CN-122020579-A - Short-circuit parameter change trend prediction method, system, equipment and medium

CN122020579ACN 122020579 ACN122020579 ACN 122020579ACN-122020579-A

Abstract

The invention discloses a short-circuit parameter change trend prediction method a system, a device and a medium for the same, relates to the technical field of operation and maintenance of a power grid, the system comprises a data acquisition and processing module, a model construction module, a characteristic analysis module, a fusion prediction module and a monitoring deduction module, and the method comprises the steps of data acquisition and processing, model construction, characteristic analysis, fusion prediction and monitoring deduction. According to the invention, nonlinear characteristics of the short-circuit parameters and complex time sequence dependency relations are effectively captured through a multi-model fusion strategy, and the prediction precision and robustness are remarkably improved. And a multi-level abnormality detection mechanism is also established, various abnormality modes are accurately identified, hidden danger of equipment is early warned, and preventive maintenance is assisted. Meanwhile, by utilizing online learning and a dynamic weight adjustment algorithm, the model is ensured to have excellent environmental adaptability and real-time computing capability. The method can provide scientific basis for operation and maintenance decisions of the power grid, optimize maintenance strategies, reduce operation cost and improve equipment reliability.

Inventors

HE SIYANG
XIE YANGHUA
YANG JUNQI
Min xun
ZHAO PENGCHENG
WANG HUAIYUAN
LI JINJUN
Rong Weikai
Wan Zongxu
HE ZIWEI
WU JINCHENG
YU HAILIN
HE GUANGLU
WANG SHUAI
PAN FUXIANG
HUANG JUNCHENG
JIN JUFENG
ZHANG CHANGZI
ZHAO SHIQIN

Assignees

贵州电网有限责任公司

Dates

Publication Date: 20260512
Application Date: 20251128

Claims (10)

1. The short-circuit parameter change trend prediction method is characterized by comprising the following steps of: Short-circuit parameter data of power grid equipment are obtained, and a multidimensional characteristic time sequence is constructed; processing the multi-dimensional characteristic time sequence to obtain a stabilized input sample; Constructing a heterogeneous feature prediction model system comprising a linear statistical evolution model and a nonlinear intelligent mapping model based on the stabilized input sample; Analyzing linear autocorrelation characteristics of the multi-dimensional characteristic time sequence through the linear statistical evolution model, and capturing long-term and short-term dependence and high-dimensional nonlinear characteristics of the multi-dimensional characteristic time sequence through the nonlinear intelligent mapping model; obtaining a basic prediction set according to the linear autocorrelation characteristic, the long-short-term dependence relationship and the high-dimensional nonlinear characteristic; Calculating a real-time prediction residual error when the linear statistical evolution model and the nonlinear intelligent mapping model output the basic prediction set, dynamically updating a weight coefficient corresponding to the linear statistical evolution model and the nonlinear intelligent mapping model according to the real-time prediction residual error, and carrying out weighted fusion on the basic prediction set by utilizing the updated weight coefficient to obtain a high-precision trend prediction value; Calculating an anomaly score based on the deviation of the high-precision trend predicted value and the actual observed value and combining a statistical process control theory to quantify the parameter deviation degree and identify an anomaly mode; And synthesizing the high-precision trend predicted value and the anomaly score, generating a parameter evolution panoramic map, deducing the future trend of the short-circuit parameter and outputting predicted information.
2. The short-circuit parameter variation trend prediction method according to claim 1, wherein: said processing said multi-dimensional feature time series to obtain smoothed input samples comprises: calculating statistical characteristic parameters of the multidimensional characteristic time sequence, and judging the stability of the sequence based on a comparison result of the statistical characteristic parameters and a theoretical critical value; Constructing a differential iteration logic, and if the judging result is non-stable, executing differential operation with increasing order on the multi-dimensional characteristic time sequence; And checking the sequence statistical characteristics after differential operation until a stable sequence meeting the stability judging condition is obtained, and taking the stable sequence as the stabilization input sample.
3. The short-circuit parameter variation trend prediction method according to claim 1, wherein: the analyzing the linear autocorrelation characteristic of the multi-dimensional characteristic time sequence through the linear statistical evolution model comprises the following steps: constructing an autoregressive polynomial logic, and quantifying the linear dependency strength between the current moment value and the historical moment value of the multidimensional feature time sequence; Constructing a moving average polynomial logic, describing the hysteresis effect of a prediction error and the linear correction effect of the hysteresis effect of the described prediction error on a current moment value; coefficients of the autoregressive polynomial logic and the moving average polynomial logic are solved to establish a mathematical expression describing the linear autocorrelation characteristics.
4. The short-circuit parameter variation trend prediction method according to claim 1, wherein: Capturing the long-term dependency relationship and the high-dimensional nonlinear characteristics of the multi-dimensional characteristic time sequence through the nonlinear intelligent mapping model comprises the following steps: constructing an information screening mechanism, processing historical state information and current input information through nonlinear transformation, and calculating memory retention strength to regulate and control the transmission proportion of the historical information; constructing a state updating mechanism, and updating the time sequence memory state in the nonlinear intelligent mapping model by combining the screened current input characteristics so as to accumulate long-term dependence information; And constructing an output mapping mechanism, and generating high-dimensional characteristic output to represent the high-dimensional nonlinear characteristic based on the updated time sequence memory state and mapping weight.
5. The short-circuit parameter variation trend prediction method according to claim 1, wherein: the dynamically updating the weight coefficients corresponding to the linear statistical evolution model and the nonlinear intelligent mapping model according to the real-time prediction residual error comprises the following steps: Constructing an error feedback mechanism, and calculating a weight adjustment step length according to the amplitude of the real-time prediction residual error; executing dynamic correction operation, and executing punishment or rewarding update on the weight coefficient at the last moment based on the weight adjustment step length; and carrying out normalization constraint processing, and mapping the updated linear statistical evolution model and the weight coefficient of the nonlinear intelligent mapping model to a unit interval to ensure that the probability sum of the fusion weights is one.
6. The short-circuit parameter variation trend prediction method according to claim 1, wherein: The calculating the anomaly score based on the deviation of the high-precision trend predictive value and the actual observed value and combining with the statistical process control theory comprises the following steps: calculating a residual sequence between the actual observed value and the high-precision trend predicted value, and counting the distribution characteristic value of the residual sequence; Constructing a dynamic control interval, and setting a fluctuation range comprising a statistical upper limit and a statistical lower limit according to the distribution characteristic value; and executing deviation degree quantization calculation, measuring and calculating the deviation degree of the current residual error value relative to the dynamic control interval, and mapping the deviation degree into the normalized anomaly score.
7. The short-circuit parameter variation trend prediction method according to claim 1, wherein: the deducting the future trend of the short-circuit parameter comprises the following steps: constructing non-parameter trend test statistics, and analyzing the rank relation of the high-precision trend predicted value in the time dimension; calculating a standardized trend index, and carrying out standardized transformation on data distribution by combining the variance of the non-parametric trend test statistic; and judging the significance of the trend, and determining the monotone direction and the confidence level of the change of the short circuit parameter according to the positive and negative signs and the numerical values of the standardized trend index.
8. A short-circuit parameter trend prediction system, applying the method according to any one of claims 1 to 7, comprising: The data acquisition and processing module is used for acquiring short-circuit parameter data of the power grid equipment, constructing a multidimensional characteristic time sequence and processing to obtain a stabilized input sample; The model construction module is used for constructing a heterogeneous characteristic prediction model system comprising a linear statistical evolution model and a nonlinear intelligent mapping model; The feature analysis module is used for analyzing the linear autocorrelation features through the linear statistical evolution model and capturing long-short-term dependency and high-dimensional nonlinear features through the nonlinear intelligent mapping model; the fusion prediction module is used for generating a basic prediction set, calculating a real-time prediction residual error and dynamically updating a weight coefficient, and obtaining a high-precision trend prediction value by using the updated weight coefficient; and the monitoring deduction module is used for calculating the anomaly score to identify an anomaly mode, deducting the future trend of the short-circuit parameter by combining the parameter evolution panoramic spectrum and outputting prediction information.
9. An electronic device, comprising: A memory and a processor; The memory is configured to store computer-executable instructions, and the processor is configured to execute the computer-executable instructions, which when executed by the processor, implement the steps of the short-circuit parameter variation trend prediction method of any one of claims 1 to 7.
10. A computer-readable storage medium storing computer-executable instructions which, when executed by a processor, implement the steps of the short-circuit parameter variation trend prediction method of any one of claims 1 to 7.

Description

Short-circuit parameter change trend prediction method, system, equipment and medium Technical Field The invention relates to the technical field of operation and maintenance of a power grid, in particular to a short-circuit parameter change trend prediction method systems, devices, and media. Background Along with the continuous expansion of the scale of the power system and the increase of the operation life, the factors such as the aging of the power grid equipment, the change of the network topology and the like cause the slow change of the short-circuit parameters, and the accuracy of the power grid protection set value and the safe and stable operation of the system are greatly influenced. The traditional short-circuit parameter calculation method is mainly based on theoretical calculation and periodic test, cannot track the parameter change trend in real time, and is difficult to predict the future parameter evolution law. The existing parameter monitoring system mainly focuses on the transient state and lacks the analysis and prediction capability of long-term change trend. The prior art has a plurality of defects in processing the time series data of the short-circuit parameters. First, traditional statistical methods such as linear regression cannot effectively capture nonlinear characteristics and complex time-dependent relationships of parameter changes. Then, the existing method lacks the capability of identifying abnormal change modes, and can not discover hidden trouble of equipment in time. Moreover, the single model has limited prediction precision and robustness, and is difficult to cope with complex and changeable power grid operation environments. In terms of data processing, the prior art has insufficient processing capacity for noise data and missing values, and influences the accuracy of a prediction model. In terms of model selection, the lack of a dedicated predictive model for short-circuit parameter characteristics has limited applicability to general time series models. In the aspect of system integration, the existing method lacks effective connection with a power grid operation management system, and real-time early warning and decision support are difficult to realize. Disclosure of Invention Therefore, the invention aims to solve the technical problems that the traditional method cannot track the short-circuit parameter change in real time, the prediction precision is low and the abnormality detection capability is weak. The technical problems are solved by the following technical scheme: a short-circuit parameter change trend prediction method comprises data acquisition and processing, model construction, feature analysis, fusion prediction and monitoring deduction. In a preferred implementation mode of the short-circuit parameter change trend prediction method, short-circuit parameter data of power grid equipment are obtained, a multi-dimensional characteristic time sequence is built, the multi-dimensional characteristic time sequence is processed to obtain a stabilized input sample, a heterogeneous characteristic prediction model system comprising a linear statistical evolution model and a nonlinear intelligent mapping model is built based on the stabilized input sample, linear autocorrelation characteristics of the multi-dimensional characteristic time sequence are analyzed through the linear statistical evolution model, long-short-period dependency relation and high-dimensional nonlinear characteristics of the multi-dimensional characteristic time sequence are captured through the nonlinear intelligent mapping model, a basic prediction set is obtained according to the linear autocorrelation characteristics, the long-short-period dependency relation and the high-dimensional nonlinear characteristics, real-time prediction residual errors are calculated when the linear statistical evolution model and the nonlinear intelligent mapping model output the basic prediction set, weight coefficients corresponding to the linear statistical evolution model and the nonlinear intelligent mapping model are dynamically updated according to the real-time prediction residual errors, the updated weight coefficients are used for weighting the basic prediction set, a high-short-circuit prediction set is obtained, a high-short-circuit value is obtained according to the linear autocorrelation characteristics, the long-short-time dependency relation and the high-time nonlinear characteristic time sequence is obtained, an abnormal-short-circuit parameter change trend is calculated, an abnormal trend is calculated, and an abnormal trend is calculated and a high-precision prediction value is calculated and is generated according to a combined with an abnormal parameter, and a high-precision prediction trend prediction value is obtained, and a high-precision prediction value is calculated. In a preferred embodiment of the short-circuit parameter variation trend prediction method, the processing the