the branch valve opening of the cryogenic nitrogen generator at the waste heat supply end to the maximum and reduce the frequency of its precooler inverter; Adjust the power of the electric heater of the faulty unit to supplement only the heat demand not covered by the waste heat supply end.

8. The cryogenic nitrogen generator cluster control method as described in claim 1, characterized in that, When adjusting the frequency of the inverter in the precooling machine, an adaptive correction based on ambient temperature is introduced, and the steps include: The current ambient temperature is collected, and when the current ambient temperature is greater than the preset high temperature ambient value, the frequency adjustment of the precooler inverter is increased linearly. When the current ambient temperature is greater than the preset low temperature ambient value but less than or equal to the preset high temperature ambient value, the original adjustment amount is maintained. When the current ambient temperature is lower than the preset low temperature ambient value, the frequency adjustment amount of the precooler inverter is linearly reduced.

9. The cryogenic nitrogen generator cluster control method as described in claim 1, characterized in that, It also includes steps to correct the molecular sieve regeneration effect: After the molecular sieve adsorber has been regenerated, the dew point temperature of its outlet air is collected to reflect the adsorption efficiency. If the dew point temperature is higher than the target dew point temperature, then calculate the insufficient regeneration heat: Q target To regenerate the target heat, Q contrib Q represents the dynamic heat contribution value. heater The heat provided to the electric heater; based on the insufficient regenerated heat ΔQ lack Increase the branch valve opening and electric heater power during the next regeneration until the dew point temperature reaches the target value.

10. The cryogenic nitrogen generator cluster control method as described in claim 4, characterized in that, The step of pre-adjusting the opening of the branch valve from the air compressor outlet to the molecular sieve electric heater includes: Based on the predicted start time of regeneration demand, Δt = 2γ + T is made in advance. sample Adjust the opening of the branch valve; linearly adjust the valve opening from the current value to the target value to ensure that the amount of waste heat recovered is equal to the amount of regeneration demand when the regeneration demand is started.

11. The method for controlling a cluster of cryogenic nitrogen generators as described in claim 1, characterized in that, The multi-objective optimization model also employs a forgetting factor recursive least squares algorithm to update model parameters, including the following steps: The latest operating data of the cluster is collected at preset intervals, including the power consumption of electric heaters, power consumption of precoolers and molecular sieve adsorption efficiency. The target weight coefficients in the model are corrected using the forgetting factor recursive least squares algorithm to adapt to changes in the operating status of the cluster and stabilize the optimization effect.

12. The cryogenic nitrogen generator cluster control method as described in claim 1, characterized in that, It also includes segmented linkage control of the precooler frequency converter, which divides the opening degree of the branch valve into multiple intervals and sets different frequency adjustment strategies for each interval.

13. The cryogenic nitrogen generator cluster control method as described in claim 12, characterized in that, Multiple ranges include a low range, a middle range, and a high range. In the low range, the frequency of the precooler inverter is adjusted linearly proportionally, and the frequency change is proportional to the valve opening change. In the middle range, the frequency of the precooler inverter is adjusted nonlinearly, and the frequency change decreases rapidly as the valve opening increases. In the high range, the frequency of the precooler inverter is adjusted according to saturation limits, and the frequency is not lower than a preset percentage of the rated frequency to avoid unit shutdown.

14. A cryogenic nitrogen generator cluster control system, used to implement the cryogenic nitrogen generator cluster control method as described in any one of claims 1-13, characterized in that, include: The data acquisition module is used to collect the air compressor outlet parameters, molecular sieve status parameters, and precooler operating parameters of each unit in the cluster; The demand forecasting module is used to predict future recycling demand based on time series models. The dynamic waste heat calculation module is used to calculate the dynamic heat contribution value of the waste heat at the air compressor outlet. The valve adjustment module is used to adjust the opening of branch valves in advance based on the dynamic waste heat calculation results; The precooler hysteresis compensation module and linkage module are used to calculate the compensated frequency regulation of the frequency converter based on the branch valve opening adjustment amount and the hysteresis characteristics of the precooler, and to control the precooler in a linkage manner. The multi-objective optimization module is used to optimize the power consumption of electric heaters, power consumption of precoolers, and regeneration efficiency through non-dominated sorting genetic algorithm II, and dynamically adjust the operating parameters of each unit. The cluster load transfer and redundancy module is used to distribute the load to other cryogenic nitrogen generators and maintain cluster stability when one of the cryogenic nitrogen generators is overloaded or malfunctions. The regeneration effect correction module is used to adjust the operating parameters for the next regeneration based on the adsorption efficiency of the molecular sieve after regeneration.

15. A cryogenic nitrogen generator cluster control device, comprising the cryogenic nitrogen generator cluster control system as described in claim 14, characterized in that, Also includes: The edge-side sensor preprocessing unit is integrated with the air compressor outlet parameter sensor of the data acquisition module. It has a built-in hardware-level filtering and temperature-flow correlation feature extraction circuit, preprocesses the high-frequency sampling data of the air compressor outlet and outputs the dynamic coupling feature value to the dynamic waste heat calculation module. A time-sensitive network bus interaction unit connects the data acquisition module, the dynamic waste heat calculation module, and the valve adjustment module, and is used to control the transmission delay from the dynamic waste heat result to the valve adjustment command. The precooler inverter drive compensation circuit connects the precooler lag compensation module and the inverter. It has a built-in hardware PID compensation algorithm to correct the inverter frequency according to the precooler heat dissipation lag time in order to perform hardware-level lag compensation. The cluster redundancy switching unit connects the cluster load transfer module and the unit load sensor. It implements the load-redundancy matrix through FPGA hardware logic and triggers load transfer when the unit load is too high or a fault occurs. The molecular sieve adsorption efficiency calibration unit connects the regeneration effect correction module and the molecular sieve outlet infrared sensor. It has a built-in hardware-level gas concentration signal processing circuit to calculate the nitrogen purity-adsorption efficiency correlation value in real time, and outputs the calibrated regeneration parameter threshold to the dynamic waste heat calculation and multi-objective optimization module.

16. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the cryogenic nitrogen generator cluster control method as described in any one of claims 1 to 13.

Images