Electric energy quality monitoring method and device based on AI engine optimization

Abstract

The invention discloses a power quality monitoring method and device based on AI engine optimization, and the method comprises the steps: S1, obtaining voltage waveform data collected by monitoring equipment, and obtaining the sequence distribution of voltage waveform short-time window data; s2, obtaining sequence distribution of key power quality parameter structured coding vectors; s3, obtaining a key power quality parameter time sequence context coding vector; s4, generating reconstructed voltage waveform data; and S5, based on the reconstruction error, determining whether there is an abnormal precursor. According to the AI engine optimization-based electric energy quality monitoring method and device disclosed by the invention, the learning ability of the AI model on the normal electric energy state waveform characteristics is utilized, and early warning on the electric energy quality disturbance event can be realized, so that preventive maintenance and active intervention are facilitated.

Description

Technical Field

[0001] The present invention belongs to the technical field of power quality monitoring, and specifically relates to a power quality monitoring method based on AI engine optimization and a power quality monitoring device based on AI engine optimization. Background Art

[0002] Power quality is a key indicator of the quality of power supply in a power system, directly impacting the safe and stable operation of the power system and the efficiency and lifespan of users' electrical equipment. With the increasing demands for power supply reliability and power purity in modern industry, commerce, and information technology, power quality issues such as voltage sags, swells, harmonics, frequency deviations, and three-phase imbalance can cause malfunctions in sensitive equipment, interrupt production processes, damage equipment, and even lead to safety accidents, resulting in significant economic losses. Therefore, developing efficient and accurate power quality monitoring solutions to promptly detect and warn of potential power quality disturbances is crucial for ensuring grid security, improving power supply service levels, and protecting user interests.

[0003] Traditional power quality monitoring solutions typically rely on dedicated monitoring equipment deployed at key nodes to measure and record a variety of power quality parameters according to preset standards (such as IEC 61000-4-30). However, these methods often focus on recording and classifying disturbance events that have already occurred and reached a clear threshold. This poses significant storage and transmission pressures, as well as heavy computational and analytical burdens, when processing massive amounts of continuous voltage waveform data. More importantly, traditional methods are limited in their ability to capture subtle, gradual abnormal signs before a disturbance occurs. They also struggle to effectively utilize the complex temporal correlation information contained in the data for early warning. Problems are often detected only after they have significantly deteriorated, limiting the potential for preventive maintenance and proactive intervention.

[0004] Therefore, further improvements are made to the above problems. Summary of the Invention

[0005] The main purpose of the present invention is to provide a power quality monitoring method and device based on AI engine optimization, which divides the voltage waveform data into short-term windows, extracts multivariate key power quality parameters from the voltage waveform data of each window, and constructs a time-series distribution of multidimensional power quality parameters to characterize the evolution of the power state. Then, by introducing a deep learning algorithm, the key power quality parameters of each short-term window are feature extracted and time-series context modeled to learn the time-series evolution law of the power quality state. On this basis, the voltage waveform data is reconstructed using the prior knowledge learned by the decoder under normal power quality data training, and then the abnormal precursors are identified based on the reconstruction error between the reconstructed voltage waveform data and the original voltage waveform data. This method utilizes the AI ​​model's ability to learn the waveform characteristics of normal power states, and can achieve early warning of power quality disturbance events to facilitate preventive maintenance and proactive intervention.

[0006] To achieve the above objectives, the present invention provides a power quality monitoring method based on AI engine optimization, including the following methods: Step S1: obtaining voltage waveform data collected by a (PQ) monitoring device, and performing short-time window segmentation on the voltage waveform data to obtain a sequence distribution of the voltage waveform short-time window data; Step S2: extracting the power quality parameter sequence of each voltage waveform short-time window data in the sequence distribution of the voltage waveform short-time window data to obtain the sequence distribution of the key power quality parameter structured coding vector; Step S3: performing context coding based on temporal counter-repulsion guidance on the sequence distribution of the key power quality parameter structured coding vector to obtain a key power quality parameter temporal context coding vector; Step S4: generating reconstructed voltage waveform data based on the key power quality parameter time series context coding vector; Step S5: determining whether there is an abnormality precursor based on a reconstruction error between the voltage waveform data and the reconstructed voltage waveform data.

[0007] As a further preferred technical solution of the above technical solution, in step S2, the power quality parameter sequence of each voltage waveform short-time window data in the sequence distribution is extracted to obtain a sequence distribution of the key power quality parameter sequence; and the key power quality parameter embedding coding matrix is ​​used to perform structured coding on each key power quality parameter sequence in the sequence distribution of the key power quality parameter sequence to obtain a sequence distribution of the key power quality parameter structured coding vector.

[0008] As a further preferred technical solution of the above technical solution, step S3 is specifically implemented as the following steps: Step S3.1: extracting a current key power quality parameter structured code vector from the sequence distribution of the key power quality parameter structured code vectors and defining other key power quality parameter structured code vectors in the sequence distribution of the key power quality parameter structured code vectors as key power quality parameter structured code vectors to be propagated, so as to obtain a sequence distribution of the key power quality parameter structured code vectors to be propagated; Step S3.2: Based on the temporal association of each key power quality parameter structured code vector to be propagated in the sequence distribution of the key power quality parameter structured code vector to be propagated relative to the current key power quality parameter structured code vector, performing feature propagation modulation on each key power quality parameter structured code vector to be propagated in the sequence distribution of the key power quality parameter structured code vector to be propagated, so as to obtain a set of key power quality parameter message dynamic code vectors; Step S3.3: Fusing the set of key power quality parameter message dynamic coding vectors and the current key power quality parameter structured coding vector to obtain the key power quality parameter temporal context coding vector.

[0009] As a further preferred technical solution of the above technical solution, step S3.2 is specifically implemented as follows: Step S3.2.1: Calculating the dynamic transmission potential energy of each key power quality parameter structured code vector to be propagated in the sequence distribution of the key power quality parameter structured code vector to be propagated relative to the current key power quality parameter structured code vector, so as to obtain a sequence distribution of the key power quality parameter node dynamic transmission potential energy code vector; Step S3.2.2: Calculating the message propagation repulsion coefficient of each key power quality parameter structured code vector to be propagated in the sequence distribution of the key power quality parameter structured code vector to be propagated relative to the current key power quality parameter structured code vector, so as to obtain the sequence distribution of the message propagation repulsion coefficient of the key power quality parameter node to be propagated; Step S3.2.3: Based on the sequence distribution of the message propagation repulsive force coefficients of the key power quality parameter nodes to be propagated and the sequence distribution of the dynamic transmission potential energy coding vectors of the key power quality parameter nodes to be propagated, dynamically adaptive message propagation coding is performed on each key power quality parameter structured coding vector to be propagated in the sequence distribution of the key power quality parameter structured coding vectors to be propagated, so as to obtain a set of key power quality parameter message dynamic coding vectors; The specific implementation of step S3.3 is as follows: Step S3.3.1: Calculate the positional sum of the set of dynamic code vectors of the key power quality parameter message to obtain a dynamic propagation aggregate code vector of the key power quality parameter message; Step S3.3.2: weighted fusion is performed on the key power quality parameter message dynamic propagation aggregation coding vector and the current key power quality parameter structured coding vector to obtain the key power quality parameter temporal context coding vector.

[0010] As a further preferred technical solution of the above technical solution, step S3.3.2 is specifically implemented as follows: Dynamic-static feature interactive perception is performed on the key power quality parameter message dynamic propagation aggregate coding vector and the current key power quality parameter structured coding vector to obtain a fusion weight coefficient; the fusion weight coefficient is used as the weight of the current key power quality parameter structured coding vector, and the difference between one and the fusion weight coefficient is used as the weight of the key power quality parameter message dynamic propagation aggregate coding vector, and the key power quality parameter message dynamic propagation aggregate coding vector and the current key power quality parameter structured coding vector are weightedly fused to obtain the key number timing context coding vector.

[0011] As a further preferred technical solution of the above technical solution, in step S4, the key power quality parameter timing context coding vector is subjected to feature renormalization based on power quality parameter node collaboration to obtain an optimized key power quality parameter timing context coding vector; the optimized key power quality parameter timing context coding vector is input into a decoder-based voltage waveform reconstruction module to obtain the reconstructed voltage waveform data.

[0012] As a further preferred technical solution of the above technical solution, in step S5, the reconstruction error is the root mean square error between the voltage waveform data and the reconstructed voltage waveform data; based on the comparison between the reconstruction error and a preset threshold, it is determined whether there is an abnormal precursor.

[0013] To achieve the above objectives, the present invention further provides a power quality monitoring device based on AI engine optimization, comprising: A voltage waveform data acquisition module is used to acquire voltage waveform data collected by the monitoring equipment; A voltage waveform short-time window segmentation module is used to segment the voltage waveform data into short-time windows to obtain a sequence distribution of the voltage waveform short-time window data; A power quality feature extraction module is used to extract the power quality features of each voltage waveform short-time window data in the sequence distribution of the voltage waveform short-time window data to obtain the sequence distribution of the key power quality parameter structured coding vector; A power quality parameter context coding module is used to perform context coding based on time series counter-repulsion guidance on the sequence distribution of the key power quality parameter structured coding vector to obtain a key power quality parameter time series context coding vector; A reconstructed voltage waveform data generating module, configured to generate reconstructed voltage waveform data based on the key power quality parameter time series context coding vector; The abnormality precursor monitoring module is used to determine whether there is an abnormality precursor based on a reconstruction error between the voltage waveform data and the reconstructed voltage waveform data. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a schematic diagram of the apparatus of the present invention. DETAILED DESCRIPTION

[0015] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are for illustrative purposes only, and those skilled in the art will readily appreciate other obvious variations. The basic principles of the present invention defined in the following description may be applied to other embodiments, variations, improvements, equivalents, and other technical solutions that do not depart from the spirit and scope of the present invention.

[0016] In the preferred embodiment of the present invention, those skilled in the art should note that the PQ monitoring device and the like involved in the present invention may be regarded as prior art.

[0017] Preferred embodiment.

[0018] The present invention discloses a power quality monitoring method based on AI engine optimization, including the following methods: Step S1: obtaining voltage waveform data collected by a (PQ) monitoring device, and performing short-time window segmentation on the voltage waveform data to obtain a sequence distribution of the voltage waveform short-time window data; Step S2: extracting the power quality parameter sequence of each voltage waveform short-time window data in the sequence distribution of the voltage waveform short-time window data to obtain the sequence distribution of the key power quality parameter structured coding vector; Step S3: performing context coding based on temporal counter-repulsion guidance on the sequence distribution of the key power quality parameter structured coding vector to obtain a key power quality parameter temporal context coding vector; Step S4: generating reconstructed voltage waveform data based on the key power quality parameter time series context coding vector; Step S5: determining whether there is an abnormality precursor based on a reconstruction error between the voltage waveform data and the reconstructed voltage waveform data.

[0019] In step S1, first, voltage waveform data collected by PQ monitoring equipment is acquired. It should be understood that because traditional power quality monitoring relies on dedicated equipment at fixed nodes, data acquisition density and real-time performance are limited, and the massive amount of continuous waveform data leads to excessive storage and transmission costs. Therefore, to capture high-frequency disturbance characteristics in the power grid in real time and reduce data redundancy, the present invention leverages the distributed deployment of high-precision PQ monitoring equipment. This technology synchronously acquires raw three-phase voltage waveform signals at a high sampling rate (e.g., above 10 kHz), covering dynamic changes at all nodes in the power grid. Specifically, intelligent PQ monitoring equipment that supports IEC standards is installed at key power grid nodes (such as substations and important load access points). Its high-speed ADC module synchronously samples and digitizes the voltage signals, generating a timestamp-containing voltage waveform sequence. This sequence is then locally cached and preprocessed by edge computing nodes. This approach preserves detailed voltage waveform features (such as high-frequency harmonics and transient sags) while avoiding the bandwidth pressure associated with traditional centralized storage, providing a high-quality data foundation for subsequent analysis.

[0020] Secondly, given the long time series and large data volume of continuous voltage waveform data, direct global analysis leads to an exponential increase in computational complexity and makes it difficult to capture the precursory characteristics of short-term disturbances. Therefore, to preserve the local details of the voltage waveform time series while reducing the computational load, the present invention uses a sliding window mechanism to segment the voltage waveform data from a continuous waveform into multiple short-term overlapping windows (e.g., each window is 200ms long and has an overlap ratio of 30%), thereby obtaining a sequence distribution of the voltage waveform short-term window data. It should be understood that power quality precursors often manifest as localized distortions with short durations (milliseconds) and subtle amplitude changes. Short-term windows can both isolate such events and prevent the loss of critical information through window overlap. In a specific implementation, a Hamming window function can be used to window the original waveform to suppress spectral leakage, while ensuring time series continuity through phase alignment between windows. In this way, the original voltage waveform data is converted into a sequence of time-indexed windows, reducing the data size of a single processing step and providing structured input for subsequent time series modeling.

[0021] Specifically, in step S2, the power quality parameter sequence of each voltage waveform short-time window data in the sequence distribution is extracted to obtain the sequence distribution of the key power quality parameter sequence; and the key power quality parameter embedding coding matrix is ​​used to perform structured coding on each key power quality parameter sequence in the sequence distribution of the key power quality parameter sequence to obtain the sequence distribution of the key power quality parameter structured coding vector.

[0022] Since the original voltage waveform contains multi-dimensional physical characteristics (such as amplitude, frequency, and harmonics), directly inputting the model will result in redundant features and difficulty in interpretation. Therefore, in order to build a quantifiable and interpretable power state characterization system, the present invention, based on IEC standards and domain knowledge, extracts multiple types of key power quality parameters from each voltage waveform short-term window data through parallel calculation, and constructs a key power quality parameter sequence for each short-term window to comprehensively reflect the spatiotemporal evolution characteristics of the voltage waveform. Specifically, the key power quality parameter sequence includes voltage RMS, voltage deviation, crest factor, total harmonic distortion, harmonic content of each sub-harmonic, frequency deviation, and voltage imbalance, where: The effective value of voltage (RMS) reflects the average level of voltage fluctuation; The voltage deviation (percentage of the nominal value) describes the degree of steady-state deviation between the actual voltage and the rated voltage; The crest factor (peak / RMS) reveals the ratio of the peak value to the effective value of the voltage waveform, which is important for evaluating instantaneous waveform distortion; Total harmonic distortion (THD) measures the impact of harmonic components on the fundamental wave and is a key indicator for evaluating the purity of electrical energy; The harmonic content of each order (such as 3rd, 5th, and 7th) provides detailed information about the harmonic spectrum, which helps to identify specific harmonic sources; Frequency deviation (the difference from the 50Hz / 60Hz reference value) reflects the difference between the actual grid frequency and the nominal frequency and is crucial for grid stability and equipment compatibility; Voltage imbalance (proportion of negative sequence component) is used to quantify the degree of imbalance in the three-phase voltage system, which is particularly critical for protecting the normal operation of three-phase equipment.

[0023] By performing parallel extraction of the above-mentioned key power quality parameters on the short-time window data of each voltage waveform, the high-dimensional waveform data is compressed into a low-dimensional feature vector, and the sequence distribution of the key power quality parameter sequence is constructed as the "feature fingerprint" of the power quality evolution. This not only retains the abnormal indicator factors with clear physical meaning and enhances the interpretability of anomaly detection, but also simplifies the model input, providing an accurate feature basis for subsequent anomaly detection.

[0024] Considering the heterogeneity of the multivariate parameters in the key power quality parameter sequence (differing in dimension and numerical range), directly using the spliced ​​sequence of multidimensional parameters as the input of the AI ​​model will make it difficult for the model to learn the implicit associations between the parameters. Therefore, in order to map the multidimensional physical parameters to a unified feature space and capture the cross-parameter coupling relationship, the present invention is based on the representation learning principle. First, each parameter in the key power quality parameter sequence is standardized to eliminate the dimension effect. Then, a trainable key power quality parameter embedding coding matrix is ​​constructed through a deep learning framework. The standardized key power quality parameter sequences are structured and encoded. The key power quality parameter vector of each short-time window is mapped to a high-dimensional embedding vector through matrix multiplication to obtain the sequence distribution of the key power quality parameter structured coding vector. At the same time, through back-propagation optimization, the distance between the key power quality parameter structured coding vectors in the embedding space reflects the correlation of the power quality status of each short-time window, thereby capturing the dynamic evolution law of the power quality status.

[0025] More specifically, since the voltage waveform data is essentially time series data, it contains rich information about the historical evolution of the voltage waveform and clues to future trend predictions. Therefore, in order to fully explore the time series dependencies and abnormal evolution patterns in the voltage waveform data and enhance the model's sensitivity to key abnormal events, the present invention proposes a context coding method based on time series counteracting repulsion guidance, which simulates the concepts of potential energy and repulsion in physics, and regards the key power quality parameter structured coding vectors of historical time nodes as "sources" with different influences on the current state. According to the distance (time interval) between the historical node and the current node and the correlation between their power quality states, the propagation potential energy and repulsion of the key power quality parameter structured coding vectors of each historical node are calculated to dynamically modulate them, thereby enhancing the model's sensitivity to abnormal patterns. By aggregating the context information of the global nodes, a time series context coding vector of key power quality parameters that integrates historical evolution laws and future trend prediction clues is obtained, providing a more refined feature representation for the subsequent reconstruction of the voltage waveform data. Step S3 is specifically implemented as follows: Step S3.1: Extract the current key power quality parameter structured code vector from the sequence distribution of the key power quality parameter structured code vectors and define other key power quality parameter structured code vectors in the sequence distribution of the key power quality parameter structured code vectors as key power quality parameter structured code vectors to be propagated, so as to obtain the sequence distribution of the key power quality parameter structured code vectors to be propagated. Specifically, it is expressed as follows: ; ; ; in, is the sequence distribution of the key power quality parameter structured coding vector, is a sequence distribution Key power quality parameter structured coding vector, represents the set of real numbers, is the number of structured encoding vectors of key power quality parameters, Indicates the length of the structured encoding vector of key power quality parameters, Represents the structured encoding vector of the current key power quality parameters, Represents the sequence distribution of the structured encoding vector of the key power quality parameters to be transmitted.

[0026] Step S3.2: Based on the temporal association of each key power quality parameter structured code vector to be propagated in the sequence distribution of the key power quality parameter structured code vector to be propagated relative to the current key power quality parameter structured code vector, performing feature propagation modulation on each key power quality parameter structured code vector to be propagated in the sequence distribution of the key power quality parameter structured code vector to be propagated, so as to obtain a set of key power quality parameter message dynamic code vectors; Step S3.3: Fusing the set of key power quality parameter message dynamic coding vectors and the current key power quality parameter structured coding vector to obtain the key power quality parameter temporal context coding vector.

[0027] Furthermore, step S3.2 is specifically implemented as follows: Step S3.2.1: Calculate the dynamic transmission potential energy of each key power quality parameter structured code vector to be propagated in the sequence distribution of the key power quality parameter structured code vector to be propagated relative to the current key power quality parameter structured code vector to obtain the sequence distribution of the key power quality parameter node dynamic transmission potential energy code vector to be propagated. Specifically, it is expressed as follows: ; ; ; in, represents the first pre-trained weight matrix, represents the pre-trained second weight matrix, represents the pre-trained weight parameters, represents matrix multiplication, 、 and Represent dot product, dot addition and dot subtraction respectively. The first sequence distribution of the structured coding vector of the key power quality parameters to be transmitted Key power quality parameter structured coding vector, for Relative to The key power quality parameter node dynamic transmission potential energy encoding vector to be propagated, Indicates calculating the square of the eigenvalues ​​at each position in the vector; Step S3.2.1: Calculate the message propagation repulsion coefficient of each key power quality parameter structured code vector to be propagated in the sequence distribution of the key power quality parameter structured code vector to be propagated relative to the current key power quality parameter structured code vector to obtain the sequence distribution of the key power quality parameter node message propagation repulsion coefficient. Specifically, it is expressed as follows: ; in, is the timestamp extraction function, represents matrix multiplication, is the temporal proximity sensitivity coefficient, express Relative to The node message propagation repulsion coefficient of the key power quality parameter to be propagated; Step S3.2.3: Based on the sequence distribution of the message propagation repulsive force coefficient of the key power quality parameter node to be propagated and the sequence distribution of the dynamic transmission potential energy coding vector of the key power quality parameter node to be propagated, dynamic adaptive message propagation coding is performed on each key power quality parameter structured coding vector to be propagated in the sequence distribution of the key power quality parameter structured coding vector to be propagated to obtain a set of key power quality parameter message dynamic coding vectors. Specifically, it is expressed as follows: ; ; in, and Represents point multiplication and point subtraction by position, represents matrix multiplication, represents the sigmoid activation function, represents the linear rectification function, represents the third weight matrix of pre-training, express The corresponding key power quality parameter message dynamic encoding vector.

[0028] The specific implementation of step S3.3 is as follows: Step S3.3.1: Calculate the positional sum of the set of dynamic code vectors of the key power quality parameter message to obtain a dynamic propagation aggregate code vector of the key power quality parameter message; Step S3.3.2: weighted fusion is performed on the key power quality parameter message dynamic propagation aggregation coding vector and the current key power quality parameter structured coding vector to obtain the key power quality parameter temporal context coding vector.

[0029] Furthermore, step S3.3.2 is specifically implemented as follows: Dynamic-static feature interactive perception is performed on the key power quality parameter message dynamic propagation aggregated coding vector and the current key power quality parameter structured coding vector to obtain a fusion weight coefficient. Specifically, it is expressed as follows: ; in, represents the fourth weight matrix of pre-training, Represents the normalized exponential function, through Function operation can obtain a binary vector. represents vector concatenation, represents the fusion weight coefficient; The fusion weight coefficient is used as the weight of the current key power quality parameter structured coding vector, and the difference between one and the fusion weight coefficient (i.e., 1 minus the fusion weight coefficient) is used as the weight of the key power quality parameter message dynamic propagation aggregate coding vector. The key power quality parameter message dynamic propagation aggregate coding vector and the current key power quality parameter structured coding vector are weightedly fused to obtain the key number temporal context coding vector. Specifically, it is expressed as follows: ; in, Represents the time series context coding vector of key power quality parameters.

[0030] At the same time, for the set of dynamic encoding vectors of key power quality parameter messages, due to the adaptability of effect propagation under nonlinear representation, each dynamic encoding vector of the key power quality parameter message actually constitutes a decoupling vector local propagation grid. Therefore, when aggregating each local propagation grid, in addition to vector summation, a local propagation grid propagation path correction that conforms to the sequence propagation characteristics can also be introduced.

[0031] Specifically, a vector product state is introduced to simulate the auxiliary propagation path representation of the local propagation grid point, so that the propagation behavior of the local grid point in the sequence path is reflected by the product structure, so that the synergistic effect can emerge naturally. Specifically, it is expressed as follows: ; The key power quality parameter timing context coding vector is optimized based on this. Specifically, it is expressed as follows: ; Thus, the propagation synergy effect is enhanced by introducing an auxiliary propagation path of the local propagation grid, and the aggregation effect of the dynamic timing information of the key power quality parameter message dynamic coding vector at each time point by the key power quality parameter timing context coding vector is improved.

[0032] Preferably, in step S4, the key power quality parameter timing context coding vector is subjected to feature renormalization based on power quality parameter node collaboration to obtain an optimized key power quality parameter timing context coding vector. The optimized key power quality parameter timing context coding vector is input into a decoder-based voltage waveform reconstruction module to obtain the reconstructed voltage waveform data. Specifically, during the training phase, the decoder is trained using power quality data under normal conditions (i.e., key power quality parameter timing context coding vectors extracted based on voltage waveform data without significant disturbances or abnormal precursors) to learn the potential representation of normal power quality conditions. By optimizing the decoder parameters, the decoder learns the inherent, typical, and regular feature patterns of normal power quality data, thereby accurately reconstructing the original voltage waveform data from the key power quality parameter timing context coding vector under normal voltage conditions. During the inference phase, when the input is the key power quality parameter timing context coding vector under normal conditions, the decoder can effectively utilize its learned prior knowledge to generate normal voltage waveform data that corresponds to the input and is as close to the original as possible. When the input is a time-series context encoding vector of key power quality parameters under a significant abnormal state, the decoder attempts to reconstruct the voltage waveform based on the learned normal state latent representation. However, due to the difference in feature space between the abnormal and normal states, the reconstructed voltage waveform deviates from the actual observed waveform. This deviation can be used as an indicator for anomaly detection; a larger deviation indicates a higher degree of anomaly in the current voltage waveform data.

[0033] Preferably, in step S5, the reconstruction error is the root mean square error between the voltage waveform data and the reconstructed voltage waveform data; based on the comparison between the reconstruction error and the preset threshold, it is determined whether there is an abnormal precursor. It should be understood that since the reconstruction error directly reflects the degree of deviation between the input voltage waveform data and the normal mode learned by the decoding model, the present invention uses it as an abnormality criterion, and calculates the root mean square error between the voltage waveform data and the reconstructed voltage waveform data to quantify the overall deviation between the actual observed waveform and the reconstructed waveform, which serves as a quantitative basis for power quality abnormality detection.

[0034] When the reconstruction error exceeds a preset threshold, it indicates a significant anomaly in the current voltage waveform data, potentially indicating a fault or instability in the power grid. The system immediately issues an alarm, indicating a potential power quality issue, enabling real-time monitoring and early warning of power quality. This approach enables automatic and sensitive detection of precursors to power quality anomalies, achieving early warning and providing strong support for preventive maintenance to ensure stable grid operation.

[0035] like Figure 1 As shown, the present invention also discloses a power quality monitoring device based on AI engine optimization, comprising: A voltage waveform data acquisition module is used to acquire voltage waveform data collected by a monitoring device; A voltage waveform short-time window segmentation module is used to segment the voltage waveform data into short-time windows to obtain a sequence distribution of the voltage waveform short-time window data; A power quality feature extraction module is used to extract the power quality features of each voltage waveform short-time window data in the sequence distribution of the voltage waveform short-time window data to obtain the sequence distribution of the key power quality parameter structured coding vector; A power quality parameter context coding module is used to perform context coding based on time series counter-repulsion guidance on the sequence distribution of the key power quality parameter structured coding vector to obtain a key power quality parameter time series context coding vector; A reconstructed voltage waveform data generating module, configured to generate reconstructed voltage waveform data based on the key power quality parameter time series context coding vector; The abnormality precursor monitoring module is used to determine whether there is an abnormality precursor based on a reconstruction error between the voltage waveform data and the reconstructed voltage waveform data.

[0036] Compared with the prior art, the present invention proposes the following innovations: 1) The voltage waveform data is divided into short-term windows, and the multi-dimensional key power quality parameters are extracted from the voltage waveform data of each window. The time-series distribution of multi-dimensional power quality parameters is constructed to characterize the evolution of the power state. Then, by introducing a deep learning algorithm, the key power quality parameters of each short-term window are feature extracted and time-series context modeled to learn the time-series evolution of the power quality state. On this basis, the voltage waveform data is reconstructed using the prior knowledge learned by the decoder under normal power quality data training. The abnormal precursors are then identified based on the reconstruction error between the reconstructed voltage waveform data and the original voltage waveform data. This method uses the AI ​​model's ability to learn the waveform characteristics of normal power states to achieve early warning of power quality disturbance events, facilitating preventive maintenance and proactive intervention.

[0037] 2) Since voltage waveform data is essentially time series data, it contains rich information about the historical evolution of the voltage waveform and clues for future trend prediction. Therefore, in order to fully explore the time series dependencies and abnormal evolution patterns in voltage waveform data and enhance the model's sensitivity to key abnormal events, this paper proposes a context encoding method based on time series counter-repulsion guidance. This method simulates the concepts of potential energy and repulsion in physics, and regards the structured encoding vectors of key power quality parameters at historical time nodes as "sources" with different influences on the current state. Based on the distance (time interval) between the historical node and the current node and the correlation between their power quality states, the propagation potential energy and repulsion of the structured encoding vectors of key power quality parameters at each historical node are calculated. This is used to dynamically modulate the vectors and enhance the model's sensitivity to abnormal patterns. By aggregating the context information of the global node, a time series context encoding vector of key power quality parameters that integrates historical evolution laws and future trend prediction clues is obtained, providing a more refined feature representation for subsequent voltage waveform data reconstruction.

[0038] It is worth mentioning that the technical features such as the PQ monitoring equipment involved in the patent application of this invention should be regarded as prior art. The specific structure, working principle and possible control method and spatial layout method of these technical features can be selected by conventional means in the field and should not be regarded as the inventive point of the patent of this invention. The patent of this invention will not be further elaborated.

[0039] For those skilled in the art, it is still possible to modify the technical solutions described in the aforementioned embodiments, or to make equivalent replacements for some of the technical features therein. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A power quality monitoring method based on AI engine optimization, characterized in that: This includes the following methods: Step S1: obtaining voltage waveform data collected by a monitoring device, and performing short-time window segmentation on the voltage waveform data to obtain a sequence distribution of the voltage waveform short-time window data; Step S2: extracting the power quality parameter sequence of each voltage waveform short-time window data in the sequence distribution of the voltage waveform short-time window data to obtain the sequence distribution of the key power quality parameter structured coding vector; Step S3: performing context coding based on temporal counter-repulsion guidance on the sequence distribution of the key power quality parameter structured coding vector to obtain a key power quality parameter temporal context coding vector; Step S4: generating reconstructed voltage waveform data based on the key power quality parameter time series context coding vector; Step S5: determining whether there is an abnormality precursor based on a reconstruction error between the voltage waveform data and the reconstructed voltage waveform data.

2. The power quality monitoring method based on AI engine optimization according to claim 1, characterized in that: In step S2, the power quality parameter sequence of each voltage waveform short-time window data in the sequence distribution is extracted to obtain a sequence distribution of the key power quality parameter sequence; and the key power quality parameter embedding coding matrix is ​​used to perform structured coding on each key power quality parameter sequence in the sequence distribution of the key power quality parameter sequence to obtain a sequence distribution of the key power quality parameter structured coding vector.

3. The power quality monitoring method based on AI engine optimization according to claim 2, characterized in that: Step S3 is specifically implemented as follows: Step S3.1: extracting a current key power quality parameter structured code vector from the sequence distribution of the key power quality parameter structured code vectors and defining other key power quality parameter structured code vectors in the sequence distribution of the key power quality parameter structured code vectors as key power quality parameter structured code vectors to be propagated, so as to obtain a sequence distribution of the key power quality parameter structured code vectors to be propagated; Step S3.2: Based on the temporal association of each key power quality parameter structured code vector to be propagated in the sequence distribution of the key power quality parameter structured code vector to be propagated relative to the current key power quality parameter structured code vector, performing feature propagation modulation on each key power quality parameter structured code vector to be propagated in the sequence distribution of the key power quality parameter structured code vector to be propagated, so as to obtain a set of key power quality parameter message dynamic code vectors; Step S3.3: Fusing the set of key power quality parameter message dynamic coding vectors and the current key power quality parameter structured coding vector to obtain the key power quality parameter temporal context coding vector.

4. The power quality monitoring method based on AI engine optimization according to claim 3 is characterized in that: The specific implementation of step S3.2 is as follows: Step S3.2.1: Calculating the dynamic transmission potential energy of each key power quality parameter structured code vector to be propagated in the sequence distribution of the key power quality parameter structured code vector to be propagated relative to the current key power quality parameter structured code vector, so as to obtain a sequence distribution of the key power quality parameter node dynamic transmission potential energy code vector; Step S3.2.2: Calculating the message propagation repulsion coefficient of each key power quality parameter structured code vector to be propagated in the sequence distribution of the key power quality parameter structured code vector to be propagated relative to the current key power quality parameter structured code vector, so as to obtain the sequence distribution of the message propagation repulsion coefficient of the key power quality parameter node to be propagated; Step S3.2.3: Based on the sequence distribution of the message propagation repulsive force coefficients of the key power quality parameter nodes to be propagated and the sequence distribution of the dynamic transmission potential energy coding vectors of the key power quality parameter nodes to be propagated, dynamically adaptive message propagation coding is performed on each key power quality parameter structured coding vector to be propagated in the sequence distribution of the key power quality parameter structured coding vectors to be propagated, so as to obtain a set of key power quality parameter message dynamic coding vectors; The specific implementation of step S3.3 is as follows: Step S3.3.1: Calculate the positional sum of the set of dynamic code vectors of the key power quality parameter message to obtain a dynamic propagation aggregate code vector of the key power quality parameter message; Step S3.3.2: weighted fusion is performed on the key power quality parameter message dynamic propagation aggregation coding vector and the current key power quality parameter structured coding vector to obtain the key power quality parameter temporal context coding vector.

5. The power quality monitoring method based on AI engine optimization according to claim 4 is characterized in that: The specific implementation of step S3.3.2 is as follows: Dynamic-static feature interactive perception is performed on the key power quality parameter message dynamic propagation aggregate coding vector and the current key power quality parameter structured coding vector to obtain a fusion weight coefficient; the fusion weight coefficient is used as the weight of the current key power quality parameter structured coding vector, and the difference between one and the fusion weight coefficient is used as the weight of the key power quality parameter message dynamic propagation aggregate coding vector, and the key power quality parameter message dynamic propagation aggregate coding vector and the current key power quality parameter structured coding vector are weightedly fused to obtain the key number timing context coding vector.

6. The power quality monitoring method based on AI engine optimization according to claim 5, characterized in that: In step S4, the key power quality parameter timing context coding vector is subjected to feature renormalization based on power quality parameter node collaboration to obtain an optimized key power quality parameter timing context coding vector; the optimized key power quality parameter timing context coding vector is input into a decoder-based voltage waveform reconstruction module to obtain the reconstructed voltage waveform data.

7. The power quality monitoring method based on AI engine optimization according to claim 6, characterized in that: In step S5, the reconstruction error is a root mean square error between the voltage waveform data and the reconstructed voltage waveform data; and based on a comparison between the reconstruction error and a preset threshold, it is determined whether there is an abnormal precursor.

8. A power quality monitoring device based on AI engine optimization, applied to a power quality monitoring method based on AI engine optimization according to any one of claims 1 to 7, characterized in that: include: A voltage waveform data acquisition module is used to acquire voltage waveform data collected by a monitoring device; A voltage waveform short-time window segmentation module is used to segment the voltage waveform data into short-time windows to obtain a sequence distribution of the voltage waveform short-time window data; A power quality feature extraction module is used to extract the power quality features of each voltage waveform short-time window data in the sequence distribution of the voltage waveform short-time window data to obtain the sequence distribution of the key power quality parameter structured coding vector; A power quality parameter context coding module is used to perform context coding based on time series counter-repulsion guidance on the sequence distribution of the key power quality parameter structured coding vector to obtain a key power quality parameter time series context coding vector; A reconstructed voltage waveform data generating module, configured to generate reconstructed voltage waveform data based on the key power quality parameter time series context coding vector; The abnormality precursor monitoring module is used to determine whether there is an abnormality precursor based on a reconstruction error between the voltage waveform data and the reconstructed voltage waveform data.

Images