A method and apparatus for accurate disturbance prediction and compensation in a networked control system

Abstract

This invention discloses a method and apparatus for accurate disturbance prediction and compensation in networked control systems. It includes treating the impact of time delay on the networked control system and external disturbances together as a lumped disturbance, and establishing an equivalent input disturbance state-space model for the networked control system. This method and apparatus target networked control systems with external disturbances, time delays, and packet loss. Through an EID estimator, it effectively reduces the impact of time delay on the performance of the networked control system. A disturbance predictor based on multiple LSTM neural networks is designed to achieve accurate multi-step disturbance prediction, improving the disturbance suppression capability of the networked control system. Furthermore, a networked predictive controller is employed to compensate for the impact of packet loss on the performance of the networked control system in real time, even with continuous packet loss, using multi-step control input prediction values. This improves the control accuracy of the networked control system and meets the requirements of high-performance networked control systems.

Description

Technical Field

[0001] This invention belongs to the field of anti-interference technology, specifically relating to a method and apparatus for accurate compensation of disturbance prediction in a networked control system. Background Technology

[0002] With the development of network communication technology, networked control systems have been widely applied in high-end equipment such as intelligent robots, CNC machine tools, and multi-axis cutting machines. However, the introduction of networks has also brought new challenges, such as communication constraints like network-induced latency and packet loss. These problems can reduce system control performance and even lead to system instability. Simultaneously, the influence of various disturbances such as system uncertainty, measurement noise, and external disturbances during the control process further weakens the control accuracy of networked control systems. Therefore, effectively suppressing the negative impacts of network communication constraints and various disturbances has become crucial for improving the stability and control capabilities of networked control systems and meeting the high-precision requirements of high-end manufacturing equipment.

[0003] In past research, scholars have proposed various disturbance observer-based techniques to simultaneously handle system delay and external disturbances, including designing generalized proportional-integral (PI) observers, generalized extended state observers, and equivalent-input-disturbance (EID) estimators. These methods typically treat external disturbances and time delays as a unified lumped disturbance, then estimate and suppress them by designing a disturbance observer. Among these, the EID method, due to its simple structure and ability to effectively suppress various disturbances and time-varying delays, is widely used in networked control systems. However, this approach, which treats delay as part of the lumped disturbance, fails to fully consider the disturbance compensation lag effect caused by the delay itself, thus limiting system performance improvement. Subsequently, researchers have proposed various disturbance predictors, such as predictors based on Taylor polynomials, predictors based on Newton series, and predictors based on extended state observers. However, these methods are generally only applicable to systems with known constant delays and fail to effectively handle more complex network conditions such as packet loss. Therefore, for networked control systems with high precision requirements, existing disturbance suppression methods are still insufficient to further improve performance when external disturbances, time-varying delays, and data packet loss coexist. Summary of the Invention

[0004] The purpose of this invention is to address the problems raised in the background art by proposing a method for accurate disturbance prediction and compensation in networked control systems.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] This invention proposes a method for accurate disturbance prediction and compensation in a networked control system, comprising:

[0007] The impact of time delay on the networked control system and external disturbances to the networked control system are treated as lumped disturbances, and an equivalent input disturbance state space model of the networked control system is established.

[0008] Based on the equivalent input disturbance state-space model, an EID estimator is designed, which includes a state observer and a filter. Based on the state observer, the state estimate and EID estimate of the networked control system are obtained.

[0009] The EID estimate is obtained by passing the EID estimate through a filter, and is called the filtered EID estimate.

[0010] For each sampling time:

[0011] Current sampling time and the current sampling time The former The filtered EID estimate at each sampling time is used as the input to the disturbance predictor to obtain the sampling time. After The predicted value of the EID estimate at each sampling time, and the perturbation predictor includes A series of sequentially connected LSTM neural networks;

[0012] Current sampling time The state estimate of the networked control system is used as the input to the state feedback controller to obtain the current sampling time. The output of the state feedback controller;

[0013] Current sampling time The output of the state feedback controller and the current sampling time After The predicted value of the EID estimate at each sampling time, and the current sampling time. The filtered EID estimate is used as the input to the networked predictive controller, which outputs the predicted control input value at each sampling time to compensate the networked control system.

[0014] Preferably, the current sampling time The former The filtered EID estimates at each sampling time point constitute the EID estimate dataset. For the EID estimation dataset The standardization process is performed to obtain the standardized EID estimate dataset. The standardized EID estimate dataset Each data point in the dataset is used as a training sample to train the LSTM neural network, resulting in a trained LSTM neural network.

[0015] Each LSTM neural network in the perturbation predictor is a trained LSTM neural network.

[0016] Preferably, the step of obtaining the current sampling time After The predicted values ​​of the EID estimates at each sampling time point include:

[0017] Current sampling time and the current sampling time The former The filtered EID estimates at each sampling time constitute the first sequence arranged in chronological order of the sampling time. Each data point in the first sequence is then standardized.

[0018] The first sequence after standardization is used as the input to the first LSTM neural network in the perturbation predictor. Then, the output of the first LSTM neural network is subjected to corresponding destandardization to obtain the current sampling time. For sampling time The predicted value of the EID estimate ;

[0019] sampling time The predicted value of the EID estimate The filtered EID estimate, which was first sampled in the first sequence, was removed to form a second sequence arranged in chronological order of the sampling times.

[0020] The standardized second sequence is used as the input to the second LSTM neural network in the perturbation predictor. The output of the second LSTM neural network is then subjected to corresponding destandardization to obtain the current sampling time. For sampling time The predicted value of the EID estimate ;

[0021] And so on, repeating. Next, and the first Next time, the standardized first time The sequence is used as the first in the perturbation predictor The input of the first LSTM neural network, and then the first... The outputs of each LSTM neural network are denormalized to obtain the current sampling time. For sampling time The predicted value of the EID estimate Thus, the current sampling time is obtained. After The predicted value of the EID estimate at each sampling time.

[0022] Preferably, the networked predictive controller includes a predictive control sequence generator and a network packet loss compensator.

[0023] Preferably, for the predictive control sequence generator, the current sampling time is... The output of the state feedback controller serves as the input to the predictive control sequence generator, which in turn applies the sampling time... The current sampling time is obtained by predicting the output of the state feedback controller. After The output of the state feedback controller at each sampling time;

[0024] Current sampling time and the current sampling time After The output of the state feedback controller at each sampling moment corresponds sequentially to the current sampling moment. The filtered EID estimate and the current sampling time After The difference between the predicted values ​​of the EID estimates at each sampling time is used to obtain the current sampling time. and the current sampling time After The control input prediction value at each sampling time.

[0025] Preferably, for the network packet loss compensator, the current sampling time is... and the current sampling time After The control input prediction values ​​at each sampling time are collectively used as the current sampling time. The data packets are input to the network packet loss compensator, and the data in the data packets are sorted in the order of the sampling time from first to last. When the network packet loss compensator receives a new data packet, it overwrites the original data packet with the new data packet.

[0026] When the networked control system is at the current sampling time When packet loss occurs, it means that the network packet loss compensator is currently storing the data from the previous sampling time. If the data packet is lost, the network packet loss compensator will adjust the previous sampling time. The second data point in the data packet is used as the output, and the second data point is the current sampling time. The predicted control input value is then used to adjust the current sampling time of the networked control system. Provide compensation;

[0027] When the networked control system is at the current sampling time If no data packet loss occurs, it means that the network packet loss compensator is currently saving the data at the sampling time. If the data packet is lost, the network packet loss compensator will adjust the current sampling time. The first data in the data packet is used as the output, and the first data is the current sampling time. The predicted control input value is then used to adjust the current sampling time of the networked control system. Compensation will be provided.

[0028] A disturbance prediction and accurate compensation device for a networked control system includes a processor and a memory storing a plurality of computer instructions, wherein the computer instructions, when executed by the processor, implement the steps of a disturbance prediction and accurate compensation method for a networked control system.

[0029] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0030] This method and apparatus for accurate disturbance prediction and compensation in networked control systems addresses the challenges of external disturbances, time delays, and packet loss. By employing an EID estimator, it effectively reduces the impact of time delays on the performance of the networked control system. Furthermore, a disturbance predictor based on multiple LSTM neural networks is designed, enabling multi-step accurate disturbance prediction and enhancing the disturbance suppression capability of the networked control system. A networked predictive controller is also incorporated, allowing for real-time compensation of the impact of packet loss on the networked control system's performance in the event of continuous packet loss, thereby improving the control accuracy and meeting the requirements of high-performance networked control systems. Attached Figure Description

[0031] Figure 1 This is a block diagram of the disturbance prediction and accurate compensation method and device for the networked control system of the present invention.

[0032] Figure 2 This is a schematic diagram of the time delay in the experimental environment of this invention;

[0033] Figure 3 This is a schematic diagram of data packet loss in the experimental environment of this invention;

[0034] Figure 4 The figure shows the experimental results of the present invention compared with the prior art in a networked control system with a 20% packet loss probability.

[0035] Figure 5 The figure shows the experimental results of the present invention compared with the prior art in a networked control system with a 40% packet loss probability. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention.

[0038] like Figure 1 As shown, a method for accurate disturbance prediction and compensation in a networked control system is provided, comprising:

[0039] Step 1: Combine the impact of time delay on the networked control system (a typical networked control system includes sensors, controllers (providing decisions and commands), actuators (executing control commands), and communication networks) (i.e., time delay disturbance) with the external disturbances of the networked control system as a lumped disturbance, and establish an equivalent input disturbance state space model of the networked control system.

[0040] Step 1.1: First, establish the state-space model of the networked control system:

[0041] (1);

[0042] in, for The status of the networked control system in real time. for The derivative, for Control inputs of a real-time networked control system for The output of the real-time networked control system System matrix of networked control system for External interference to the networked control system at all times For the time delay of networked control systems, The sampling period of the sensor in a networked control system. for Packet loss indicators for real-time networked control systems;

[0043] Step 1.2, adjust the delay The impact on the networked control system is modeled as a time-delay induced disturbance, which yields the discrete-time state-space model of the networked control system based on formula (1):

[0044] (2);

[0045] in,

[0046] ;

[0047] ;

[0048] ;

[0049] ;

[0050] ;

[0051] ;

[0052] ;

[0053] ;

[0054] ;

[0055] in, Sampling time (i.e., the first) The state of the networked control system at each sampling time. Sampling time The output of the networked control system Sampling time Packet loss metrics for networked control systems Sampling time Control inputs of a networked control system Sampling time Lumped disturbances in networked control systems. Sampling time External interference to networked control systems Sampling time Delay disturbances in networked control systems , , , , , and These are all intermediate parameters. This is the average delay. For the uncertain component of time delay, It is the integral variable.

[0056] Step 1.3: As can be seen from the concept of EID, there exists a control input signal at the control input. Its output to the networked control system Impact and lumped disturbance Output of networked control system The effects are exactly the same, at this time it is called For the EID of the networked control system, the equivalent input disturbance state space model of the networked control system based on formula (2) can be obtained:

[0057] (3).

[0058] Step 2: Based on the equivalent input disturbance state space model, design an EID estimator, which includes a state observer and a filter. Based on the state observer, obtain the state estimate and EID estimate of the networked control system.

[0059] The state observer is used for observing and estimating the state of the networked control system, and is expressed by the following formula:

[0060]

[0061] in, Sampling time Status of networked control systems The estimated value, Sampling time Networked control system output The estimated value, For the state observer gain, Sampling time The output of the state feedback controller (input to the state controller);

[0062] Among them, sampling time The EID estimate is obtained based on the state observer. :

[0063] (4);

[0064] in, for The generalized inverse.

[0065] Step 3: Pass the EID estimate through a filter to obtain the filtered EID estimate, which is called the filtered EID estimate.

[0066] It should be noted that, according to formula (4), the sampling time is... EID estimate Calculation and sampling time Control input of networked control system The correlation raises causal issues; therefore, a first-order low-pass filter is introduced. right Filtering is performed to address causality issues, resulting in a filtered EID estimate, which is called the filtered EID estimate. ,and , yes Z-transform, yes The Z-transform, in which a first-order low-pass filter Satisfying the formula: ,in, It is a time constant. Let be a complex variable in the Z-transform.

[0067] For each sampling time (steps 4-6 are for each sampling time):

[0068] Step 4: Set the current sampling time and the current sampling time The former The filtered EID estimate at each sampling time (i.e., the current sampling time) The filtered EID estimate and the current sampling time The former The filtered EID estimates at each sampling time point, totaling... The filtered EID estimate of the step is used as the input to the disturbance predictor to obtain the sampling time. After The predicted value of the EID estimate at each sampling time (i.e., the history of the perturbation predictor) Step-filtered EID estimates predict future The predicted value of the EID estimate of the step), and the perturbation predictor includes A series of sequentially connected LSTM neural networks (in this embodiment) );

[0069] Step 4.1: Each LSTM neural network in the perturbation predictor is a trained LSTM neural network. When training the LSTM neural network, the current sampling time is used. The former The filtered EID estimates at each sampling time point constitute the EID estimate dataset. For the EID estimation dataset The standardization process is performed to obtain the standardized EID estimate dataset. (in, , for The average of all filtered EID estimates. for The standard deviation of all filtered EID estimates in the dataset will be used to standardize the EID estimate dataset. Each data point is used as a training sample to train the LSTM neural network, resulting in a trained LSTM neural network. During training, each sample is used as input to the LSTM neural network, and forward propagation yields the predicted value for each sample. The predicted value of each sample is then compared with the true value of the sample (i.e., the sample itself) to calculate the loss function. , for The number of training samples in the middle, and The first The system calculates the true and predicted values ​​of each sample. Then, using backpropagation, it calculates the gradient of the loss function with respect to the weights and biases of the LSTM neural network. Subsequently, the Adam optimizer updates the LSTM neural network parameters based on the calculated gradients and the set learning rate, gradually reducing the prediction error. Finally, based on the training results, the LSTM neural network with the smallest prediction error is selected as the trained network, and its standardized parameters are saved. and (used in the construction of the perturbation predictor).

[0070] Step 4.2: Set the current sampling time and the current sampling time The former The filtered EID estimate at each sampling time (i.e., the current sampling time) The filtered EID estimate and the current sampling time The former The filtered EID estimate at each sampling time is used as the input to the disturbance predictor to obtain the current sampling time. After The predicted values ​​of the EID estimates at each sampling time point include:

[0071] Current sampling time and the current sampling time The former The filtered EID estimates at each sampling time constitute the first sequence, arranged chronologically by sampling time (each sequence is ordered from back to front and from left to right). , The current sampling time The filtered EID estimate (i.e. ), The current sampling time The former The filtered EID estimate at the sampling time is used to standardize each data point in the first sequence (the standardized parameters are stored in the trained LSTM neural network). and );

[0072] The first sequence after standardization is used as the input to the first LSTM neural network in the perturbation predictor. Then, the output of the first LSTM neural network is subjected to corresponding destandardization to obtain the current sampling time. For sampling time The predicted value of the EID estimate ;

[0073] sampling time The predicted value of the EID estimate Add it to the first sequence, and remove the earliest filtered EID estimate from the first sequence (that is, remove the earliest filtered EID estimate from the first sequence to ensure that the number of data in each sequence is 1). (number of samples), forming a second sequence arranged in chronological order of sampling time (the second sequence is...). );

[0074] The standardized second sequence is used as the input to the second LSTM neural network in the perturbation predictor. The output of the second LSTM neural network is then subjected to the corresponding denormalization process (i.e., the denormalization process corresponding to the normalization process) to obtain the current sampling time. For sampling time The predicted value of the EID estimate ;

[0075] sampling time The predicted value of the EID estimate The filtered EID estimate, which was first obtained at the earliest sampling time, was added to the second sequence, and the second sequence was then removed to form a third sequence arranged in chronological order of sampling time (the third sequence is...). );

[0076] And so on, repeating. The new sequence obtained is used as the input to the next LSTM neural network, and the... Next time, the standardized first time Sequence (No. The sequence is ) as the first in the disturbance predictor The input of the first LSTM neural network, and then the first... The outputs of each LSTM neural network are denormalized to obtain the current sampling time. For sampling time The predicted value of the EID estimate Thus, the current sampling time is obtained. After The predicted value of the EID estimate at each sampling time.

[0077] Step 5: Set the current sampling time The state estimate of the networked control system is used as the input to the state feedback controller to obtain the current sampling time. The output of the state feedback controller is given by the following formula:

[0078]

[0079] in, For state feedback gain, in this embodiment, It is obtained through the pole placement method.

[0080] Step 6: Set the current sampling time The output of the state feedback controller and the current sampling time After The predicted value of the EID estimate at each sampling time, and the current sampling time. The filtered EID estimate is used as the input to the networked predictive controller. The networked predictive controller outputs the predicted control input value at each sampling time to compensate the networked control system, specifically as follows:

[0081] Step 6.1: The networked predictive controller includes a control sequence prediction generator (CSPG) and a network packet loss compensator (NPLC).

[0082] Step 6.2: For the predictive control sequence generator, set the current sampling time... The output of the state feedback controller serves as the input to the predictive control sequence generator, which in turn applies the sampling time... The output of the state feedback controller is used for prediction (using a network packet loss compensator, NPC) to obtain the current sampling time. After The output of the state feedback controller at each sampling time;

[0083] Current sampling time (Output of the state feedback controller) and the current sampling time After The output of the state feedback controller at each sampling moment corresponds sequentially to the current sampling moment. The filtered EID estimate and the current sampling time After The difference between the predicted values ​​of the EID estimates at each sampling time is used to obtain the current sampling time. Control input predicted value and current sampling time After Predicted control input values ​​at each sampling time;

[0084] The current sampling time The output of the state feedback controller and the current sampling time After The outputs of the state feedback controller at each sampling time are combined into a set. express: , The current sampling time Predicting the first The output of the state feedback controller at each sampling time;

[0085] Among them, the current sampling time The filtered EID estimate and the current sampling time After The predicted values ​​of the EID estimates at each sampling time point are used together in a set. express: ;

[0086] Will and Perform the difference operation to obtain the current sampling time. Control input predicted value and current sampling time After The predicted control input value at each sampling time ( , The current sampling time Control input predicted value and current sampling time After (The set consisting of the control input prediction values ​​at each sampling time).

[0087] Step 6.3: For the network packet loss compensator, the current sampling time... Control input predicted value and current sampling time After The control input prediction values ​​at each sampling time are collectively used as the current sampling time. Data packets (i.e.) As of the current sampling time Data packets (of which The data in the data is from the current time of adoption. up to the sampling time The data is processed in real time and input into the network packet loss compensator (NPLC), which has a data buffer to store the received data packets. The data in the data packets are sorted in order from the earliest to the latest sampling time. When the network packet loss compensator receives a new data packet, it overwrites the original data packet with the new data packet.

[0088] When the networked control system is at the current sampling time When packet loss occurs (the network packet loss compensator cannot receive the current sampling time) The data packets indicate that the network packet loss compensator has recently saved the data from the previous sampling time. If the data packet is lost, the network packet loss compensator will adjust the previous sampling time. The second data point in the data packet is used as the output, and the second data point is the current sampling time. The predicted control input value is then used to adjust the current sampling time of the networked control system. Provide compensation;

[0089] When the networked control system is at the current sampling time When no data packet loss occurs (the network packet loss compensator receives the current sampling time) The data packets indicate that the network packet loss compensator has recently saved the data at the current sampling time. If the data packet is lost, the network packet loss compensator will adjust the current sampling time. The first data in the data packet is used as the output, and the first data is the current sampling time. The predicted control input value is then used to adjust the current sampling time of the networked control system. Provide compensation;

[0090] The output of the network packet loss compensator is expressed by the formula:

[0091] ;

[0092] in, The number of consecutively lost data packets. Sampling time Predicted control input values ​​for networked control systems Sampling time right The control input predicted value.

[0093] In the specific experiment of this embodiment, the system matrix of the networked control system is: In time An external disturbance is applied within s. The expression for the external disturbance is: The sampling period of the networked control system is set to... The time delay and data packet loss in the experimental environment are respectively as follows: Figure 2 and Figure 3 As shown, where , can be obtained Choose the poles respectively. and The observer gain can be obtained. and state feedback gain The time constant of the filter is designed to be... The LSTM neuron is configured with 5 input nodes, 2 hidden layers, and 1 output node. The perturbation predictor and CSPG will then... The system performs 5-step prediction (in this embodiment, it is 5-step prediction) and then compensates for the impact of packet loss in real time using NPLC.

[0094] like Figure 4 and Figure 5 This application (hereinafter referred to as the EID+LSTM method) compares its superiority with the traditional equivalent input disturbance method (EID method) and an equivalent input disturbance method incorporating networked predictive control (hereinafter referred to as the EID+NPC method). The EID+NPC method uses the NPC method to perform a 5-step prediction of the state feedback controller output. Two sets of experiments were conducted with packet loss rates of 20% and 40%, respectively. The output of the networked control system in this application is compared with that of the two existing technologies (i.e., the EID method and the EID+NPC method). Figure 4 and Figure 5 As shown (where the y-axis represents the output of the networked control system) As can be seen, in the case of packet loss, the overall disturbance suppression effect of the method proposed in this application is better than that of the EID method and the EID+NPC method. As the packet loss probability increases, the disturbance suppression effect of the EID method and the EID+NPC method deteriorates, while the method proposed in this application still maintains good disturbance suppression performance, which verifies the effectiveness of the method proposed in this application and its robustness to network communication constraints.

[0095] In another embodiment, the present invention also provides a disturbance prediction and accurate compensation device for a networked control system, including a processor and a memory storing a plurality of computer instructions, wherein the computer instructions, when executed by the processor, implement the steps of the disturbance prediction and accurate compensation method for the networked control system.

[0096] For specific limitations on the disturbance prediction and accurate compensation device for networked control systems, please refer to the limitations on the disturbance prediction and accurate compensation method for networked control systems mentioned above, which will not be repeated here.

[0097] The memory and processor are electrically connected directly or indirectly to enable data transmission or interaction. For example, these components can be electrically connected to each other via one or more communication buses or signal lines. The memory stores a computer program that can run on the processor, and the processor implements the method of the present invention by running the computer program stored in the memory.

[0098] The memory may be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The memory is used to store programs, which the processor executes upon receiving execution instructions.

[0099] The processor may be an integrated circuit chip with data processing capabilities. The aforementioned processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor.

[0100] This method and apparatus for accurate disturbance prediction and compensation in networked control systems addresses the challenges of external disturbances, time delays, and packet loss. By employing an EID estimator, it effectively reduces the impact of time delays on the performance of the networked control system. Furthermore, a disturbance predictor based on multiple LSTM neural networks is designed, enabling multi-step accurate disturbance prediction and enhancing the disturbance suppression capability of the networked control system. A networked predictive controller is also incorporated, allowing for real-time compensation of the impact of packet loss on the networked control system's performance in the event of continuous packet loss, thereby improving the control accuracy and meeting the requirements of high-performance networked control systems.

[0101] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0102] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A method for accurate disturbance prediction and compensation in a networked control system, characterized in that: The disturbance prediction and accurate compensation method for the networked control system includes: The impact of time delay on the networked control system and external disturbances to the networked control system are treated as lumped disturbances, and an equivalent input disturbance state space model of the networked control system is established. Based on the equivalent input disturbance state-space model, an EID estimator is designed, which includes a state observer and a filter. Based on the state observer, the state estimate and EID estimate of the networked control system are obtained. The EID estimate is obtained by passing the EID estimate through a filter, and is called the filtered EID estimate. For each sampling time: Current sampling time and the current sampling time The former The filtered EID estimate at each sampling time is used as the input to the disturbance predictor to obtain the sampling time. After The predicted value of the EID estimate at each sampling time, and the perturbation predictor includes A series of sequentially connected LSTM neural networks; Current sampling time The state estimate of the networked control system is used as the input to the state feedback controller to obtain the current sampling time. The output of the state feedback controller; Current sampling time The output of the state feedback controller and the current sampling time After The predicted value of the EID estimate at each sampling time, and the current sampling time. The filtered EID estimate is used as the input to the networked predictive controller, which outputs the predicted control input value at each sampling time to compensate the networked control system.

2. The method for accurate disturbance prediction and compensation in a networked control system as described in claim 1, characterized in that: The current sampling time The former The filtered EID estimates at each sampling time point constitute the EID estimate dataset. For the EID estimation dataset The standardization process is performed to obtain the standardized EID estimate dataset. The standardized EID estimate dataset Each data point in the dataset is used as a training sample to train the LSTM neural network, resulting in a trained LSTM neural network. Each LSTM neural network in the perturbation predictor is a trained LSTM neural network.

3. The method for accurate disturbance prediction and compensation in a networked control system as described in claim 1, characterized in that: The current sampling time is obtained. After The predicted values ​​of the EID estimates at each sampling time point include: Current sampling time and the current sampling time The former The filtered EID estimates at each sampling time constitute the first sequence arranged in chronological order of the sampling time. Each data point in the first sequence is then standardized. The first sequence after standardization is used as the input to the first LSTM neural network in the perturbation predictor. Then, the output of the first LSTM neural network is subjected to corresponding destandardization to obtain the current sampling time. sampling time The predicted value of the EID estimate ; sampling time The predicted value of the EID estimate The filtered EID estimate, which was first sampled in the first sequence, was removed to form a second sequence arranged in chronological order of the sampling times. The standardized second sequence is used as the input to the second LSTM neural network in the perturbation predictor. The output of the second LSTM neural network is then subjected to corresponding destandardization to obtain the current sampling time. sampling time The predicted value of the EID estimate ; And so on, repeating. Next, and the first Next time, the standardized first time The sequence is used as the first in the perturbation predictor The input to the LSTM neural network, and then the LSTM neural network... The outputs of each LSTM neural network are denormalized to obtain the current sampling time. sampling time The predicted value of the EID estimate Thus, the current sampling time is obtained. After The predicted value of the EID estimate at each sampling time.

4. The method for accurate disturbance prediction and compensation in a networked control system as described in claim 1, characterized in that: The networked predictive controller includes a predictive control sequence generator and a network packet loss compensator.

5. The method for accurate disturbance prediction and compensation in a networked control system as described in claim 4, characterized in that: For the predictive control sequence generator, the current sampling time The output of the state feedback controller serves as the input to the predictive control sequence generator, which in turn generates the predictive control sequence based on the sampling time. The current sampling time is obtained by predicting the output of the state feedback controller. After The output of the state feedback controller at each sampling time; Current sampling time and the current sampling time After The output of the state feedback controller at each sampling moment corresponds sequentially to the current sampling moment. The filtered EID estimate and the current sampling time After The difference between the predicted values ​​of the EID estimates at each sampling time is used to obtain the current sampling time. and the current sampling time After The predicted control input value at each sampling time.

6. The method for accurate disturbance prediction and compensation in a networked control system as described in claim 5, characterized in that: For the network packet loss compensator, the current sampling time will be used. and the current sampling time After The control input prediction values ​​at each sampling time are collectively used as the current sampling time. The data packets are input to the network packet loss compensator, and the data in the data packets are sorted in the order of the sampling time from first to last. When the network packet loss compensator receives a new data packet, it overwrites the original data packet with the new data packet. When the networked control system is at the current sampling time When packet loss occurs, it means that the network packet loss compensator is currently storing the data from the previous sampling time. If the data packet is lost, the network packet loss compensator will adjust the previous sampling time. The second data point in the data packet is used as the output, and the second data point is the current sampling time. The predicted control input value is then used to adjust the current sampling time of the networked control system. Provide compensation; When the networked control system is at the current sampling time If no data packet loss occurs, it means that the network packet loss compensator is currently saving the data at the sampling time. If the data packet is lost, the network packet loss compensator will adjust the current sampling time. The first data in the data packet is used as the output, and the first data is the current sampling time. The predicted control input value is then used to adjust the current sampling time of the networked control system. Compensation will be provided.

7. A disturbance prediction and accurate compensation device for a networked control system, comprising a processor and a memory storing a plurality of computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 6.

Images