Power quality disturbance identification method based on data driving

Abstract

The invention provides an electric energy quality disturbance identification method based on data driving, and belongs to the technical field of electric energy quality disturbance identification, and the method comprises the following steps: S1, obtaining an electric energy quality disturbance voltage signal in a set time period of a power distribution network; s2, performing data enhancement on the power quality disturbance voltage signal based on Hilbert transform, and amplifying disturbance data in the power quality disturbance voltage signal; s3, using an MLSTM-FCNs model, inputting the power quality disturbance voltage signal subjected to data enhancement into the MLSTM-FCNs model to pre-train the power quality disturbance voltage signal, and finally obtaining and storing an optimal training model; and S4, taking the optimal training model as a feature extractor, establishing and training a width learning classification model based on a width learning algorithm, classifying power quality disturbance voltage signals through the width learning classification model, and outputting an identification result of disturbance data. And the method has better robustness.

Description

Technical Field

[0001] The present invention belongs to the technical field of power quality disturbance identification, and in particular relates to a power quality disturbance identification method based on data driving. Background Art

[0002] Power quality refers to the quality of the power provided by the power system to the user's equipment, including the stability and accuracy of parameters such as voltage, frequency, and waveform. High-quality power is essential for the reliable operation of the power system and the normal operation of end-user equipment. When disturbance problems occur, it will not only cause failure and damage to user equipment, but also reduce the efficiency of the power system and increase operating costs.

[0003] Common power quality disturbances include voltage sag, voltage swell, harmonics, voltage flicker, transient overvoltage, etc. These disturbances will have different degrees of impact on the power system and user equipment. For example, voltage sag may cause industrial equipment to shut down, harmonics may cause overheating and damage to electrical equipment, and voltage flicker may cause flickering and instability of lighting equipment. The identification and management of these disturbances are of great significance to ensure the reliable operation of the power system and the normal use of user equipment.

[0004] In recent years, the power quality disturbance identification technology has made significant progress. At present, the power quality disturbance identification method includes two major steps, the first is feature extraction, and then the disturbance classification is performed according to the extracted features. According to the different feature extraction methods, power quality disturbances can be divided into algorithms based on manual feature extraction and algorithms based on automatic feature extraction. Traditional disturbance identification methods mainly rely on manual feature extraction, using experience-based rules and signal processing techniques such as Fourier transform and wavelet transform. These methods can effectively identify power quality disturbances to a certain extent, but they have limitations in complex power systems and large-scale data processing. In addition, algorithms based on automatic feature extraction methods such as support vector machines (SVM), artificial neural networks (ANN) and deep learning (DL) have gradually been applied to the field of power quality disturbance identification and have achieved remarkable results. These methods based on automatic feature extraction can handle complex nonlinear data relationships and improve the accuracy and efficiency of disturbance identification, but they require a large amount of data, the algorithm calculation is complex, and the cost of retraining after the data set changes is high. Therefore, it is urgent to solve the diversity and complexity of the power quality composite disturbance problem and design a high-precision, scalable and low-computation identification technology solution. Summary of the invention

[0005] In order to solve the above problems existing in the prior art, a data-driven power quality disturbance identification method is provided.

[0006] The technical solution adopted by the present invention to solve its technical problem is: The present invention proposes a data-driven power quality disturbance identification method, comprising the following steps: S1: Obtain the power quality disturbance voltage signal within the set time period of the distribution network; S2: Performing data enhancement on the power quality disturbance voltage signal based on Hilbert transform to amplify the disturbance data in the power quality disturbance voltage signal; S3: Use the MLSTM-FCNs model and input the power quality disturbance voltage signal after data enhancement into the MLSTM-FCNs model for pre-training, and finally obtain and save the best training model; S4: Use the best training model as a feature extractor to extract voltage disturbance features, establish and train a width learning classification model based on the width learning algorithm, classify the power quality disturbance voltage signal through the width learning classification model, and output the recognition result of the disturbance data.

[0007] Furthermore, in S1, the power quality disturbance voltage signal includes a basic disturbance signal and a composite disturbance signal. The basic disturbance signals include: disturbance signals of negligible disturbance, disturbance signals of temporary voltage drop, disturbance signals of temporary voltage rise, disturbance signals of temporary power outage, disturbance signals of harmonic influence, disturbance signals of fluctuation, disturbance signals of transient oscillation, disturbance signals of transient impact, disturbance signals of periodic gap, and disturbance signals of periodic peak; The composite disturbance signal includes: a disturbance signal of voltage drop and harmonic composite, a disturbance signal of voltage rise and harmonic composite, a disturbance signal of temporary power outage and harmonic composite, a disturbance signal of voltage temporary drop and fluctuation composite, a disturbance signal of voltage rise and fluctuation composite, and a disturbance signal of harmonic and fluctuation composite.

[0008] Furthermore, in the basic disturbance signal: The disturbance signal with negligible disturbance is generated by the microgrid voltage parameter equation with negligible disturbance, which is as follows: (1); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α A is the disturbance fluctuation coefficient, α A Randomly generated in the range (0, 0.1), t 1. t 2 are the start and end times of the disturbance,y [ i ] is the generated power quality disturbance data; The disturbance signal of temporary voltage drop is generated by the voltage parameter equation of the microgrid with temporary voltage drop, and the formula is as follows: (2); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α B is the voltage drop amplitude coefficient, α B Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data; The disturbance signal of temporary voltage rise is generated by the voltage parameter equation of the microgrid with temporary voltage rise, and the formula is as follows: (3); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α C is the voltage rise amplitude coefficient, α C Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data; The disturbance signal of temporary power outage is generated by the voltage parameter equation of the microgrid with temporary power outage, and the formula is as follows: (4); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α D is the voltage drop amplitude coefficient,α D Randomly generated in the range [0.9, 1.0], t 1. t 2 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data; The disturbance signal affected by harmonics is generated by the microgrid voltage parameter equation affected by harmonics, and the formula is as follows: (5); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively, a 3 、a 5 、a 7 is randomly generated in the range [0.05, 0.15], y [ i ] is the generated power quality disturbance data; The disturbance signal that generates fluctuations is generated by the voltage parameter equation of the microgrid where the fluctuation occurs, and the formula is as follows: (6); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. α E is the voltage amplitude fluctuation coefficient, which is randomly generated in the range of [0.1, 0.2], is the voltage frequency fluctuation coefficient, which is randomly generated in the range of [0.1, 0.4], y [ i ] is the generated power quality disturbance data; The disturbance signal of transient oscillation is generated by the microgrid voltage parameter equation of transient oscillation, and the formula is as follows: (7); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, t 3 , t 4are the start and end times of the disturbance, T n is the period of transient oscillation, T n Randomly generated in the range [0.0011, 0.0033], τ is the time constant, τ Randomly generated in the range [150, 1000], k 1 is the amplitude coefficient of transient oscillation, k 1 Randomly generated in the range [0.1, 0.8], y [ i ] is the generated power quality disturbance data; The disturbance signal of transient impact is generated by the microgrid voltage parameter equation of transient impact, and the formula is as follows: (8); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α F is the amplitude coefficient of voltage transient impact, α F Randomly generated in the range [0, 0.414], t 5 , t 6 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data; The disturbance signal of periodic gap is generated by the voltage parameter equation of microgrid with periodic gap, and the formula is as follows: (9); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, sign(t) is the sign function, k 2 is the amplitude coefficient of the voltage periodic gap, which is randomly generated in the range of [0.1, 0.4]. t 7 , t 8 are the start and end times of the disturbance, y[ i ] is the generated power quality disturbance data; The disturbance signal of periodic spikes is given by the microgrid voltage parameter equation of periodic spikes, which is as follows: (10); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, sign(t) is the sign function, k 2 is the amplitude coefficient of the voltage periodic gap, which is randomly generated in the range of [0.1, 0.4]. t 7 , t 8 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data.

[0009] Furthermore, in the composite disturbance signal: The disturbance signal of voltage drop and harmonic composite is generated by the voltage parameter equation of the microgrid with temporary voltage drop and the voltage parameter equation of the microgrid affected by harmonics. The formula is as follows: (11); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α B is the voltage drop amplitude coefficient, α B Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively, a 3 、a 5 、a 7 is randomly generated in the range [0.05, 0.15], y [ i ] is the generated power quality disturbance data; The disturbance signal of voltage rise and harmonic composite is generated by the voltage parameter equation of the microgrid with temporary voltage rise and the voltage parameter equation of the microgrid affected by harmonics. The formula is as follows: (12); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α C is the voltage rise amplitude coefficient, α C Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively, a 3 、a 5 、a 7 is randomly generated in the range [0.05, 0.15], y [ i ] is the generated power quality disturbance data; The disturbance waveform signal of temporary power outage and harmonic composite is generated by the voltage parameter equation of the microgrid affected by temporary power outage and the voltage parameter equation of the microgrid affected by harmonics. The formula is as follows: (13); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α D is the voltage drop amplitude coefficient, α D Randomly generated in the range [0.9, 1.0], t 1. t 2 are the start and end times of the disturbance, a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively, a 3 、a 5 、a7 is randomly generated in the range [0.05, 0.15], y [ i ] is the generated power quality disturbance data; The disturbance signal of voltage temporary drop and fluctuation is generated by the voltage parameter equation of the microgrid with temporary voltage drop and the voltage parameter equation of the microgrid with fluctuation. The formula is as follows: (14); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α B is the voltage drop amplitude coefficient, α B Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, α E is the voltage amplitude fluctuation coefficient, which is randomly generated in the range of [0.1, 0.2], is the frequency fluctuation coefficient of the voltage, which is randomly generated in the range of [0.1, 0.4], y [ i ] is the generated power quality disturbance data; The disturbance signal of voltage rise and fluctuation is generated by the voltage parameter equation of the microgrid with temporary voltage rise and the voltage parameter equation of the microgrid with fluctuation. The formula is as follows: (15); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α C is the voltage rise amplitude coefficient, α C Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, α E is the voltage amplitude fluctuation coefficient, which is randomly generated in the range of [0.1, 0.2], is the frequency fluctuation coefficient of the voltage, which is randomly generated in the range of [0.1, 0.4],y [ i ] is the generated power quality disturbance data; The disturbance signal of harmonic and fluctuation composite is generated by the composite voltage parameter equation of the microgrid affected by the harmonic and the voltage parameter equation of the microgrid where the fluctuation occurs. The formula is as follows: (16); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. α E is the voltage amplitude fluctuation coefficient, which is randomly generated in the range of [0.1, 0.2], a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively, a 3 、a 5 、a 7 is randomly generated in the range [0.05, 0.15], is the frequency fluctuation coefficient of the voltage, which is randomly generated in the range of [0.1, 0.4], y [ i ] is the generated power quality disturbance data.

[0010] Furthermore, the microgrid voltage fundamental period T The value includes 0.02s, the number of points that the analog-to-digital converter can collect in 1 second N The values ​​of include 3200, and i is 0, 1, 2, …, 10NT-1 respectively.

[0011] Furthermore, t 1 , t 2 、t 3 , t 4 、t 5 , t 6 are all random integers in [0, 10NT-1], t 7 , t 8 are all random integers in [0, NT], t 1 , t 2 Satisfy NT ≤ t 2 - t1 ≤ 9NT, t 3 , t 4 Satisfy 0.5NT ≤ t 4 - t 3 ≤ 3NT, t 5 , t 6 Satisfy 0.05NT ≤ t 6 - t 5 ≤ 0.1NT, t 7 , t 8 Satisfy 0.01NT ≤ t 8 - t 7 ≤ 0.05NT.

[0012] Furthermore, a 1、 a 3 、a 5 、a 7. Satisfy between: (17); In the formula a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively.

[0013] Furthermore, in S2, the data enhancement process includes the following steps: The power quality disturbance voltage signal is normalized, and the formula is as follows: (18); In the formula, X is the original sample data, X´ is the normalized sample data, X max and X min are the maximum and minimum values ​​of the original sample data respectively; The Hilbert transform of the normalized power quality disturbance voltage signal is as follows: (19); In the formula, f(t) represents the normalized power quality disturbance voltage signal, Express f (t) signal after Hilbert transformation; Introducing analytical signals f a (t), the formula is as follows: (20); In the formula, f a (t) means f (t) The corresponding analytical signal; Using Euler's formula f a (t) is transformed and obtained f The envelope sequence and phase of (t) are as follows: (twenty one); (twenty two); (twenty three); In the formula, I A (t) means f (t) is the envelope sequence, Iθ (t) represents the phase; The envelope signal is used as a weight to perform dot product calculation with the normalized power quality disturbance voltage signal to complete data enhancement. The formula is as follows: (twenty four); In the formula, I A (t) means f (t) is the envelope sequence, y (t) represents the enhanced signal.

[0014] Furthermore, in S3, the judgment criteria for the best training model of MLSTM-FCNs include: after the power quality disturbance voltage signal after data enhancement is input into the current MLSTM-FCNs model, the accuracy on the verification set is greater than or equal to 90%, which meets the pre-training requirements, and the current model is the best training model.

[0015] Furthermore, in S4, the width learning classification model training process includes the following steps: The data passed through the feature extractor and its corresponding labels are input into the width learning classification model, and the feature mapping layer of the width learning classification model is trained. The formula is as follows: (25); In the formula, Z i For thei Group mapping features, For the i Group random mapping, X For input data, W ei For the i The random weight coefficients with appropriate dimensions corresponding to the group enhancement nodes, β ei For the i The bias of the group enhancement node; The enhancement layer of the width learning classification model is trained as follows: (26); In the formula, H j For the j Group Enhancement Node, is a nonlinear mapping on the linear combination result, Z m For the m Group mapping features, W hj For the j The random weight coefficients with appropriate dimensions corresponding to the group enhancement layer, β hj For the j Bias of group enhancement layer; Update the mapping function Y from input to output, the formula is as follows: (27); In the formula, Z m Before m Group all mapped features, Before n Group all enhanced nodes, W Represents the weight of each layer.

[0016] Compared with the prior art, the present invention has the following advantages: This application combines deep learning and width learning, uses deep learning algorithms as feature extractors, uses width learning algorithms as classifiers, and combines them with traditional signal processing technology to achieve data-driven power quality disturbances. Compared with traditional processing methods, the power quality disturbance data set does not need to be manually designed for feature extraction, does not rely on personal experience, avoids interference from artificial factors, and has better robustness. In addition, since the width learning algorithm is used as a classifier at the same time, compared with the traditional deep learning algorithm alone, the recognition method proposed in this scheme has the characteristic of continuous learning from new data. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which: Figure 1 It is the overall flow chart of the present invention; Figure 2 is a data enhancement flow chart of the present invention; Figure 3 It is a comparison diagram of the original signal and the enhanced signal of the present invention; Figure 4 It is a structural diagram of the feature extraction module of the present invention. DETAILED DESCRIPTION

[0018] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0019] like Figure 1-4 As shown, this embodiment proposes a data-driven power quality disturbance identification method, wherein: The following steps are involved: S1: Obtain the power quality disturbance voltage signal within the set time period of the distribution network; S2: Performing data enhancement on the power quality disturbance voltage signal based on Hilbert transform to amplify the disturbance data in the power quality disturbance voltage signal; S3: Use the MLSTM-FCNs model and input the power quality disturbance voltage signal after data enhancement into the MLSTM-FCNs model for pre-training, and finally obtain and save the best training model; S4: Use the best training model as a feature extractor to extract voltage disturbance features, establish and train a width learning classification model based on the width learning algorithm, classify the power quality disturbance voltage signal through the width learning classification model, and output the recognition result of the power quality disturbance.

[0020] For the problem of composite power quality disturbance, there is a problem that public data sets are difficult to obtain. There are two main solutions to obtain public data sets. The first solution is to use sensors to record data on site. The advantage of this solution is that the data is authentic and reliable, but the equipment and time costs are high, and it is difficult to cover composite disturbance types. The second solution is to use parametric equations to synthesize disturbance waveforms. Based on the correct derivation formula, massive and reliable data can be generated at almost no cost, and all disturbance categories can be covered. This embodiment adopts the second solution and uses the microgrid voltage parameter equation to batch generate disturbance waveforms, namely power quality disturbance voltage signals.

[0021] It should be noted that, in this embodiment, the power quality disturbance voltage signal includes a basic disturbance signal and a composite disturbance signal. The basic disturbance signals include: disturbance signals of negligible disturbance, disturbance signals of temporary voltage drop, disturbance signals of temporary voltage rise, disturbance signals of temporary power outage, disturbance signals of harmonic influence, disturbance signals of fluctuation, disturbance signals of transient oscillation, disturbance signals of transient impact, disturbance signals of periodic gap, and disturbance signals of periodic peak; Composite disturbance signals include: disturbance signals of voltage drop and harmonics, disturbance signals of voltage rise and harmonics, disturbance signals of temporary power outage and harmonics, disturbance signals of temporary voltage drop and fluctuation, disturbance signals of voltage rise and fluctuation, and disturbance signals of harmonics and fluctuations.

[0022] In summary, in this embodiment, the power quality disturbance voltage signal refers to the normal voltage signal superimposed with disturbance data, and the disturbance data specifically includes negligible disturbance, temporary voltage drop disturbance, temporary voltage rise disturbance, temporary power outage disturbance, harmonic disturbance, transient oscillation disturbance, transient impulse disturbance, periodic notch disturbance, periodic spike disturbance, voltage drop and harmonic composite disturbance, voltage rise and harmonic composite disturbance, temporary power outage and harmonic composite disturbance, voltage temporary drop and fluctuation composite disturbance, voltage rise and fluctuation composite disturbance, harmonic and fluctuation composite disturbance.

[0023] Among them, in the basic disturbance signal: The disturbance signal with negligible disturbance is generated by the microgrid voltage parameter equation with negligible disturbance, which is as follows: (1); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α A is the disturbance fluctuation coefficient, α A Randomly generated in the range (0, 0.1), t 1. t 2 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data; The disturbance signal of temporary voltage drop is generated by the voltage parameter equation of the microgrid with temporary voltage drop, and the formula is as follows: (2); In the formula, Tis the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α B is the voltage drop amplitude coefficient, α B Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data; The disturbance signal of temporary voltage rise is generated by the voltage parameter equation of the microgrid with temporary voltage rise, and the formula is as follows: (3); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α C is the voltage rise amplitude coefficient, α C Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data; The disturbance signal of temporary power outage is generated by the voltage parameter equation of the microgrid with temporary power outage, and the formula is as follows: (4); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α D is the voltage drop amplitude coefficient, α D Randomly generated in the range [0.9, 1.0], t 1. t 2 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data; The disturbance signal affected by harmonics is generated by the microgrid voltage parameter equation affected by harmonics, and the formula is as follows: (5); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively, a 3 、a 5 、a 7 is randomly generated in the range [0.05, 0.15], y [ i ] is the generated power quality disturbance data; The disturbance signal that generates fluctuations is generated by the voltage parameter equation of the microgrid where the fluctuation occurs, and the formula is as follows: (6); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. α E is the voltage amplitude fluctuation coefficient, which is randomly generated in the range of [0.1, 0.2], is the voltage frequency fluctuation coefficient, which is randomly generated in the range of [0.1, 0.4], y [ i ] is the generated power quality disturbance data; The disturbance signal of transient oscillation is generated by the microgrid voltage parameter equation of transient oscillation, and the formula is as follows: (7); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, t 3 , t 4 are the start and end times of the disturbance, T n is the period of transient oscillation, T n Randomly generated in the range [0.0011, 0.0033], τ is the time constant, τ Randomly generated in the range [150, 1000],k 1 is the amplitude coefficient of transient oscillation, k 1 Randomly generated in the range [0.1, 0.8], y [ i ] is the generated power quality disturbance data; The disturbance signal of transient impact is generated by the microgrid voltage parameter equation of transient impact, and the formula is as follows: (8); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α F is the amplitude coefficient of voltage transient impact, α F Randomly generated in the range [0, 0.414], t 5 , t 6 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data; The disturbance signal of periodic gap is generated by the voltage parameter equation of microgrid with periodic gap, and the formula is as follows: (9); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, sign(t) is the sign function, k 2 is the amplitude coefficient of the voltage periodic gap, which is randomly generated in the range of [0.1, 0.4]. t 7 , t 8 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data; The disturbance signal of periodic spikes is given by the microgrid voltage parameter equation of periodic spikes, which is as follows: (10); In the formula, T is the fundamental period of the microgrid voltage,N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, sign(t) is the sign function, k 2 is the amplitude coefficient of the voltage periodic gap, which is randomly generated in the range of [0.1, 0.4]. t 7 , t 8 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data.

[0024] Among them, in the composite disturbance signal: The disturbance signal of voltage drop and harmonic composite is generated by the voltage parameter equation of the microgrid with temporary voltage drop and the voltage parameter equation of the microgrid affected by harmonics. The formula is as follows: (11); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α B is the voltage drop amplitude coefficient, α B Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively, a 3 、a 5 、a 7 is randomly generated in the range [0.05, 0.15], y [ i ] is the generated power quality disturbance data; The disturbance signal of voltage rise and harmonic composite is generated by the voltage parameter equation of the microgrid with temporary voltage rise and the voltage parameter equation of the microgrid affected by harmonics. The formula is as follows: (12); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [t ] is the unit step response function, α C is the voltage rise amplitude coefficient, α C Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively, a 3 、a 5 、a 7 is randomly generated in the range [0.05, 0.15], y [ i ] is the generated power quality disturbance data; The disturbance waveform signal of temporary power outage and harmonic composite is generated by the voltage parameter equation of the microgrid affected by temporary power outage and the voltage parameter equation of the microgrid affected by harmonics. The formula is as follows: (13); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α D is the voltage drop amplitude coefficient, α D Randomly generated in the range [0.9, 1.0], t 1. t 2 are the start and end times of the disturbance, a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively, a 3 、a 5 、a 7 is randomly generated in the range [0.05, 0.15], y [ i ] is the generated power quality disturbance data; The disturbance signal of voltage temporary drop and fluctuation is generated by the voltage parameter equation of the microgrid with temporary voltage drop and the voltage parameter equation of the microgrid with fluctuation. The formula is as follows: (14); In the formula, Tis the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α B is the voltage drop amplitude coefficient, α B Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, α E is the voltage amplitude fluctuation coefficient, which is randomly generated in the range of [0.1, 0.2], is the frequency fluctuation coefficient of the voltage, which is randomly generated in the range of [0.1, 0.4], y [ i ] is the generated power quality disturbance data; The disturbance signal of voltage rise and fluctuation is generated by the voltage parameter equation of the microgrid with temporary voltage rise and the voltage parameter equation of the microgrid with fluctuation. The formula is as follows: (15); In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α C is the voltage rise amplitude coefficient, α C Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, α E is the voltage amplitude fluctuation coefficient, which is randomly generated in the range of [0.1, 0.2], is the frequency fluctuation coefficient of the voltage, which is randomly generated in the range of [0.1, 0.4], y [ i ] is the generated power quality disturbance data; The disturbance signal of harmonic and fluctuation composite is generated by the composite voltage parameter equation of the microgrid affected by the harmonic and the voltage parameter equation of the microgrid where the fluctuation occurs. The formula is as follows: (16); In the formula, T is the fundamental period of the microgrid voltage, Nis the number of points that the analog-to-digital converter can collect in 1 second. α E is the voltage amplitude fluctuation coefficient, which is randomly generated in the range of [0.1, 0.2], a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively, a 3 、a 5 、a 7 is randomly generated in the range [0.05, 0.15], is the frequency fluctuation coefficient of the voltage, which is randomly generated in the range of [0.1, 0.4], y [ i ] is the generated power quality disturbance data.

[0025] The power quality disturbance data generated by all the above microgrid voltage parameter equations together constitute the power quality disturbance data set.

[0026] It should be noted that the fundamental period of the microgrid voltage T The value of includes 0.02s, the number of points that the analog-to-digital converter can collect in 1 second N The values ​​of include 3200, and i is 0, 1, 2, …, 10NT-1 respectively.

[0027] t 1 , t 2 、t 3 , t 4 、t 5 , t 6 are all random integers in [0, 10NT-1], t 7 , t 8 are all random integers in [0, NT], t 1 , t 2 Satisfy NT ≤ t 2 - t 1 ≤ 9NT, t 3 , t 4 Satisfy 0.5NT ≤ t 4 - t3 ≤ 3NT, t 5 , t 6 Satisfy 0.05NT ≤ t 6 - t 5 ≤ 0.1NT, t 7 , t 8 Satisfy 0.01NT ≤ t 8 - t 7 ≤ 0.05NT.

[0028] a 1、 a 3 、a 5 、a 7. Satisfy between: (17); In the formula a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively.

[0029] This application adopts the method of synthesizing disturbance waveforms through microgrid voltage parameter equations. Compared with traditional acquisition and processing methods, the power quality disturbance data set does not need to be manually designed for feature extraction, does not rely on experience, avoids interference from artificial factors, and has better robustness.

[0030] like Figure 2 As shown in Figure 1, in S2, Hilbert transform is used in the data enhancement process. Hilbert transform is a common method in signal processing. Hilbert transform converts the signal into a complex form, including real and imaginary parts. The real part is equivalent to the original signal, and the imaginary part leads the phase by 90 degrees. Therefore, Hilbert transform is also called 90-degree phase shift transform. Compared with FFT, Hilbert transform does not cause energy loss, and can be used to draw the envelope of the signal and display the unilateral spectrum.

[0031] The data augmentation process consists of the following steps: The power quality disturbance voltage signal is normalized, and the formula is as follows: (18); In the formula, X is the original sample data, X´ is the normalized sample data, X max andX min are the maximum and minimum values ​​of the original sample data respectively; The Hilbert transform of the normalized power quality disturbance voltage signal is as follows: (19); In the formula, f (t) represents the normalized power quality disturbance voltage signal, Express f (t) signal after Hilbert transformation; Introducing analytical signals f a (t), the formula is as follows: (20); In the formula, f a (t) means f (t) The corresponding analytical signal; Using Euler's formula f a (t) is transformed and obtained f The envelope sequence and phase of (t) are as follows: (twenty one); (twenty two); (twenty three); In the formula, I A (t) means f (t) is the envelope sequence, Iθ (t) represents the phase; The envelope signal is used as a weight to perform dot product calculation with the normalized power quality disturbance voltage signal to complete data enhancement. The formula is as follows: (twenty four); In the formula, I A (t) means f (t) is the envelope sequence, y (t) represents the enhanced signal.

[0032] Please refer to Figure 2 and Figure 3In this embodiment, Python is used in combination with the drawing library provided by Python for simulation. The original waveform signal is Hilbert transformed by using formula (20) to formula (23). The original waveform signal is the 16 types of power quality disturbance voltage signals mentioned in this paragraph. The power quality disturbance voltage signal includes both normal sinusoidal voltage signals and local disturbances. The disturbance voltage signal is a disturbance voltage signal superimposed on the sinusoidal wave. The disturbance voltage signal will change on the basis of the sinusoidal wave, such as a drop or a sudden rise.

[0033] After data enhancement processing of the power quality disturbance voltage signal is performed through Hilbert transform, the normal part of the power quality disturbance voltage signal will remain almost unchanged, while the local disturbance part, that is, the non-sinusoidal wave band, will be amplified. Therefore, the enhanced data can accurately observe the local amplitude changes of the voltage signal.

[0034] The MLSTM-FCNs model is used to extract features from the power quality disturbance voltage signal after data enhancement. In this embodiment, the MLSTM-FCNs model adopts the existing technology. Please refer to Figure 4 ,The MLSTM-FCNs model contains a fully convolutional block and an LSTM block, which combines the ,advantages of the long short-term memory network (LSTM) and the fully convolutional ,network (FCN), which can not only effectively capture the temporal dependencies in time ,series data, but also extract rich local and global features.

[0035] Among them, the long short-term memory network (LSTM) is a special recursive neural network. Through its unique memory and forgetting mechanism, it can effectively capture the time dependency in sequence data and extract time-dependent features from the input data.

[0036] The fully convolutional network contains three temporal convolution blocks. As a feature extractor, each temporal convolution block contains a one-dimensional convolution layer with 128, 256, and 128 filters and 8, 5, and 3 kernel sizes, respectively. Each convolutional layer is followed by batch normalization (BN), which is used to normalize the output of each layer, reduce internal covariate shift, and improve the training efficiency and stability of the model. After normalization, the ReLU activation function is used. The ReLU activation function introduces nonlinearity, allowing the model to fit more complex functional relationships. The SE module further enhances the feature extraction capability of the model and improves the network's attention to important features by reweighting the feature channels. In addition, a squeeze-excitation block is followed by the first two convolution blocks. Finally, the temporal convolution block is followed by an average pooling layer. The introduction of the global pooling layer helps to reduce the number of model parameters, reduce the risk of overfitting, and effectively integrate global information. Through global pooling, each feature map is converted to a single value, thereby retaining the global characteristics of the feature map. Global pooling not only simplifies the network structure, but also enhances the generalization ability of the model. This squeeze-activation block is a supplement to the FCN block and is used to adaptively recalibrate the weight value of the feature map.

[0037] The MLSTM-FCNs model needs to be pre-trained before use, and the best training The model is saved as a basic feature extractor. In S3, when the MLSTM-FCNs model is pre-trained, the above 16 power quality disturbance data sets synthesized by the microgrid voltage parameter equation are input into the MLSTM-FCNs model after data enhancement, and the model parameters are adjusted. When the accuracy on the validation set reaches 90%, it is considered that the pre-training requirements are met, and the current model is the best training model. Save the current model.

[0038] After the pre-training of the MLSTM-FCNs model is completed and the best training model is obtained, the best training model can be used for feature extraction. It should be noted that in S4, the above 16 power quality disturbance data sets synthesized using the microgrid voltage parameter equation are input into the best training model operation after data enhancement, and the result of the operation is used as the input for the next step. The parameters in the MLSTM-FCNs model are not changed during feature extraction, and the feature extraction process also belongs to the prior art, which will not be repeated here.

[0039] In S4, the width learning classification model training process includes the following steps: The data passed through the feature extractor and its corresponding labels are input into the width learning classification model, and the feature mapping layer of the width learning classification model is trained. The formula is as follows: (25); In the formula, Z i For the i Group mapping features, For the i Group random mapping, X For input data, W ei For the i The random weight coefficients with appropriate dimensions corresponding to the group enhancement nodes, β ei For the i The bias of the group enhancement node; The enhancement layer of the width learning classification model is trained as follows: (26); In the formula, H j For the j Group Enhancement Node, is a nonlinear mapping on the linear combination result, Z m For the m Group mapping features, W hj For the j The random weight coefficients with appropriate dimensions corresponding to the group enhancement layer, β hj For the j Bias of group enhancement layer; Update the mapping function Y from input to output, the formula is as follows: (27); In the formula, Z m Before m Group all mapped features, Before n Group all enhanced nodes, W Represents the weight of each layer.

[0040] After the width learning classification model is trained, it can be used to accurately identify the power quality disturbance data of the actual power signal in the power system. When new data belonging to the category of the original data arrives, it is only necessary to simply expand the input data into the proposed algorithm to obtain the classification result of the added data. When the category of the added data does not belong to the category of the original data, the initial weights need to be adjusted. It is expanded to accommodate matrix operations, and then the weights are updated to calculate the recognition accuracy.

[0041] By combining deep learning, width learning and traditional signal processing technology, data-driven power quality disturbance is realized. Compared with traditional processing methods, a width learning algorithm is used as a classifier, so that the recognition method proposed in this scheme has the characteristic of continuous learning from new data compared to traditional deep learning algorithms.

[0042] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

Claims

1. A data-driven power quality disturbance identification method, characterized in that: The following steps are involved: S1: Obtain the power quality disturbance voltage signal within the set time period of the distribution network; S2: Performing data enhancement on the power quality disturbance voltage signal based on Hilbert transform to amplify the disturbance data in the power quality disturbance voltage signal; S3: Use the MLSTM-FCNs model and input the power quality disturbance voltage signal after data enhancement into the MLSTM-FCNs model for pre-training, and finally obtain and save the best training model; S4: Use the best training model as a feature extractor to extract voltage disturbance features, establish and train a width learning classification model based on the width learning algorithm, classify the power quality disturbance voltage signal through the width learning classification model, and output the recognition result of the disturbance data.

2. The data-driven power quality disturbance identification method according to claim 1, characterized in that: In S1, the power quality disturbance voltage signal includes a basic disturbance signal and a composite disturbance signal. The basic disturbance signals include: disturbance signals of negligible disturbance, disturbance signals of temporary voltage drop, disturbance signals of temporary voltage rise, disturbance signals of temporary power outage, disturbance signals of harmonic influence, disturbance signals of fluctuation, disturbance signals of transient oscillation, disturbance signals of transient impact, disturbance signals of periodic gap, and disturbance signals of periodic peak; The composite disturbance signal includes: a disturbance signal of voltage drop and harmonic composite, a disturbance signal of voltage rise and harmonic composite, a disturbance signal of temporary power outage and harmonic composite, a disturbance signal of voltage temporary drop and fluctuation composite, a disturbance signal of voltage rise and fluctuation composite, and a disturbance signal of harmonic and fluctuation composite.

3. The data-driven power quality disturbance identification method according to claim 2 is characterized in that: In the basic disturbance signal: The disturbance signal with negligible disturbance is generated by the microgrid voltage parameter equation with negligible disturbance, which is as follows: （1）； In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α A is the disturbance fluctuation coefficient, α A Randomly generated in the range (0, 0.1), t 1. t 2 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data; The disturbance signal of temporary voltage drop is generated by the voltage parameter equation of the microgrid with temporary voltage drop, and the formula is as follows: （2）； In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α B is the voltage drop amplitude coefficient, α B Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data; The disturbance signal of temporary voltage rise is generated by the voltage parameter equation of the microgrid with temporary voltage rise, and the formula is as follows: （3）； In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α C is the voltage rise amplitude coefficient, α C Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data; The disturbance signal of temporary power outage is generated by the voltage parameter equation of the microgrid with temporary power outage, and the formula is as follows: （4）； In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α D is the voltage drop amplitude coefficient, α D Randomly generated in the range [0.9, 1.0], t 1. t 2 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data; The disturbance signal affected by harmonics is generated by the microgrid voltage parameter equation affected by harmonics, and the formula is as follows: （5）； In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively, a 3 、a 5 、a 7 is randomly generated in the range [0.05, 0.15], y [ i ] is the generated power quality disturbance data; The disturbance signal that generates fluctuations is generated by the voltage parameter equation of the microgrid where the fluctuation occurs, and the formula is as follows: （6）； In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. α E is the voltage amplitude fluctuation coefficient, which is randomly generated in the range of [0.1, 0.2], is the voltage frequency fluctuation coefficient, which is randomly generated in the range of [0.1, 0.4], y [ i ] is the generated power quality disturbance data; The disturbance signal of transient oscillation is generated by the microgrid voltage parameter equation of transient oscillation, and the formula is as follows: （7）； In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, t 3 , t 4 are the start and end times of the disturbance, T n is the period of transient oscillation, T n Randomly generated in the range [0.0011, 0.0033], τ is the time constant, τ Randomly generated in the range [150, 1000], k 1 is the amplitude coefficient of transient oscillation, k 1 Randomly generated in the range [0.1, 0.8], y [ i ] is the generated power quality disturbance data; The disturbance signal of transient impact is generated by the microgrid voltage parameter equation of transient impact, and the formula is as follows: （8）； In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α F is the amplitude coefficient of voltage transient impact, α F Randomly generated in the range [0, 0.414], t 5 , t 6 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data; The disturbance signal of periodic gap is generated by the voltage parameter equation of microgrid with periodic gap, and the formula is as follows: （9）； In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, sign(t) is the sign function, k 2 is the amplitude coefficient of the voltage periodic gap, which is randomly generated in the range of [0.1, 0.4]. t 7 , t 8 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data; The disturbance signal of periodic spikes is given by the microgrid voltage parameter equation of periodic spikes, which is as follows: （10）； In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, sign(t) is the sign function, k 2 is the amplitude coefficient of the voltage periodic gap, which is randomly generated in the range of [0.1, 0.4]. t 7 , t 8 are the start and end times of the disturbance, y [ i ] is the generated power quality disturbance data.

4. The data-driven power quality disturbance identification method according to claim 3 is characterized in that: In the composite disturbance signal: The disturbance signal of voltage drop and harmonic composite is generated by the voltage parameter equation of the microgrid with temporary voltage drop and the voltage parameter equation of the microgrid affected by harmonics. The formula is as follows: （11）； In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α B is the voltage drop amplitude coefficient, α B Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively, a 3 、a 5 、a 7 is randomly generated in the range [0.05, 0.15], y [ i ] is the generated power quality disturbance data; The disturbance signal of voltage rise and harmonic composite is generated by the voltage parameter equation of the microgrid with temporary voltage rise and the voltage parameter equation of the microgrid affected by harmonics. The formula is as follows: （12）； In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α C is the voltage rise amplitude coefficient, α C Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively, a 3 、a 5 、a 7 is randomly generated in the range [0.05, 0.15], y [ i ] is the generated power quality disturbance data; The disturbance waveform signal of temporary power outage and harmonic composite is generated by the voltage parameter equation of the microgrid affected by temporary power outage and the voltage parameter equation of the microgrid affected by harmonics. The formula is as follows: （13）； In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α D is the voltage drop amplitude coefficient, α D Randomly generated in the range [0.9, 1.0], t 1. t 2 are the start and end times of the disturbance, a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively, a 3 、a 5 、a 7 is randomly generated in the range [0.05, 0.15], y [ i ] is the generated power quality disturbance data; The disturbance signal of voltage temporary drop and fluctuation is generated by the voltage parameter equation of the microgrid with temporary voltage drop and the voltage parameter equation of the microgrid with fluctuation. The formula is as follows: （14）； In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α B is the voltage drop amplitude coefficient, α B Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, α E is the voltage amplitude fluctuation coefficient, which is randomly generated in the range of [0.1, 0.2], is the frequency fluctuation coefficient of the voltage, which is randomly generated in the range of [0.1, 0.4], y [ i ] is the generated power quality disturbance data; The disturbance signal of voltage rise and fluctuation is generated by the voltage parameter equation of the microgrid with temporary voltage rise and the voltage parameter equation of the microgrid with fluctuation. The formula is as follows: （15）； In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. u [ t ] is the unit step response function, α C is the voltage rise amplitude coefficient, α C Randomly generated in the range [0.1, 0.9], t 1. t 2 are the start and end times of the disturbance, α E is the voltage amplitude fluctuation coefficient, which is randomly generated in the range of [0.1, 0.2], is the frequency fluctuation coefficient of the voltage, which is randomly generated in the range of [0.1, 0.4], y [ i ] is the generated power quality disturbance data; The disturbance signal of harmonic and fluctuation composite is generated by the composite voltage parameter equation of the microgrid affected by the harmonic and the voltage parameter equation of the microgrid where the fluctuation occurs. The formula is as follows: （16）； In the formula, T is the fundamental period of the microgrid voltage, N is the number of points that the analog-to-digital converter can collect in 1 second. α E is the voltage amplitude fluctuation coefficient, which is randomly generated in the range of [0.1, 0.2], a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively, a 3 、a 5 、a 7 is randomly generated in the range [0.05, 0.15], is the frequency fluctuation coefficient of the voltage, which is randomly generated in the range of [0.1, 0.4], y [ i ] is the generated power quality disturbance data.

5. The data-driven power quality disturbance identification method according to claim 4 is characterized in that: The microgrid voltage fundamental period T The value of includes 0.02s, the number of points that the analog-to-digital converter can collect in 1 second N The values ​​of include 3200, and i is 0, 1, 2, …, 10NT-1 respectively.

6. A data-driven power quality disturbance identification method according to claim 4, characterized in that: Said t 1 , t 2 、t 3 , t 4 、t 5 , t 6 are all random integers in [0, 10NT-1], t 7 , t 8 are all random integers in [0, NT]. t 1 , t 2 Satisfy NT ≤ t 2 - t 1 ≤ 9NT, t 3 , t 4 Satisfy 0.5NT ≤ t 4 - t 3 ≤ 3NT, t 5 , t 6 Satisfy 0.05NT ≤ t 6 - t 5 ≤ 0.1NT, t 7 , t 8 Satisfy 0.01NT ≤ t 8 - t 7 ≤ 0.05NT.

7. The data-driven power quality disturbance identification method according to claim 4 is characterized in that: Said a 1、 a 3 、a 5 、a 7. Satisfy between: （17）； In the formula a 1 is the fundamental amplitude, a 3 、a 5 、a 7 are the third, fifth and seventh harmonic amplitudes respectively.

8. The data-driven power quality disturbance identification method according to claim 1, characterized in that: In S2, the data enhancement process includes the following steps: The power quality disturbance voltage signal is normalized, and the formula is as follows: （18）； In the formula, X is the original sample data, X´ is the normalized sample data, X max and X min are the maximum and minimum values ​​of the original sample data respectively; The Hilbert transform of the normalized power quality disturbance voltage signal is as follows: （19）； In the formula, f (t) represents the normalized power quality disturbance voltage signal, Express f (t) signal after Hilbert transformation; Introducing analytical signals f a (t), the formula is as follows: （20）； In the formula, f a (t) means f (t) the corresponding analytical signal; Using Euler's formula f a (t) is transformed and obtained f The envelope sequence and phase of (t) are as follows: （21）； （22）； （23）； In the formula, I A (t) means f (t) is the envelope sequence, Iθ (t) represents the phase; The envelope signal is used as a weight to perform dot product calculation with the normalized power quality disturbance voltage signal to complete data enhancement. The formula is as follows: （24）； In the formula, I A (t) means f (t) is the envelope sequence, y (t) represents the enhanced signal.

9. The data-driven power quality disturbance identification method according to claim 1, characterized in that: In S3, the judgment criteria for the best training model of MLSTM-FCNs include: after the power quality disturbance voltage signal after data enhancement is input into the current MLSTM-FCNs model, the accuracy on the verification set is greater than or equal to 90%, which meets the pre-training requirements, and the current model is the best training model.

10. The data-driven power quality disturbance identification method according to claim 1, characterized in that: In S4, the width learning classification model training process includes the following steps: The data passed through the feature extractor and its corresponding labels are input into the width learning classification model, and the feature mapping layer of the width learning classification model is trained. The formula is as follows: （25）； In the formula, Z i For the i Group mapping features, For the i Group random mapping, X For input data, W ei For the i The random weight coefficients with appropriate dimensions corresponding to the group enhancement nodes, β ei For the i The bias of the group enhancement node; The enhancement layer of the width learning classification model is trained as follows: （26）； In the formula, H j For the j Group Enhancement Node, is a nonlinear mapping on the linear combination result, Z m For the m Group mapping features, W hj For the j The random weight coefficients with appropriate dimensions corresponding to the group enhancement layer, β hj For the j Bias of group enhancement layer; Update the mapping function Y from input to output, the formula is as follows: （27）； In the formula, Z m Before m Group all mapped features, Before n Group all enhanced nodes, W Represents the weight of each layer.

Images