CN-122000951-A - Time constant-based quantitative analysis system for reactive transient stability of camera

CN122000951ACN 122000951 ACN122000951 ACN 122000951ACN-122000951-A

Abstract

The application relates to the technical field of cameras, in particular to a time constant-based reactive transient stability quantitative analysis system for a camera. The simulation test device comprises a data acquisition device, a data enhancement device, a simulation test device, a model parameter correction device and a time constant calculation device. The scheme carries out intelligent amplification on sparse actually measured disturbance data (power angle/rotating speed deviation curve) through an countermeasure network model to generate an extended sample set covering multiple working conditions, and drives a simulation model to carry out closed-loop optimization based on the extended sample set, thereby obviously improving the identification precision of a damping coefficient (D), further ensuring the reliability of time constant calculation and finally providing an authoritative analysis report with high precision and low data dependence for the quantitative evaluation of the reactive transient stability of the phase regulator.

Inventors

YANG YUNLI
ZHANG QIANG
LIU GUANSONG
ZHAO JUN
MA GUANGJUN

Assignees

中电投新疆能源化工集团托里有限责任公司

Dates

Publication Date: 20260508
Application Date: 20260116

Claims (10)

1. The utility model provides a camera reactive transient stability quantization analysis system based on time constant which characterized in that includes: the data acquisition device is used for acquiring a corresponding power angle deviation curve and a rotor rotating speed deviation curve of the phase modulation unit under different disturbance information to generate an actual sample set; the data enhancement device is used for carrying out sample amplification on the actual sample set based on the countermeasure network model to generate an extended sample set; The simulation test device is internally provided with a phase modulation unit simulation model, and generates a simulation result based on disturbance information in the expansion sample set, wherein the simulation result comprises a simulated power angle deviation curve and a simulated rotor rotating speed deviation curve; The model parameter correction device is used for adjusting simulation parameters of a simulation model of the phase modulation unit by taking a simulation result and a best fitting result of a real power angle deviation curve and a real rotor rotating speed deviation curve in an expansion sample set as targets, and extracting damping coefficients of the phase modulation unit from the simulation parameters; and a time constant calculating device for calculating a time constant from the damping coefficient and the inertia constant and generating a performance report of the camera based on the time constant.
2. The system for quantized analysis of reactive transient stability of a time constant-based rectifier according to claim 1, wherein the simulation parameters include damping coefficient, synchronous torque coefficient, AVR gain, AVR time constant, PSS gain, PSS lead time constant, and PSS lag time constant.
3. The quantitative analysis system for reactive transient stability of a phase modulator based on time constants of claim 1, wherein the disturbance information is the variation information of the reference voltage of an automatic voltage regulator of the phase modulator group; The single disturbance information is any one of step disturbance, pulse disturbance and pseudo-random disturbance; ; where t represents time, t 0 represents injection time of disturbance information, Representing the reference voltage before the disturbance, Representing the step amplitude; When the disturbance information V ref (t) is a step disturbance: ; representing the step amplitude; when the disturbance information V ref (t) is a pulse disturbance: ; Representing the amplitude of the pulse, Representing pulse width; when the disturbance information V ref (t) is a pseudo-random disturbance: , Represents the kth frequency amplitude, N represents the number of frequency components, k represents the frequency component index, Representing the kth frequency phase.
4. The time constant-based phase modulator reactive transient stability quantitative analysis system of claim 1, wherein the phase modulator set simulation model is: ; ; Wherein D represents a damping coefficient, K represents a synchronous torque coefficient, K A represents an AVR gain, T A represents an AVR time constant, K PSS represents a PSS gain, T 1 represents a PSS lead time constant, T 2 represents a PSS lag time constant, Represents the synchronous angular velocity, M represents the inertial time constant, H represents the inertial constant, Indicating the deviation of the electromagnetic torque, The voltage at the machine end is represented by the voltage at the machine end, The PSS washing and filtering time constant is expressed, A state variable is represented and a state variable is represented, Represents the power angle of the generator, Indicating the actual rotational speed of the rotor, The excitation voltage is indicated as such, Representing the PSS output signal, Representing disturbance signals, f 1 、f 2 、f 3 、f 4 are the inverse of the state variables, Indicating an initial steady state power angle.
5. The time constant-based camera reactive transient stability quantitative analysis system according to any one of claims 1 to 4, wherein the data enhancement device comprises: the data classification module is used for acquiring an actual sample set, and dividing disturbance information in the actual sample set division into a step disturbance information sample set, a pulse disturbance sample set and a pseudo-random disturbance sample set according to the type of the disturbance information; the sample expansion module is internally provided with an countermeasure network model; the training control module trains the countermeasure network model by adopting the step disturbance information sample set to obtain step model parameters; The training control module trains the countermeasure network model by adopting the pulse disturbance information sample set to obtain pulse model parameters; The training control module trains the countermeasure network model by adopting the pseudo-random disturbance information sample set to obtain pseudo-random model parameters; The method comprises the steps of loading step model parameters to an countermeasure network model to expand a step disturbance information sample set; loading pulse model parameters against the network model to expand a pulse-loading disturbance information sample set; The pseudo-random model parameters are loaded against the network model to expand the pseudo-random disturbance information sample set.
6. The time constant based camera reactive transient stability quantitative analysis system of claim 5, wherein the countermeasure network model comprises: a generator for generating an extended sample according to the inputted random information; The discriminator is used for randomly inputting a real sample and an expanded sample and outputting the probability that the input information is the real sample; The generator and the discriminator are both deep neural network models, and training is performed synchronously; wherein the loss function of the discriminator is : ; The loss function of the generator is : ; Representing the i-th real sample, Representing the discrimination probability of the discriminator for the real sample, Representing the i-th expansion sample, Representing the discrimination probability of the discriminator on the extended samples, and m represents the number of samples.
7. The time constant-based phase-change machine reactive transient stability quantitative analysis system according to claim 5, wherein: The random information is the random information of the d dimension, and each dimension obeys the normal distribution of the standard; the dimension of random information of the generated step disturbance information sample set, the pulse disturbance information sample set and the pseudo-random disturbance information sample set is higher and higher.
8. The time constant based phase-change memory reactive transient stability quantitative analysis system of claim 5, wherein the generator comprises: the input layer is used for inputting the noise vector so as to remodel the noise vector to obtain remolded characteristics; an upsampling layer upsampling the plastic-heavy characteristic to obtain an upsampled characteristic; a normalization layer for normalizing the up-sampling feature to obtain a normalized feature; an activation function layer, which activates the normalized feature RelU to generate an activation feature; The convolution layer is provided with a plurality of layers, and the convolution operation is repeated on the activation characteristic to obtain a convolution characteristic; The output layer is used for activating the convolution characteristics by adopting a tanh function to obtain output characteristics, and performing inverse normalization on the output characteristics to generate an expansion sample; The discriminator includes: an information input layer for inputting the extended samples or the real samples to generate initial features; The information convolution layer is used for carrying out convolution operation on the initial characteristics to generate convolution characteristics; The information pooling layer is used for carrying out global average pooling on the convolution characteristics to generate pooled characteristics; the full connection layer maps the pooling features to a low-dimensional space to generate low-dimensional features; And the information output layer is used for generating the sample true probability by adopting a Sigmoid function on the low-dimensional features.
9. The time constant based phase-change machine reactive transient stability quantitative analysis system according to claim 8, wherein the loss function of the countermeasure network in joint training is: ; wherein G represents a generator, D represents a discriminator, Representing the distribution of the real data and, The probability distribution is represented by a graph of the probability distribution, Representation of The average value over the distribution of the real data, Representation of An average value on the probability distribution, x representing the input of the arbiter, z representing the random information; Representing a core cost function; In training the countermeasure network model, the following training cycle is performed: fixed generator G, update arbiter D, maximize ; Fixed arbiter D, update generator G, minimize ; When the discriminator D is updated by the fixed generator G, generating an extended sample by adopting a plurality of noise information, mixing the extended sample with a plurality of real samples and conveying the mixed sample to the discriminator D; For each sample of input discriminators D, a loss function is calculated According to the loss function The back propagation updates the weight parameters inside the discriminant D; Fixed arbiter D, update generator G, minimize When (1): Sampling a plurality of new random information, generating a plurality of extended samples by using a generator G, mixing the extended samples with a plurality of real samples, and conveying the mixed samples to a discriminator D; For each sample, a loss function of the generator is calculated According to the loss function The back propagation updates the weight parameters inside the generator.
10. The system of claim 9, wherein each training of the generator G is performed at the beginning of each cycle, judging whether each neuron in the hidden layer is hidden according to a preset random probability h, freezing the hidden neurons to generate a new hidden layer structure, training the generator G, updating the unfrozen weight parameters in the hidden layer, and recovering the frozen neurons when the cycle is finished.

Description

Time constant-based quantitative analysis system for reactive transient stability of camera Technical Field The application relates to the technical field of cameras, in particular to a time constant-based reactive transient stability quantitative analysis system for a camera. Background The time constant tau is a core comprehensive parameter for representing the dynamic response characteristic of the rotor of the camera, and directly determines the speed of recovering the synchronous speed and the attenuation rate of the power angle oscillation after the system is disturbed. The smaller the tau value is, the stronger the synergistic effect of the rotor kinetic energy and the damping moment is, the more outstanding the capability of the unit to maintain synchronous operation under the condition of power grid voltage fluctuation or fault disturbance is, and the better the reactive support performance (such as voltage recovery speed and oscillation inhibition efficiency) is in the transient process. Thus, τ is a key indicator to evaluate its transient stability margin. The physical basis of the time constant τ depends on two core parameters, the rotor inertia constant (H) and the damping coefficient (D). The inherent capability of the rotor of the unit for storing kinetic energy is determined by the mechanical structure of the equipment, and can be directly obtained as a factory calibration value, while the damping efficiency of the system for restraining oscillation is reflected by D, and the essence of the damping efficiency is the comprehensive embodiment of electric and mechanical damping. Because D is affected by the actual operation condition, the on-site identification and calculation are required to be carried out through dynamic test. The current dynamic identification method (such as frequency domain response method, prony analysis or least square fitting) of the damping coefficient D needs to rely on high-precision synchronously collected multidimensional operation data, including but not limited to rotor power angle, machine end voltage/current, active/reactive power and excitation system variables. However, such high-dimensional data available in practical power grids tend to be sparse and discrete, limited by measurement device placement, communication cost, and data storage capacity, and difficult to fully cover with continuous, complete disturbance process data. Disclosure of Invention The summary of the application is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description. The summary of the application is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter. Some embodiments of the present application provide a time constant-based quantitative analysis system for transient stability of a tuner reactive power, which solves the technical problems mentioned in the background section above. As a first aspect of the present application, some embodiments of the present application provide a time constant-based quantization analysis system for transient stability of a dimmer, comprising: the data acquisition device is used for acquiring a corresponding power angle deviation curve and a rotor rotating speed deviation curve of the phase modulation unit under different disturbance information to generate an actual sample set; the data enhancement device is used for carrying out sample amplification on the actual sample set based on the countermeasure network model to generate an extended sample set; The simulation test device is internally provided with a phase modulation unit simulation model, and generates a simulation result based on disturbance information in the expansion sample set, wherein the simulation result comprises a simulated power angle deviation curve and a simulated rotor rotating speed deviation curve; The model parameter correction device is used for adjusting simulation parameters of a simulation model of the phase modulation unit by taking a simulation result and a best fitting result of a real power angle deviation curve and a real rotor rotating speed deviation curve in an expansion sample set as targets, and extracting damping coefficients of the phase modulation unit from the simulation parameters; and a time constant calculating device for calculating a time constant from the damping coefficient and the inertia constant and generating a performance report of the camera based on the time constant. According to the scheme, sparse actually measured disturbance data (power angle/rotating speed deviation curve) are intelligently amplified through the countermeasure network model, an expansion sample set covering multiple working conditions is generated, closed-loop optimization is driven based on the expansion sample set, identification accuracy of a damping coefficient (D) is remarkably improv