Speed control method, control system and refrigeration device
By combining the LADRC controller with reinforcement learning algorithms, the compressor control parameters are optimized, which solves the shortcomings of PI control at low speeds of refrigerator compressors, achieves more stable speed and torque control, and improves operational stability and user experience.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- MIDEA BIOMEDICAL CO LTD
- Filing Date
- 2026-05-09
- Publication Date
- 2026-07-14
Smart Images
Figure CN122170578B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of refrigerator control technology, and in particular to a speed control method, control system and refrigeration equipment. Background Technology
[0002] In related technologies, the compressor speed control scheme of refrigerators is based on a PI controller. For example, the speed adjustment amount is determined according to the deviation between the set temperature and the actual temperature, thereby dynamically adjusting the compressor speed so that the actual temperature of the refrigerator compartment approaches and stabilizes at the set temperature.
[0003] However, in low-speed compressor scenarios, PI control schemes suffer from limitations such as bandwidth constraints, phase lag, fixed parameters, and insufficient disturbance rejection capabilities. This leads to periodic fluctuations in the compressor's load torque during low-speed operation. These periodic load disturbances cause significant speed fluctuations, resulting in increased mechanical vibration and noise, accelerated wear of mechanical components, and severely impacting compressor operational stability and user comfort. Summary of the Invention
[0004] This application provides a speed control method, a control system, and a refrigeration device, which can improve the compressor's operating capability at low speeds.
[0005] In a first aspect, embodiments of this application provide a compressor speed control method applied to a control system of a refrigeration device. The control system includes an LADRC controller and a parameter adjustment module. The method includes:
[0006] Acquire observation data for multiple control cycles from the linear extended state observer output of the LADRC controller, and acquire tracking errors for multiple control cycles;
[0007] Based on the observed data, determine the rotational speed fluctuation index and disturbance statistical characteristics, and based on the tracking error, determine the time-domain index of the tracking error;
[0008] The speed fluctuation index, the disturbance statistical characteristics, and the time-domain index are input into the parameter adjustment module to obtain the parameter adjustment amount of the LADRC control parameters output by the parameter adjustment module; the parameter adjustment module is equipped with a parameter adjustment model or an adjustment strategy corresponding to the parameter adjustment model;
[0009] The LADRC controller is optimized based on the parameter adjustment amount of the LADRC control parameters, and the compressor speed is controlled by the optimized LADRC controller.
[0010] The parameter adjustment model is obtained by training a preset model using a preset reinforcement learning algorithm. The state space of the reinforcement learning algorithm includes speed fluctuation index, disturbance statistical features, and time-domain index. The action space of the reinforcement learning algorithm includes the adjustment amount of LADRC control parameters. The reward function of the reinforcement learning algorithm optimizes the response speed and regulation stability of the LADRC controller's speed control.
[0011] In some embodiments, the observation data includes speed estimates and disturbance estimates, the speed fluctuation index includes speed fluctuation amplitude and speed fluctuation frequency data, and the disturbance statistical characteristics include disturbance estimate mean and disturbance estimate variance;
[0012] The step of determining the speed fluctuation index and disturbance statistical characteristics based on the observed data includes:
[0013] The maximum speed, minimum speed, and average speed are determined based on the speed estimates from multiple control cycles.
[0014] The speed fluctuation amplitude is determined based on the difference between the maximum speed and the minimum speed.
[0015] The fluctuation component is obtained by subtracting the average speed from several consecutive estimated speed values, and the fluctuation component is subjected to a fast Fourier transform to obtain the frequency data of the speed fluctuation.
[0016] The mean disturbance estimate is determined based on the disturbance estimates from multiple control cycles;
[0017] The variance of the disturbance estimate is determined based on the disturbance estimates from multiple control cycles and the mean of the disturbance estimates.
[0018] In some embodiments, subtracting the average speed from a plurality of consecutive speed estimates to obtain the fluctuation component includes:
[0019] A speed smoothing period is determined based on the speed estimates from multiple control cycles; the volatility of the speed estimates during the speed smoothing period is less than a preset volatility.
[0020] The fluctuation component is obtained by subtracting the average speed from a number of consecutive estimated speed values within the smoothing time period.
[0021] In some embodiments, obtaining the tracking error across multiple control cycles includes:
[0022] The reference rotational speeds of the tracking differentiator output of the LADRC controller for multiple control cycles, and the estimated rotational speeds of the linearly extended state observer output for multiple control cycles are obtained.
[0023] The tracking error for multiple control cycles is determined based on the reference speed and the estimated speed.
[0024] In some embodiments, the time-domain metrics include overshoot and settling time of the control cycle; determining the time-domain metrics for the tracking error based on the tracking error includes:
[0025] The overshoot is determined based on the speed variation of the tracking error over multiple control cycles and the reference speed.
[0026] The adjustment time is determined based on the speed change of the tracking error over multiple control cycles and the stable speed range of the compressor; the stable speed range represents the speed range in which the compressor speed converges to the reference speed.
[0027] In some embodiments, the LADRC control parameters include at least one of the bandwidth of the linear extended state observer, the bandwidth of the LADRC controller, and the disturbance compensation coefficient.
[0028] In some embodiments, the reward function of the reinforcement learning algorithm is constructed in the following manner: a first penalty term is constructed based on the tracking error, a second penalty term is constructed based on the overshoot of the rotational speed, a third penalty term is constructed based on the difference in control output between two adjacent control cycles, and a fourth penalty term is constructed based on the settling time.
[0029] In some embodiments, the method for deploying the parameter tuning model to the parameter tuning module includes:
[0030] The parameter adjustment model is then ported to the storage space of the control system.
[0031] Alternatively, the parameter adjustment model can be ported to the edge computing device of the control system, wherein the control system is used to send the speed fluctuation index, the disturbance statistical characteristics and the time domain index to the edge computing device, and to receive the parameter adjustment amount of the LADRC control parameters returned by the edge computing device;
[0032] The method for deploying the adjustment strategy corresponding to the parameter adjustment model to the parameter adjustment module includes:
[0033] The parameter adjustment model is converted into a decision tree; the input of the decision tree is the speed fluctuation index, the disturbance statistical characteristics and the time domain index, and the output is the parameter adjustment amount of the LADRC control parameters. The decision tree is converted into executable code or an executable program to be embedded into the control system.
[0034] Secondly, embodiments of this application also provide a control system, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the compressor speed control method as described in the first aspect.
[0035] Thirdly, embodiments of this application also provide a computer-readable storage medium storing computer-executable instructions for causing a control system to perform the compressor speed control method as described in the first aspect.
[0036] Fourthly, embodiments of this application also provide a computer program product, including a computer program or computer instructions, characterized in that the computer program or computer instructions are stored in a computer-readable storage medium, a processor of a refrigeration device reads the computer program or computer instructions from the computer-readable storage medium, and the processor executes the computer program or computer instructions to cause the refrigeration device to perform the compressor speed control method as described in the first aspect.
[0037] Fifthly, embodiments of this application also provide a refrigeration device, including a compressor and a control system as described in the second aspect.
[0038] The speed control method, control system, and refrigeration equipment of this application embodiment have at least the following beneficial effects: The control system of the refrigeration equipment is based on a LADRC controller. It uses observation data from multiple control cycles output by a linearly extended state observer and determines the tracking error of multiple control cycles through the LADRC controller. Then, it performs signal processing on the observation data and tracking error to obtain speed fluctuation index, disturbance statistical characteristics, and time-domain index of tracking error. These data constitute the state space features of the reinforcement learning algorithm. The adjustment amount of LADRC parameters is used as the action space features of the reinforcement learning algorithm. A reward function is set with the response speed and regulation stability of the LADRC controller's speed control as the optimization objective. A parameter adjustment model is trained to obtain the parameter adjustment amount. Using this parameter adjustment model, the refrigeration equipment can provide real-time observation data and tracking error during operation to obtain the adjustment amount of LADRC parameters, thereby continuously optimizing the parameters of the LADRC controller, improving the accuracy of the LADRC controller parameters, reducing compressor speed fluctuation, and suppressing compressor torque pulsation under low-speed conditions.
[0039] Other features and advantages of this application will be set forth in the following description and will be apparent in part from the description or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the description and the accompanying drawings. Attached Figure Description
[0040] Figure 1 A flowchart illustrating the overall speed control method for a compressor provided in one embodiment of this application;
[0041] Figure 2 A flowchart for determining the speed fluctuation index provided in one embodiment of this application;
[0042] Figure 3 A flowchart for determining statistical characteristics of disturbances provided in one embodiment of this application;
[0043] Figure 4 This is a flowchart illustrating the determination of tracking error according to an embodiment of this application;
[0044] Figure 5 This is a flowchart illustrating the determination of overshoot and settling time according to an embodiment of this application;
[0045] Figure 6 This is a schematic diagram of the connection relationship of a control system provided in one embodiment of this application. Detailed Implementation
[0046] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. Furthermore, the features, operations, or characteristics described in the specification can be combined in any suitable manner to form various implementations. Simultaneously, the steps or actions described in the method description can be rearranged or adjusted in a manner readily apparent to those skilled in the art. Therefore, the various orders in the specification and drawings are merely for the clear description of a particular embodiment and do not imply a mandatory order, unless otherwise stated that a particular order must be followed.
[0047] In the description of this application, "several" means one or more, "more than" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.
[0048] The serial numbers assigned to components in this document, such as "first" and "second," are used only to distinguish the described objects and have no sequential or technical meaning. The terms "connection" and "linkage" used in this application, unless otherwise specified, include both direct and indirect connections (linkages).
[0049] Currently, the mainstream speed loop control scheme in industrial applications still relies on PI controllers. However, in low-speed compressor operation scenarios, PI control schemes suffer from limitations such as bandwidth constraints, phase lag, fixed parameters, and insufficient disturbance rejection capabilities. This leads to periodic fluctuations in the compressor's load torque during low-speed operation. These periodic load disturbances cause significant speed fluctuations, resulting in increased mechanical vibration and noise, accelerated wear of mechanical components, and severely impacting compressor operational stability and user comfort. Some related technologies utilize LADRC (Linear Active Disturbance Rejection Control) controllers to perform system tracking control and improve the system's dynamic performance. For example, related technologies mention maximum power point tracking control of wind power systems based on improved LADRC, using the tip speed ratio method to obtain the optimal speed, improving LADRC as the speed controller and PI as the current controller, and adopting a dual closed-loop vector control method with an outer speed loop and an inner current loop. In terms of LADRC improvement, an improved linear extended state observer is introduced, using the error between each state variable and its observed value to adjust the derivative of each observed value, that is, to construct a new disturbance observation error. By adopting the new disturbance observation error, the mathematical model of the improved linear extended state observer LESO can be obtained.
[0050] The above scheme is designed for maximum power point tracking control of wind power systems, but not for low-speed control scenarios. Although new disturbance observation errors are introduced, some parameters of the LADRC controller still rely on experience for tuning, including the bandwidth of the linear extended state observer, the bandwidth of the LADRC controller, and the disturbance compensation coefficient, which affect the accuracy of the LADRC controller.
[0051] Based on this, this application provides a speed control method, a control system, and a refrigeration device. The control system of the refrigeration device is based on a LADRC controller. It uses observation data from multiple control cycles output by a linearly extended state observer and determines the tracking error of multiple control cycles through the LADRC controller. Then, it performs signal processing on the observation data and tracking error to obtain speed fluctuation index, disturbance statistical characteristics, and time-domain index of tracking error. These data constitute the state space features of the reinforcement learning algorithm. The adjustment amount of LADRC parameters is used as the action space features of the reinforcement learning algorithm. A reward function is set with the response speed and regulation stability of the LADRC controller's speed control as the optimization objective. A parameter adjustment model is trained to obtain the parameter adjustment amount. Using this parameter adjustment model, the refrigeration device can provide real-time observation data and tracking error during operation to obtain the adjustment amount of LADRC parameters, thereby continuously optimizing the parameters of the LADRC controller, improving the accuracy of the LADRC controller parameters, reducing compressor speed fluctuation, and suppressing compressor torque pulsation under low-speed conditions.
[0052] The LADRC controller consists of three parts: a tracking differentiator (TD), a linear extended state observer (LESO), and a state error feedback (SEF). The tracking differentiator (TD) smooths the reference input signal, generating a smooth reference signal. In compressor applications of refrigeration equipment, it generates a smooth reference speed or target speed. The linear extended state observer (LESO) estimates the system state and total disturbance. In compressor applications of refrigeration equipment, it can estimate the compressor speed, load torque, parameter drift, unmodeled dynamics, and other total disturbances in real time. The state error feedback (SEF) combines the state and disturbance estimated by LESO to calculate the control output. In compressor applications of refrigeration equipment, it can generate the final current command to drive the compressor based on the error between the reference speed and the actual / estimated speed, combined with the compensation amount of LESO.
[0053] The control parameters of the LADRC controller include the bandwidth of the linear extended state observer, the bandwidth of the LADRC controller, and the disturbance compensation coefficient. These parameters usually need to be set based on experience, and then adjusted again through trial and error during actual operation. Therefore, the tuning method of LADRC control parameters results in low accuracy and cannot be applied to some dynamically changing scenarios, such as the speed control of the compressor in refrigeration equipment, which covers both high-speed and low-speed conditions, and the speed control of the motor is more difficult under low-speed conditions.
[0054] The various embodiments of the compressor speed control method of this application will be further described below with reference to the accompanying drawings.
[0055] Reference Figure 1 As shown, Figure 1 This is an overall flowchart of the compressor speed control method provided in this application embodiment. The speed control method is applied to the control system of refrigeration equipment. The control system includes a LADRC controller and a parameter adjustment module. The speed control method includes, but is not limited to, the following steps:
[0056] Step S110: Obtain observation data for multiple control cycles from the linear extended state observer output of the LADRC controller and obtain the tracking error for multiple control cycles.
[0057] Step S120: Determine the rotational speed fluctuation index and disturbance statistical characteristics based on the observed data, and determine the time-domain index of the tracking error based on the tracking error;
[0058] Step S130: Input the speed fluctuation index, disturbance statistical characteristics and time domain index into the parameter adjustment module to obtain the parameter adjustment amount of the LADRC control parameters output by the parameter adjustment module; the parameter adjustment module is equipped with a parameter adjustment model or the adjustment strategy corresponding to the parameter adjustment model.
[0059] Step S140: Optimize the LADRC controller according to the parameter adjustment amount of the LADRC control parameters, and control the compressor speed through the optimized LADRC controller;
[0060] The parameter adjustment model is obtained by training a preset model using a preset reinforcement learning algorithm. The state space of the reinforcement learning algorithm includes speed fluctuation index, disturbance statistical characteristics, and time domain index. The action space of the reinforcement learning algorithm includes the adjustment amount of LADRC control parameters. The reward function of the reinforcement learning algorithm optimizes the response speed and regulation stability of the LADRC controller's speed control.
[0061] This application embodiment performs speed control based on a parameter adjustment model or the corresponding adjustment strategy deployed by the parameter adjustment module. The control system of the refrigeration system adapts to the state-space input of the parameter adjustment model, processes the collected signals to obtain the speed fluctuation index, disturbance statistical characteristics, and time-domain index of tracking error, and inputs them into the parameter adjustment model or the corresponding adjustment strategy to obtain the parameter adjustment amount of the LADRC control parameters. The LADRC controller is optimized using the parameter adjustment amount of the LADRC control parameters, realizing online continuous optimization of the LADRC controller and improving the control capability of the LADRC controller. Specifically, during the operation of the refrigeration equipment, based on the observation data of multiple control cycles output by LESO in the LADRC controller, which typically includes the speed estimate and disturbance estimate output by LESO, the speed fluctuation index and disturbance statistical characteristics are determined based on these parameters. The refrigeration equipment also determines the time-domain index of tracking error based on the tracking error of multiple control cycles.
[0062] It is worth noting that the parameter adjustment model is obtained through the fusion training of a reinforcement learning algorithm and an LADRC controller. Specifically, it adapts to the data provided by the LADRC controller, constructs the state space of the reinforcement learning algorithm, and sets the corresponding adjustment amounts of the LADRC control parameters (as initial training samples, which can be obtained experimentally by controlling the compressor in a laboratory, or simulated by establishing a compressor simulation system) to construct the action space of the reinforcement learning algorithm. Finally, the reward function of the reinforcement learning algorithm is set with the response speed and regulation stability of speed control as optimization objectives. The parameter adjustment model is obtained by training the preset model in the above manner. This overcomes the adaptability limitations of fixed-parameter LADRC controllers under variable operating conditions in practical applications.
[0063] Therefore, it can be seen that by pre-training the parameter adjustment model through the fusion of reinforcement learning algorithm and LADRC controller, and then deploying the parameter adjustment model or the adjustment strategy corresponding to the parameter adjustment model to the control system of refrigeration equipment, the parameters of LADRC controller can be continuously optimized during the online operation of refrigeration equipment equipped with LADRC controller. This allows for adaptive optimization of disturbance suppression capability and dynamic response speed to adapt to complex working conditions, and also suppresses the fluctuation of compressor speed and torque at low speeds.
[0064] In some embodiments, the observation data includes speed estimates and disturbance estimates.
[0065] Speed fluctuation indicators include speed fluctuation amplitude and speed fluctuation frequency data;
[0066] Reference Figure 2 As shown, the step S120 above, which determines the rotational speed fluctuation index based on the observed data, includes:
[0067] Step S210: Determine the maximum speed, minimum speed, and average speed based on the speed estimates from multiple control cycles;
[0068] Step S220: Determine the speed fluctuation amplitude based on the difference between the maximum and minimum speed values;
[0069] Step S230: Subtract the average speed from several consecutive speed estimates to obtain the fluctuation component, and perform a fast Fourier transform on the fluctuation component to obtain the frequency data of the speed fluctuation.
[0070] The LADRC controller's LESO function outputs a speed estimate in each control cycle. Based on this speed estimate, it determines the speed fluctuation amplitude and frequency data. Specifically, multiple control cycles can be consecutive or spaced out (e.g., a refrigeration unit switching between different operating conditions within multiple control cycles of a given condition). Each control cycle yields a speed estimate, and then the maximum, minimum, and average speed values are determined based on these multiple estimates. The speed estimate is used... This indicates that multiple speed estimates are expressed as... , , , ,......, n is a positive integer, and the maximum and minimum speeds are represented as follows: and The average speed is obtained by averaging the above multiple speed estimates. .
[0071] The rotational speed fluctuation amplitude in this embodiment = The amplitude of rotational fluctuations can also be obtained by calculating the peak-to-peak value of the estimated rotational speed between two control cycles.
[0072] In this embodiment, the fluctuation component is calculated by subtracting the average speed from a series of consecutive speed estimates. The frequency data of the rotational speed fluctuations can be determined by performing a fast Fourier transform on these fluctuation components. For example, the maximum frequency domain amplitude of each fluctuation component can be used.
[0073] Several consecutive speed estimates can be selected from the aforementioned multiple speed estimates in different ways. For example, a speed smoothing period can be determined based on the speed estimates from multiple control cycles. During this smoothing period, the fluctuation range of the speed estimates is small; for instance, the volatility of the speed estimates during this smoothing period is all less than a preset volatility (the volatility of a speed estimate is based on the speed estimate and the average speed). (The ratio is obtained). Taking several consecutive speed estimates within this smoothing period is used to calculate the fluctuation component and perform spectral analysis. This avoids the impact of large rotational fluctuations on the accuracy of frequency analysis and improves the accuracy of the state space data in the reinforcement learning algorithm. If the speed estimates for multiple control cycles do not meet the requirements within a smoothing period, the preset fluctuation rate can be increased to achieve dynamic adaptive adjustment.
[0074] The statistical characteristics of the disturbance include the mean and variance of the disturbance estimate.
[0075] Reference Figure 3 As shown, step S120 above, which determines the statistical characteristics of the disturbance based on the observed data, includes:
[0076] Step S310: Determine the mean disturbance estimate based on the disturbance estimates from multiple control cycles;
[0077] Step S320: Determine the disturbance estimation variance based on the disturbance estimates and the mean disturbance estimates for multiple control cycles.
[0078] The LADRC controller's LESO outputs a disturbance estimate in each control cycle, and determines the disturbance estimate mean and variance for each disturbance estimate. A disturbance estimate is obtained for each control cycle, and then the average of multiple disturbance estimates is calculated to obtain the disturbance estimate mean. Finally, the variance between each disturbance estimate and the disturbance estimate mean is calculated, reflecting the degree of fluctuation in the disturbance estimate output in each control cycle.
[0079] Reference Figure 4 As shown, in some embodiments, in step S110 above, obtaining the tracking error of multiple control cycles includes:
[0080] Step S410: Obtain the reference rotational speed of multiple control cycles output by the tracking differentiator of the LADRC controller, and the estimated rotational speed of multiple control cycles output by the linear extended state observer.
[0081] Step S420: Determine the tracking error for multiple control cycles based on the reference speed and the speed estimate.
[0082] The LADRC controller's TD outputs a smooth reference speed in each control cycle. Combined with the speed estimate output by LESO (Also known as speed observation), the tracking error for each control cycle is calculated. Furthermore, the LADRC controller can calculate the control quantity for the control cycle based on the tracking error e. .
[0083] Reference Figure 5 As shown, in some embodiments, the time-domain indicators include the overshoot and settling time of the control cycle; the time-domain indicators of the tracking error determined in step S120 above include:
[0084] Step S510: Determine the overshoot based on the speed change of the tracking error in multiple control cycles and the reference speed;
[0085] Step S520: Determine the adjustment time based on the speed change of the tracking error in multiple control cycles and the speed stability range of the compressor; the speed stability range characterizes the speed range at which the compressor speed converges to the reference speed.
[0086] Overshoot and settling time are calculated by accumulating data over multiple control cycles; these two parameters constitute the time-domain indices of the tracking error. The overshoot can be calculated cycle by cycle using the tracking error of each control cycle. The TD output provides a smooth reference speed. As a steady-state value, it is based on the estimated maximum speed over multiple control cycles. Calculate the overshoot value for the entire adjustment process (corresponding to the multiple control cycles mentioned above), that is, calculate... If the calculated result is greater than 0, it indicates that the overshoot value is equal to 0. If the calculation result is less than or equal to 0, the overshoot value is set to 0. The settling time can be the settling time of multiple control cycles mentioned above (i.e., the settling time corresponding to completing one full speed convergence process). Speed convergence is determined based on the speed stability range, which can be determined based on the reference speed. To determine, for example, by setting a stable speed range and taking a reference speed. ±5% is considered to indicate speed convergence when the tracking error remains within the speed stability range for multiple consecutive control cycles. The time from the start of the first control cycle to the moment of speed convergence is the adjustment time.
[0087] In some embodiments, the reward function of the reinforcement learning algorithm is constructed in the following ways: a first penalty term is constructed based on the tracking error, a second penalty term is constructed based on the overshoot of the rotational speed, a third penalty term is constructed based on the difference in control output between two adjacent control cycles, and a fourth penalty term is constructed based on the settling time.
[0088] The reward function guides the model to learn and make decisions towards the desired goal. In the above application, it is necessary to consider the response speed and regulation stability of the compressor speed control of the refrigeration equipment. The output of the reward function can be determined by setting a reward term or a penalty term, guiding the model to achieve a balance between response speed and regulation stability. For example, taking the setting of penalty terms as an example, in terms of response speed, the first penalty term is determined based on the tracking error (e.g., the negative of the square of the tracking error), with the goal of making the tracking error converge as quickly as possible. In terms of regulation stability, the second penalty term is determined based on the overshoot (e.g., the negative of the square of the overshoot), with the goal of preventing the speed from exceeding the limit and causing problems such as noise and vibration in the compressor. In terms of regulation stability, the third penalty term is also determined based on the difference in control output (e.g., the negative of the square of the difference between the control output of the current cycle and the control output of the previous cycle), with the goal of smoothing the control process and avoiding drastic changes in the regulation process that would affect the normal operation of the compressor. In terms of response speed, the fourth penalty term is also determined based on the adjustment time (e.g., the negative of the square of the adjustment time), with the goal of reducing the length of the adjustment time and speeding up the adjustment process. The above four penalty terms are added together to construct a reward function, and corresponding weight coefficients can be set for each penalty term. For example, the first weight coefficient can be multiplied by the first penalty term, the second weight coefficient can be multiplied by the second penalty term, the third weight coefficient can be multiplied by the third penalty term, and the fourth weight coefficient can be multiplied by the fourth penalty term. Based on this, additional reward items can be added to the above reward function. For example, a first reward item can be set, which takes effect when the ratio of tracking error to average speed is less than a certain value (e.g., less than 5%). A second reward item can be set, which takes effect when the adjustment time is less than a certain duration, and so on.
[0089] In some embodiments, the trained parameter tuning model can be deployed in various ways. For example, directly deploying the trained model to the parameter tuning module specifically includes:
[0090] The parameter tuning model is ported to the storage space of the control system;
[0091] Alternatively, the parameter adjustment model can be ported to the edge computing device of the control system, which is used to send speed fluctuation indicators, disturbance statistics, and time-domain indicators to the edge computing device, as well as to receive the parameter adjustment amount of the LADRC control parameters returned by the edge computing device.
[0092] The first approach involves directly porting the parameter tuning model. A large parameter tuning model is stored in the control system's memory. The control system runs the model and performs local calculations based on the acquired and processed data (speed fluctuation indicators, disturbance statistical characteristics, and time-domain indicators of tracking error), outputting the adjustment amount of the LADRC control parameters. The second approach involves directly porting the parameter tuning model to an edge computing device. This is suitable for lightweight parameter tuning models. The control system sends the acquired and processed data to the edge computing device, which uses this data to perform calculations and returns the adjustment amount of the LADRC control parameters to the control system. Based on this, the edge computing device can continuously train the parameter tuning model using the received data. This approach emphasizes online, continuous learning and optimization, rather than offline, fixed model deployment, and possesses strong environmental adaptability and long-term optimization potential.
[0093] For example, deploying the tuning strategy corresponding to the parameter tuning model to the parameter tuning module specifically includes:
[0094] The parameter adjustment model is converted into a decision tree. The input of the decision tree is the speed fluctuation index, disturbance statistical characteristics and time domain index, and the output is the parameter adjustment amount of LADRC control parameters. The decision tree is converted into executable code or executable program to be embedded into the control system.
[0095] The above approach involves converting the parameter adjustment model into a decision tree, then transforming the decision tree into code or an executable program that can be directly run on the microcontroller. This code or executable program is then embedded into the control system, which makes local decisions based on the acquired and processed data, outputting the adjustment amount of the LADRC control parameters. This embedding method is suitable for scenarios where the microcontroller's computing power is relatively low.
[0096] In summary, the control system of the refrigeration equipment is based on the LADRC controller. It utilizes observation data from multiple control cycles output by a linearly extended state observer and determines the tracking error for these cycles using the LADRC controller. Signal processing of the observation data and tracking error yields speed fluctuation indicators, disturbance statistical characteristics, and time-domain indicators of the tracking error. These data constitute the state-space features of the reinforcement learning algorithm. The adjustment of the LADRC parameters serves as the action-space feature of the algorithm. A reward function is set with the response speed and regulation stability of the LADRC controller's speed control as optimization objectives. This allows for the training of a parameter adjustment model. Using this model, the refrigeration equipment can obtain real-time observation data and tracking errors during operation, leading to the determination of LADRC parameter adjustments. This continuously optimizes the LADRC controller parameters, improving their accuracy, reducing compressor speed fluctuations, and suppressing compressor torque pulsation under low-speed conditions.
[0097] The methods of this application will be explained in detail below through some specific examples.
[0098] This example proposes an adaptive optimization method for the control parameters of a compressor's LADRC controller. By obtaining compressor data through the LADRC controller's observer, the observer bandwidth, controller bandwidth, and compensation coefficient of the LADRC controller are adjusted online based on this data. This suppresses the fluctuations in compressor speed and torque at low speeds, solves the problem of insufficient suppression of periodic disturbances by traditional PI control, and overcomes the adaptability of fixed-parameter LADRC controllers under variable operating conditions.
[0099] A LADRC controller is used to replace the traditional PI controller for compressor speed control. The core is to use a data-driven approach to compensate for the experience-based dependence of model-driven LADRC in parameter tuning. Through reinforcement learning, the LADRC controller learns to dynamically adjust its parameters based on data such as speed fluctuations. The algorithm architecture consists of three layers: the device layer, the signal processing layer, and the decision layer, forming a closed-loop optimization structure of "execution-perception-decision".
[0100] The underlying layer is the LADRC controller, which is responsible for real-time control. It estimates and compensates for internal and external disturbances through a Linear Extended State Observer (LESO). Its parameters (such as observer bandwidth, controller bandwidth, compensation coefficients, etc.) directly affect the disturbance suppression capability and dynamic response speed.
[0101] The middle layer is the signal processing layer, which extracts key features from the system output (such as motor speed), including:
[0102] (1) Amplitude and frequency components of rotational speed fluctuation;
[0103] (2) Statistical characteristics of the disturbance estimates (such as mean and variance);
[0104] (3) Time-domain indicators of tracking error (such as overshoot and settling time).
[0105] These features constitute the state space input for reinforcement learning, providing quantitative basis for the upper decision-making layer.
[0106] The upper layer is the decision layer, which uses reinforcement learning algorithms (such as DDPG, PPO, or TD3) to dynamically adjust the LADRC control parameters based on the current state (i.e., the characteristics of the signal processing layer). Its core design includes:
[0107] (1) State space: including speed fluctuation characteristics, statistical characteristics of disturbance estimates, time-domain indices, historical parameters, etc.;
[0108] (2) Action space: the adjustment amount of the LADRC parameters;
[0109] (3) Reward function: guides the design of LADRC controller to balance the two aspects of "speed" and "stability".
[0110] The above-mentioned reinforcement learning and LADRC fusion strategy training modes are divided into two types. One is to pre-train the reinforcement learning model using historical data or simulation environment, and adjust the strategy by learning parameters through a large number of trial and error to form an initial agent, which can significantly reduce the online computing burden. The second is to deploy the trained model to the actual system and fine-tune the parameters in real time according to the speed fluctuation. An online learning mode can also be adopted to continuously adapt to the time-varying characteristics of the system.
[0111] By introducing reinforcement learning algorithms, the parameters of the LADRC controller can be optimized online, which solves the problem of insufficient suppression of periodic disturbances by traditional PI control. At the same time, it overcomes the adaptability limitations of fixed-parameter LADRC under varying operating conditions, reduces compressor speed fluctuations, and suppresses compressor torque pulsation under low-speed conditions.
[0112] Based on the compressor speed control methods of the above embodiments, the following presents various embodiments of the control system, refrigeration equipment, computer-readable storage medium, and computer program product of this application.
[0113] like Figure 6 As shown, Figure 6 This is a schematic diagram of a control system for executing a compressor speed control method according to an embodiment of this application. The control system 100 implemented in this application includes: a processor 110, a memory 120, and a computer program stored in the memory 120 and executable on the processor 110, wherein... Figure 6 The example uses a processor 110 and a memory 120.
[0114] Processor 110 and memory 120 can be connected via a bus or other means. Figure 6 Taking the example of a connection between China and Israel via a bus.
[0115] Memory 120, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory 120 may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 120 may optionally include remotely located memories 120 relative to processor 110, which can be connected to the control system 100 via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.
[0116] Those skilled in the art will understand that Figure 6 The device structure shown does not constitute a limitation on the control system 100 and may include more or fewer components than shown, or combine certain components, or have different component arrangements.
[0117] exist Figure 6 In the control system 100 shown, the processor 110 can be used to call the control program stored in the memory 120 to implement the compressor speed control method described above. Specifically, the non-transient software program and instructions required to implement the compressor speed control method of the above embodiment are stored in the memory 120. When executed by the processor 110, the compressor speed control method of the above embodiment is executed.
[0118] It is worth noting that since the control system 100 of this application embodiment can execute the compressor speed control method of any of the above embodiments, the specific implementation and technical effects of the control system 100 of this application embodiment can refer to the specific implementation and technical effects of the compressor speed control method of any of the above embodiments.
[0119] In addition, one embodiment of this application also provides a refrigeration device, including a compressor and the aforementioned control system.
[0120] Furthermore, one embodiment of this application provides a computer-readable storage medium storing computer-executable instructions for performing the above-described compressor speed control method. Exemplarily, the method steps of the compressor speed control method described above are executed.
[0121] It is worth noting that, since the computer-readable storage medium of this application embodiment is capable of executing the compressor speed control method of any of the above embodiments, the specific implementation and technical effects of the computer-readable storage medium of this application embodiment can be referred to the specific implementation and technical effects of the compressor speed control method of any of the above embodiments.
[0122] Furthermore, one embodiment of this application also provides a computer program product, including a computer program or computer instructions, which are stored in a computer-readable storage medium. A processor of a computer device reads the computer program or computer instructions from the computer-readable storage medium and executes the computer program or computer instructions, causing the computer device to perform the aforementioned compressor speed control method. Exemplarily, the method steps of the compressor speed control method described above are executed.
[0123] It is worth noting that, since the computer program product of this application embodiment can execute the compressor speed control method of any of the above embodiments, the specific implementation method and technical effect of the computer program product of this application embodiment can refer to the specific implementation method and technical effect of the compressor speed control method of any of the above embodiments.
[0124] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network nodes. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.
[0125] It will be understood by those skilled in the art that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.
[0126] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.
[0127] In the several embodiments provided in this application, it should be understood that the disclosed systems, instruments, and methods can be implemented in other ways. For example, the instrument embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between instruments or units may be electrical, mechanical, or other forms. Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, i.e., they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0128] It should also be understood that the various implementation methods provided in this application can be combined arbitrarily to achieve different technical effects.
[0129] The above is a detailed description of the preferred embodiments of this application. However, this application is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.
Claims
1. A method for controlling the speed of a compressor, characterized in that, A control system for refrigeration equipment, the control system including an LADRC controller and a parameter adjustment module, the method comprising: Acquire observation data for multiple control cycles from the linear extended state observer output of the LADRC controller, and acquire tracking errors for multiple control cycles; The speed fluctuation index and disturbance statistical characteristics are determined based on the observed data, and the time-domain index of the tracking error is determined based on the tracking error; the observed data includes speed estimate and disturbance estimate, the speed fluctuation index includes speed fluctuation amplitude and speed fluctuation frequency data, and the disturbance statistical characteristics include disturbance estimate mean and disturbance estimate variance; The speed fluctuation index, the disturbance statistical characteristics, and the time-domain index are input into the parameter adjustment module to obtain the parameter adjustment amount of the LADRC control parameters output by the parameter adjustment module; the parameter adjustment module is equipped with a parameter adjustment model or an adjustment strategy corresponding to the parameter adjustment model; The LADRC controller is optimized based on the parameter adjustment amount of the LADRC control parameters, and the compressor speed is controlled by the optimized LADRC controller. The parameter adjustment model is obtained by training a preset model using a preset reinforcement learning algorithm. The state space of the reinforcement learning algorithm includes speed fluctuation index, disturbance statistical features, and time-domain index. The action space of the reinforcement learning algorithm includes the adjustment amount of LADRC control parameters. The reward function of the reinforcement learning algorithm optimizes the response speed and regulation stability of the LADRC controller's speed control.
2. The method according to claim 1, characterized in that, The step of determining the speed fluctuation index and disturbance statistical characteristics based on the observed data includes: The maximum speed, minimum speed, and average speed are determined based on the speed estimates from multiple control cycles. The speed fluctuation amplitude is determined based on the difference between the maximum speed and the minimum speed. The fluctuation component is obtained by subtracting the average speed from several consecutive estimated speed values, and the fluctuation component is subjected to a fast Fourier transform to obtain the frequency data of the speed fluctuation. The mean disturbance estimate is determined based on the disturbance estimates from multiple control cycles; The variance of the disturbance estimate is determined based on the disturbance estimates from multiple control cycles and the mean of the disturbance estimates.
3. The method according to claim 2, characterized in that, The step of subtracting the average speed from a series of consecutive speed estimates to obtain the fluctuation component includes: A speed smoothing period is determined based on the speed estimates from multiple control cycles; the volatility of the speed estimates during the speed smoothing period is less than a preset volatility. The fluctuation component is obtained by subtracting the average speed from a number of consecutive estimated speed values within the smoothing time period.
4. The method according to claim 1, characterized in that, The acquisition of tracking error over multiple control cycles includes: The reference rotational speeds of the tracking differentiator output of the LADRC controller for multiple control cycles are obtained, as well as the estimated rotational speeds of the linearly extended state observer output for multiple control cycles. The tracking error for multiple control cycles is determined based on the reference speed and the estimated speed.
5. The method according to claim 4, characterized in that, The time-domain indicators include the overshoot and settling time of the control cycle; the time-domain indicators for determining the tracking error based on the tracking error include: The overshoot is determined based on the speed variation of the tracking error over multiple control cycles and the reference speed. The adjustment time is determined based on the speed change of the tracking error over multiple control cycles and the stable speed range of the compressor; the stable speed range represents the speed range in which the compressor speed converges to the reference speed.
6. The method according to claim 1, characterized in that, The LADRC control parameters include at least one of the following: the bandwidth of the linear extended state observer, the bandwidth of the LADRC controller, and the disturbance compensation coefficient.
7. The method according to claim 1, characterized in that, The reward function of the reinforcement learning algorithm is constructed in the following way: a first penalty term is constructed based on the tracking error, a second penalty term is constructed based on the overshoot of the rotational speed, a third penalty term is constructed based on the difference in control output between two adjacent control cycles, and a fourth penalty term is constructed based on the settling time.
8. The method according to claim 1, characterized in that, The method for deploying the parameter tuning model to the parameter tuning module includes: The parameter adjustment model is then ported to the storage space of the control system. Alternatively, the parameter adjustment model can be ported to the edge computing device of the control system, wherein the control system is used to send the speed fluctuation index, the disturbance statistical characteristics and the time domain index to the edge computing device, and to receive the parameter adjustment amount of the LADRC control parameters returned by the edge computing device; The method for deploying the adjustment strategy corresponding to the parameter adjustment model to the parameter adjustment module includes: The parameter adjustment model is converted into a decision tree; the input of the decision tree is the speed fluctuation index, the disturbance statistical characteristics and the time domain index, and the output is the parameter adjustment amount of the LADRC control parameters. The decision tree is converted into executable code or an executable program to be embedded into the control system.
9. A control system, characterized in that, It includes at least one processor and a memory for communicatively connecting to the at least one processor; the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method as described in any one of claims 1 to 8.
10. A refrigeration device, characterized in that, It includes a compressor and a control system as described in claim 9.