Search

CN-121978440-A - Intelligent voltage stabilizer detection and compensation method and system based on data deviation analysis

CN121978440ACN 121978440 ACN121978440 ACN 121978440ACN-121978440-A

Abstract

The invention discloses a data deviation analysis-based intelligent detection and compensation method and system for a voltage stabilizer, and relates to the technical field of detection of voltage stabilizers, wherein the method comprises the steps of acquiring multi-source data through topology and working condition requirements of a voltage detection system, and constructing a full-working condition effective data set through cooperative preprocessing; the method comprises the steps of quantifying total voltage deviation, establishing a full-working-condition deviation matrix, decomposing the total voltage deviation into static and dynamic deviation, constructing a deviation-influencing factor dynamic correlation model, determining deviation characteristics, constructing a fusion compensation model which fuses a static compensation sub-model and an attention mechanism LSTM dynamic compensation sub-model, executing real-time compensation by combining real-time acquisition data, outputting compensated data and running state labels, and constructing a closed-loop optimization link of passive correction and active pre-judgment based on deviation data and trend prediction. The invention has the advantages of realizing accurate adaptation of all working conditions, having high dynamic response speed, actively pre-judging the working conditions and element changes, and effectively counteracting fixed and real-time variation deviation.

Inventors

  • PENG JIANHONG

Assignees

  • 昆山市宝应源电子有限公司

Dates

Publication Date
20260505
Application Date
20260129

Claims (10)

  1. 1. The intelligent detection and compensation method for the voltage stabilizer based on the data deviation analysis is characterized by comprising the following steps of: acquiring multi-source data comprising original voltage detection data, baseline reference data, environment interference data and working condition characteristic data through the topology of a voltage detection system and the working condition requirements; collaborative preprocessing is carried out on the multi-source data, and an all-condition effective data set is constructed; Quantizing the total voltage deviation, establishing a full-working-condition deviation matrix, decomposing the full-working-condition deviation matrix into static deviation and dynamic deviation, and constructing a dynamic correlation model of deviation-influencing factors; Determining static deviation characteristics and dynamic deviation characteristics, constructing a fusion compensation model of a fusion static compensation sub-model and an attention mechanism LSTM dynamic compensation sub-model, and realizing accurate offset of fixed deviation and real-time-varying deviation; acquiring real-time acquisition data, combining a fusion compensation model, executing real-time compensation operation of voltage detection data, and outputting compensated data and a voltage stabilizer running state label; And constructing a closed-loop optimization link of passive correction and active prediction by combining deviation real-time data and trend prediction, and continuously updating model parameters.
  2. 2. The intelligent detection and compensation method for a voltage stabilizer based on data deviation analysis according to claim 1, wherein the obtaining multi-source data through topology and working condition requirements of a voltage detection system specifically comprises: collecting original voltage detection data at deployed detection nodes, wherein the collection frequency is at least 10 times of the power frequency of a power grid, so that dynamic working condition details are ensured to be captured, and the original voltage detection data comprise an input side voltage instantaneous value/effective value of a voltage stabilizer, an output side voltage instantaneous value/effective value, a collection time stamp, a voltage stabilizer number, load working condition information and a voltage stabilizer working state; Acquiring the voltage range and the precision requirement of a detection node, and selecting a standard voltage source which is matched with the rated output voltage range of the voltage stabilizer and has the precision grade not lower than a preset precision grade as a datum reference source; Under the same load working condition and acquisition frequency, acquiring reference voltage data through a reference source, and ensuring that acquisition time sequences of the reference voltage data and original voltage detection data are completely synchronous, wherein the reference voltage data are dynamic reference ranges of +/-5% of rated output voltage of the voltage stabilizer, and the coverage working conditions comprise steady-state working conditions of no-load, 25%/50%/75%/rated load, and dynamic working conditions of abrupt load change and fluctuation of input voltage; When the original data is acquired, synchronously acquiring environment interference data, internal state data and working condition characteristic data; And (3) performing time sequence alignment and validity verification on the multi-source data, removing time sequence dislocation and acquisition failure data, and establishing a data ledger classified according to the number, the working condition type and the acquisition time of the voltage stabilizer.
  3. 3. The intelligent detection and compensation method for the voltage stabilizer based on the data deviation analysis of claim 2, wherein the collaborative preprocessing is performed on the multi-source data to construct an all-condition effective data set, and the method specifically comprises the following steps: Extracting original voltage data, reference data, interference and working condition characteristic data based on the synchronous data set, and respectively constructing an original data set, a reference data set and an auxiliary data set; Presetting a significance level, identifying and removing abnormal data points of an original data set, setting the size of a sliding window to be 50 data points, detecting data mutation in the window in real time, marking the data mutation as abnormal points of a dynamic working condition, and storing the abnormal points independently; 4 layers of wavelet decomposition is carried out on the original data set with the outlier removed, a high-frequency noise component and a low-frequency effective component are separated, high-frequency noise is further filtered, and preprocessed original detection data is obtained through reconstruction; Performing the same abnormal value rejection and noise reduction processing on the reference data set and the auxiliary data set, ensuring the consistency of the data, and reconstructing the preprocessed reference data; And performing secondary verification on the preprocessed original detection data and the preprocessed reference data, ensuring the consistency of response time sequences of the input voltage and the output voltage of the voltage stabilizer, determining the noise reduction efficiency of the original detection data and the preprocessed reference data, confirming that the preprocessing is effective when the noise reduction efficiency is greater than or equal to a preset noise reduction efficiency threshold value, and outputting an all-condition effective data set, otherwise, adjusting the number of wavelet transformation decomposition layers to reprocess.
  4. 4. The intelligent detection and compensation method for voltage stabilizer based on data deviation analysis according to claim 3, wherein the quantifying total voltage deviation, establishing a full-condition deviation matrix, decomposing the full-condition deviation matrix into static deviation and dynamic deviation, and establishing a dynamic correlation model of deviation-influencing factors, specifically comprising: determining total voltage deviation at time sequence points one by one based on the preprocessed original detection data and reference data, and constructing a full-working-condition deviation matrix of working-condition type-time-deviation values, wherein the total voltage deviation is the difference value between the detection voltage at the output side of the voltage stabilizer and the rated output reference voltage; Extracting deviation data of steady-state working conditions in a full-working-condition deviation matrix, acquiring the average value of the fixed deviation data of output voltage under each steady-state working condition, and obtaining static deviation by adopting weighted average to represent inherent errors of elements and circuit drift; determining dynamic deviation based on the total voltage deviation and the static deviation, constructing a dynamic deviation data set, and classifying and labeling according to the type of the dynamic working condition, wherein the dynamic deviation is a time-varying deviation caused by input voltage fluctuation, load mutation and internal temperature rise after the static deviation is removed from the total deviation; Based on the environmental interference data, the internal state data, the working condition characteristic data and the dynamic deviation data set, determining the correlation coefficient of each environmental factor and the dynamic deviation; According to the absolute value sequence of the correlation coefficient, selecting environmental factors which are larger than or equal to a preset correlation coefficient threshold as key influence factors, establishing a time sequence correlation model of the key influence factors and the dynamic deviation, and determining the action weight of each factor on the dynamic deviation.
  5. 5. The intelligent detection and compensation method for a voltage stabilizer based on data deviation analysis according to claim 4, wherein the determining of the static deviation feature and the dynamic deviation feature constructs a fusion compensation model that fuses a static compensation sub-model and an attention mechanism LSTM dynamic compensation sub-model, and specifically comprises: Based on the static deviation and the steady state data under all working conditions, fitting the linear relation between the static deviation and the preprocessed detection data to obtain a static compensation coefficient, and further determining a static compensation submodel by solving the static compensation coefficient and a static constant term which minimize the secondary linear function value; Dividing a training set and a testing set based on a key influence factor and a dynamic deviation data set, wherein the training set accounts for 70 percent, the testing set accounts for 30 percent, taking the key influence factor, the working condition type code and the deviation change rate as input characteristics, taking the dynamic deviation as an output label, and constructing a dynamic compensation submodel; based on the static compensation sub-model and the dynamic compensation sub-model, constructing a fusion compensation model, and determining a final fusion compensation formula to obtain compensated data; Screening a verification set from the preprocessed reference data, verifying the fusion compensation model, and obtaining a fusion compensation model with the verification completed; and inputting the preprocessed detection data and the key influence factors into a fusion compensation model after verification, determining the deviation between the compensated data and the preprocessed reference data, and confirming that the model is constructed effectively when the mean value of the absolute values of the deviation is smaller than or equal to a preset reference threshold value.
  6. 6. The intelligent detection and compensation method for voltage regulator based on data deviation analysis according to claim 5, wherein the acquiring real-time collected data, combining with the fusion compensation model, performing real-time compensation operation of voltage detection data and outputting compensated data and operation state label of the voltage regulator specifically comprises: acquiring original voltage real-time data and corresponding environment interference real-time data in real time through a sensor of a detection node and a data acquisition terminal; performing outlier rejection and wavelet noise reduction treatment on the original voltage real-time data to obtain preprocessed original voltage real-time data, and synchronously preprocessing the environment interference real-time data, rejecting outliers and normalizing; inputting the original voltage real-time data into a static compensation sub-model to obtain a real-time static compensation value, thereby obtaining a static compensation real-time coefficient; Inputting the preprocessed environment interference real-time data into a dynamic compensation sub-model to obtain a real-time dynamic deviation predicted value; based on a final fusion compensation formula, acquiring a voltage value after real-time compensation, and simultaneously recording matched record information, wherein the matched record information comprises a compensation time stamp, a sensor number, corresponding working condition information and a voltage stabilizer running state label; Synchronously transmitting the voltage value after real-time compensation and matched record information to a voltage monitoring platform; and generating a real-time compensation ledger on the monitoring platform, and synchronously displaying the original data, the preprocessing data, the compensation data and the deviation value.
  7. 7. The intelligent detection and compensation method for voltage stabilizer based on data deviation analysis according to claim 6, wherein the steps of comparing the compensated data with the reference data, iteratively optimizing compensation model parameters, and forming a continuously optimized compensation link comprise: Extracting a real-time compensated voltage value from a real-time compensation ledger, and synchronously acquiring reference voltage data under the same time sequence and the same working condition by combining a reference source; acquiring the compensated deviation, counting the absolute value distribution of the voltage value after real-time compensation, and setting a compensation precision threshold; judging whether the absolute value of the compensated deviation exceeds the compensation precision threshold value, if not, maintaining the parameters of the current fusion compensation model unchanged, and continuously executing real-time compensation; If the data exceeds the threshold value and the data proportion is more than or equal to five percent, triggering a model optimization flow, extracting original voltage real-time data exceeding the threshold value period, environment interference data, reference voltage data and deviation data under the same working condition, and supplementing the original voltage real-time data, the environment interference data and the reference voltage data and the deviation data to an effective data set and a total deviation data set; Retraining the static compensation sub-model, updating a static compensation coefficient and a static constant term, retraining the dynamic compensation sub-model, optimizing model parameters, reducing loss function values, and obtaining an optimized fusion compensation model; And replacing the original model with the optimized fusion compensation model, putting into a real-time compensation process, and recording model optimization time and accuracy comparison data before and after optimization.
  8. 8. A voltage regulator intelligent detection and compensation system based on data deviation analysis, for implementing the compensation method according to any one of claims 1 to 7, comprising: The multi-source data acquisition module is used for acquiring multi-source data comprising original voltage detection data, datum reference data, environment interference data and working condition characteristic data through the topology and working condition requirements of the voltage detection system; the multi-source data collaborative preprocessing module is used for collaborative preprocessing of multi-source data and constructing an all-condition effective data set; the deviation analysis and dynamic association module is used for quantifying total voltage deviation, establishing a full-working-condition deviation matrix, decomposing the full-working-condition deviation matrix into static deviation and dynamic deviation, and constructing a dynamic association model of deviation-influencing factors; The fusion compensation model construction module is used for determining static deviation characteristics and dynamic deviation characteristics, constructing a fusion compensation model of a fusion static compensation sub-model and an attention mechanism LSTM dynamic compensation sub-model, and realizing accurate offset of fixed deviation and real-time variation deviation; The real-time compensation and closed-loop optimization module is used for acquiring real-time acquisition data, combining the fusion compensation model, executing real-time compensation operation of voltage detection data, outputting compensated data and a voltage stabilizer running state label, and simultaneously combining deviation real-time data and trend prediction to construct a closed-loop optimization link with passive correction and active pre-judgment, and continuously updating model parameters.
  9. 9. The intelligent data bias analysis-based voltage regulator detection and compensation system of claim 8, wherein the bias analysis and dynamic association module comprises: The total voltage deviation quantization unit is used for quantizing the total voltage deviation and establishing a full-working-condition deviation matrix; the static and dynamic deviation decomposition unit is used for decomposing the all-condition deviation matrix into static deviation and dynamic deviation; The dynamic association model building unit is used for building a dynamic association model of deviation-influence factors; And the deviation feature extraction unit is used for extracting static deviation features and dynamic deviation features and providing data support for subsequent model construction.
  10. 10. The intelligent detection and compensation system of a voltage regulator based on data deviation analysis of claim 8, wherein the fusion compensation model building module comprises: The static compensation sub-model building unit is used for building a static compensation sub-model based on the static deviation characteristics; The dynamic compensation sub-model building unit is used for building an attention mechanism LSTM dynamic compensation sub-model based on the dynamic deviation characteristics; The fusion model integration unit is used for fusing the static compensation submodel and the attention mechanism LSTM dynamic compensation submodel to form a fusion compensation model; The model verification unit is used for verifying the validity of the fusion compensation model and ensuring that the fusion compensation model can realize accurate offset of fixed deviation and real-time-varying deviation.

Description

Intelligent voltage stabilizer detection and compensation method and system based on data deviation analysis Technical Field The invention relates to the technical field of voltage stabilizer detection, in particular to a voltage stabilizer intelligent detection and compensation method and system based on data deviation analysis. Background The voltage stabilizer is a core voltage stabilizing component of a power system and industrial equipment, and the voltage stabilizing precision of the voltage stabilizer directly influences the operation safety of downstream equipment. In actual operation, the voltage stabilizer is easy to cause problems of increased output voltage deviation, delayed dynamic adjustment response and the like due to factors such as internal element aging, input voltage fluctuation, load mutation, environmental temperature and humidity change and the like. The existing intelligent voltage detection compensation method and system based on data deviation analysis are not combined with the characteristics of voltage regulator input and output voltage coupling and steady state/dynamic working condition adaptation, and only aiming at general voltage data compensation, the actual running state (such as failure precursor of a voltage regulator module and load adaptation deviation) of the voltage regulator is difficult to accurately reflect, so that the compensation pertinence is insufficient and the long-term stability is poor. Disclosure of Invention In order to solve the technical problems, the technical scheme provides a voltage stabilizer intelligent detection and compensation method and system based on data deviation analysis, and solves the problems that the existing voltage detection intelligent compensation method and system based on data deviation analysis provided in the background art does not combine the characteristics of voltage stabilizer input and output voltage coupling and steady state/dynamic working condition adaptation, only aims at general voltage data compensation, and is difficult to accurately reflect the actual running state of the voltage stabilizer (such as failure precursors of a voltage stabilizing module and load adaptation deviation), and the compensation pertinence is insufficient and the long-term stability is poor. In order to achieve the above purpose, the invention adopts the following technical scheme: the intelligent voltage stabilizer detection and compensation method based on data deviation analysis comprises the following steps: acquiring multi-source data comprising original voltage detection data, baseline reference data, environment interference data and working condition characteristic data through the topology of a voltage detection system and the working condition requirements; collaborative preprocessing is carried out on the multi-source data, and an all-condition effective data set is constructed; Quantizing the total voltage deviation, establishing a full-working-condition deviation matrix, decomposing the full-working-condition deviation matrix into static deviation and dynamic deviation, and constructing a dynamic correlation model of deviation-influencing factors; Determining static deviation characteristics and dynamic deviation characteristics, constructing a fusion compensation model of a fusion static compensation sub-model and an attention mechanism LSTM dynamic compensation sub-model, and realizing accurate offset of fixed deviation and real-time-varying deviation; acquiring real-time acquisition data, combining a fusion compensation model, executing real-time compensation operation of voltage detection data, and outputting compensated data and a voltage stabilizer running state label; And constructing a closed-loop optimization link of passive correction and active prediction by combining deviation real-time data and trend prediction, and continuously updating model parameters. In an optional embodiment, the acquiring the multi-source data through the topology and the working condition requirement of the voltage detection system specifically includes: collecting original voltage detection data at deployed detection nodes, wherein the collection frequency is at least 10 times of the power frequency of a power grid, so that dynamic working condition details are ensured to be captured, and the original voltage detection data comprise an input side voltage instantaneous value/effective value of a voltage stabilizer, an output side voltage instantaneous value/effective value, a collection time stamp, a voltage stabilizer number, load working condition information and a voltage stabilizer working state; Acquiring the voltage range and the precision requirement of a detection node, and selecting a standard voltage source which is matched with the rated output voltage range of the voltage stabilizer and has the precision grade not lower than a preset precision grade as a datum reference source; Under the same load working condition and acquisi