Search

CN-121983986-A - CPPS small-interference stable control method based on LSTM fuzzy self-adaptive PID

CN121983986ACN 121983986 ACN121983986 ACN 121983986ACN-121983986-A

Abstract

The invention provides a CPPS small interference stable control method based on LSTM fuzzy self-adaptive PID, belonging to the technical field of electric power and small interference stable control, comprising the steps of collecting historical network delay data to train an LSTM model to obtain an LSTM prediction model; the method comprises the steps of collecting the rotating speed deviation (error) and the rotating speed deviation change rate (error change rate) of a synchronous generator, predicting network delay at the current moment by utilizing an LSTM prediction model, dynamically setting the proportional coefficient, the integral coefficient and the differential coefficient of a PID controller through fuzzy logic decision, inputting the set proportional coefficient, integral coefficient and differential coefficient into the PID controller to generate control signals, enabling the control signals to act on the synchronous generator of an information physical power system, and realizing the small-interference stable control of the information physical power system. The invention solves the problem of system instability caused by network delay randomness in CPPS.

Inventors

  • WANG LIWEI
  • CHEN SHI

Assignees

  • 四川大学

Dates

Publication Date
20260505
Application Date
20251209

Claims (10)

  1. 1. The CPPS small-interference stable control method based on LSTM fuzzy self-adaptive PID is characterized by comprising the following steps: acquiring historical network delay data generated in the operation process of an information physical power system of a stand-alone infinite power system with network communication delay, and training an LSTM model based on the historical network delay data to obtain an LSTM prediction model capable of predicting future network delay; Collecting the rotating speed deviation and the rotating speed deviation change rate of the synchronous generator through a monitoring unit of the information physical power system, defining the rotating speed deviation as an error, and defining the rotating speed deviation change rate as an error change rate; Predicting network delay at the current moment by utilizing an LSTM prediction model, combining errors and error change rates, dynamically setting the proportional coefficient, the integral coefficient and the differential coefficient of the PID controller through a fuzzy logic decision, and outputting the set proportional coefficient, integral coefficient and differential coefficient according to a preset proportional coefficient control rule, an integral coefficient control rule and a differential coefficient control rule; and inputting the set proportional coefficient, integral coefficient and differential coefficient to a PID controller, and generating a control signal by the PID controller according to the error and the error change rate, wherein the control signal acts on a synchronous generator of the information physical power system to realize the small-interference stable control of the information physical power system.
  2. 2. The LSTM-based fuzzy self-adaptive PID CPPS small-interference stable control method is characterized in that an information physical power system of a single-machine infinite power system with network communication delay is built, a TrueTime module is adopted in the information physical power system to simulate network delay, core parameters of the single-machine infinite power system are determined, the core parameters comprise synchronous generator capacity, machine end voltage, excitation comprehensive amplification coefficient, data transmission rate, minimum data frame, initial active input and simulation total duration, and small disturbance occurrence time and active input values after disturbance are set in the simulation total duration.
  3. 3. The LSTM-based fuzzy adaptive PID CPPS small interference stabilization control method of claim 1, wherein the LSTM model is trained based on historical network delay data, in particular comprising: Denoising and normalizing the historical network delay data, splitting the preprocessed historical network delay data into an input sequence and a corresponding target sequence according to a preset time step, and forming a supervised learning pair; Initializing parameters of an LSTM model and an Adam optimizer, wherein the parameters of the LSTM model comprise a weight matrix and an offset, and the parameters of the Adam optimizer comprise a learning rate, a first-order momentum attenuation rate, a second-order momentum attenuation rate and a numerical stability constant; inputting the supervised learning pair into the initialized LSTM model, carrying out forward propagation calculation step by step, and outputting a network delay predicted value; Calculating the error between the network delay predicted value and the real network delay in the target sequence by taking the mean square error as a loss function, and transmitting an error gradient along the network structure of the LSTM model through a back propagation algorithm; And updating parameters of the LSTM model by using an Adam optimizer according to the error gradient, iterating until the iteration number reaches the preset iteration number or the loss function value is smaller than a preset error threshold value, and stopping training to obtain the LSTM prediction model.
  4. 4. The CPPS small-interference stable control method based on the LSTM fuzzy self-adaptive PID is characterized in that a monitoring unit of an information physical power system is used for collecting the rotating speed deviation and the rotating speed deviation change rate of a synchronous generator, specifically, the monitoring unit is used for collecting the actual operating rotating speed of the synchronous generator in real time, calculating the difference value between the actual operating rotating speed and the rated rotating speed to obtain the rotating speed deviation, continuously sampling the rotating speed deviation at the time interval of 10ms, and calculating the ratio of the difference value between two adjacent sampling values to the time interval to obtain the rotating speed deviation change rate.
  5. 5. The LSTM-based CPPS small-interference stable control method based on fuzzy self-adaptive PID is characterized in that core parameters of a single-machine infinite power system are specifically that the capacity of a synchronous generator is 325.5MW, the voltage of a machine end is 10.5kV, the excitation comprehensive amplification coefficient=10, the data transmission rate is 80kb/s, the minimum data frame is 5 bytes, the initial active input is 0.7376pu, the simulation total duration is 30s, the occurrence time of small disturbance is 15s, and the active input value after disturbance is 1.7pu.
  6. 6. The CPPS small interference stable control method based on LSTM fuzzy self-adaptive PID according to claim 1, wherein the method is characterized by outputting the set proportional coefficient, integral coefficient and differential coefficient according to the preset proportional coefficient control rule, integral coefficient control rule and differential coefficient control rule, and specifically comprises the steps of mapping the actual values of errors and error change rates to corresponding integers in a fuzzy logic decision theory domain, inquiring the proportional coefficient adjustment coefficient corresponding to the integer combination according to the proportional coefficient control rule, inquiring the integral coefficient adjustment coefficient corresponding to the integer combination according to the integral coefficient control rule, inquiring the differential coefficient adjustment coefficient corresponding to the integer combination according to the differential coefficient control rule, multiplying the proportional coefficient adjustment coefficient, the integral coefficient adjustment coefficient and the differential coefficient adjustment coefficient with the initial proportional coefficient, the initial integral coefficient and the initial differential coefficient of a PID controller respectively to obtain the set proportional coefficient, the integral coefficient and the differential coefficient, wherein the initial proportional coefficient of the PID controller is 0.5, the initial integral coefficient is 0.1, and the initial differential coefficient is 0.6.
  7. 7. The CPPS small interference stable control method based on LSTM fuzzy self-adaptive PID according to claim 2, wherein TrueTime modules are adopted to simulate network delay, specifically, a TrueTime2.0 module is embedded in a communication link between an information acquisition node and a control node of an information physical power system, the TrueTime2.0 module calculates time consumption of data packing, transmission and unpacking according to a set data transmission rate and a minimum data frame, simulates signal acquisition delay and communication transmission delay, and outputs current network delay data to a control unit of the information physical power system in real time for acquisition of historical network delay data.
  8. 8. The CPPS small-disturbance stable control method based on the LSTM fuzzy self-adaptive PID according to claim 1, wherein the monitoring unit of the information physical power system is used for collecting the power angle, the rotating speed and the active output data of the synchronous generator in real time, calculating the overshoot of the rotating speed and the active output after small disturbance, and judging that the small-disturbance stable control effect reaches the standard when the rotating speed overshoot is less than or equal to 2.4% and the active output overshoot is less than or equal to 100.1%, wherein the calculation mode of the overshoot is (data maximum value-steady state value)/steady state value multiplied by 100%.
  9. 9. The CPPS small interference stable control method based on LSTM fuzzy self-adaptive PID according to claim 1, wherein the historical network delay data is divided into a training set and a test set according to the ratio of 7:3, after the LSTM model training is completed by the training set, the test set is input into the LSTM prediction model, the root mean square error and the average absolute error of the predicted value and the true value of the test set are calculated, and when the average absolute error of the test set is less than or equal to 0.006 and the root mean square error is less than or equal to 0.008, the LSTM prediction model is judged to meet the use requirement.
  10. 10. The CPPS small-interference stable control method based on the LSTM fuzzy self-adaptive PID is characterized in that a PID controller generates control signals according to errors and error change rates, specifically, the PID controller carries out integral operation on the errors to obtain integral terms, carries out differential operation on the error change rates to obtain differential terms, adds the proportional terms, namely the proportional coefficient multiplied by the errors, the integral terms and the differential terms to obtain the amplitude of the control signals, adjusts the excitation voltage of a synchronous generator according to the amplitude of the control signals, and further adjusts the electromagnetic power of the generator to realize small-interference stable control.

Description

CPPS small-interference stable control method based on LSTM fuzzy self-adaptive PID Technical Field The invention relates to the technical field of stable control of electric power and small interference, in particular to a CPPS small interference stable control method based on LSTM fuzzy self-adaptive PID. Background Because of the difficulty in avoiding small interference in the operation process of the power system, a system with unstable small interference is difficult to normally operate in practice. Thus, small disturbance stable control of an electrical power system is the most basic and important control task of the electrical power system. The interdependence relationship between the information network and the physical power network brings convenience and also brings new challenges. From the data transmission dimension analysis, the randomness problem of network delay including signal acquisition and communication delay exists, and the control unit response abnormality may be caused, and the stable operation of the system is destroyed when serious, so that the control unit has become one of key bottlenecks for restricting the development of CPPS (information physical power system (Cyber-Physical Power Systems)). The intelligent optimization control theory provides a new technical path for the stable operation of the system by fusing an advanced algorithm, so that a stronger control scheme is designed by combining an intelligent optimization technology and a time sequence prediction method on the basis of fully considering the time delay dynamic random characteristic, and the intelligent optimization control theory has important engineering application value for realizing the stable operation of the power system. The small disturbance most frequently encountered in the operation process of the power system is easy to cause the low-frequency electromechanical oscillation phenomenon with the frequency of 0.1-2.5 Hz. To suppress low frequency oscillations, improve small disturbance stability, it is often employed to reduce line impedance, or employ power system stabilizers (Power System Stabilizer, PSS) to rapidly respond to generator terminal voltage changes, providing additional damping torque. Under ideal time-lapse-free conditions, PSS has gained widespread use in engineering practice, by virtue of its excellent economy and engineering applicability, recognized as a key device to enhance electromechanical oscillation damping. Aiming at network delay, the excitation compensation control strategy for wide-area time-lag scenes based on the unified Smith predictor can eliminate the influence of time lags on a system through the Smith predictor, can completely compensate fixed time lags theoretically, and is suitable for a first-order time lag system with accurate model. However, sensitive to model accuracy, it is difficult to cope with time-varying lags or complex nonlinear systems. A wide-area damping control strategy integrating fuzzy logic and a time-lag compensation mechanism improves dynamic performance through rolling optimization and feedback control. The method relies on remote signal transmission, and if the communication delay exceeds the estimated range, control failure can be caused, and the model complexity is high. The time lag, parameter uncertainty and external disturbance are processed through Lyapunov-Krasovskii (L-K) functional, linear matrix inequality and other technologies, and the method is suitable for complex scenes such as a mixed time lag varying system and the like. The LMI (linear matrix inequality) or the functional with complex structure is needed to be solved on line, the calculation complexity is high, and the real-time performance is poor. In the aspect of delay robust controller design, a Pade approximation method is adopted to approximate a delay link and convert the delay link into a transfer function form, so that a WAMS (wide area measurement system) -based wide area controller considering delay is designed, and the method is a damping design aiming at fixed delay time. There is no small interference stable control method for network randomness. Disclosure of Invention The invention provides a CPPS small-interference stable control method based on LSTM fuzzy self-adaptive PID, which solves the problem of system instability caused by network delay randomness in CPPS. In order to achieve the above purpose, the invention adopts the following technical scheme: A CPPS small interference stable control method based on LSTM fuzzy self-adaptive PID comprises the following steps: acquiring historical network delay data generated in the operation process of an information physical power system of a stand-alone infinite power system with network communication delay, and training an LSTM model based on the historical network delay data to obtain an LSTM prediction model capable of predicting future network delay; Collecting the rotating speed deviation and the rotating speed