Method for predicting and inhibiting harmonic waves generated by photovoltaic grid connection

Abstract

The invention discloses a method for predicting and suppressing harmonic waves generated by photovoltaic grid connection, which relates to the technical field of harmonic wave prediction, and comprises the following steps: detecting a three-phase current signal of a photovoltaic grid connection circuit by using an ip-iq detection method, and decomposing the three-phase current into an active component and a reactive component through Clarke conversion and Park conversion; according to the method, high-precision detection and separation of harmonic waves in the photovoltaic grid-connected system are realized by combining an ip-iq method with Clarke and Park transformation, dynamic modeling and harmonic signal prediction are performed by using the LSTM model, the defect of poor adaptability of a traditional method to nonlinear loads is overcome, and the prediction error is as low as 0.1%. Through combination of LSTM prediction and APF dynamic compensation, the compensation current opposite to the harmonic current is generated in real time, the harmonic suppression efficiency and the power grid stability are significantly improved, the electric energy quality is optimized, and the dynamic change of a complex load is adapted.

Description

Technical Field

[0001] The present invention relates to the technical field of harmonic prediction, and in particular to a method for predicting and suppressing harmonics generated by photovoltaic grid connection. Background Art

[0002] With the large-scale construction of photovoltaic power stations and their access to the power grid, the intermittent, random and instability of photovoltaic power generation have significantly affected the power quality of the power grid, especially the harmonic problem. Harmonics mainly come from nonlinear components such as inverters in photovoltaic systems, which will lead to increased losses in power generation, transmission, substation and distribution, thereby reducing the efficiency of the entire power system. In addition, harmonics can also cause overheating, insulation aging and abnormal operation of power supply equipment, greatly reducing the utilization rate of equipment. In the context of increasingly stringent requirements for power quality of industrial production equipment, the harmonic problem caused by photovoltaic grid connection has become one of the main technical bottlenecks restricting the stability and reliability of the power grid.

[0003] Current harmonic prediction research focuses on the use of traditional mathematical models or experience-based algorithms. However, these methods are often difficult to handle time series harmonic data with strong periodicity and short-term interference fluctuation characteristics in photovoltaic grid-connected systems, and the real-time and accuracy of predictions are insufficient. In addition, existing harmonic suppression methods are mostly based on active power filters (APFs), but their harmonic detection and compensation capabilities are limited under complex load conditions, and it is difficult to dynamically adapt to random changes in grid loads. At the same time, with the further popularization of photovoltaic power generation, the scale of harmonic-related data continues to increase, and existing methods are difficult to solve the data explosion problem of power quality indicators when large-scale photovoltaic grid-connected systems are connected. These technical limitations significantly hinder the efficient operation of photovoltaic power generation systems connected to the grid and the improvement of power quality in the grid.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0005] The purpose of the present invention is to provide a method for predicting and suppressing harmonics generated by photovoltaic grid connection. The three-phase signal in the photovoltaic grid connection system is detected by combining the ip-iq method with the Clarke transformation and the Park transformation. The phase-locked loop (PLL) is used to ensure that the fundamental voltage is synchronized with the grid phase, and the fundamental and harmonics are accurately separated by a low-pass filter to improve the real-time and accuracy of the detection. The long short-term memory network (LSTM) is introduced to model and predict harmonic data. The LSTM captures the long-term dependence and short-term fluctuation characteristics in the harmonic signal through a gating mechanism. There is no need to preset the signal distribution, and high-precision prediction is achieved, with the error controlled within 0.1%. Finally, the LSTM prediction model is combined with the active power filter (APF), and the harmonics are dynamically offset by generating a compensation current in the opposite direction of the harmonic current in real time, which significantly improves the efficiency and flexibility of harmonic suppression, effectively improves the stability of the grid and the quality of power, so as to solve the problems in the above-mentioned background technology.

[0006] In order to achieve the above object, the present invention provides the following technical solution: a method for predicting and suppressing harmonics generated by photovoltaic grid connection, comprising the following steps:

[0007] The ip-iq detection method is used to detect the three-phase current signal of the photovoltaic grid-connected circuit, and the three-phase current is decomposed into active and reactive components through Clarke transformation and Park transformation, and the DC component and harmonic component are extracted respectively by using a low-pass filter;

[0008] Based on the harmonic detection results, a long short-term memory network model is constructed. By extracting features from the time series information of the harmonic data, the periodicity and short-term interference characteristics in the data are captured, and a deep learning model for harmonic prediction is generated.

[0009] The acquired harmonic data is input into the trained long short-term memory network model to predict the harmonic component values ​​in the future in real time to improve the accuracy and real-time performance of harmonic prediction;

[0010] Combined with the control strategy of the active power filter, a compensation current equal to and opposite to the load harmonic current is generated through a compensation current generator according to the predicted harmonic component value;

[0011] The compensation current is injected into the grid to offset the harmonic components, dynamically suppress the harmonics generated by photovoltaic grid connection, reduce the total harmonic distortion rate of the grid and improve the power quality.

[0012] Preferably, the IP-IQ harmonic detection method is implemented by the following steps:

[0013] The reference phase of the voltage signal of the photovoltaic grid-connected system is obtained through a phase-locked loop;

[0014] The three-phase current signal is converted into α and β components in a two-phase stationary coordinate system using Clarke transformation;

[0015] The α and β components are converted into active current component Id and reactive current component Iq in the rotating coordinate system by using Park transformation;

[0016] Id and Iq are processed by a low-pass filter to extract the DC component and AC component, and the fundamental signal and harmonic signal are separated by inverse transformation.

[0017] Preferably, the training method of the long short-term memory network model comprises the following steps:

[0018] The harmonic signal extracted by the ip-iq harmonic detection method is input into the LSTM network;

[0019] The input gate is used to perform weighted update on the harmonic signal input at the current moment. The input gate state update formula is:

[0020] i t =σ(W i ·[h t-1 , x t ]+b i ),

[0021] Among them, i t Indicates the input gate state, W i is the input gate weight matrix, h t-1 is the output of the previous moment, x t is the current input signal, b i Bias for input gate;

[0022] The forget gate is used to update the cell state at the previous moment. The forget gate state update formula is:

[0023] f t =σ(W f ·[h t-1 , x t ]+b f ),

[0024] Among them, f t Represents the state of the forget gate, W f is the forget gate weight matrix, b f Bias for the forget gate;

[0025] The output gate is used to update the current cell state. The output gate state update formula is:

[0026] o t =σ(W o ·[h t-1 , x t ]+bo ),

[0027] Among them, t Indicates the output gate state, W o is the output gate weight matrix, b o Bias for the output gate;

[0028] The cell state and hidden state are finally updated through the gating mechanism, and the feature vector is extracted to build a model for harmonic prediction.

[0029] Preferably, the LSTM model updates the cell state through the Hadamard product, and the update formula is:

[0030] ,

[0031] Among them, C t is the current cell state, C t-1 is the cell state at the last moment, ⊙ is the Hadamard product operation, is the candidate cell state, and the current output is calculated by the formula. The calculation expression is:

[0032] h t =o t ⊙tanh(C t ),

[0033] Among them, h t Output for the current moment.

[0034] Preferably, the active power filter comprises the following modules:

[0035] Harmonic detection module: extracts the harmonic components in the load current in real time based on the ip-iq detection method;

[0036] Compensation current generator: generates a compensation current equal to and opposite to the load harmonic current according to the predicted harmonic components;

[0037] Compensation current injection module: injects compensation current into the power grid to achieve real-time suppression of harmonics.

[0038] Preferably, the control strategy of the active power filter is implemented based on the following formula:

[0039] I x =-I h

[0040] , where I x is the compensation current, I h For harmonic current, the compensation current is injected in reverse to offset the harmonic current, dynamically filter out the harmonics, and optimize the power quality of the power grid.

[0041] Preferably, the simulation experiment of prediction and compensation adopts the following parameters:

[0042] The power supply voltage is 220V and the frequency is 50Hz;

[0043] The A-phase load current is a sinusoidal AC signal, including fundamental current and odd-order harmonic currents.

[0044] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0045] The present invention uses the ip-iq method combined with Clarke transformation and Park transformation to detect and separate harmonics of three-phase signals in photovoltaic grid-connected systems, effectively solving the problems of poor real-time performance and insufficient precision in traditional detection methods. The reference phase of the fundamental voltage is obtained through a phase-locked loop (PLL) to ensure that the system is synchronized with the grid phase during the transformation process, thereby improving the stability and accuracy of the detection. The three-phase signal is converted into components in a two-phase stationary coordinate system using Clarke transformation, and then the components are mapped to a rotating coordinate system through Park transformation, which can accurately extract active and reactive components, and separate the fundamental and harmonic components through a low-pass filter.

[0046] The present invention innovatively introduces the long short-term memory network (LSTM) to model and predict harmonic data, overcoming the shortcomings of traditional prediction methods that are poorly adaptable to nonlinear loads and complex harmonic signals. Through the gating mechanism, the LSTM model can capture the long-term dependencies and short-term fluctuation characteristics in the harmonic data, thereby generating high-precision prediction results. Compared with traditional mathematical models, LSTM does not need to pre-assume the signal distribution, can dynamically learn the law of harmonic changes based on historical data, and has stronger generalization capabilities. In the simulation experiment, the harmonic prediction error of the LSTM model was controlled within 0.1%, and the prediction time step was extended to hundreds of milliseconds. It can accurately provide the future trend of harmonic component changes, providing an accurate and reliable reference for dynamic compensation.

[0047] The present invention realizes the dynamic suppression of harmonics and real-time stable optimization of the power grid by combining the LSTM prediction model with the active power filter (APF). The APF uses the harmonic reference value generated by the prediction model to generate a compensation current equal to the load harmonic current but opposite in direction in real time through the compensation current generator, and injects it into the power grid to offset the actual harmonic current. Unlike the traditional fixed compensation strategy, the dynamic compensation function of the present invention can adapt to load changes in real time, improving the efficiency and flexibility of harmonic suppression. Experimental data show that after APF compensation, the total harmonic distortion (THD) of the system is reduced from 5.99% to 1.2%, which significantly improves the power quality of the photovoltaic grid-connected system, reduces the loss of harmonics to power grid equipment, and improves the stability of system operation and equipment life. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0049] Figure 1 The present invention is a flow chart of a method for predicting and suppressing harmonics generated by photovoltaic grid connection.

[0050] Figure 2 This is the structure diagram of IP-IQ harmonic detection.

[0051] Figure 3 This is a diagram of the structure of long-term and short-term cellular memory.

[0052] Figure 4 This is the schematic diagram of the parallel active power filter.

[0053] Figure 5 This is the A phase load current diagram.

[0054] Figure 6 This is the FFT analysis diagram of phase A current.

[0055] Figure 7 This is the FFT analysis diagram after phase A elimination. DETAILED DESCRIPTION

[0056] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0057] The present invention provides Figure 1-7 The method for predicting and suppressing harmonics generated by photovoltaic grid connection shown includes the following steps:

[0058] The ip-iq detection method is used to detect the three-phase current signal of the photovoltaic grid-connected circuit, and the three-phase current is decomposed into active and reactive components through Clarke transformation and Park transformation, and the DC component and harmonic component are extracted respectively by using a low-pass filter;

[0059] First, the harmonic data in the photovoltaic grid-connected system is accurately separated by the IP-IQ detection method. The three-phase current signal of the photovoltaic system is converted into a two-phase stationary coordinate system signal through Clarke transformation, and then the active and reactive components are obtained through Park transformation. Subsequently, a low-pass filter is used to extract the DC component, from which the fundamental signal and the harmonic signal are separated. This method can make full use of the frequency domain and time domain characteristics of the power grid signal to achieve accurate detection of the harmonic components and provide high-quality basic data input for subsequent harmonic prediction and suppression. Compared with the traditional Fourier transform method, the IP-IQ detection method has the advantages of strong real-time performance and high precision.

[0060] Based on the harmonic detection results, a long short-term memory network (LSTM) model is constructed to extract features from the time series information of the harmonic data, capture the periodicity and short-term interference characteristics in the data, and generate a deep learning model for harmonic prediction.

[0061] The long short-term memory network (LSTM) is applied to the extraction and learning of harmonic features. The LSTM model effectively captures the long-term dependencies and short-term interference characteristics of time series data through a gating mechanism, and is particularly suitable for the complex data characteristics of photovoltaic grid-connected harmonics with strong periodicity and random fluctuations. In this step, the extracted harmonic data is converted into an input feature set suitable for deep learning processing, and the training process is used to optimize the model's parameter settings, thereby generating a high-precision harmonic prediction model.

[0062] The ip-iq harmonic detection method is implemented through the following steps:

[0063] The reference phase of the voltage signal of the photovoltaic grid-connected system is obtained through a phase-locked loop (PLL);

[0064] Clarke transformation is used to transform the three-phase current signal (I a ,I b ,I c ) is converted into α and β components in the two-phase stationary coordinate system;

[0065] The α component reflects the component of the three-phase signal in the horizontal axis direction of the stationary coordinate system, which is directly equal to the A-phase current signal.

[0066] The β component reflects the component of the three-phase signal in the vertical axis direction of the stationary coordinate system and is a combination of the three-phase currents.

[0067] By decomposing the α and β components, the analysis of the three-phase current signal is simplified to the processing of two-phase signals, making the subsequent harmonic detection and separation more convenient.

[0068] The α and β components are further converted into active current component Id and reactive current component Iq in the rotating coordinate system by using Park transformation;

[0069] The active current component Id and the reactive current component Iq are processed by a low-pass filter to extract the DC component (fundamental wave) and the AC component (harmonic), and the fundamental signal and the harmonic signal are separated by inverse transformation.

[0070] The training method of the LSTM model includes the following steps:

[0071] The harmonic signal extracted by the ip-iq harmonic detection method is input into the LSTM network;

[0072] The input gate is used to perform weighted update on the harmonic signal input at the current moment. The input gate state update formula is:

[0073] i t =σ(W i ·[h t-1 , x t ]+b i ),

[0074] Among them, i t Indicates the input gate state, W i is the input gate weight matrix, h t-1 is the output of the previous moment, x t is the current input signal, b i Bias for the input gate.

[0075] The forget gate is used to update the cell state at the previous moment. The forget gate state update formula is:

[0076] f t =σ(W f ·[h t-1 , x t ]+b f ),

[0077] Among them, f t Represents the state of the forget gate, W f is the forget gate weight matrix, b f Bias for the forget gate.

[0078] The output gate is used to update the current cell state. The output gate state update formula is:

[0079] o t =σ(W o ·[h t-1 , x t ]+b o ),

[0080] Among them, t Indicates the output gate state, W o is the output gate weight matrix, b oBias for the output gate.

[0081] The cell state and hidden state are finally updated through the gating mechanism, and the feature vector is extracted to build a model for harmonic prediction.

[0082] The LSTM model updates the cell state through the Hadamard product. The update formula is:

[0083] ,

[0084] Among them, C t is the current cell state, C t-1 is the cell state at the last moment, ⊙ is the Hadamard product operation, is the candidate cell state, and the current output is calculated by the formula. The calculation expression is:

[0085] h t =o t ⊙tanh(C t ),

[0086] Among them, h t The output is the current moment. This process improves the model's ability to learn the features of complex harmonic sequences by capturing long-term dependencies in harmonic data.

[0087] The acquired harmonic data is input into the trained LSTM model to predict the harmonic component values ​​in the future in real time to improve the accuracy and real-time performance of harmonic prediction;

[0088] Based on feature extraction, the real-time detected harmonic data is input into the trained LSTM model to predict the harmonic value in the future. Through deep learning methods, the model can adapt to the harmonic change trend under different grid conditions, improving the real-time and accuracy of the prediction. The significant feature of this step is that its prediction results can dynamically reflect the load changes of the photovoltaic system, providing a reliable basis for subsequent harmonic compensation, especially suitable for large-scale photovoltaic grid-connected environments.

[0089] Combined with the control strategy of active power filter (APF), a compensation current equal to and opposite to the load harmonic current is generated through a compensation current generator according to the predicted harmonic component value;

[0090] By embedding the prediction results into the control strategy of the active power filter (APF), a compensation current equal to the harmonic current but opposite in direction is generated in real time. Based on the predicted harmonic component values, the APF generates a reverse compensation current through its internal compensation current generator and injects it into the power grid to offset the load harmonic current. Compared with the traditional APF based on a fixed compensation strategy, the dynamic adjustment of the compensation current is achieved by using the harmonic prediction results, which effectively improves the accuracy and adaptability of the compensation.

[0091] The active power filter (APF) includes the following modules:

[0092] Harmonic detection module: extracts the harmonic components in the load current in real time based on the ip-iq detection method;

[0093] Compensation current generator: generates compensation current equal to and opposite to the load harmonic current according to the predicted harmonic components;

[0094] Compensation current injection module: injects compensation current into the power grid to achieve real-time suppression of harmonics.

[0095] Through the coordinated work of the above modules, the total harmonic distortion (THD) of the power grid is effectively reduced.

[0096] The control strategy of the active power filter (APF) is implemented based on the following formula:

[0097] I x =-I h ,

[0098] Among them, I x is the compensation current, I h The compensation current is injected in reverse to offset the harmonic current, thereby dynamically filtering out the harmonics and optimizing the power quality of the power grid.

[0099] Injecting compensation current into the grid, dynamically suppressing the harmonics generated by photovoltaic grid connection by offsetting the harmonic components, reducing the total harmonic distortion (THD) of the grid and improving the power quality;

[0100] After the compensation current is injected into the grid, it is offset by the actual harmonic current to achieve the effect of dynamically suppressing harmonics. It can not only significantly reduce the total harmonic distortion (THD) of the photovoltaic grid-connected system, thereby improving the power quality, but also effectively reduce the loss and interference of harmonics on grid equipment, and improve the operating efficiency and equipment life of the photovoltaic grid-connected system. In actual applications, through experimental verification, the total harmonic distortion rate can be reduced from the original 5.99% to 1.2%, significantly improving the stability and reliability of the grid.

[0101] The following parameters are used in the simulation experiment of prediction and compensation:

[0102] The power supply voltage is 220V and the frequency is 50Hz;

[0103] The A-phase load current is a sinusoidal AC signal, including fundamental current and odd-order harmonic currents;

[0104] By analyzing the harmonic components and distribution through Fourier transform, the total current distortion (THD) was reduced from 5.99% to 1.2% during the experiment.

[0105] Specific implementation 1: Harmonic detection based on ip-iq method;

[0106] Through the IP-IQ harmonic detection method, the three-phase current signal in the photovoltaic grid-connected system is detected and separated in real time, providing accurate data support for subsequent harmonic prediction and compensation. Specifically, the IP-IQ method mainly consists of the following steps:

[0107] First, the voltage reference phase of the photovoltaic grid-connected system is obtained through a phase-locked loop (PLL) to ensure the phase synchronization between the grid voltage and the photovoltaic system load current. The phase-locked loop accurately captures the phase information of the fundamental signal by detecting the zero point of the grid voltage, thereby providing a reliable reference signal for subsequent harmonic detection.

[0108] Then, the three-phase current signal (I a ,I b ,I c ) is converted into α and β components in a two-phase stationary coordinate system through Clarke transformation. Clarke transformation is a linear transformation used to convert a three-phase signal into a two-phase orthogonal signal in order to simplify the mathematical representation of a three-phase circuit. Specifically, the α component is directly equal to the A-phase current signal, that is, I α =I α , I α is the three-phase current signal I a ,I b ,I c The α component obtained by Clarke transformation has the physical meaning of the projection of the three-phase current signal on the horizontal axis (α axis) in the stationary coordinate system; the β component combines the current signals of phase A and phase B and is expressed as It reflects the projection of the signal in the vertical axis direction. This transformation not only simplifies the processing of three-phase signals, but also provides input data for the subsequent Park transformation.

[0109] Next, the Park transform is used to map the α and β components in the stationary coordinate system to the rotating coordinate system to obtain the active current component (Id) and the reactive current component (Iq). The Id component represents the DC component of the current, which is mainly related to the fundamental wave; the Iq component represents the AC component of the current, which is mainly related to the harmonics. The Park transform is implemented through a rotation matrix based on the fundamental phase of the power grid, projecting the signal components of the stationary coordinate system into the rotating coordinate system, so that the fundamental wave and harmonics can be separated more easily in frequency.

[0110] Subsequently, the Id and Iq components are processed by a low-pass filter to extract the components of the fundamental and harmonic signals. The low-pass filter can effectively filter out the high-frequency components, retaining only the DC signal of the fundamental part, while separating the AC components of the harmonics. Through inverse transformation, the filtered Id and Iq components are remapped back to the stationary coordinate system and the three-phase coordinate system, thereby completing the separation of the fundamental and harmonics.

[0111] Finally, the harmonic signal obtained by this method has high precision and high real-time performance, which lays the foundation for the subsequent harmonic prediction based on deep learning model and harmonic compensation of active filter. The experimental results show that this method can accurately extract the harmonic components of photovoltaic grid-connected system, and the harmonic detection accuracy reaches more than 99%, which greatly improves the reliability of harmonic data analysis and significantly reduces the computational complexity and data processing delay.

[0112] Specific implementation 2: Harmonic prediction based on LSTM model;

[0113] The long short-term memory network (LSTM) is applied to the harmonic prediction of photovoltaic grid-connected systems. The LSTM model is a deep learning method optimized for time series data, which can effectively capture the long-term dependencies and short-term fluctuation characteristics in the data. Harmonic data, as a signal with significant time correlation, is very suitable for modeling and prediction using the LSTM model.

[0114] First, the input data for harmonic prediction comes from the harmonic signal extracted by the IP-IQ detection method. Through data preprocessing, the harmonic signal is segmented by time series and normalized into a format suitable for deep learning model training. The LSTM model structure consists of input layers, hidden layers, and output layers of multiple time steps. The core of the hidden layer is a combination of input gates, forget gates, and output gates. The flow and storage of input information are dynamically adjusted through the gating mechanism.

[0115] Specifically, the input gate is based on the current input signal (x t ) and the previous hidden state (h t-1 ) determines the amount of information input, and its update formula is: t =σ(W i ·[h t-1 , x t ]+b i ); The forget gate controls which historical information needs to be retained, and its formula is: t =σ(W f ·[h t-1 , x t ]+b f ); the output gate is through the formula: t =σ(W o ·[ht-1 , x t ]+b o ) determines the output state at the current moment. At the same time, the cell state is updated by the following formula: Here, ⊙ represents the Hadamard product, which is used to perform element-wise operations.

[0116] After multiple rounds of training, the LSTM model can accurately fit the changing pattern of harmonic signals and predict future harmonic values ​​based on historical harmonic data. Compared with traditional linear prediction methods, the LSTM model has higher prediction accuracy and stronger generalization ability. Experiments show that in practical applications, the model can control the prediction error of the total harmonic distortion (THD) within 0.1%, providing a reliable reference for subsequent compensation strategies.

[0117] Specific implementation 3: Dynamic compensation of harmonics based on APF;

[0118] The present invention further combines an active power filter (APF) to realize a harmonic dynamic compensation function based on a prediction model. The core function of the APF is to generate a compensation current that is equal to and opposite to the load harmonic current in real time to dynamically offset the harmonic component, thereby optimizing the power quality.

[0119] First, the harmonic detection module of APF uses the ip-iq method to detect the load harmonic current in real time, and predicts the harmonic value in the future through the LSTM model. The prediction result is received by the compensation current generator and used to dynamically generate the compensation current that matches the predicted harmonic current. Specifically, the compensation current generation formula is: x =-I h , where I x is the compensation current, I h is the load harmonic current.

[0120] The compensation current is fed back to the grid through the injection module of the APF, and after being superimposed with the actual harmonic current, it cancels each other out, eliminating the interference of the harmonic components. The simulation experiment shows that after the APF is connected to the photovoltaic grid-connected system, the system total harmonic distortion (THD) drops significantly from 5.99% to 1.2%. At the same time, the experiment also shows that the APF can adapt to the environment of complex loads and random fluctuations of the photovoltaic system, realize dynamic adjustment of the amplitude and phase of the compensation current, and ensure complete suppression of harmonics.

[0121] By combining LSTM prediction with APF compensation, the present invention achieves full chain optimization from harmonic detection, prediction to suppression, greatly improving the power quality and operation stability of photovoltaic grid-connected systems. Experiments also show that in scenarios with large load changes, the generation response time of compensation current is less than 5 milliseconds, with excellent real-time performance.

[0122] The present invention uses the ip-iq method combined with Clarke transformation and Park transformation to detect and separate harmonics of three-phase signals in photovoltaic grid-connected systems, effectively solving the problems of poor real-time performance and insufficient precision in traditional detection methods. The reference phase of the fundamental voltage is obtained through a phase-locked loop (PLL) to ensure that the system is synchronized with the grid phase during the transformation process, thereby improving the stability and accuracy of the detection. The three-phase signal is converted into components in a two-phase stationary coordinate system using Clarke transformation, and then the components are mapped to a rotating coordinate system through Park transformation, which can accurately extract active and reactive components, and separate the fundamental and harmonic components through a low-pass filter.

[0123] The present invention innovatively introduces the long short-term memory network (LSTM) to model and predict harmonic data, overcoming the shortcomings of traditional prediction methods that are poorly adaptable to nonlinear loads and complex harmonic signals. Through the gating mechanism, the LSTM model can capture the long-term dependencies and short-term fluctuation characteristics in the harmonic data, thereby generating high-precision prediction results. Compared with traditional mathematical models, LSTM does not need to pre-assume the signal distribution, can dynamically learn the law of harmonic changes based on historical data, and has stronger generalization capabilities. In the simulation experiment, the harmonic prediction error of the LSTM model was controlled within 0.1%, and the prediction time step was extended to hundreds of milliseconds. It can accurately provide the future trend of harmonic component changes, providing an accurate and reliable reference for dynamic compensation.

[0124] The present invention realizes the dynamic suppression of harmonics and real-time stable optimization of the power grid by combining the LSTM prediction model with the active power filter (APF). The APF uses the harmonic reference value generated by the prediction model to generate a compensation current equal to the load harmonic current but opposite in direction in real time through the compensation current generator, and injects it into the power grid to offset the actual harmonic current. Unlike the traditional fixed compensation strategy, the dynamic compensation function of the present invention can adapt to load changes in real time, improving the efficiency and flexibility of harmonic suppression. Experimental data show that after APF compensation, the total harmonic distortion (THD) of the system is reduced from 5.99% to 1.2%, which significantly improves the power quality of the photovoltaic grid-connected system, reduces the loss of harmonics to power grid equipment, and improves the stability of system operation and equipment life.

[0125] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0126] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0127] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0128] In the several embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0129] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0130] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0131] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage media include: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks or optical disks.

[0132] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for predicting and suppressing harmonics generated by photovoltaic grid connection, characterized in that: The following steps are involved: The ip-iq detection method is used to detect the three-phase current signal of the photovoltaic grid-connected circuit, and the three-phase current is decomposed into active and reactive components through Clarke transformation and Park transformation, and the DC component and harmonic component are extracted respectively by using a low-pass filter; Based on the harmonic detection results, a long short-term memory network model is constructed. By extracting features from the time series information of the harmonic data, the periodicity and short-term interference characteristics in the data are captured, and a deep learning model for harmonic prediction is generated. The acquired harmonic data is input into the trained long short-term memory network model to predict the harmonic component values ​​in the future in real time to improve the accuracy and real-time performance of harmonic prediction; Combined with the control strategy of the active power filter, a compensation current equal to and opposite to the load harmonic current is generated through a compensation current generator according to the predicted harmonic component value; The compensation current is injected into the grid to offset the harmonic components, dynamically suppress the harmonics generated by photovoltaic grid connection, reduce the total harmonic distortion rate of the grid and improve the power quality.

2. The method for predicting and suppressing harmonics generated by photovoltaic grid connection according to claim 1, characterized in that: The ip-iq harmonic detection method is implemented through the following steps: The reference phase of the voltage signal of the photovoltaic grid-connected system is obtained through a phase-locked loop; The three-phase current signal is converted into α and β components in a two-phase stationary coordinate system using Clarke transformation; The α and β components are converted into active current component Id and reactive current component Iq in the rotating coordinate system by using Park transformation; Id and Iq are processed by a low-pass filter to extract the DC component and AC component, and the fundamental signal and harmonic signal are separated by inverse transformation.

3. The method for predicting and suppressing harmonics generated by photovoltaic grid connection according to claim 1, characterized in that: The training method of the long short-term memory network model includes the following steps: The harmonic signal extracted by the ip-iq harmonic detection method is input into the LSTM network; The input gate is used to perform weighted update on the harmonic signal input at the current moment. The input gate state update formula is: i t =σ(W i ·[h t-1 ,x t ]+b i ), Among them, i t Indicates the input gate state, W i is the input gate weight matrix, h t-1 is the output of the previous moment, x t is the current input signal, b i Bias for input gate; The forget gate is used to update the cell state at the previous moment. The forget gate state update formula is: f t =σ(W f ·[h t-1 ,x t ]+b f ), Among them, f t Represents the state of the forget gate, W f is the forget gate weight matrix, b f Bias for the forget gate; The output gate is used to update the current cell state. The output gate state update formula is: the t =σ(W o ·[h t-1 ,x t ]+b o ), Among them, t Indicates the output gate state, W o is the output gate weight matrix, b o Bias for the output gate; The cell state and hidden state are finally updated through the gating mechanism, and the feature vector is extracted to build a model for harmonic prediction.

4. The method for predicting and suppressing harmonics generated by photovoltaic grid connection according to claim 3, characterized in that: The LSTM model updates the cell state through the Hadamard product. The update formula is: ， Among them, C t is the current cell state, C t-1 is the cell state at the last moment, ⊙ is the Hadamard product operation, is the candidate cell state, and the current output is calculated by the formula. The calculation expression is: h t =o t ⊙tanh(C t ), Among them, h t Output for the current moment.

5. The method for predicting and suppressing harmonics generated by photovoltaic grid connection according to claim 1, characterized in that: The active power filter includes the following modules: Harmonic detection module: extracts the harmonic components in the load current in real time based on the ip-iq detection method; Compensation current generator: generates compensation current equal to and opposite to the load harmonic current according to the predicted harmonic components; Compensation current injection module: injects compensation current into the power grid to achieve real-time suppression of harmonics.

6. The method for predicting and suppressing harmonics generated by photovoltaic grid connection according to claim 1, characterized in that: The control strategy of the active power filter is based on the following formula: I x =-I h , Among them, I x is the compensation current, I h For harmonic current, the compensation current is injected in reverse to offset the harmonic current, dynamically filter out the harmonics, and optimize the power quality of the power grid.

7. The method for predicting and suppressing harmonics generated by photovoltaic grid connection according to claim 1, characterized in that: The following parameters are used in the simulation experiment of prediction and compensation: The power supply voltage is 220V and the frequency is 50Hz; The A-phase load current is a sinusoidal AC signal, including fundamental current and odd-order harmonic currents.

Images