Multi-dimensional time sequence photovoltaic grid-connected harmonic current prediction control method

Abstract

The application discloses a multi-dimensional time sequence photovoltaic grid-connected harmonic current prediction control method and particularly relates to the technical field of power system control; original multi-dimensional electrical signals of grid connection are collected and are decomposed into multi-dimensional frequency signals by using Fourier transform, the characteristic components of independent harmonic sources are separated by using blind source separation and non-negative matrix decomposition technology, and the frequency and amplitude characteristics thereof are extracted; subsequently, a multi-dimensional time sequence prediction model is constructed through a recurrent neural network, the dynamic behaviors of multiple harmonic sources are simultaneously predicted, high-frequency compensation and multi-scale analysis are combined, frequency domain data is reconstructed, and frequency aliasing errors are corrected; finally, the prediction result is input into a photovoltaic inverter control system, a compensation current opposite to a target harmonic is dynamically generated, the harmonic is effectively offset, the multi-source harmonic problem in a complex power grid environment can be accurately processed, and the harmonic suppression capability, power quality and stability of system operation of the photovoltaic grid-connected system are greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system control, in particular to a multi-dimensional time sequence photovoltaic grid-connected harmonic current prediction control method. BACKGROUND

[0002] The multi-dimensional time sequence photovoltaic grid-connected harmonic current prediction control is a kind of advanced control technology specially used for photovoltaic power generation system. Photovoltaic system converts direct current into alternating current through inverter and connects to power grid, but harmonic current may be generated in this process, which affects the power supply quality of power grid. The multi-dimensional time sequence prediction control technology aims to predict the future behavior of harmonic current in photovoltaic power generation system in advance and control it optimally. This method uses the dynamic correlation of multi-dimensional variables (such as current, voltage, power factor, etc.) and time sequence data, accurately predicts and adjusts harmonic characteristics through mathematical modeling and optimization algorithm, so as to realize the stable operation of photovoltaic grid-connected system.

[0003] Compared with traditional harmonic suppression technology, the multi-dimensional time sequence prediction control has significant advantages. First, it can predict the future harmonic change trend based on the current system state and historical data, so that the control decision is more forward-looking. Second, with the help of multi-dimensional data analysis, it can consider the influence of multiple factors on harmonic current and optimize the control strategy to improve the prediction accuracy. In addition, by updating the prediction model in real time, the multi-dimensional time sequence control can adapt to the dynamic changes of photovoltaic power generation environment, such as light intensity and temperature fluctuation.

[0004] The prior art has the following deficiencies:

[0005] There may be multiple independent harmonic sources (such as industrial loads, grid-connected energy storage devices, etc.) in the power grid, which may be superimposed together at different frequencies and amplitudes to form complex waveforms. If the model cannot distinguish the independent components of these harmonics, it may mistakenly consider the superposition effect as a single harmonic characteristic for prediction and control. And this superposition effect may cause frequency aliasing, so that the input data received by the prediction model seems to conform to a certain error pattern, thereby causing error control and leading to system oscillation or overload. SUMMARY

[0006] The purpose of the present application is to provide a multi-dimensional time sequence photovoltaic grid-connected harmonic current prediction control method to solve the deficiencies in the background art.

[0007] In order to achieve the above purpose, the present application provides the following technical solution: a multi-dimensional time sequence photovoltaic grid-connected harmonic current prediction control method, comprising the following steps:

[0008] S1: Collecting original multi-dimensional electrical signals of photovoltaic grid-connected point, the original multi-dimensional electrical signals including current, voltage and frequency, using Fourier transform to decompose the original multi-dimensional electrical signals into multi-dimensional frequency signals;

[0009] S2: Using blind source separation algorithm and non-negative matrix factorization technology to separate the characteristic components of independent harmonic sources from the multi-dimensional frequency signals, extracting the frequency and amplitude characteristics of each independent harmonic source, and reducing the prediction deviation caused by superposition effect;

[0010] S3: Based on the separated independent harmonic source characteristics, using recurrent neural network to construct a time series prediction model, and processing multiple harmonic sources simultaneously through the multi-dimensional prediction model to predict the future characteristics of the overall harmonic source;

[0011] S4: Combining the separated harmonic source data, performing high-frequency compensation and reconstruction on the frequency domain data, and further correcting the frequency aliasing error through multi-scale analysis to ensure the accuracy of the frequency separation result;

[0012] S5: Inputting the multi-dimensional prediction result into the control system of the photovoltaic inverter to dynamically adjust the output current of the inverter to offset the target harmonic component in the power grid.

[0013] Preferably, in S1, Fourier transform is used for frequency analysis of the multi-dimensional time domain electrical signals of the photovoltaic grid-connected point, and time domain discrete current signal, voltage signal and frequency signal data are collected. Discrete Fourier transform is applied to each signal to convert it from time domain to frequency domain, extract the corresponding frequency spectrum information including frequency component and amplitude characteristics, and combine the frequency spectrum data of current, voltage and frequency signals to form a multi-dimensional frequency matrix.

[0014] Preferably, in S2, the multi-dimensional frequency signal matrix is composed of linear superposition of frequency components of multiple independent harmonic sources. The signal matrix is processed by blind source separation to whiten and eliminate the correlation between signals, so that the components of each harmonic source are independent of each other.

[0015] In the blind source separation process, the separation matrix is solved by optimization algorithm, and the separation result is verified to meet the harmonic characteristics, and then the frequency components of each independent harmonic source are obtained. Non-negative matrix factorization is used to decompose the separation result, non-negative constraint is introduced to ensure that the separated frequency and amplitude are positive values. The weight matrix and basis component matrix of each independent harmonic source are obtained by NMF, and the main frequency characteristics and amplitude characteristics of the harmonic source are extracted. The main frequency and its amplitude of each independent harmonic source are calculated to form the frequency and amplitude characteristic vector of the independent harmonic source.

[0016] Preferably, in S3, the independent characteristic vector T r = [f r , A r], r = 1, 2, …, R; the time series characteristics of each independent harmonic source are expressed as: X r = {T r (t1), T r (t2), …, T r (t N )}; in the formula, T r (t i ) = [f r (t i ), A r (t i )] represents the frequency and amplitude characteristics at time t i , r = 1, 2, …, R represents the harmonic source number, and N is the time series length; the input data is normalized, and a prediction model is constructed using a recurrent neural network; the model input is a multi-dimensional characteristic time series, and the output is a predicted value at a future time;

[0017] The data ET r (t i ) input at each time step is the frequency and amplitude characteristics after normalization; a recurrent structure is used to capture the long and short term dependencies of the time series; the state update formula of the hidden layer is: h t = φ(W h h t-1 +W x x t +b h ); h t is the hidden layer state, x t is the input characteristic vector, W h , W x , and b h are hidden layer parameters, and φ(·) is an activation function; the output layer predicts the future frequency and amplitude characteristics, and the output is: With the time series {T r (t1), T r (t2), …, T r (t N )} as input, the RNN model predicts the characteristics at a future time t N+1 : For multiple harmonic sources r = 1, 2, …, R, multiple sub-models are trained simultaneously, and a loss function is used to evaluate the error between the predicted results and the true values The expression is: Where: T r (t i ) is the true value, is the predicted value; the time series of each harmonic source is divided into a sliding window to form a training sample set; the training set and the validation set are divided: 80% of the data is divided into a training set, and 20% is divided into a validation set.

[0018] In S3, the current time sequence {T r (t1),T r (t2),…,T r (t N )} is input into the model to predict the characteristics at the next time t N+1 : Based on the single-step prediction result, the predicted value T is taken as the next input to iteratively predict the future multiple time steps: The predicted harmonic source characteristic matrix is: M represents the number of predicted time steps; The frequency and amplitude characteristics of the harmonic source r at future time are included.

[0019] In S4, based on the obtained harmonic source frequency and amplitude data, a spectrum diagram is drawn, the high-frequency components are compensated by an exponential model using the amplitude decay with frequency characteristic, the compensated high-frequency harmonic amplitude is generated, the high-frequency harmonic signal is reconstructed using the inverse Fourier transform combined with the phase information; the wavelet transform is used for multi-scale decomposition, the signal is divided into low-frequency and high-frequency components, the abnormal frequency components can be identified and corrected through multi-scale decomposition, the low-frequency information and high-frequency components are retained, and the aliasing part is removed; the corrected high-frequency and low-frequency harmonic signals are merged to form a complete harmonic signal.

[0020] In S5, the multi-dimensional prediction result is input into the control system of the photovoltaic inverter to dynamically adjust the inverter output current to offset the target harmonic component in the power grid, specifically:

[0021] For the rth independent harmonic source, the harmonic component at future time t is represented as: i r (t) = A r sin(2πf r t + φ r ); A r is the predicted harmonic amplitude, f r is the predicted harmonic frequency, and φ r is the predicted harmonic phase; the target harmonic signal in the power grid is the superposition of independent harmonic components: The photovoltaic inverter needs to output a compensation current i comp (t) equal and opposite to the target harmonic to make the total harmonic current tend to zero: The inverter generates a discrete-time signal i T s : sampling period of the control system; n: index of sampling time, the control system calculates the reference current i ref [n] according to the compensation current formula: i ref [n] = i comp [n]; iref [n] is the ideal current that the inverter needs to output.

[0022] Preferably, pulse width modulation (PWM) technology is adopted, and the actual output current is tracked by adjusting the duty cycle of the inverter switching device. ref [n], the duty cycle of the PWM signal is calculated as follows: D[n] : PWM duty cycle; I max : the maximum output current of the inverter, which is adjusted in real time through closed-loop control; the actual output current i out [n] is measured in real time: e[n] = i ref [n] - i out [n] ; a correction signal is generated using a proportional-integral controller to adjust the output of the inverter: K p is the proportional gain; K i is the integral gain; u[n] is the control signal used to correct the PWM duty cycle.

[0023] The PWM duty cycle is corrected according to the control signal: D[n] = D[n-1] + u[n] ; the compensation current i total (t) output by the inverter is shown as follows: total (t) = i harm (t) + i out (t) ; i out (t) = i comp (t) = -i harm (t) is substituted, the total current becomes: i total (t) = i harm (t) - i harm (t) = 0 ; harmonic cancellation is achieved.

[0024] In the above technical solution, the technical effects and advantages provided by the present application are as follows:

[0025] 1. The present application performs spectral decomposition on the original multi-dimensional electrical signals (current, voltage, frequency) of the photovoltaic grid-connected point through Fourier transform, constructs a multi-dimensional frequency signal matrix, and realizes harmonic characteristic extraction. Combined with blind source separation and non-negative matrix factorization technology, the frequency and amplitude characteristics of independent harmonic sources are accurately separated, and the interference of superposition effect is eliminated; a recurrent neural network is used to construct a time series prediction model, which simultaneously processes the dynamic characteristics of multiple harmonic sources and predicts future harmonic behavior; through high-frequency compensation and multi-scale analysis technology of wavelet transform, the frequency aliasing error is corrected to ensure the accuracy of the separation and prediction results; finally, the prediction results are input into the photovoltaic inverter control system to dynamically generate a compensation current to cancel the target harmonic in the power grid.

[0026] 2、The application can accurately separate and predict the characteristics of multiple harmonic sources in a complex power grid environment, dynamically adjust the output current of the photovoltaic inverter, realize efficient suppression of harmonics and improvement of power quality of the power grid. Through multi-dimensional data processing and intelligent prediction technology, the method greatly reduces the influence of harmonic superposition and frequency aliasing on system stability, ensures the reliability, flexibility and robustness of the photovoltaic grid-connected system operation, and provides important technical support for realizing high-quality operation of green energy grid-connected. BRIEF DESCRIPTION OF DRAWINGS

[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0028] Figure 1 The method flowchart of the present application. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical scheme and advantages of the embodiments of the present application more clear, the technical scheme in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0030] Embodiment, please refer to Figure 1 The multi-dimensional time sequence photovoltaic grid-connected harmonic current prediction control method described in the present embodiment includes the following steps:

[0031] S1: Collecting the original multi-dimensional electrical signals of the photovoltaic grid-connected point, the original multi-dimensional electrical signals including current, voltage and frequency, and using Fourier transform to decompose the original multi-dimensional electrical signals into multi-dimensional frequency signals;

[0032] S2: Using blind source separation algorithm and non-negative matrix decomposition technology to separate the characteristic components of independent harmonic sources from the multi-dimensional frequency signals, extracting the frequency and amplitude characteristics of each independent harmonic source, and reducing the prediction deviation caused by superposition effect;

[0033] S3: Based on the separated independent harmonic source characteristics, using recursive neural network to construct a time sequence prediction model, and processing multiple harmonic sources simultaneously through the multi-dimensional prediction model to predict the future characteristics of the overall harmonic source;

[0034] S4: Combine the separated harmonic source data, perform high-frequency compensation and reconstruction on the frequency domain data, and further correct the frequency aliasing error through multi-scale analysis to ensure the accuracy of the frequency separation result;

[0035] S5: Input the multi-dimensional prediction result into the control system of the photovoltaic inverter, dynamically adjust the inverter output current, and offset the target harmonic component in the power grid.

[0036] In S1, the original multi-dimensional electrical signals of the photovoltaic grid-connected point are collected, including current, voltage and frequency, and the original multi-dimensional electrical signals are decomposed into multi-dimensional frequency signals by Fourier transform, specifically:

[0037] Install a current transformer (CT) or Hall current sensor at the grid-connected point to detect the grid-connected current waveform in real time, covering the basic frequency and main harmonic frequency band (e.g. 50Hz and its integer multiples). Install a voltage transformer (VT) or voltage divider at the grid-connected point to collect the grid-connected voltage signal and ensure that harmonic fluctuations are captured. Real-time detection of grid frequency fluctuations is achieved through a dedicated frequency measurement module (e.g. a frequency meter based on a phase-locked loop).

[0038] Convert the analog electrical signals to digital signals, with a sampling frequency set to more than twice the maximum frequency component of the target signal (Nyquist sampling theorem), recommended at 10kHz or higher. Use high-precision clocks or GPS synchronization technology to ensure consistency of current, voltage and frequency data on the time axis. Use high-speed data buses (such as SPI, CAN or Ethernet) to transmit digital signals to a central processing unit (CPU) or FPGA for real-time processing. Store the original data in real time for subsequent analysis and modeling.

[0039] The collected current and voltage signals are time-domain discrete data: current signal i[n] = i(n·T s ), n = 0, 1, 2, …, N-1; voltage signal v[n] = v(n·T s ), n = 0, 1, 2, …, N-1; where, f s is the sampling frequency; for time-domain signal x(t) (such as current i(t) or voltage v(t)), the expression of discrete Fourier transform (DFT) is: k = 0, 1, 2, …, N-1; in the formula, X[k] is the frequency spectrum value of the signal at the kth frequency component, and the frequency is x[n] is the time-domain discrete signal, is a complex exponential base function used for frequency analysis;

[0040] The discrete Fourier transform is applied to the current signal i[n] with the expression: k = 0, 1, 2,..., N-1; where I[k] is the spectral value of the frequency domain current signal at frequency index k;

[0041] The Fourier transform is applied to the voltage signal v[n] with the expression: k = 0, 1, 2,..., N-1; where V[k] is the spectral value of the frequency domain voltage signal;

[0042] The Fourier transform is applied to the frequency signal f[n] with the expression: k = 0, 1, 2,..., N-1; where F[k] is the spectral value of the frequency domain frequency signal;

[0043] The frequency domain results I[k], V[k], F[k] of the current, voltage, and frequency signals are combined to form a multi-dimensional frequency signal matrix S[k]: S[k] = [I[k], V[k], F[k]]; S[k] is the feature vector of the multi-dimensional frequency signal at frequency index k.

[0044] S2: Using blind source separation algorithms and non-negative matrix factorization techniques, the characteristic components of independent harmonic sources are separated from the multi-dimensional frequency signal, and the frequency and amplitude characteristics of each independent harmonic source are extracted, reducing the prediction bias caused by the superposition effect.

[0045] It is assumed that the multi-dimensional frequency signal matrix S[k] is composed of the frequency components H[k] of M independent harmonic sources through a linear combination of a mixing matrix A: S[k] = A · H[k]; where: S[k] ∈ R N×M : observed signal (frequency signal); H[k] ∈ R M ×M : frequency component of independent harmonic source; A ∈ R N×M : mixing matrix, representing the linear superposition weight of each harmonic source. The goal of blind source separation is to solve H[k] and A from the observed signal S[k], so that each harmonic source Hi[k] is statistically independent; the mixing matrix A remains sparse and non-negative.

[0046] S[k] is preprocessed to have a mean of 0 and a covariance matrix of the unit matrix, eliminating the correlation between data: W = whitening matrix; independence maximization: using a method that maximizes non-Gaussianity (such as the FastICA algorithm), the separation matrix W is found through iterative optimization to obtain statistically independent signal components: Separation result verification: verify whether the frequency, amplitude, etc. of each signal in H[k] conforms to the harmonic characteristics. Through blind source separation, the frequency signal component H[k] of each independent harmonic source is obtained, and preliminary separation is realized. In order to further improve the physical interpretability of the separation result, non-negative matrix factorization (NMF) is used to decompose H[k]: H[k] = W H ·H′; wherein: W H ∈R M×M : weight matrix, representing the frequency weight of the harmonic source; H′ ∈ R M×M : basis component matrix, containing the main frequency characteristics of each harmonic source; R is the number of basis frequency components, usually R < M.

[0047] Since both the frequency and the amplitude are non-negative quantities, NMF adds a non-negativity constraint to improve the physical meaning of the decomposition result: non-negativity constraint: W H ≥ 0, H′ ≥ 0; the optimization objective function is: where represents the Frobenius norm of the matrix. Through NMF, the frequency and amplitude characteristics of the independent harmonic sources are extracted, making the separation result more sparse and easy to interpret. For each independent harmonic source H′ r ∈ H′, its main frequency component f r is calculated: f r = argmax k |H′ r [k] |; where H′ r [k] represents the spectral value of the harmonic source r at frequency index k.

[0048] For each independent harmonic source H′ r , its amplitude characteristic is: A r = ||H r ′||1; through blind source separation and non-negative matrix factorization, the frequency and amplitude characteristic vectors of multiple independent harmonic sources are obtained: characteristic vector T r = [f r , A r ], r = 1, 2, …, R; these characteristic vectors can be input into a time series prediction model to provide high-precision data basis for subsequent prediction and control.

[0049] S3: Based on the separated independent harmonic source characteristics, a time series prediction model is constructed using a recurrent neural network, and multiple harmonic sources are simultaneously processed through a multi-dimensional prediction model to predict the future characteristics of the overall harmonic source.

[0050] From step S2, the independent characteristic vectors T r = [f r , A r ], r = 1, 2, …, R (frequency fr and amplitude A r The characteristics of each harmonic source are represented by a time series dataset. This dataset contains information about the dynamic changes of the harmonic sources over time, and recurrent neural networks (RNNs) are needed to model their temporal relationships to predict future frequency and amplitude characteristics. RNNs are ideal tools for time series data prediction because their recurrent structure in hidden layers preserves historical state information, making them suitable for capturing the dynamic characteristics of time series data.

[0051] The time-series characteristics of each independent harmonic source are represented as: X r ={T r (t1),T r (t2),…,T r (t N In the formula, T r (t i )=[f r (t i ),A r (t i )] represents time t i The frequency and amplitude characteristics are given by r = 1, 2, ..., R, representing the harmonic source number, and N is the time series length. The input data is standardized, and a prediction model is constructed using a recurrent neural network (RNN). The model input is a multidimensional characteristic time series, and the output is the predicted value for future times.

[0052] Input layer: Inputs the data ET for each time step. r (t i ), representing the frequency and amplitude characteristics after standardization;

[0053] Hidden layer: Employing a recursive structure, such as LSTM or GRU, to capture long-short-term dependencies in time series data. The state update formula for the hidden layer is: h t =φ(W h h t-1 +W x x t +b h );h t For the hidden layer state, x t W is the input feature vector. h W x ,b h φ(·) represents the hidden layer parameters and the activation function.

[0054] Output layer: Predicts future frequency and amplitude characteristics, outputting:

[0055] With time series {T r (t1),Tr (t2),…,T r (t N )} as input, the RNN model predicts the characteristics of the next time step t N+1 For multiple harmonic sources r = 1, 2, …, R, multiple sub-models are trained simultaneously, or the data of all harmonic sources is spliced into a multi-dimensional input matrix: X = [X1, X2, …, X R All harmonic sources are processed by sharing the RNN network, improving computational efficiency and modeling consistency.

[0056] The loss function is used to evaluate the error between the predicted results and the true values, and the mean square error (MSE) is commonly used: Where: T r (t i ) is the true value, is the predicted value. The stochastic gradient descent (SGD) or Adam optimization algorithm is used to adjust the model parameters W h ,W x ,b h to minimize the loss function.

[0057] The time series of each harmonic source is divided into sliding windows to form a training sample set. The training set and the validation set are divided: 80% of the data is divided into the training set, and 20% is divided into the validation set to prevent overfitting.

[0058] The current time series {T r (t1), T r (t2), …, T r (t N )} is input into the model to predict the characteristics of the next time step t N+1 Based on the single-step prediction results, the predicted value is used as the next input to iteratively predict multiple time steps in the future: The predicted harmonic source characteristic matrix is: M represents the number of predicted time steps; Contains the frequency and amplitude characteristics of the harmonic source r at the future time.

[0059] S4: Combine the separated harmonic source data, perform high-frequency compensation and reconstruction on the frequency domain data, and further correct the frequency aliasing error through multi-scale analysis to ensure the accuracy of the frequency separation result.

[0060] The obtained harmonic source frequency data f r [k] and amplitude A r ​​[k], plot the spectral graph: observe the attenuation characteristics of harmonic amplitudes with frequency, identify frequency bands that may be affected by high-frequency loss. High-frequency harmonic component amplitudes usually decay exponentially with frequency, estimate the neglected high-frequency components: where: is the high-frequency compensated harmonic amplitude; a is the attenuation factor, obtained by fitting the harmonic frequency characteristics data; f cutoff is the Nyquist frequency of the sampling frequency.

[0061] Compensation step: compensate the frequency components of f k > f cutoff , recalculate the high-frequency amplitudes using the formula; superimpose the compensated frequency components to form the complete high-frequency compensated harmonic spectrum.

[0062] Use the compensated spectrum and phase information φ r [k], reconstruct the high-frequency components into the time domain through inverse Fourier transform: is the high-frequency compensated harmonic source signal.

[0063] Use the harmonic separation result to focus on the analysis of frequency bands that may exist aliasing (such as f k close to f cutoff band). Detect the main features of aliasing: sudden enhancement of harmonic components or abnormal frequency drift; frequency intermodulation that does not conform to physical laws (such as non-integer multiple frequency).

[0064] Use wavelet transform for multi-scale decomposition to separate the low-frequency and high-frequency components of the signal: continuous wavelet transform: convert the harmonic source signal h r (t) to the time-scale domain: where: W(a,b) is the wavelet coefficient, representing the signal component corresponding to scale a and time b; ψ * is the mother wavelet function (such as Morlet wavelet or Haar wavelet); a is the scale factor, controlling the frequency resolution; b is the translation factor, controlling the time resolution. Through multi-scale decomposition, the signal is divided into different frequency bands, and the possible aliasing frequencies are identified and corrected: retain the components of low-scale (high-frequency) signal; use high-scale (low-frequency) components to correct the aliasing part. Merge the low-frequency and medium-frequency signals after multi-scale correction to obtain the complete harmonic source signal: Compare the reconstructed signal spectrum with the separation result to verify whether the corrected signal conforms to the actual harmonic characteristics.

[0065] ​S5: Input the multidimensional prediction results into the control system of the photovoltaic inverter to dynamically adjust the inverter output current and offset the target harmonic components in the power grid.

[0066] For the r-th independent harmonic source, the harmonic component at time t in the future is represented as: i r (t)=A r sin(2πf r t+φ r A r For the predicted harmonic amplitude, f r For the predicted harmonic frequency, φ r This represents the predicted harmonic phase. The target harmonic signal in the power grid is the superposition of these independent harmonic components: The photovoltaic inverter needs to output a compensation current i that is equal to but opposite to the target harmonic. comp (t), which makes the total harmonic current approach zero: The inverter generates discrete-time signals through the control system. T s : Sampling period of the control system; n: Index of the sampling time. The control system calculates the reference current i according to the compensation current formula. ref [n]:i ref [n] = i comp [n];i ref [n] is the ideal current that the inverter needs to output.

[0067] By employing pulse width modulation (PWM) technology, the duty cycle of the inverter's switching devices is adjusted to ensure that the actual output current tracks the reference current i. ref [n], Formula for calculating the duty cycle of the PWM signal: D[n]: PWM duty cycle; I max The inverter's maximum output current. The actual inverter output current may deviate from the reference value due to system dynamics or external disturbances. Closed-loop control adjusts the output in real time: the actual output current i is measured in real time. out [n], the error in calculation with reference value: e[n]=i ref [n]-i out [n]; A proportional-integral (PI) controller is used to generate a correction signal to adjust the inverter output: K p For proportional gain; K i is the integral gain; u[n] is the control signal used to correct the PWM duty cycle.

[0068] The PWM duty cycle is corrected according to the control signal: D[n] = D[n-1] + u[n]; the compensation current i output by the inverter is... total (t), the total current at the grid connection point is expressed as: i total (t)=iharm (t)+i out (t); i out (t) = i comp (t) = -i harm (t) into, the total current becomes: i total (t) = i harm (t)-i harm (t) = 0; harmonic cancellation is achieved.

[0069] In the embodiment, first, the original multi-dimensional electrical signals (including current, voltage, frequency) are collected at the photovoltaic grid-connected point, and the Fourier transform is used to decompose the original multi-dimensional electrical signals into multi-dimensional frequency signals; then, the blind source separation algorithm and the non-negative matrix decomposition technology are used to separate the characteristic components of independent harmonic sources from the frequency signals, and the frequency and amplitude characteristics of the independent harmonic sources are extracted to reduce the prediction deviation caused by the superposition effect. Subsequently, based on the separated harmonic source characteristics, a multi-dimensional time series prediction model is constructed by using a recurrent neural network to accurately predict the future behavior of each harmonic source. On this basis, in combination with the data of the separated harmonic sources, high-frequency compensation and reconstruction are performed on the frequency domain, multi-scale analysis is used to correct the frequency aliasing error, and the accuracy of the frequency separation result is ensured. Finally, the prediction result is input into the photovoltaic inverter control system, the output current is dynamically adjusted, the reverse harmonic compensation signal is generated, the effective cancellation of the target harmonic of the power grid is achieved, and thus the power quality and the stability of the grid-connected system are improved.

[0070] The above formulas are dimensionless numerical calculations, the formulas are obtained by software simulation of a large amount of data to obtain a formula of the most recent real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0071] The above-described embodiments can be implemented in part or in whole through software, hardware, firmware or any combination thereof. When implemented in software, the above-described embodiments can be implemented using one or more computer programs written in any suitable programming language. The computer programs can be stored in one or more computer-readable storage media, such as a memory, a magnetic disk, an optical disk, a hard disk, a floppy disk, a magnetic tape, a memory card, a ROM, a DVD, a Blu-ray Disc, a CD, a semiconductor memory, a flash memory, or the like. The computer programs can be loaded into a computer, a server, a computer network, or the like, and executed. The computer programs can be distributed to computer systems connected to a network, and executed in parallel. The computer programs can be executed by a computer system that is capable of accessing a network, such as the Internet, and executed in parallel.

[0072] It should be understood that the term "and / or" in this document is merely used to describe associated objects, and can represent the three conditions of "A", "B", and "A and / or B". For example, "A and / or B" can represent the three conditions of "A alone", "B alone", and "A and B together". In addition, the character " / " in this document generally represents an "or" relationship between the associated objects, but can also represent an "and / or" relationship. Those skilled in the art can understand the "and / or" relationship according to the context. Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed in this document can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized by hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0073] The above description is merely a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application.

Claims

1. A multi-dimensional time-series photovoltaic grid-connected harmonic current prediction control method, characterized in that: Includes the following steps; S1: Collect the original multidimensional electrical signals of the photovoltaic grid connection point. The original multidimensional electrical signals include current, voltage and frequency. Use Fourier transform to decompose the original multidimensional electrical signals into multidimensional frequency signals. S2: Using blind source separation algorithm and non-negative matrix decomposition technology, the characteristic components of independent harmonic sources are separated from multi-dimensional frequency signals. The frequency and amplitude characteristics of each independent harmonic source are extracted to reduce the prediction deviation caused by superposition effect. S3: Based on the characteristics of the isolated independent harmonic sources, a time-series prediction model is constructed using a recurrent neural network. The multidimensional prediction model processes multiple harmonic sources simultaneously to predict the future characteristics of the overall harmonic source. S4: Combining the separated harmonic source data, high-frequency compensation and reconstruction are performed on the frequency domain data. Multi-scale analysis is used to further correct the frequency aliasing error and ensure the accuracy of the frequency separation results. In S4, based on the obtained harmonic source frequency and amplitude data, a spectrum diagram is plotted. Utilizing the characteristic that amplitude decays with frequency, the high-frequency components are compensated using an exponential model to generate compensated high-frequency harmonic amplitudes. Combined with phase information, the high-frequency harmonic signal is reconstructed using inverse Fourier transform. Wavelet transform is used for multi-scale decomposition to divide the signal into low-frequency and high-frequency components. Through multi-scale decomposition, abnormal frequency components can be identified and corrected, preserving low-frequency information and high-frequency components while removing aliasing. The corrected high-frequency and low-frequency harmonic signals are combined to form a complete harmonic signal; S5: Input the multi-dimensional prediction results into the control system of the photovoltaic inverter to dynamically adjust the inverter output current and cancel the target harmonic components in the power grid; Specifically, in S5, inputting the multi-dimensional prediction results into the control system of the photovoltaic inverter to dynamically adjust the inverter output current and cancel the target harmonic components in the power grid involves: For the rth independent harmonic source, the harmonic component at future time t is represented as: ; is the predicted harmonic amplitude, is the predicted harmonic frequency, is the predicted harmonic phase; the target harmonic signal in the grid is the superposition of independent harmonic components: ; the photovoltaic inverter needs to output a compensation current equal and opposite to the target harmonic so that the total harmonic current tends to zero: ; Inverter generates discrete time signals through control system : ; : Sampling period of control system; n: index of sampling time, the control system calculates the reference current according to the compensation current formula : ; is the ideal current required by the inverter to output.

2. The multi-dimensional timing photovoltaic grid-connected harmonic current predictive control method according to claim 1, characterized in that: In S1, frequency analysis of multidimensional time-domain electrical signals at the photovoltaic grid connection point is performed through Fourier transform. Discrete current, voltage, and frequency signal data in the time domain are collected. Discrete Fourier transform is applied to each signal to convert it from the time domain to the frequency domain, and the corresponding spectrum information, including frequency components and amplitude characteristics, is extracted. The spectrum data of current, voltage, and frequency signals are combined to form a multidimensional frequency matrix.

3. The multi-dimensional timing photovoltaic grid-connected harmonic current predictive control method according to claim 2, characterized in that: In S2, the multidimensional frequency signal matrix is ​​set to be composed of the linear superposition of frequency components of multiple independent harmonic sources. The signal matrix is ​​processed by blind source separation to whiten and eliminate the correlation between signals, so that the components of each harmonic source are independent of each other. In the blind source separation process, the separation matrix is ​​solved by optimization algorithm, and the separation result is verified to meet the harmonic characteristics. Then, the frequency components of each independent harmonic source are obtained. The separation result is decomposed by non-negative matrix factorization, and non-negativity constraint is introduced to ensure that the separated frequency and amplitude are positive values. The weight matrix and fundamental component matrix of each independent harmonic source are obtained by NMF, and the main frequency characteristics and amplitude characteristics of the harmonic source are extracted. The main frequency and amplitude of each independent harmonic source are calculated to form the frequency and amplitude characteristic vector of the independent harmonic source.

4. The multi-dimensional time-series photovoltaic grid-connected harmonic current prediction and control method according to claim 3, characterized in that: S3, extract independent feature vector ; the time series characteristics of each independent harmonic source are represented as: ; in the formula, , the frequency and amplitude characteristics at time , , represents the harmonic source number, N is the length of the time series, the input data is normalized, and a prediction model is constructed using a recurrent neural network; the model input is a multi-dimensional characteristic time series, and the output is a predicted value at a future time Input data for each time step For the normalized frequency and amplitude characteristics; a recursive structure is used to capture the long and short term dependencies of the time series; The state update formula of the hidden layer is: ; is the hidden layer state, is the input characteristic vector, is the hidden layer parameter, is the activation function; the output layer is to predict the future frequency and amplitude characteristics, and the output is: ; taking the time series as input, the RNN model predicts the characteristics of the future time : ; for multiple harmonic sources r = 1, 2, …, R, multiple sub-models are trained at the same time, and the loss function is used to evaluate the error between the predicted results and the true values , the expression is: ; wherein: is the true value, is the predicted value; the time series of each harmonic source is divided into a sliding window to form a training sample set; the training set and the validation set are divided: 80% of the data is divided into the training set, and 20% is divided into the validation set.

5. The multi-dimensional time-series photovoltaic grid-connected harmonic current prediction and control method according to claim 4, characterized in that: In S3, the current time series Input the model to predict the next time step. Its characteristics: Based on the single-step prediction results, the predicted values ​​are... As input for the next step, iteratively predict multiple future time steps: The predicted harmonic source characteristic matrix is: M represents the prediction time step; It includes the frequency and amplitude characteristics of the harmonic source r at future times.

6. The multi-dimensional time-series photovoltaic grid-connected harmonic current prediction and control method according to claim 1, characterized in that: By employing pulse width modulation (PWM) technology, the duty cycle of the inverter's switching devices is adjusted to ensure that the actual output current tracks the reference current. PWM signal duty cycle calculation formula: ; D[n]: PWM duty cycle; The inverter's maximum output current is adjusted in real time via closed-loop control; the actual output current is measured in real time. Error in calculation compared to reference value: ; A proportional-integral controller is used to generate a correction signal to adjust the inverter's output. ; For proportional gain; For integral gain; u[n] is a control signal used to correct the PWM duty cycle; Correct the PWM duty cycle based on the control signal: Compensation current output by the inverter The total current at the grid connection point is expressed as: ;Will Substituting, the total current becomes: Harmonic cancellation was achieved.

Images