Photovoltaic grid-connected harmonic current prediction method based on deep learning

Abstract

The invention discloses a photovoltaic grid-connected harmonic current prediction method based on deep learning, and relates to the technical field of power systems and photovoltaic grid connection. According to the method, three-phase load current and grid-connected point voltage data in a photovoltaic grid-connected system are obtained, the three-phase load current is converted into a two-phase direct-current component through Clark and Park conversion, fundamental current is extracted through a low-pass filter, and a harmonic current component is obtained. Time domain and frequency domain characteristics of grid-connected point voltage data are extracted through fast Fourier transform, correlation between voltage harmonic waves and load current harmonic waves is analyzed, and grid-connected point voltage characteristics are extracted. Local features of harmonic current components are extracted through a convolutional neural network, time sequence features are captured through a bidirectional long-short-term memory network, and a harmonic current prediction model is constructed. And finally, inputting the preprocessed current and voltage characteristics, and generating a harmonic current predicted value. And the problems of nonlinearity and uncertainty in the photovoltaic grid-connected system can be more effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of power system and photovoltaic grid-connected technology, and specifically to a photovoltaic grid-connected harmonic current prediction method based on deep learning. Background Art

[0002] With the rapid development of photovoltaic power generation grid-connected systems, the use of a large number of power electronic devices has caused photovoltaic grid-connected systems to generate more harmonics during operation, resulting in power quality problems. At the same time, with the exponential increase in nonlinear loads, the harmonic problem of the power grid system during operation has intensified, which may lead to a decrease in system power factor, energy loss, accelerated insulation aging of generators and transformers, and even malfunction of relays, causing fires and other dangerous accidents. While photovoltaic power generation is booming, harmonic problems should not be underestimated.

[0003] With the vigorous promotion of the "dual carbon" policy, new energy sources are gradually rising, but their nonlinear operating characteristics make traditional harmonic detection technology appear exhausted and unable to meet the requirements. At present, traditional harmonic detection technology has shown fatigue and cannot meet the requirements of harmonic analysis and prediction in modern complex power systems. Therefore, how to accurately and effectively predict harmonic currents has become the focus of research. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a photovoltaic grid-connected harmonic current prediction method based on deep learning, which solves the problems of the above-mentioned background technology.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a photovoltaic grid-connected harmonic current prediction method based on deep learning, comprising the following steps: S1. Acquire three-phase load current and grid-connected point voltage data in a photovoltaic grid-connected system; S2. Convert the three-phase load current into a two-phase DC component through Clark transform and Park transform, extract the fundamental current of the two-phase DC component through a low-pass filter, calculate the difference between the total current of the three-phase load current and the fundamental current, obtain the harmonic current component, and pre-process the harmonic current component; S3. Extract the time domain and frequency domain characteristics of the grid-connected point voltage data through fast Fourier transform, analyze the amplitude and frequency distribution of the voltage harmonic component, evaluate the correlation between the voltage harmonic and the load current harmonic, and obtain the grid-connected point voltage characteristics; S4. Extract local features of the harmonic current component through a convolutional neural network, capture the time series characteristics of the harmonic current component through a bidirectional long short-term memory network, and construct a harmonic current prediction model; S5. Input the pre-processed harmonic current component and the grid-connected point voltage characteristics into the wave current prediction model to generate a harmonic current prediction value.

[0006] Furthermore, the three-phase load current is converted into a two-phase DC component through Clark transformation and Park transformation, and the specific process of extracting the fundamental current of the two-phase DC component through a low-pass filter is as follows: the three-phase load current is converted into a two-phase DC component through Clark transformation, and the three-phase AC current is projected into a stationary coordinate system; the two-phase DC component is converted into a DC component in a rotating coordinate system through Park transformation to obtain a DC component of the two-phase AC current; the fundamental current component in the DC component of the two-phase AC current is extracted through a low-pass filter to remove the high-frequency harmonic component.

[0007] Furthermore, the specific process of subtracting the total current of the three-phase load current from the fundamental current to obtain the harmonic current component is as follows: the total current of the three-phase load current is calculated based on the three-phase load current; the fundamental component is extracted from the DC component of the two-phase AC current, and the calculated total current of the three-phase load current is subtracted from the extracted fundamental current to obtain the harmonic current component.

[0008] Furthermore, the specific process of preprocessing the harmonic current components is as follows: denoising the harmonic current components, using filters to remove high-frequency noise components; standardizing and smoothing the denoised harmonic current components to remove mutation values ​​and abnormal fluctuations; interpolating and filling data according to the time series characteristics of the harmonic current components, processing missing values, and ensuring data integrity; extracting features within the local time period through the window sliding method to generate the time series data format required for training.

[0009] Furthermore, the time domain and frequency domain characteristics of the grid connection point voltage data are extracted by fast Fourier transform, and the specific process of analyzing the amplitude and frequency distribution of the voltage harmonic components is as follows: fast Fourier transform is performed on the grid connection point voltage data, the time domain data is converted into frequency domain data, the frequency components in the grid connection point voltage signal are extracted, and the main frequency peaks of the voltage signal are identified; the amplitude of each frequency component is calculated, and the contribution of different frequencies is evaluated; the harmonic components of the voltage signal are analyzed, including the amplitude and frequency distribution of the fundamental wave and each order harmonic, to obtain the harmonic characteristics of the voltage signal; according to the frequency domain analysis results, the energy distribution of each frequency band in the voltage signal is determined.

[0010] Furthermore, the correlation between voltage harmonics and load current harmonics is evaluated, and the specific process of obtaining the voltage characteristics of the grid-connected point is as follows: by analyzing the harmonic components of the voltage signal and the harmonic components of the load current, the similarities between the voltage signal and the load current in the frequency domain are compared, and the relationship between the two in different frequency bands is evaluated; the correlation coefficient is used to quantify the degree of correlation between the voltage harmonics and the load current harmonics, and based on the correlation analysis results, the correlation characteristics between the voltage signal and the load current signal are extracted to obtain the characteristic vector of the grid-connected point voltage.

[0011] Furthermore, the specific process of extracting local features of harmonic current components through convolutional neural networks is as follows: the time series data of harmonic current components is input into the convolutional neural network, and the features of the input harmonic current components are extracted through the convolution layer; the local spatial features of the harmonic current components are captured by sliding the convolution kernel in the local area, and the data dimension is reduced through the pooling layer to reduce the computational complexity; the local features of the harmonic current components at different levels are gradually extracted through the convolution layer and the pooling layer, and the temporal feature learning is represented.

[0012] Furthermore, the specific process of capturing the timing characteristics of harmonic current components through a bidirectional long short-term memory network is as follows: the local characteristics of the harmonic current components extracted by the convolutional neural network are input into the bidirectional long short-term memory network; the bidirectional long short-term memory network captures the timing characteristics in the time series by modeling the timing data of the harmonic current components; the bidirectional long short-term memory network learns the forward and reverse timing information during the forward and backward propagation processes, respectively, and generates a forward hidden state vector and a reverse hidden state vector; based on the forward hidden state vector and the reverse hidden state vector, a hidden state matrix of the bidirectional timing information is generated.

[0013] Furthermore, the specific process of generating harmonic current prediction values ​​is as follows: the preprocessed harmonic current components and grid-connected point voltage characteristics are input into the wave current prediction model, the input harmonic current components and grid-connected point voltage characteristics are time-series modeled, and the dynamic change characteristics of current and voltage are integrated to generate harmonic current prediction values.

[0014] The present invention has the following beneficial effects:

[0015] (1) This deep learning-based photovoltaic grid-connected harmonic current prediction method provides an accurate basic data source for harmonic current prediction by obtaining the three-phase load current and grid-connected point voltage data of the photovoltaic grid-connected system. The three-phase load current is converted into a two-phase DC component by Clark transform and Park transform, and the fundamental current is extracted by combining a low-pass filter to effectively separate the harmonic current component. By eliminating the interference of the fundamental current, the accuracy of harmonic current component extraction is improved, providing purer harmonic feature data for subsequent model training. By extracting the time domain and frequency domain characteristics of the grid-connected point voltage through fast Fourier transform, not only can the amplitude and frequency distribution of voltage harmonics be quantified, but also its correlation with load current harmonics can be evaluated. By comprehensively analyzing the grid-connected point voltage characteristics, it is helpful to reveal the formation mechanism of harmonic current and its relationship with the power grid. The ability to identify complex harmonic components is improved, providing key input for the prediction model, and improving its adaptability to nonlinearity and time variability.

[0016] (2) This photovoltaic grid-connected harmonic current prediction method based on deep learning extracts local features of harmonic current components through convolutional neural networks, and then combines bidirectional long short-term memory networks to capture time series features, thereby achieving deep fusion of multi-dimensional features. Convolutional neural networks focus on spatial feature extraction, while bidirectional long short-term memory networks accurately capture dynamic changes in time series, thereby improving the ability to characterize complex characteristics of harmonic currents. This solves the problem that traditional prediction methods cannot fully utilize time series information, making the model more able to reflect the evolution law of actual harmonic currents and improving the accuracy and robustness of prediction results. The preprocessed harmonic current components and grid-connected point voltage characteristics are input into the harmonic current prediction model, and multi-feature fusion is used to improve the prediction accuracy. By combining current and voltage characteristic information, the model can comprehensively reflect the grid operation status and load characteristics, and dynamically generate accurate harmonic current prediction values. This significantly improves the stability and safety of system operation.

[0017] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flow chart of the photovoltaic grid-connected harmonic current prediction method based on deep learning of the present invention. DETAILED DESCRIPTION

[0019] The embodiments of the present application use a photovoltaic grid-connected harmonic current prediction method based on deep learning to solve the problems of low harmonic current prediction accuracy, poor real-time performance, and inability to effectively capture complex timing characteristics in existing photovoltaic grid-connected systems.

[0020] The overall idea of ​​the solution in the embodiments of this application is as follows:

[0021] Obtain three-phase load current and grid connection point voltage data in the photovoltaic grid-connected system.

[0022] The three-phase load current is converted into a two-phase DC component through Clark transformation and Park transformation. The fundamental current of the two-phase DC component is extracted through a low-pass filter. The total current of the three-phase load current is subtracted from the fundamental current to obtain the harmonic current component, and the harmonic current component is preprocessed.

[0023] The time domain and frequency domain characteristics of the grid connection point voltage data are extracted through fast Fourier transform, the amplitude and frequency distribution of the voltage harmonic components are analyzed, the correlation between the voltage harmonics and the load current harmonics is evaluated, and the voltage characteristics of the grid connection point are obtained.

[0024] The local features of harmonic current components are extracted through convolutional neural networks, and the time series characteristics of harmonic current components are captured through bidirectional long short-term memory networks to build a harmonic current prediction model.

[0025] The preprocessed harmonic current components and grid-connected point voltage characteristics are input into the harmonic current prediction model to generate harmonic current prediction values.

[0026] See also Figure 1 , an embodiment of the present invention provides a technical solution: a photovoltaic grid-connected harmonic current prediction method based on deep learning, comprising the following steps: S1. Acquire three-phase load current and grid-connected point voltage data in a photovoltaic grid-connected system; S2. Convert the three-phase load current into a two-phase DC component through Clark transform and Park transform, extract the fundamental current of the two-phase DC component through a low-pass filter, calculate the difference between the total current of the three-phase load current and the fundamental current, obtain the harmonic current component, and pre-process the harmonic current component; S3. Extract the time domain and frequency domain characteristics of the grid-connected point voltage data through fast Fourier transform, analyze the amplitude and frequency distribution of the voltage harmonic component, evaluate the correlation between the voltage harmonic and the load current harmonic, and obtain the grid-connected point voltage characteristics; S4. Extract local features of the harmonic current component through a convolutional neural network, capture the time series features of the harmonic current component through a bidirectional long short-term memory network, and construct a harmonic current prediction model; S5. Input the pre-processed harmonic current component and the grid-connected point voltage characteristics into the wave current prediction model to generate a harmonic current prediction value.

[0027] In this implementation scheme, step S1: the core input data of the system is obtained by collecting the three-phase load current and grid-connected point voltage data in the photovoltaic grid-connected system. The three-phase load current reflects the dynamic changes of the load in the grid-connected system, while the grid-connected point voltage data reflects the overall characteristics of the grid operation. By obtaining these data, a basic guarantee is provided for the subsequent extraction and feature analysis of harmonic current components. Step S2: The three-phase load current is converted into a two-phase DC component by Clark transform and Park transform. This processing method can simplify the feature expression of the three-phase signal and facilitate subsequent calculations. At the same time, the fundamental current of the two-phase DC component is extracted by a low-pass filter, and the total current of the three-phase load current is subtracted from the fundamental current, so as to accurately separate the harmonic current component. Step S3: The time domain and frequency domain features are extracted from the grid-connected point voltage data using the fast Fourier transform (FFT), which can comprehensively analyze the harmonic components in the voltage signal. Specifically, the amplitude and frequency distribution of the harmonics can be quantified by FFT, and the correlation between the voltage harmonics and the load current harmonics can be further evaluated. These features provide the model with important information reflecting the operating status of the power grid, improving the globality and accuracy of the prediction. Step S4: The local features of the harmonic current components are extracted using a convolutional neural network (CNN). CNN can identify the spatial distribution characteristics in harmonic signals. Combined with a bidirectional long short-term memory network (Bi-LSTM), the time series characteristics of the harmonic current components are captured. Bi-LSTM learns the dynamic characteristics of the time series through forward and backward propagation. Step S5: The preprocessed harmonic current components and grid-connected point voltage features are used as inputs to the constructed harmonic current prediction model. The model generates a predicted value of the harmonic current by fusing multiple features. This predicted value provides an important basis for subsequent harmonic governance and system optimization, and helps to discover harmonic problems in advance and take effective measures. Clark transform is a mathematical method used to convert three-phase AC signals (such as current or voltage) from a three-dimensional spatial coordinate system to a two-dimensional plane coordinate system. This transformation simplifies the originally complex three-phase signal into two components, which is more convenient for subsequent calculation and analysis. In power systems, Clark transform is widely used to process symmetrical and asymmetrical three-phase signals in order to more intuitively analyze signal characteristics. Park transform is a method developed based on Clark transform, which is used to further transform the signal in the two-dimensional plane coordinate system into the signal in the rotating coordinate system. Through this transformation, the change of the AC signal can be described as a steady-state DC signal, which is particularly suitable for extracting the fundamental component in the signal. This method is very important in the field of power grid and motor control, and can effectively distinguish between harmonic components and fundamental components. Fast Fourier transform is an efficient computational method used to convert time domain signals into frequency domain signals. Through this transformation, the frequency composition of the signal can be analyzed and the fundamental and harmonic components can be identified.Fast Fourier transform is very important in harmonic analysis. It can reveal the frequency distribution and amplitude characteristics of the signal, thereby evaluating the harmonic characteristics of the voltage and current in the power grid. Convolutional neural network is a deep learning model that can automatically extract features from data. In the present invention, CNN is used to process harmonic current components and extract local features therein. Compared with traditional feature extraction methods, CNN does not require manual design of features, but automatically captures patterns and details in the signal through learning, making the processing process more efficient and accurate. Bidirectional long short-term memory network is a deep learning model for processing time series data, which can simultaneously capture the forward and reverse characteristics of the signal in time. Its advantage is that it can make full use of historical information and future information to model the dynamic changes of the signal. In harmonic current prediction, Bi-LSTM can accurately capture the timing characteristics of the signal, which helps to improve the prediction accuracy.

[0028] Specifically, the three-phase load current is converted into a two-phase DC component through Clark transformation and Park transformation, and the specific process of extracting the fundamental current of the two-phase DC component through a low-pass filter is as follows: the three-phase load current is converted into a two-phase DC component through Clark transformation, and the three-phase AC current is projected into a stationary coordinate system; the two-phase DC component is converted into a DC component in a rotating coordinate system through Park transformation to obtain the DC component of the two-phase AC current; the fundamental current component in the DC component of the two-phase AC current is extracted through a low-pass filter to remove the high-frequency harmonic component.

[0029] In this embodiment, Clark transformation: The purpose of Clark transformation is to convert three-phase alternating current (such as I a ,I b ,I c ) By projecting from the symmetrical three-phase coordinate system to the α-β stationary coordinate system, the complex three-phase signal can be converted into a two-phase signal, which is convenient for subsequent calculation and analysis. The formula is expressed as: Formula parameter explanation: I a ,I b ,I c : Three-phase current signal, representing the instantaneous current of the load in the photovoltaic grid-connected system. I α ,I β : The two-phase current components in the stationary coordinate system represent the projections on the α axis and the β axis respectively. Projection matrix: maps the three-phase system to a two-dimensional plane (stationary coordinate system). This transformation is essentially the projection of the three-phase signal from the rotating space to the stationary coordinate system. α ,I βThe signal is still an AC signal, but it has a relatively simplified mathematical characteristic, which is convenient for subsequent processing. Park transform: The function of Park transform is to transform the α-β signal in the stationary coordinate system into a dq signal in the rotating coordinate system, where the d-axis signal is the DC component and the q-axis signal reflects the rotation component. The formula is expressed as: Formula parameter explanation: θ: The phase angle at the current moment, synchronized with the angular frequency of the grid voltage, used to define the direction of the rotating coordinate system I d : The component in the rotating coordinate system that is consistent with the θ direction corresponds to the DC fundamental wave component. q : The component perpendicular to θ in the rotating coordinate system mainly reflects the oscillating part of the AC. Through this transformation, the fundamental signal of the power grid is extracted to the d-axis component, which is expressed as an approximately unchanged DC value, while the high-order harmonics and noise components are still reflected as oscillating signals, concentrated in the q-axis component.

[0030] Specifically, the specific process of subtracting the total current of the three-phase load current from the fundamental current to obtain the harmonic current component is as follows: according to the three-phase load current, the total current of the three-phase load current is calculated; the fundamental component is extracted from the DC component of the two-phase AC current, and the calculated total current of the three-phase load current and the extracted fundamental current are differenced to obtain the harmonic current component.

[0031] In this embodiment, according to the three-phase load current I a ,I b ,I c , the total current of the three-phase load current can be calculated by simple vector and operation. Assume I total Indicates the total current of the three-phase current, which is calculated as follows: Formula parameter explanation: I a ,I b ,I c : Respectively represent the instantaneous values ​​of the three-phase current. total : represents the total composite current of the three-phase current. The purpose of this step is to combine the magnitudes of the three-phase currents to provide a reference value for subsequent calculations. The three-phase current is converted into a two-phase DC component through methods such as Clark transformation and Park transformation, and then the fundamental current component is extracted from it through a low-pass filter. The fundamental current component represents the most important frequency component in the current signal, and is usually extracted synchronously with the grid frequency as I base , which can be obtained from I d The amount obtained is I d After low-pass filtering, the component represents the fundamental DC current. The total current I total And the extracted fundamental current component I base Perform difference calculation to obtain harmonic current components. This process can be expressed by the following formula:harmonic =I total -I base ; Formula parameter explanation: I total : The total current of the three-phase load current, the calculated composite current. I base : The fundamental current extracted from the DC component, the fundamental component after being processed by a low-pass filter. harmonic : The final harmonic current component obtained represents the part of the total current after removing the fundamental component.

[0032] Specifically, the specific process of preprocessing the harmonic current components is as follows: denoising the harmonic current components, using filters to remove high-frequency noise components; standardizing and smoothing the denoised harmonic current components to remove mutation values ​​and abnormal fluctuations; interpolating and filling data according to the time series characteristics of the harmonic current components, processing missing values, and ensuring data integrity; extracting features within the local time period through the window sliding method to generate the time series data format required for training.

[0033] In this embodiment, the harmonic current component may be disturbed by high-frequency noise. In order to ensure the accuracy of subsequent analysis, these noise components need to be removed. At this time, a filter is used for denoising. The function of the filter is to retain only meaningful signal components by setting a frequency range and remove high-frequency noise components. The filter can be designed as a low-pass, high-pass or band-pass filter according to requirements to eliminate unnecessary high-frequency noise. The denoised harmonic current component often has sudden changes or abnormal fluctuations, and these irregular fluctuations may affect the stability of subsequent analysis. Therefore, it is necessary to standardize it to ensure that the mean of the data is 0 and the standard deviation is 1, making it more suitable for the training of machine learning models. At the same time, smoothing is used to remove high-frequency jitter in the signal to avoid affecting the accuracy of the prediction model. The smoothing method can be a simple moving average or a more complex weighted average method to reduce data fluctuations. In the actual acquisition process, there may be missing values ​​in the harmonic current data. Missing data will affect the accuracy of subsequent analysis, so it is necessary to interpolate and fill the missing values. The interpolation method predicts missing values ​​based on the time series characteristics of the harmonic current component. Commonly used interpolation methods include linear interpolation, spline interpolation, etc. Through interpolation processing, the integrity of the time series data is ensured so that subsequent analysis can be performed on continuous data. The characteristics of the harmonic current components are not only reflected in the trend of the overall data, but the characteristics in the local time period are also very important. In order to better capture the detailed changes in the time series, the window sliding method is used to extract the features in the local time period. Specifically, the entire time series is divided into multiple small time windows, each containing a certain number of data points. In each window, features related to the prediction target are extracted, such as mean, standard deviation, maximum value, etc. These local features are crucial for model training and help improve the model's ability to capture changes in time series.

[0034] Specifically, the time domain and frequency domain characteristics of the grid connection point voltage data are extracted by fast Fourier transform, and the specific process of analyzing the amplitude and frequency distribution of the voltage harmonic components is as follows: fast Fourier transform is performed on the grid connection point voltage data, the time domain data is converted into frequency domain data, the frequency components in the grid connection point voltage signal are extracted, and the main frequency peaks of the voltage signal are identified; the amplitude of each frequency component is calculated, and the contribution of different frequencies is evaluated; the harmonic components of the voltage signal are analyzed, including the amplitude and frequency distribution of the fundamental wave and each order harmonic, to obtain the harmonic characteristics of the voltage signal; according to the frequency domain analysis results, the energy distribution of each frequency band in the voltage signal is determined.

[0035] In this embodiment, the grid connection point voltage data is converted from the time domain to the frequency domain. This is achieved by fast Fourier transform (FFT). FFT is an efficient algorithm that can convert time domain signals into frequency domain signals, so that we can observe the contribution of different frequency components in the signal. In short, FFT reveals the amplitude and phase information of the input voltage signal at different frequencies by mathematically transforming the input voltage signal. After Fourier transform, the obtained frequency domain data represents the intensity of the voltage signal at each frequency. In the frequency domain, the size (amplitude) of each frequency component corresponds to the contribution of the frequency to the voltage signal. In this way, the main frequency peak in the voltage signal, that is, the most significant frequency component in the signal, can be identified. These frequency components may represent basic signals (such as fundamental frequency) and harmonic signals in the power system. For each frequency component in the frequency domain data, we calculate its amplitude, that is, the energy of the frequency in the voltage signal. The larger the amplitude, the more significant the impact of the frequency component on the overall voltage signal. The process of calculating the amplitude is completed by the modulus length in the complex frequency domain representation obtained by Fourier transform. Amplitude information is crucial for subsequent harmonic analysis and can reveal the spectral structure of the voltage signal. Harmonic components in voltage signals refer to high-frequency components other than the fundamental frequency (i.e., power frequency), which are caused by common nonlinear factors in power systems. Through Fourier transform analysis, the voltage signal can be decomposed into fundamental and harmonic components of various orders. The fundamental is usually the main frequency of the system, while the harmonic is an integer multiple of the fundamental. These harmonic components affect the operation of power equipment. Therefore, accurately identifying and analyzing the amplitude of each order of harmonics and their frequency distribution is crucial for system optimization and fault prevention. Based on the frequency domain data obtained from Fourier transform, the energy distribution of the voltage signal in different frequency bands can be further evaluated. This means that we can identify the energy aggregation of different frequency bands in the voltage signal. The analysis of energy distribution helps us understand which frequency bands have a greater impact on voltage quality, especially in the case of severe harmonic pollution, which can provide a basis for system optimization.

[0036] Specifically, the specific process of evaluating the correlation between voltage harmonics and load current harmonics and obtaining the voltage characteristics of the grid-connected point is as follows: by analyzing the harmonic components of the voltage signal and the harmonic components of the load current, comparing the similarities between the voltage signal and the load current in the frequency domain, and evaluating the relationship between the two in different frequency bands; quantifying the degree of correlation between the voltage harmonics and the load current harmonics through the correlation coefficient, and based on the correlation analysis results, extracting the correlation characteristics between the voltage signal and the load current signal, and obtaining the characteristic vector of the grid-connected point voltage.

[0037] In this embodiment, the voltage signal and the load current signal are subjected to Fourier transform respectively to extract their respective harmonic components. The voltage signal and the load current signal usually contain fundamental components and harmonic components of various orders. In this step, we focus on the amplitude and frequency distribution of each frequency component (including fundamental and harmonic) in the voltage and load current. Then, by comparing these harmonic components, the similarity of the voltage signal and the load current signal in the frequency domain is analyzed to find out their similarities and differences in different frequency bands. By comparing the harmonic components of the voltage and load current signals in the frequency domain, we can understand the relationship between the two in different frequency ranges. For example, some frequency bands may show obvious synchronization, while other frequency bands may have large differences. The comparison process not only focuses on the alignment of frequencies, but also includes the matching of amplitudes. This analysis helps to identify the correlation between the voltage signal and the load current signal, and understand how the change of load current affects the quality of the power grid voltage, especially when the harmonic pollution is severe. In order to quantitatively evaluate the correlation between voltage harmonics and load current harmonics, the correlation coefficient is a common indicator used to measure the similarity between the two signals, and the formula is as follows: Where: ρ represents the correlation coefficient between voltage and load current; x i and i are the amplitudes of the voltage signal and the load current signal in a certain frequency band respectively; and are the mean values ​​of the voltage signal and the load current signal in the frequency band respectively; n is the number of data points. This formula calculates the correlation degree of the voltage and load current signals in each frequency band to determine whether the two have a significant linear relationship in the frequency domain. Its value range is usually between -1 and +1, where +1 indicates a complete positive correlation, -1 indicates a complete negative correlation, and 0 indicates no linear correlation. Here, we calculate the correlation coefficients of the harmonic components of the voltage signal and the load current signal, quantify the correlation between them, and evaluate the mutual influence between the voltage and load current harmonics in different frequency bands. According to the correlation coefficient analysis results, we can extract the correlation features between the voltage signal and the load current signal. For example, in some frequency bands, the change trends of the voltage signal and the load current signal may be very similar, while in other frequency bands, the correlation is low. Through the extraction of these features, we are able to construct a feature vector that describes the relationship between voltage and load current. This feature vector contains the relevant information between the voltage harmonics and the load current harmonics, and is one of the key inputs for the subsequent harmonic current prediction model. Based on the above correlation analysis, the correlation features between the voltage signal and the load current signal are obtained, thereby extracting the feature vector of the grid connection point voltage. This feature vector contains the characteristics of the voltage signal in the frequency domain, especially the relationship with the harmonic components of the load current. This feature vector will be used as input for the subsequent harmonic current prediction model to help improve the accuracy of the prediction and the responsiveness of the system.

[0038] Specifically, the specific process of extracting local features of harmonic current components through convolutional neural networks is as follows: input the time series data of harmonic current components into the convolutional neural network, and extract features of the input harmonic current components through the convolution layer; capture the local spatial features of the harmonic current components by sliding the convolution kernel in the local area, and reduce the data dimension and computational complexity through the pooling layer; gradually extract local features of the harmonic current components at different levels through the convolution layer and the pooling layer, and represent the temporal feature learning.

[0039] In this embodiment, the time series data of the harmonic current component is input into the convolutional neural network. The time series data is usually composed of multiple continuous current values, which change over time. In this process, the harmonic current value at each time point is used as an input feature for the neural network to learn and analyze. The core of the convolutional neural network is the convolution layer. The convolution layer extracts local features in the data by performing convolution operations on the input data using multiple convolution kernels (also called filters). Specifically, the convolution kernel slides on the input data (i.e., the time series of the harmonic current component) and calculates the weighted sum in each local area. In this way, the convolution layer can capture the local spatial features of the input data, such as the changing pattern of the current waveform or the behavior of certain specific frequency bands. The output of the convolution layer is processed by the pooling layer, and the role of the pooling layer is to reduce the dimensionality of the features extracted by the convolution layer to reduce the computational complexity and avoid overfitting. Pooling operations usually have two forms: maximum pooling and average pooling. Maximum pooling selects the maximum value in the local area, while average pooling selects the average value in the area. Through the pooling layer, the dimension of the data can be effectively reduced while retaining the main features of the input data and maintaining computational efficiency. In a convolutional neural network, a structure is usually composed of multiple convolutional layers and pooling layers stacked together. Each convolutional layer and pooling layer extracts features at different levels from the data. In the first convolutional layer, the network usually extracts relatively simple local features (such as edges, waveform changes, etc.); in deeper convolutional layers, the network can capture more complex features (such as periodic changes or complex patterns of harmonic components). Through this layer-by-layer feature extraction, the convolutional neural network can extract more and more abstract and recognizable features from time series data. After being processed by multiple convolutional layers and pooling layers, the network will extract rich local features from the original harmonic current data. These features not only contain the regularity of the time series, but also the frequency domain characteristics of the signal. Through such feature learning, the convolutional neural network can identify the key patterns in the time series data, which can then be used for the prediction of harmonic current in subsequent steps. These extracted time series features can effectively reflect the periodic changes and abnormal behaviors of the current waveform.

[0040] Specifically, the specific process of capturing the timing characteristics of harmonic current components through a bidirectional long short-term memory network is as follows: the local characteristics of the harmonic current components extracted by the convolutional neural network are input into the bidirectional long short-term memory network; the bidirectional long short-term memory network captures the timing characteristics in the time series by modeling the timing data of the harmonic current components; the bidirectional long short-term memory network learns the forward and reverse timing information during the forward and backward propagation processes, respectively, and generates a forward hidden state vector and a reverse hidden state vector; based on the forward hidden state vector and the reverse hidden state vector, a hidden state matrix of the bidirectional timing information is generated.

[0041] In this embodiment, the local features of the harmonic current components are extracted by a convolutional neural network. These features extracted by the convolutional network are passed as input to the bidirectional long short-term memory network. The local features here include the local change pattern of the current component in time, which can help the bidirectional long short-term memory network better understand the time series characteristics of the harmonic current components. The bidirectional long short-term memory network (BiLSTM) is a neural network that processes sequence data, which combines the advantages of traditional LSTM (long short-term memory network) and bidirectional structure. LSTM can effectively handle long-term dependencies in time series, while the bidirectional structure further enhances the ability to capture time series information by forward and reverse modeling of sequence data. In BiLSTM, each input of the time series passes through two LSTM layers: a forward LSTM layer (from front to back) and a reverse LSTM layer (from back to front). In this way, BiLSTM can capture past and future time series information at the same time, thereby improving the ability to learn time series features. During the training process, the BiLSTM network learns the information in the time series data through forward propagation and back propagation: Forward propagation (forward propagation): The forward LSTM layer processes the input time series, starting from the first time step of the sequence and propagating forward in sequence. The output of each time step depends on the hidden state of the current input and the previous time step, so it can capture the forward time series characteristics of the time series. Backward propagation: The reverse LSTM layer propagates forward from the last time step of the sequence to reversely model the information of the time series. This back propagation can capture the dependency between the current time step and the future time step. Through the two forward and reverse propagation methods, BiLSTM can learn information in time series data from two directions at the same time. The output of the bidirectional long short-term memory network is a combination of the forward hidden state and the reverse hidden state to form a bidirectional hidden state matrix. This matrix will contain both past and future time series information, thereby providing richer context for further prediction or classification tasks. Hidden state matrix (H t ): The bidirectional hidden state matrix concatenates or weighted combines the forward and reverse hidden state vectors. Assuming that the dimensions of the forward and reverse hidden states are d, the dimension of the final hidden state matrix is ​​2d. in, is the hidden state of the forward propagation, is the hidden state of back propagation, and finally generates a hidden state matrix H containing forward and reverse timing information t . Formula and parameter explanation: Hidden state h t : represents the hidden state at time step t, which is the memory of the previous and next time sequence information. The hidden states obtained by forward and backward propagation are and Weight matrix W f ,W b : In LSTM, the weight matrix controls how the input information affects the update of the hidden state. During the forward propagation process, W f controls the update of the forward propagation, and W b Controls the back-propagation updates.

[0042] Specifically, the specific process of generating harmonic current prediction values ​​is as follows: input the preprocessed harmonic current components and grid-connected point voltage characteristics into the wave current prediction model, perform time series modeling on the input harmonic current components and grid-connected point voltage characteristics, and integrate the dynamic change characteristics of current and voltage to generate harmonic current prediction values.

[0043] In this implementation scheme, the preprocessed data is input into the wave current prediction model for time series modeling. By simultaneously inputting the harmonic current components and the grid-connected point voltage characteristics into the model, the model can simultaneously learn the dynamic change characteristics of the current and voltage over time. This fusion process helps the model capture the interaction and common change pattern between current and voltage, which is crucial for accurately predicting changes in harmonic currents. Based on time series modeling and feature fusion, the model passes the final time series features to the output layer through forward propagation to generate a predicted value. This predicted value is the harmonic current value at a certain moment in the future. Output layer: The output layer of the model is usually a regression layer, which outputs a continuous numerical value representing the predicted harmonic current value. The predicted value reflects the possible changing trend of the harmonic current at future moments.

[0044] In summary, this application has at least the following effects:

[0045] The photovoltaic grid-connected harmonic current prediction method based on deep learning can effectively capture the time series characteristics and dynamic changes of harmonic current components and grid-connected point voltage by combining convolutional neural networks and bidirectional long short-term memory networks, significantly improve the prediction accuracy of harmonic current, and reduce prediction errors. By using their dynamic change characteristics for joint modeling, the model can comprehensively consider the relationship between current and voltage, thereby improving the system's sensitivity and prediction ability to harmonic current changes. Time series modeling through deep learning methods can realize automated harmonic current prediction, reflect the operating status of the power grid in real time, provide data support for dynamic monitoring and optimization of the power system, and improve real-time response capabilities. The bidirectional long short-term memory network captures the long-term and short-term dynamic changes of current and voltage, has strong adaptability, and can accurately predict harmonic current under different load conditions and power grid environments. The convolutional neural network reduces the data dimension in the feature extraction stage, thereby reducing the computational complexity of subsequent time series modeling, making the model training and prediction process more efficient.

[0046] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0047] The present invention is described with reference to flowcharts and / or block diagrams of systems, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0048] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0049] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0050] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0051] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A photovoltaic grid-connected harmonic current prediction method based on deep learning, characterized in that: The following steps are involved: S1. Obtain the three-phase load current and grid connection point voltage data in the photovoltaic grid-connected system; S2. Convert the three-phase load current into a two-phase DC component through Clark transformation and Park transformation, extract the fundamental current of the two-phase DC component through a low-pass filter, calculate the difference between the total current of the three-phase load current and the fundamental current, obtain the harmonic current component, and pre-process the harmonic current component; S3. Extract the time domain and frequency domain characteristics of the grid connection point voltage data through fast Fourier transform, analyze the amplitude and frequency distribution of the voltage harmonic components, evaluate the correlation between the voltage harmonics and the load current harmonics, and obtain the grid connection point voltage characteristics; S4. Extract local features of harmonic current components through convolutional neural network, capture the time series features of harmonic current components through bidirectional long short-term memory network, and build a harmonic current prediction model; S5. Input the preprocessed harmonic current components and grid-connected point voltage characteristics into the harmonic current prediction model to generate harmonic current prediction values.

2. The photovoltaic grid-connected harmonic current prediction method based on deep learning according to claim 1 is characterized in that: The specific process of converting the three-phase load current into a two-phase DC component through Clark transformation and Park transformation and extracting the fundamental current of the two-phase DC component through a low-pass filter is as follows: The three-phase load current is converted into a two-phase DC component through Clark transformation, and the three-phase AC current is projected into a stationary coordinate system; The two-phase DC components are converted into DC components in the rotating coordinate system by Park transformation to obtain the DC components of the two-phase AC currents; The fundamental current component in the DC component of the two-phase AC current is extracted through a low-pass filter to remove the high-frequency harmonic components.

3. The photovoltaic grid-connected harmonic current prediction method based on deep learning according to claim 2 is characterized in that: The specific process of obtaining the harmonic current components by subtracting the total current of the three-phase load current from the fundamental current is as follows: According to the three-phase load current, calculate the total current of the three-phase load current; The fundamental component is extracted from the DC component of the two-phase AC current, and the total current of the calculated three-phase load current is differenced with the extracted fundamental current to obtain the harmonic current component.

4. The photovoltaic grid-connected harmonic current prediction method based on deep learning according to claim 3 is characterized in that: The specific process of preprocessing the harmonic current components is as follows: De-noising the harmonic current components, using filters to remove high-frequency noise components; The denoised harmonic current components are standardized and smoothed to remove sudden changes and abnormal fluctuations; According to the time series characteristics of harmonic current components, data interpolation and filling are performed to process missing values ​​and ensure data integrity; The window sliding method is used to extract features within a local time period and generate the time series data format required for training.

5. The photovoltaic grid-connected harmonic current prediction method based on deep learning according to claim 4 is characterized in that: The specific process of extracting the time domain and frequency domain characteristics of the grid connection point voltage data through fast Fourier transform and analyzing the amplitude and frequency distribution of the voltage harmonic components is as follows: Perform fast Fourier transform on the grid connection point voltage data, convert the time domain data into frequency domain data, extract the frequency components in the grid connection point voltage signal, and identify the main frequency peak of the voltage signal; Calculate the amplitude of each frequency component and evaluate the contribution of different frequencies; Analyze the harmonic components of the voltage signal, including the amplitude and frequency distribution of the fundamental wave and each order harmonic, and obtain the harmonic characteristics of the voltage signal; According to the frequency domain analysis results, the energy distribution of each frequency band in the voltage signal is determined.

6. The photovoltaic grid-connected harmonic current prediction method based on deep learning according to claim 5 is characterized in that: The specific process of evaluating the correlation between voltage harmonics and load current harmonics and obtaining the voltage characteristics of the grid connection point is as follows: By analyzing the harmonic components of the voltage signal and the load current, the similarity of the voltage signal and the load current in the frequency domain is compared, and the relationship between the two in different frequency bands is evaluated; The correlation coefficient is used to quantify the correlation between voltage harmonics and load current harmonics. Based on the correlation analysis results, the correlation characteristics between the voltage signal and the load current signal are extracted to obtain the characteristic vector of the grid connection point voltage.

7. The photovoltaic grid-connected harmonic current prediction method based on deep learning according to claim 6 is characterized in that: The specific process of extracting local features of harmonic current components through convolutional neural networks is as follows: The time series data of harmonic current components are input into the convolutional neural network, and the features of the input harmonic current components are extracted through the convolutional layer; The convolution kernel slides in the local area to capture the local spatial characteristics of the harmonic current component, and the pooling layer reduces the data dimension and the computational complexity. The local features of different levels of harmonic current components are gradually extracted through convolutional layers and pooling layers to represent the temporal feature learning.

8. The photovoltaic grid-connected harmonic current prediction method based on deep learning according to claim 7 is characterized in that: The specific process of capturing the timing characteristics of harmonic current components through a bidirectional long short-term memory network is as follows: The local features of harmonic current components extracted by the convolutional neural network are input into the bidirectional long short-term memory network; The bidirectional long short-term memory network captures the temporal characteristics in the time series by modeling the time series data of the harmonic current components; During the forward and backward propagation processes, the bidirectional long short-term memory network learns the forward and reverse time sequence information, respectively, and generates the forward hidden state vector and the reverse hidden state vector; According to the forward hidden state vector and the reverse hidden state vector, a hidden state matrix of bidirectional time series information is generated.

9. The photovoltaic grid-connected harmonic current prediction method based on deep learning according to claim 8, characterized in that: The specific process of generating harmonic current prediction values ​​is as follows: The preprocessed harmonic current components and grid-connected point voltage characteristics are input into the wave current prediction model, and the input harmonic current components and grid-connected point voltage characteristics are modeled in time series. The dynamic change characteristics of current and voltage are integrated to generate the predicted value of harmonic current.

Images