A self-learning feedforward control method and system for a storage and supply system

Abstract

The application provides a self-learning feedforward control method and system of a storage and supply system, and belongs to the technical field of spacecraft propulsion. The method solves the problem that the compensation of the existing method deviates from the actual demand of the system and cannot guarantee stable supply of the flow in the whole life cycle. The method is as follows: a feedforward compensation branch is connected in parallel on the closed-loop feedback control branch of the proportional valve in the flow control module; during the system operation, after each depressurization module completes the inflation of the buffer tank, the current buffer tank pressure and the proportional valve opening degree output by the closed-loop feedback control branch at the corresponding moment are collected as a group of sample data during the subsequent deflation stage; based on the multiple groups of sample data, the parameters of the feedforward model used in the feedforward compensation branch are identified and updated online; and the feedforward model with the updated parameters is used for feedforward compensation of the opening degree of the proportional valve in the next inflation stage. The method is used in the field of engines.

Description

Technical Field

[0001] This invention belongs to the field of spacecraft propulsion technology, and in particular relates to a self-learning feedforward control method for a storage and supply system. Background Technology

[0002] In the field of spacecraft propulsion technology, the propellant supply system, as the power source, has the core function of providing a precise and stable propellant flow to the engine. The stability of the flow directly determines key performance indicators such as thrust and specific impulse. Large fluctuations in the flow will lead to uneven thrust output and may even cause instability in the propulsion system, seriously threatening the success or failure of spacecraft attitude control and orbit maintenance missions. Therefore, achieving high-precision propellant flow regulation is a cutting-edge key technology for ensuring that modern high-performance spacecraft can complete complex missions.

[0003] Currently, storage and supply systems combining Bang-Bang solenoid valves and proportional control valves are widely used due to their advantages such as simple actuators, fast response speed, and high reliability. The typical operating mode of this type of system is as follows: when the pressure in the buffer tank drops to a threshold due to working fluid consumption, the Bang-Bang valve quickly opens to inflate the tank, causing the pressure to rise rapidly. During the inflatation interval, the proportional valve uses closed-loop feedback control to finely regulate the output flow rate, maintaining downstream pressure stability. However, the discrete switching action of the Bang-Bang valve introduces a step-up disturbance in the buffer tank pressure. This disturbance, transmitted through the pipeline, creates significant "sawtooth wave" pressure fluctuations downstream of the proportional valve. These periodic, severe disturbances are difficult to suppress promptly and completely using only the proportional valve's own feedback control (such as PID control), leading to periodic fluctuations in thrust output and affecting the consistency of propulsion performance.

[0004] To address the aforementioned disturbances, existing technologies have proposed several control schemes. For example, Chinese patent CN114527640A focuses on achieving precise closed-loop control even when pressure sensors are subject to interference; other schemes, such as Chinese patent CN121317132A, aim to integrate and modularize the hardware of the storage and supply system. These existing methods improve the control accuracy or integration of the system to some extent. However, they are generally based on a key premise: that the dynamic characteristics of the controlled system, such as valve flow coefficients, flow channel resistance, and mechanism friction, are constant or change so slowly as to be negligible, and its control performance is highly dependent on a pre-established, precise mathematical model with fixed parameters. In actual long-term on-orbit missions, this premise is difficult to maintain. With the continuous consumption of the working fluid, long-term wear of the valve core, aging of the sealing material, cyclical changes in the temperature environment, and drift of sensor characteristics, the dynamic parameters of the system will undergo slow but not negligible time-varying changes. If the feedforward compensation parameters remain fixed, the compensation effect will gradually deviate from the actual system requirements, causing the ability to suppress sawtooth wave disturbances to decline with increasing mission duration, making it impossible to guarantee a stable flow supply throughout the entire lifecycle. Therefore, there is a significant deficiency in existing technologies: the lack of a feedforward compensation mechanism that can automatically track slow changes in system characteristics and self-update and correct online, so that the Bang-Bang type storage and supply system can adaptively suppress periodic inflation disturbances and maintain constant propulsion performance over a long period of time. Summary of the Invention

[0005] In view of this, the present invention aims to propose a self-learning feedforward control method and system for a storage and supply system, in order to solve the problem that the compensation of existing methods deviates from the actual needs of the system, resulting in the ability to suppress sawtooth wave disturbances declining with the increase of task time, and failing to guarantee a stable supply of flow throughout the entire life cycle.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a self-learning feedforward control method for a storage and supply system, applied to a Bang-Bang type storage and supply system including a primary step-down module and a flow control module, the method comprising: Step S1: In the flow control module, a feedforward compensation branch is set in parallel on the closed-loop feedback control branch of the proportional valve to output the feedforward compensation amount of the proportional valve opening. Step S2: During system operation, whenever the pressure reduction module completes one inflation of the buffer tank, in the subsequent deflation phase, the current buffer tank pressure P and the proportional valve opening U output by the closed-loop feedback control branch at the corresponding moment are collected as a set of sample data. Step S3: Based on multiple sets of sample data collected during a single venting stage, identify and update the parameters of the feedforward model used by the feedforward compensation branch online; Step S4: Use the updated feedforward model to perform feedforward compensation on the opening of the proportional valve in the next inflation stage.

[0007] Furthermore, a preferred approach is proposed, wherein the feedforward model is a linear mapping model between the proportional valve opening U and the reciprocal of the buffer tank pressure 1 / P, expressed as:

[0008] Where a and b are the feedforward parameters to be identified.

[0009] Furthermore, an optimal approach is proposed, which uses the least squares method to perform linear regression on the collected multiple sets of sample data to solve the feedforward parameters a and b online.

[0010] Furthermore, a preferred method is proposed, wherein the condition for triggering sample collection during the venting phase is: the pressure drop of the buffer tank since the last sampling point reaches a dynamic sampling interval.

[0011] Furthermore, a preferred method is proposed, wherein the dynamic sampling interval is dynamically calculated and determined based on the maximum pressure of the buffer tank, the inflation trigger pressure, and the preset number of sampling points during the current inflation cycle.

[0012] Furthermore, a preferred approach is proposed, in which the online identification and parameter update process in step S3 is performed only during the deflation stage and completed before the start of the next inflation stage; during the inflation stage, the parameter update of the feedforward model is frozen.

[0013] Furthermore, a preferred method is proposed: in step S3, the updated feedforward model parameters are replaced with the original parameters in a gradually changing manner and connected to the feedforward compensation branch.

[0014] Based on the same inventive concept, this invention also proposes a self-learning feedforward control system for a storage and supply system, applied to a Bang-Bang type storage and supply system, comprising: The closed-loop feedback control module is used to calculate and output the first opening command of the proportional valve based on the deviation between the downstream pressure and the target value using a PID control algorithm. The feedforward compensation module, which is connected in parallel with the closed-loop feedback control module, is used to calculate the feedforward compensation amount of the proportional valve opening based on the buffer tank pressure. The online learning module is used to collect the buffer tank pressure and the proportional valve opening output by the closed-loop feedback control module at the corresponding moment as sample data during the deflation phase after each inflation. Based on multiple sample data, it identifies and updates the parameters of the feedforward model used by the feedforward compensation module online. The output of the feedforward compensation module and the output of the closed-loop feedback control module are superimposed to serve as the final opening command of the proportional valve.

[0015] Based on the same inventive concept, the present invention also proposes a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes a self-learning feedforward control method for a storage and supply system according to any of the preceding claims.

[0016] Based on the same inventive concept, the present invention also proposes a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of a self-learning feedforward control method for a storage and supply system as described in any of the above-mentioned embodiments.

[0017] Compared with the prior art, the beneficial effects of the present invention are: Unlike existing technologies that rely on one-time ground calibration to obtain fixed parameters, the method proposed in this invention utilizes the actual output data from the system's own periodic steady-state operation phases as learning samples. The principle is as follows: During the deflation phase after inflation, the system is in a stable, closed-loop feedback regulation state. At this time, the proportional valve opening output by the controller is the correct answer after considering all current real-world characteristics such as valve wear, frictional hysteresis, and pressure changes in order to maintain the target flow rate. By collecting the pressure-opening correspondence during this phase, it is equivalent to online sampling of the system's actual fingerprint.

[0018] This invention does not establish a complex, fully physical parameter model. Instead, based on the physical mechanism of fluid throttling, it cleverly constructs a linear mapping relationship between the proportional valve opening and the reciprocal of the buffer tank pressure as a feedforward model. This model is simple in form but has clear physical meaning. More importantly, its parameters can be periodically recursively identified and updated based on the aforementioned online data using a resource-friendly embedded least squares method. Furthermore, it employs dynamic interval sampling based on the pressure drop in the buffer tank, rather than fixed-time or fixed-pressure interval sampling. This ensures that, regardless of the pressure fluctuation range, the sampling points are approximately uniformly distributed within the effective range during each inflation cycle, guaranteeing the representativeness of the samples and the stability of the identification. The parameter learning and updating process is strictly limited to the deflation phase, and learning is frozen during the inflation disturbance phase. This avoids parameter identification during periods of severe system fluctuation, prevents interference between the learning and control processes, and ensures the stability of the system's main control loop. The updated parameters are connected to the feedforward branch using a gradual change rather than a hard switch, avoiding the introduction of new shocks to the control system due to step changes in parameters.

[0019] This invention endows the gas supply system with on-orbit self-optimization capabilities. During mission cycles lasting several years or even longer, the system can autonomously sense and adapt to slowly changing time-varying factors such as valve characteristic drift, component aging, and propellant consumption, dynamically correcting the feedforward compensation. This fundamentally solves the problem of the diminishing effectiveness of fixed-parameter feedforward over mission time, reduces reliance on repeated ground calibration and manual intervention, and significantly improves the gas supply stability and mission reliability of the spacecraft propulsion system throughout the entire mission cycle.

[0020] In the embodiments of the invention, under a target pressure of 12 kPa, the maximum dynamic deviation of the downstream pressure caused by inflation is reduced to 0.3766 kPa, the dynamic deviation ratio is only 3.1%, and the disturbance recovery time is shortened to 0.2 s. In multiple consecutive inflation cycles, the maximum dynamic deviation ratio remains below 5%. This demonstrates that the present invention can effectively anticipate and counteract the sawtooth wave disturbance introduced by the inflation of the Bang-Bang valve, significantly reduce the amplitude of the downstream pressure fluctuation, and accelerate the system recovery speed, thereby providing a more stable propellant supply for the thruster and directly improving the consistency and accuracy of thrust output. Furthermore, since the feedforward parameters are derived from the system's own real-time data, rather than offline calibration results for specific hardware units under specific operating conditions, this method has stronger adaptability to different individual differences, different initial states, and slowly changing environmental conditions such as supply pressure fluctuations. This reduces the stringent dependence of the control system on the accuracy of the hardware model, making the same control algorithm more universally applicable to different storage and supply systems of the same type, reducing the cost of customized development and ground verification.

[0021] The linear model and least squares regression algorithm used in this invention have low computational cost, making them ideal for real-time operation in aerospace embedded controllers with limited computing resources. Its periodic segmented learning mechanism also distributes the computational load, avoiding any impact on the real-time performance of the main control loop. Attached Figure Description

[0022] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a schematic diagram of the Bang-Bang type storage and supply system described in this invention. Figure 2 This is a schematic diagram of the inflation process without feedforward described in this invention; Figure 3 This is a schematic diagram of the inflation interference under self-learning feedforward described in this invention; Figure 4 This is a schematic diagram showing the opening degree of the P2 / P3 proportional valve under multiple cycles of normal mode self-learning feedforward as described in this invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of the present invention can be combined with each other, and the described embodiments are only some embodiments of the present invention, not all embodiments.

[0024] Implementation Method 1, see Figure 1 This embodiment describes a self-learning feedforward control method for a storage and supply system. This method addresses the problem that existing methods, due to their compensation deviating from the actual system requirements, experience a decline in the ability to suppress sawtooth wave disturbances as the task duration increases, thus failing to guarantee a stable flow supply throughout the entire lifecycle. The method is applied to a Bang-Bang type storage and supply system including a primary step-down module and a flow control module. The method includes: Step S1: In the flow control module, a feedforward compensation branch is set in parallel on the closed-loop feedback control branch of the proportional valve to output the feedforward compensation amount of the proportional valve opening. Step S2: During system operation, whenever the pressure reduction module completes one inflation of the buffer tank, in the subsequent deflation phase, the current buffer tank pressure P and the proportional valve opening U output by the closed-loop feedback control branch at the corresponding moment are collected as a set of sample data. Step S3: Based on multiple sets of sample data collected during a single venting stage, identify and update the parameters of the feedforward model used by the feedforward compensation branch online; Step S4: Use the updated feedforward model to perform feedforward compensation on the opening of the proportional valve in the next inflation stage.

[0025] In this embodiment, the feedforward model is a linear mapping model between the proportional valve opening U and the reciprocal of the buffer tank pressure 1 / P, and its expression is:

[0026] Where a and b are the feedforward parameters to be identified.

[0027] In this embodiment, the least squares method is used to perform linear regression on the collected multiple sets of sample data in order to solve the feedforward parameters a and b online.

[0028] In step S2 of this embodiment, the condition for triggering sample collection during the venting stage is that the pressure in the buffer tank decreases by a dynamic sampling interval after the last sampling point.

[0029] Furthermore, the dynamic sampling interval is dynamically calculated and determined based on the maximum pressure of the buffer tank, the inflation trigger pressure, and the preset number of sampling points during the current inflation cycle.

[0030] In step S3 of this embodiment, the online identification and parameter update process is only performed during the deflation stage and is completed before the start of the next inflation stage; during the inflation stage, the parameter update of the feedforward model is frozen.

[0031] In step S3 of this embodiment, the updated feedforward model parameters are replaced with the original parameters in a gradually changing manner and connected to the feedforward compensation branch.

[0032] The Bang-Bang type storage and supply system described in this embodiment is mainly divided into two functional modules: a primary pressure reduction module and a flow control module, which are connected sequentially through pipelines.

[0033] Primary step-down module (high-pressure side): includes a high-pressure gas cylinder, two switching solenoid valves (i.e., Bang-Bang valves), and a small gas chamber.

[0034] like Figure 1 As shown, the outlet of the high-pressure gas cylinder is connected to the inlet of the first Bang-Bang valve via a pipeline. The outlet of this valve is then connected to the inlet of the small gas chamber. The outlet of the small gas chamber is then connected to the inlet of the buffer tank via a second Bang-Bang valve. This module is responsible for the initial depressurization and quantitative release of the gas stored in the high-pressure gas cylinder. The two Bang-Bang valves work together to control the high-pressure gas to enter the small gas chamber in batches and pulses, and then be injected into the buffer tank, realizing one depressurization and filling process.

[0035] Flow control module (low-pressure regulating side): This module includes a proportional valve and a downstream pressure sensor, forming a closed-loop feedback control unit. The outlet of the buffer tank is connected to the inlet of the proportional valve. The outlet of the proportional valve is the system's propellant output port, supplying the downstream thruster. The downstream pressure sensor accurately measures the pressure at the proportional valve outlet. This module is responsible for the accurate and stable control of the final flow rate. The proportional valve continuously adjusts its opening according to control commands. The downstream pressure sensor feeds back the measured pressure signal to the controller, comparing it with the target pressure value to form a closed-loop control, maintaining a constant output flow rate or pressure.

[0036] During normal system operation, the proportional valve in the flow control module remains in closed-loop regulation. When the pressure in the buffer tank drops to a preset inflation trigger threshold due to the consumption of the working fluid, the Bang-Bang valve in the primary pressure reduction module opens according to the command. High-pressure gas is rapidly released into the buffer tank through the small gas chamber, causing its pressure to rise rapidly—this is the inflation process. Because the opening and closing action of the Bang-Bang valve is discrete and abrupt, this inflation process introduces a step-like pressure disturbance, i.e., a sawtooth wave disturbance, into the buffer tank and downstream pipeline. This disturbance is transmitted downstream of the proportional valve, affecting the stability of the output pressure. The self-learning feedforward control method of this invention is precisely designed to compensate for the proportional valve opening in advance at this time to counteract this disturbance.

[0037] Implementation Method 2, see below Figures 2 to 4 This embodiment describes a complete implementation process of the self-learning feedforward control method for a storage and supply system described in Embodiment 1, including: A self-learning feedforward control method for a gas storage and supply system is proposed to address the problems of existing fixed-parameter feedforward systems, which struggle to adapt to long-term time-varying characteristics and continuously suppress inflation disturbances. This method, applied to a Bang-Bang type gas storage and supply system, introduces an online-updable feedforward compensation branch based on the original closed-loop feedback control of the proportional valve. By extracting the system's current operating data during the deflation phase after each inflation cycle, the mapping relationship between the buffer tank pressure and the proportional valve opening is identified online. The identification results are then used for advance compensation of subsequent inflation disturbances, thereby reducing downstream pressure fluctuations and improving gas supply stability and thrust consistency.

[0038] This embodiment addresses a Bang-Bang type gas supply system comprising a primary pressure reduction module and a flow control module. The primary pressure reduction module includes a high-pressure gas cylinder, two solenoid valves, and a small gas chamber. The flow control module includes a proportional valve and a feedback control unit consisting of a downstream pressure sensor. During system operation, the proportional valve operates in a closed-loop regulation state. When the working fluid in the buffer tank is insufficient and the pressure drops to the inflation trigger threshold, the Bang-Bang valve opens, releasing high-pressure gas into the buffer tank through the small gas chamber, causing a rapid increase in the buffer tank pressure. Because this inflation process exhibits discrete and abrupt changes, it introduces significant sawtooth-wave disturbances at the downstream pressure, which are difficult to offset promptly using only feedback control, thus affecting the downstream flow rate and thruster performance stability.

[0039] Unlike existing feedforward methods based on fixed empirical parameters, the key protection of this implementation lies in the fact that the feedforward compensation amount is not fixed for a long time based on a one-time offline calibration result, but is updated online based on the actual system data collected in each operating cycle. Because actual systems experience factors such as valve core wear, dead zone changes, friction and hysteresis changes, sensor aging, pressure environment changes, and working fluid consumption during long-term operation, fixed parameter feedforward is only effective near the calibration conditions and cannot continuously match the true characteristics of the system. This invention, through online self-learning, enables the feedforward parameters to be synchronously corrected as the system state slowly evolves, thereby achieving adaptive compensation for the entire lifecycle of operating conditions.

[0040] This implementation employs a dual-branch control architecture combining closed-loop feedback control and self-learning feedforward compensation. The original feedback control branch of the proportional valve maintains the target pressure or flow rate based on the downstream pressure deviation; the feedforward compensation branch outputs a proportional valve opening compensation amount corresponding to the inflation disturbance based on the pressure state inside the buffer tank. The outputs of the two branches are superimposed at the proportional valve control input. Thus, when a disturbance is caused by inflation of the Bang-Bang valve, the feedforward branch can quickly correct the proportional valve opening before or simultaneously with the feedback adjustment action, thereby reducing the peak value of the downstream pressure deviation and shortening the recovery time.

[0041] Furthermore, in this embodiment, the feedforward compensation branch is always connected in parallel to the PID control module. Once activated, the feedforward branch continuously participates in system operation, but its internal parameters are not updated unconditionally in real time. Instead, they are sampled, identified, and replaced at specific stages according to preset triggering logic. This structural arrangement maintains the continuity of feedforward compensation while avoiding parameter learning during severe system disturbances, thus reducing the risk of coupling between the identification and control processes.

[0042] This implementation considers the proportional valve opening output by the closed-loop controller during the venting phase as a representation of the system's current true comprehensive characteristics, and uses the data from this phase to reconstruct the feedforward mapping relationship online. Within each operating cycle, the buffer tank pressure curve includes both a rising edge during inflation and a falling edge during venting. During the venting phase after inflation, if the closed-loop feedback control is in a normal and stable operating state, the PID controller will automatically provide the proportional valve opening required to maintain flow stability based on the current true system characteristics. This opening information is not simply a theoretical value, but rather incorporates multiple factors such as valve flow characteristics, friction and hysteresis, component wear and aging, and changes in the pressure supply environment.

[0043] Therefore, this implementation does not rely on establishing an additional complex full-parameter physical model. Instead, it directly utilizes the actual correspondence between the buffer tank pressure and the proportional valve opening during the closed-loop stabilization phase as the basis for online correction of the feedforward model. In other words, the control output given by the closed-loop controller during the venting phase is reused in this implementation as an identification result of the current system state, and then applied in advance to the next venting disturbance phase through the feedforward branch. This technical approach ensures continuous consistency between the feedforward model and the real system, thereby improving the effectiveness of feedforward compensation.

[0044] This implementation establishes a linear mapping relationship between the proportional valve opening U and the reciprocal 1 / P of the buffer tank pressure P. The feedforward model is expressed as:

[0045] Where U is the proportional valve opening command, P is the buffer tank pressure, and a and b are the feedforward parameters to be identified.

[0046] The model form described above is not arbitrarily chosen, but rather based on the physical mechanism of the fluid throttling process. According to the orifice plate flow relationship, under the condition that the output flow rate Q is essentially constant, the effective opening area of ​​the proportional valve... The pressure difference, related to the valve core position, can be characterized by the proportional valve opening. However, assuming negligible changes in ambient temperature, gas density, and flow coefficient, the throttling pressure difference is primarily determined by the buffer tank pressure. Therefore, to maintain the same outlet flow rate, the proportional valve opening and the upstream pressure are inversely related; that is, the lower the buffer tank pressure, the larger the valve opening required to maintain the same flow rate. Thus, using the reciprocal of the pressure as the fitting variable is more consistent with the system mechanism and more conducive to improving the accuracy of the feedforward mapping than directly using the pressure value itself.

[0047] This implementation does not simply perform a linear regression, but rather preserves the online mapping relationship between the reciprocal of the buffer tank pressure and the proportional valve opening in the Bang-Bang type storage and supply system as a feedforward model, and continuously updates the model parameters using periodic operating data.

[0048] This implementation provides a set of initial parameters for the feedforward model during the system startup phase. These initial parameters can be obtained from ground calibration or estimated from historical operating data. After the system starts running, the initial parameters first participate in feedforward compensation, enabling the controller to have basic feedforward functions. Subsequently, during subsequent operation, the feedforward parameters are gradually corrected by the online learning module, so that the initial parameters serve only as a startup reference rather than long-term fixed values.

[0049] In practical operation, this implementation uses each Bang-Bang inflation event as the boundary of a learning cycle: after an inflation process ends, the system enters the deflation phase, at which point data acquisition begins; after sufficient samples are collected, parameter identification is performed; after the parameters are updated, sampling is paused until the next inflation cycle is completed before starting a new round of sampling and updating. This periodically segmented execution method allows the self-learning process to be stably embedded into the operation of the storage and supply system, avoiding direct overlap between the learning process and the intense inflation disturbance phase.

[0050] The method proposed in this embodiment collects sample data only during the deflation phase after inflation, rather than continuously sampling at any time. This is because the inflation phase is a highly volatile period, where both the buffer tank pressure and the downstream valve pressure may experience rapid changes. Directly using this phase for parameter identification could easily introduce transient shocks and unstable control output errors into the feedforward model, leading to fitting distortion. In contrast, the deflation phase, under closed-loop regulation, is closer to a quasi-steady state, and the recorded pressure-opening relationship better represents the system's current, realistic, and reproducible operating characteristics.

[0051] Regarding the sampling content, each sampling records at least two types of data: the current pressure P inside the buffer tank and the proportional valve opening U output by the PID controller at that moment. The pressure value serves as the basic data for fitting the independent variable, and the opening value serves as the dependent variable. Since the buffer tank pressure signal has better smoothness and stability than the proportional valve opening signal, this invention preferably uses the pressure change amplitude rather than a fixed time interval or opening change amplitude as the basis for triggering a new sampling point.

[0052] To ensure that online learning samples are reasonably distributed within the effective range, this implementation design employs a dynamic sampling triggering mechanism based on pressure drop. Specifically, after each inflation cycle, the highest pressure Ptop of the buffer tank after inflation is first recorded. Once the system enters the deflation phase, a new sampling is triggered whenever the pressure in the buffer tank drops by a dynamic sampling interval ΔPdyn relative to the previous sampling point, and P and U at that moment are recorded.

[0053] In this embodiment, ΔPdyn is not a fixed value, but is dynamically calculated based on the operating status of the current cycle. Its calculation is based on at least the following parameters: the highest pressure reached after inflation, Ptop; the pressure at which inflation is triggered, Pdown; the preset number of required sampling points, m; and the sampling margin, n. By introducing these parameters, the sampling points can be distributed approximately equidistantly within the current effective deflation range. This avoids insufficient sampling points for effective fitting when the pressure span is small, and also avoids excessive concentration of sampling points in the high-pressure area while leaving insufficient samples in the low-pressure area when the pressure span is large. Simultaneously, by setting the sampling margin, coupling with the next inflation event can be avoided due to insufficient sampling and updating time. Compared to fixed-time sampling or fixed-pressure interval sampling, this invention adaptively determines the sampling interval based on the actual pressure span of each inflation cycle, making the sample distribution more adaptable to the current operating conditions, thereby improving the stability and repeatability of the online identification results.

[0054] To ensure the feasibility of least squares fitting, this embodiment requires at least two sampling points for each round of parameter updates. Considering that too few sampling points make the fitting results susceptible to single-point fluctuations and result in insufficient feedforward accuracy, while too many sampling points increase the storage and computational burden on the controller, this embodiment preferably sets the preset number of sampling points m to 15. The main calculation principle is as follows:

[0055]

[0056] Where a and b are fitting parameters, U is the given valve opening information, and P is the pressure in the buffer chamber at this valve position.

[0057] The significance of using least squares regression lies in its low computational complexity, suitability for embedded implementation, and ability to quickly obtain parameter results even with a limited number of sample points. This makes it suitable for scenarios like Bang-Bang type storage and supply systems that require periodic online updates but are constrained by controller resources. Therefore, this implementation not only proposes an online update method but also provides a practically deployable parameter solution path on the controller.

[0058] In this embodiment, the new parameters identified online do not instantly replace the old parameters with a hard switch. Instead, a gradual intervention method is used to update the feedforward parameters. This is because the feedforward branch is directly connected in parallel to the PID module. If the old and new feedforward parameters switch abruptly at a certain moment, it is equivalent to injecting an additional step disturbance into the proportional valve control input, which may disrupt the stability of the original closed-loop system and create a transfer shock. Therefore, this embodiment requires that the feedforward parameter update be performed away from the inflation disturbance phase, and the updated feedforward compensation is gradually integrated into the control system in a gradual manner.

[0059] By employing the method proposed in this embodiment, the feedforward branch can continuously correct its compensation model based on the actual operating state of the system. This allows for timely correction of the proportional valve opening when the pressure in the buffer tank surges due to Bang-Bang valve inflation, reducing the dynamic deviation of the downstream pressure and shortening the recovery time. Experimental results show that, with a target pressure set at 12 kPa, the maximum dynamic deviation of the actual downstream pressure output during inflation using the self-learning feedforward method of this invention is 0.3766 kPa, the maximum dynamic deviation ratio is 3.1%, and the disturbance recovery time is 0.2 s. In multiple consecutive inflation cycles, the maximum dynamic deviation ratio does not exceed 5%, indicating that this invention can effectively suppress inflation interference and achieve stable gas supply.

[0060] Implementation Method 3: This implementation method proposes a self-learning feedforward control system for a storage and supply system, applied to a Bang-Bang type storage and supply system, characterized by comprising: The closed-loop feedback control module is used to calculate and output the first opening command of the proportional valve based on the deviation between the downstream pressure and the target value using a PID control algorithm. The feedforward compensation module, which is connected in parallel with the closed-loop feedback control module, is used to calculate the feedforward compensation amount of the proportional valve opening based on the buffer tank pressure. The online learning module is used to collect the buffer tank pressure and the proportional valve opening output by the closed-loop feedback control module at the corresponding moment as sample data during the deflation phase after each inflation. Based on multiple sample data, it identifies and updates the parameters of the feedforward model used by the feedforward compensation module online. The output of the feedforward compensation module and the output of the closed-loop feedback control module are superimposed to serve as the final opening command of the proportional valve.

[0061] Implementation Method 4: This implementation method proposes a computer device, including a memory and a processor. The memory stores a computer program. When the processor runs the computer program stored in the memory, the processor executes a self-learning feedforward control method for a storage and supply system according to any one of Implementation Methods 1 to 2.

[0062] Implementation Method 5: This implementation method proposes a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of a self-learning feedforward control method for a storage and supply system as described in any one of Implementation Methods 1 to 2.

[0063] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0064] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0065] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this disclosure and not to limit its protection scope. Although this disclosure has been described in detail with reference to the above embodiments, those skilled in the art should understand that after reading this disclosure, they can still make various changes, modifications or equivalent substitutions to the specific implementation of the invention, but these changes, modifications or equivalent substitutions are all within the protection scope of the pending claims.

Claims

1. A self-learning feedforward control method for a storage and supply system, applied to a Bang-Bang type storage and supply system including a primary step-down module and a flow control module, characterized in that, The method includes: Step S1: In the flow control module, a feedforward compensation branch is set in parallel on the closed-loop feedback control branch of the proportional valve to output the feedforward compensation amount of the proportional valve opening. Step S2: During system operation, whenever the pressure reduction module completes one inflation of the buffer tank, in the subsequent deflation phase, the current buffer tank pressure P and the proportional valve opening U output by the closed-loop feedback control branch at the corresponding moment are collected as a set of sample data. Step S3: Based on multiple sets of sample data collected during a single venting stage, identify and update the parameters of the feedforward model used by the feedforward compensation branch online; Step S4: Use the updated feedforward model to perform feedforward compensation on the opening of the proportional valve in the next inflation stage.

2. The self-learning feedforward control method for a storage and supply system according to claim 1, characterized in that, The feedforward model is a linear mapping model between the proportional valve opening U and the reciprocal of the buffer tank pressure 1 / P, and its expression is: Where a and b are the feedforward parameters to be identified.

3. The self-learning feedforward control method for a storage and supply system according to claim 2, characterized in that, The least squares method is used to perform linear regression on the collected sample data to solve the feedforward parameters a and b online.

4. The self-learning feedforward control method for a storage and supply system according to claim 1, characterized in that, In step S2, the condition for triggering sample collection during the venting stage is that the pressure in the buffer tank decreases by a dynamic sampling interval after the last sampling point.

5. The self-learning feedforward control method for a storage and supply system according to claim 4, characterized in that, The dynamic sampling interval is dynamically calculated and determined based on the maximum pressure of the buffer tank, the inflation trigger pressure, and the preset number of sampling points during the current inflation cycle.

6. The self-learning feedforward control method for a storage and supply system according to claim 1, characterized in that, In step S3, the online identification and parameter update process is performed only during the deflation stage and is completed before the start of the next inflation stage; during the inflation stage, the parameter update of the feedforward model is frozen.

7. The self-learning feedforward control method for a storage and supply system according to claim 1, characterized in that, In step S3, the updated feedforward model parameters are replaced with the original parameters in a gradually changing manner and connected to the feedforward compensation branch.

8. A self-learning feedforward control system for a storage and supply system, applied to a Bang-Bang type storage and supply system, characterized in that, include: The closed-loop feedback control module is used to calculate and output the first opening command of the proportional valve based on the deviation between the downstream pressure and the target value using a PID control algorithm. The feedforward compensation module is connected in parallel with the closed-loop feedback control module and is used to calculate the feedforward compensation amount of the proportional valve opening based on the buffer tank pressure. The online learning module is used to collect the buffer tank pressure and the proportional valve opening output by the closed-loop feedback control module at the corresponding moment as sample data during the deflation phase after each inflation. Based on multiple sample data, it identifies and updates the parameters of the feedforward model used by the feedforward compensation module online. The output of the feedforward compensation module and the output of the closed-loop feedback control module are superimposed to serve as the final opening command of the proportional valve.

9. A computer device, characterized in that: The system includes a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes a self-learning feedforward control method for a storage and supply system according to any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of a self-learning feedforward control method for a storage and supply system as described in any one of claims 1-8.

Images