An influencing factor analysis and suppression method of power quality of an inverter grid-connected device and related apparatus

Abstract

The application provides an influence factor analysis and suppression method for power quality of an inverter grid-connected device and related devices, acquires electrical quantity data and external disturbance data on the grid-connected side of the inverter, and calculates power quality characteristic parameters to form a multi-dimensional data matrix. A power quality multi-factor analysis model based on a Bayesian network structure is used to identify the dominant factor of reducing power quality; a power quality dynamic prediction model based on a neural network structure is used to predict the target power quality index value. The control parameters and strategies of the grid-connected inverter are determined according to the dominant factor, the control adjustment timing is determined according to the power quality state change, the grid-connected inverter is controlled based on this, and the reduction of power quality is suppressed. The application identifies the dominant factor through multi-dimensional data, predicts the index value in advance, dynamically determines the control parameters, strategies and adjustment timing to control the grid-connected inverter, and improves the accuracy, real-time performance, intelligence and system generalizability of the power quality control of the grid-connected inverter.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of inverter grid-connected, and particularly relates to an influence factor analysis and suppression method for power quality of an inverter grid-connected device and related devices. BACKGROUND

[0002] With the rapid development of new energy technologies such as photovoltaic power generation and wind power generation, the inverter, as the core interface device between new energy and the power grid, plays a crucial role in converting DC to AC, regulating output power quality, and synchronizing grid connection. Especially in the low-voltage distributed generation scenario, the number of inverters is large and widely distributed, and their operating state has a non-negligible impact on the power quality of the distribution network.

[0003] However, due to the use of pulse width modulation (PWM) technology based on switching devices by inverters, a large number of high-order harmonic components are inevitably contained in the output voltage and current. At the same time, affected by factors such as grid fluctuations, load disturbances, environmental temperature changes, and grid impedance, the inverter is prone to cause typical power quality problems such as voltage distortion (e.g., increased harmonic content, distorted waveform), voltage flicker and fluctuation, three-phase imbalance, reactive power fluctuation, and frequency deviation during operation.

[0004] Although there are many solutions for inverter grid-connected power quality control, there are still many key defects and deficiencies. For example, the existing technology has insufficient ability to identify the factors affecting power quality, which affects the accuracy of the control strategy; the regulation strategy is rigid and cannot dynamically adjust the operating mode according to the grid state, which may cause system problems due to delayed response; it relies on manual operation and maintenance intervention, which is high in maintenance cost and low in efficiency; the hardware system structure is complex, the integration difficulty is high, the cost is rising, the maintenance is complex, the adaptability is poor, and it is not suitable for large-scale promotion. SUMMARY

[0005] Therefore, the present application provides an influence factor analysis and suppression method for power quality of an inverter grid-connected device and related devices, which aims to at least solve one of the above-mentioned deficiencies of the existing inverter grid-connected power quality control method.

[0006] To achieve the above-mentioned purpose, the technical solutions provided by the present application are as follows:

[0007] In a first aspect, the present application provides an influence factor analysis and suppression method for power quality of an inverter grid-connected device, comprising the following steps:

[0008] The electrical quantity data of the grid-connected side of the inverter and external disturbance data that will affect power quality are collected, power quality characteristic parameters for measuring power quality are calculated based on the electrical quantity data, and a multi-dimensional data matrix is formed according to the collected and calculated data;

[0009] According to the multi-dimensional data matrix, a pre-trained power quality multi-factor analysis model is used to identify the dominant factor that reduces power quality; the power quality multi-factor analysis model is based on a Bayesian network structure and is used to identify the factor that causes the reduction of power quality through the causal relationship of the Bayesian network;

[0010] According to the multi-dimensional data matrix, a pre-trained power quality dynamic prediction model is used to predict the target power quality index value; the power quality dynamic prediction model is based on a neural network structure and is used to mine the time series characteristics in the multi-dimensional data matrix and predict the target power quality index value;

[0011] According to the dominant factor, the control parameters and control strategy of the grid-connected inverter are determined, and according to the target power quality index value, the control adjustment time of the grid-connected inverter is determined, and the control of the grid-connected inverter is performed based on the control adjustment time and the control parameters and control strategy to suppress the reduction of power quality.

[0012] Further, the multi-dimensional data matrix is formed according to the collected and calculated data, including:

[0013] The collected electrical quantity data and external disturbance data, and the calculated power quality characteristic parameters are preprocessed to remove noise interference and abnormal measurement errors in the data, and all preprocessed data are interpolated and aligned based on a unified time reference to obtain multi-class parameter data corresponding to the same time point;

[0014] The multi-class parameter data is stored in a data matrix indexed by a timestamp, thereby forming a multi-dimensional data matrix.

[0015] Further, the pre-trained power quality multi-factor analysis model is used to identify the dominant factor that reduces power quality, including:

[0016] An improved principal component analysis method is used to reduce the dimension of the multi-dimensional data matrix to obtain a plurality of features after dimension reduction; in the dimension reduction process, the improved principal component analysis method performs standardization processing on the data in the multi-dimensional data matrix according to the following formula:

[0017]

[0018]

[0019] In the formula, denotes the standardized result of the i th sample, the j th feature in the multi-dimensional data matrix, denotes the original value of the i th sample, the j th feature, denotes the original data mean of the j th feature, denotes the weight of the j th feature, denotes the standard deviation of the j th feature; denotes the coefficient of variation of the j th feature dimension, denotes the correlation coefficient of the j th feature dimension and the target variable;

[0020] The covariance matrix is also updated according to the following formula:

[0021]

[0022] In the formula, denotes the covariance matrix at time t, denotes the size of the sliding window for selecting data from the multi-dimensional data matrix, denotes the normalized data at time n; denotes the mean of the normalized data in the sliding window;

[0023] The principal component selection weight is also determined according to the following formula:

[0024]

[0025] In the formula, denotes the k th principal component selection weight, denotes the eigenvalue of the k th principal component, denotes the original variable index set contained in the k th principal component, is the causal influence degree of the j th original variable on the target power quality index;

[0026] The reduced number of features are input into the pre-trained Bayesian network, the influence degree of each feature on the target power quality index is quantified, and a factor contribution rate matrix is generated;

[0027] Based on the factor contribution rate matrix, the influencing factors are arranged in descending order of contribution degree, and the first m factors with cumulative contribution degree exceeding a set threshold are selected as the dominant factors, and m is a set value.

[0028] Further, the neural network is a long short-term memory neural network, and the input of the power quality dynamic prediction model is the data in the multi-dimensional data matrix, and the output is the predicted value of the target power quality index.

[0029] Further, the control parameters and control strategies of the grid-connected inverter are determined according to the dominant factors, including:

[0030] According to the dominant factor, the controller parameter needing adjustment is determined;

[0031] The adjustment value of the controller parameter needing adjustment is determined by using the pre-constructed fuzzy control rule;

[0032] According to the dominant factor, whether the power grid is in a special scene outside the normal working condition is identified, if yes, the control strategy is switched to a more robust control strategy based on the fuzzy rule.

[0033] Further, it further comprises:

[0034] The change trend of the target power quality index is monitored in real time;

[0035] If the change trend indicates that the control effect is not up to standard, the control parameter re-tuning and control strategy re-selection are triggered;

[0036] If the change trend indicates that the system reaches a steady state, the current control parameter and control strategy are kept running, and the data of the current scene are recorded as training samples of each model.

[0037] Further, it further comprises: a visual interface is adopted to display the current power quality state, control action history, prediction and warning information and equipment running state in real time.

[0038] In the second aspect, the application provides an influence factor analysis and suppression device for power quality of an inverter grid-connected equipment, comprising:

[0039] A data acquisition module is used to acquire electrical quantity data of the inverter grid-connected side and external disturbance data that will affect power quality, and calculate power quality characteristic parameters for measuring power quality based on the electrical quantity data, and form a multi-dimensional data matrix according to the acquired and calculated data;

[0040] A factor analysis module is used to identify the dominant factor of reducing power quality by using a pre-trained power quality multi-factor analysis model according to the multi-dimensional data matrix; the power quality multi-factor analysis model is based on a Bayesian network structure and is used to identify the factor causing the reduction of power quality through the causal relationship of the Bayesian network;

[0041] A quality prediction module is used to predict the target power quality index value by using a pre-trained power quality dynamic prediction model according to the multi-dimensional data matrix; the power quality dynamic prediction model is based on a neural network structure and is used to mine the time sequence characteristics in the multi-dimensional data matrix and predict the target power quality index value;

[0042] The inhibition module is configured to determine the control parameter and the control strategy of the grid-connected inverter according to the dominant factor, determine the control adjustment time of the grid-connected inverter according to the target power quality index value, and control the grid-connected inverter based on the control adjustment time and the control parameter and the control strategy, so as to inhibit the reduction of the power quality.

[0043] In a third aspect, the present application provides a computer device, which comprises a processor and a memory:

[0044] The memory is configured to store a computer program and send instructions of the computer program to the processor.

[0045] The processor is configured to execute the instructions of the computer program to perform the method for analyzing and inhibiting the factors affecting the power quality of the grid-connected inverter device according to the first aspect.

[0046] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method for analyzing and inhibiting the factors affecting the power quality of the grid-connected inverter device according to the first aspect.

[0047] In summary, the present application provides a method for analyzing and inhibiting the factors affecting the power quality of the grid-connected inverter device and related devices, which comprises collecting electrical quantity data on the grid-connected side of the inverter and external disturbance data that will affect the power quality, calculating power quality characteristic parameters for measuring the power quality based on the electrical quantity data, and forming a multi-dimensional data matrix according to the collected and calculated data; identifying the dominant factor reducing the power quality by using a pre-trained power quality multi-factor analysis model based on the multi-dimensional data matrix; the power quality multi-factor analysis model is based on a Bayesian network structure and is used to identify the factors causing the reduction of the power quality by the causal relationship of the Bayesian network; predicting the target power quality index value by using a pre-trained power quality dynamic prediction model based on the multi-dimensional data matrix; the power quality dynamic prediction model is based on a neural network structure and is used to mine the time series characteristics in the multi-dimensional data matrix and predict the target power quality index value; determining the control parameter and the control strategy of the grid-connected inverter according to the dominant factor, determining the control adjustment time of the grid-connected inverter according to the state change of the power quality, and controlling the grid-connected inverter based on the control adjustment time and the control parameter and the control strategy, so as to inhibit the reduction of the power quality. The present application identifies the dominant factor affecting the power quality by the multi-dimensional data, predicts the target power quality index value in advance, and then dynamically determines the control parameter, the strategy and the adjustment time to control the grid-connected inverter according to these results, so as to overcome the problems of insufficient identification of the factors affecting the power quality, rigid control strategy and dependence on manual operation in the prior art, and effectively improve the accuracy, real-time performance, intelligence and system generalizability of the control of the grid-connected inverter. BRIEF DESCRIPTION OF DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0049] Figure 1 A flow chart of an influence factor analysis and suppression method of power quality of an inverter grid-connected device provided for an embodiment of the present application;

[0050] Figure 2 A composition block diagram of an influence factor analysis and suppression device of power quality of an inverter grid-connected device provided for an embodiment of the present application;

[0051] Figure 3 A composition block diagram of a computer device provided for an embodiment of the present application. DETAILED DESCRIPTION

[0052] In order to make the objectives, features and advantages of the present application more obvious and easy to understand, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings of the embodiments of the present application. Obviously, the following described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0053] Please refer to Figure 1 The embodiment of the present application provides an influence factor analysis and suppression method of power quality of an inverter grid-connected device, which comprises the following steps:

[0054] S11: Collecting electrical quantity data of an inverter grid-connected side and external disturbance data which will affect power quality, and calculating power quality characteristic parameters for measuring power quality based on the electrical quantity data, and forming a multi-dimensional data matrix according to the collected and calculated data.

[0055] It should be noted that the electrical quantity data refers to voltage, current, frequency, active power, reactive power and other electrical parameters directly reflecting the operating state of a power system at the connection side of the inverter and the power grid.

[0056] The external disturbance data refers to interference data from outside the power grid and which will affect power quality.

[0057] Power quality characteristic parameters are indicators for quantitatively describing the pros and cons of power quality, such as total harmonic distortion (measuring the degree of voltage or current waveform distortion), voltage deviation (the difference between actual voltage and rated voltage), three-phase imbalance (the degree of asymmetry of three-phase voltage or current), frequency deviation, etc.

[0058] The multi-dimensional data matrix is a multi-column and multi-row data set composed of collected electrical quantity data, external disturbance data, and calculated power quality characteristic parameters. Each column represents a data dimension, and each row represents a group of data at a specific time or operating condition.

[0059] This step can collect electrical quantity data on the grid side of the inverter in real time through sensors and other devices, while collecting external disturbance data that may affect power quality. Then, using relevant calculation formulas and standards, power quality characteristic parameters are calculated based on the collected electrical quantity data, such as total harmonic distortion calculated from voltage waveform data. Finally, these raw data and calculated characteristic parameters are sorted according to certain rules to form a multi-dimensional data matrix containing multiple dimensional information.

[0060] S12: According to the multi-dimensional data matrix, the pre-trained power quality multi-factor analysis model is used to identify the dominant factor that reduces power quality; the power quality multi-factor analysis model is based on a Bayesian network structure and is used to identify the factors that cause power quality reduction through the causal relationship of the Bayesian network.

[0061] It should be noted that the power quality multi-factor analysis model is a model based on a Bayesian network structure, which analyzes the relationship between multiple factors and power quality, and then identifies the key factors that cause power quality to decline.

[0062] Bayesian networks are a probabilistic graphical model composed of nodes (representing variables) and directed edges (representing causal relationships between variables), which can infer the probability of unknown variables from the probability distribution of known variables, and are particularly suitable for handling uncertainty problems and causal relationship inference.

[0063] In this step, the pre-trained power quality multi-factor analysis model has learned the causal relationship patterns between various factors and power quality. The multi-dimensional data matrix is input into the model, which uses Bayesian network principles to analyze the probability relationships and causal connections between variables in the data, and uses Bayesian inference algorithms to identify the factors that play a dominant role in reducing power quality from among the many influencing factors. For example, when a certain external disturbance data has a strong causal relationship with the deterioration of power quality characteristic parameters and a high probability of occurrence, the disturbance may be identified as the dominant factor.

[0064] S13: predicting the target power quality index value by using the pre-trained power quality dynamic prediction model according to the multi-dimensional data matrix; the power quality dynamic prediction model is based on a neural network structure and is used for mining time sequence features in the multi-dimensional data matrix and predicting the target power quality index value.

[0065] It should be noted that the power quality dynamic prediction model is a model based on a neural network structure, which is used for predicting the target power quality index in the future according to historical and current multi-dimensional data, and has the ability to mine time sequence features.

[0066] In this step, the power quality dynamic prediction model processes the multi-dimensional data matrix through the multi-layer structure of the neural network. The model first extracts time sequence features from the data, i.e., the rules and trends of data changes over time, such as the change patterns of electrical quantity data at different times, the time distribution characteristics of external disturbances, etc. Then, by using these time sequence features and combining the nonlinear mapping ability of the neural network, the target power quality index value at a future time (such as the total harmonic distortion rate in the next 5 seconds) is predicted, providing a basis for taking control measures in advance.

[0067] S14: determining the control parameters and control strategies of the grid-connected inverter according to the dominant factors, and determining the control adjustment time of the grid-connected inverter according to the target power quality index value, and controlling the grid-connected inverter based on the control adjustment time and the control parameters and control strategies to suppress the reduction of power quality.

[0068] It should be noted that the control parameters refer to specific numerical values used to adjust the operating state of the grid-connected inverter, such as the switching frequency of PWM (pulse width modulation), the proportional integral parameters of the current loop and the voltage loop, etc.

[0069] The control strategy is a method and rule for guiding the grid-connected inverter to adjust to improve power quality, such as a control strategy based on harmonic compensation, a low voltage ride through control strategy during voltage drop, etc.

[0070] The control adjustment time refers to the best moment for controlling and adjusting the grid-connected inverter, which is usually determined according to the state changes of power quality, such as when it is predicted that the power quality index will exceed the allowed range or has shown a downward trend.

[0071] The step determines the control parameters and control strategy of the grid-connected inverter according to the dominant factor. For example, if the dominant factor is harmonic interference, the control strategy can adopt harmonic compensation, and the control parameters are set to the corresponding compensation coefficients, etc. At the same time, the power quality dynamic prediction result and the real-time monitored power quality state change are used to determine whether the power quality is about to deteriorate or is in a state that needs to be adjusted, so as to determine the control adjustment time. Finally, at the determined time, the set control parameters and control strategy are applied to control the grid-connected inverter, such as adjusting the output voltage, compensating for reactive power, etc., to suppress the reduction of power quality.

[0072] The embodiment provides an influence factor analysis and suppression method for power quality of an inverter grid-connected device. The method integrates electrical quantity data, external disturbance data and power quality characteristic parameters by constructing a multi-dimensional data matrix, realizes comprehensive capture of factors affecting power quality, and breaks through the limitation of single data dimension of traditional methods. Secondly, the Bayesian network and the neural network are combined, the dominant factor is accurately identified by using the causal relationship deduction ability of the Bayesian network, and the dynamic prediction is realized by using the time sequence feature mining ability of the neural network. The advantages of the two models are complementary, and the accuracy and reliability of analysis and prediction are improved. Finally, the control parameters, strategy and adjustment time are determined based on the dominant factor and the dynamic prediction result, forming a closed-loop suppression system of "data acquisition-factor identification-quality prediction-accurate control", realizing active, accurate and timely suppression of the reduction of power quality, overcoming the lag and blindness of traditional control methods, and significantly improving the management level of power quality of the inverter grid-connected device.

[0073] In an embodiment of the present application, a multi-dimensional data matrix is formed according to the collected and calculated data, including:

[0074] S21: The collected electrical quantity data and external disturbance data, and the calculated power quality characteristic parameters are preprocessed to remove noise interference and abnormal measurement errors in the data, and all the preprocessed data is interpolated and aligned based on a unified time reference to obtain multi-class parameter data corresponding to the same time point.

[0075] It should be noted that the electrical quantity data and the external disturbance data include but are not limited to the following data: three-phase voltage, current, active power, reactive power, harmonic voltage / current (including 2-50 times), voltage fluctuation rate, voltage unbalance degree, temperature, load change, grid impedance change and other external environmental parameters. The data sampling frequency is not less than 5 kHz, which ensures complete capture of dynamic characteristics. Specifically, a plurality of channel high-speed measurement sensors and analog / digital conversion units are deployed on the grid-connected side of the inverter to simultaneously collect the following physical quantities:

[0076] Voltage data: three-phase AC voltage (A / B / C), line voltage, effective value, instantaneous waveform;

[0077] Current data: three-phase output current, peak value, current distortion, etc.

[0078] Power parameters: active power (P), reactive power (Q), apparent power (S);

[0079] Harmonic parameters: harmonic voltage and harmonic current spectrum, with a collection range of 2 to 50 harmonics;

[0080] Voltage quality parameters: voltage fluctuation rate, voltage change rate (dv / dt), short-time voltage interruption;

[0081] Three-phase balance: voltage unbalance (Unb), current unbalance;

[0082] External disturbance parameters: ambient temperature (T), load current change rate (dI / dt), grid impedance (Zg) dynamic change trend.

[0083] The preprocessing operation specifically includes the following procedures:

[0084] (1) De-noising filter processing:

[0085] For high-speed sampling signals such as voltage and current, a band-pass FIR (Finite Impulse Response) filter (typical passband is 45~2500Hz) is used to filter out high-frequency interference and fundamental DC offset; harmonic data is reconstructed using windowed Fast Fourier Transform (FFT) to improve spectral identification accuracy.

[0086] (2) Outlier identification and correction:

[0087] The sampling points are processed by sliding window (such as 5~20 points) and the local mean and variance are calculated; if a sampling point deviates from the local mean by more than a certain threshold (such as 3σ principle), it is determined as an outlier, and is corrected by using the average, median interpolation and other methods.

[0088] (3) Synchronization correction mechanism:

[0089] All sampling channels use internal clock synchronization (or external clock source such as GPS) to ensure that the time references of various data are consistent; the collected data is interpolated / aligned to ensure that multiple parameters at the same time point are one-to-one corresponding.

[0090] S22: Store multiple parameter data into a data matrix indexed by time stamp, thereby forming a multi-dimensional data matrix.

[0091] It should be noted that the unified data matrix structure constructed with the timestamp may be as follows:

[0092]

[0093] Among them: represents the unified data matrix with the timestamp as the primary key; : A-phase voltage at time t; : A-phase current at time t; : Total harmonic distortion; : Ambient temperature; : Grid point impedance; & represents a data splicing symbol, that is, splicing multiple types of parameter data to form the unified data matrix.

[0094] All data form a wide format table or tensor, supporting horizontal analysis and longitudinal sequence modeling. In the process of forming a multi-dimensional data matrix, for electrical quantity data (such as voltage, current), high-frequency sampling (10 kHz) is adopted, and after downsampling every 100 points, it is stored in the matrix; for environmental quantities (such as temperature, humidity, grid impedance), low-frequency sampling (1~10 Hz) is adopted, and interpolation is matched to the main timestamp. Aligning data from different sources and different frequencies generates complete time series samples.

[0095] In addition, a data cache and interface service are designed to interface with other processing procedures. For example, for the cache mechanism, all preprocessed data is cached in a ring queue according to the second-level timestamp, and the recent 5~15 minute window is retained; for the data interface, a unified data calling interface is provided to the upper module, supporting time period and feature vector dimension query; for the storage mode, local database (SQLite) and edge / cloud synchronous transmission are supported.

[0096] In an embodiment of the present application, a pre-trained power quality multi-factor analysis model is used to identify the dominant factor that reduces power quality, including:

[0097] S31: The improved principal component analysis method is used to reduce the dimension of the multi-dimensional data matrix, and a plurality of characteristics after dimension reduction are obtained.

[0098] To adapt to the multi-dimensional, nonlinear and time-varying characteristics of power quality data in the inverter grid-connected scene, the present application improves the traditional principal component analysis (PCA) algorithm in the following three aspects, so as to realize the power quality multi-factor analysis with more practicality and discriminability, and the specific improvements are as follows:

[0099] (1) Improvement I: Adaptive weighted standardization processing.

[0100] Traditional PCA uses zero mean, unit variance normalization for all input variables, but in the grid-connected inverter scenario, the fluctuation degree of different data dimensions (such as harmonics, voltage, current, impedance, etc.) at different times and the significant influence on system disturbance are different, which leads to distortion of dimension reduction by unified standardization. The improved principal component analysis in this step performs standardization processing on the data in the multi-dimensional data matrix according to the following formula during dimension reduction processing:

[0101]

[0102]

[0103] In the formula, denotes the standardized result of the i-th sample, the j-th feature in the multi-dimensional data matrix, denotes the original value of the i-th sample, the j-th feature, denotes the original data mean of the j-th feature, denotes the weight of the j-th feature, which is automatically updated according to the historical data fluctuation rate and correlation contribution, denotes the standard deviation of the j-th feature; denotes the coefficient of variation (reflecting instability) of the j-th feature dimension, denotes the correlation coefficient between the j-th feature dimension and the target variable;

[0104] (2) Improvement II: Sliding window covariance matrix update.

[0105] Traditional PCA needs to use all samples to calculate the covariance matrix at one time, which does not have real-time performance in the data stream scenario. This step uses a sliding window to update the covariance matrix online, dynamically updating at the last sampling time:

[0106]

[0107] In the formula, denotes the covariance matrix at time t, denotes the sliding window size for selecting data from the multi-dimensional data matrix, denotes the standardized data at time n; denotes the mean of the standardized data in the sliding window, and T denotes the transpose operation.

[0108] (3) Improvement III: Principal component selection rule fusion with causal explanation ability.

[0109] In traditional PCA, the principal component selection is usually based on the empirical rule that the cumulative contribution rate is greater than 85%, and the low-variance but high-causal-contribution factors are easily ignored. This step combines the contribution rate correction selection mechanism of the causal structure of the Bayesian network, specifically:

[0110] A. Calculate the contribution rate of each principal component using the traditional method;

[0111] B. Based on the trained Bayesian network, quantify the causal strength score of each original variable on the target index (such as harmonic distortion, frequency shift) ;

[0112] C. Correct the principal component selection weight:

[0113]

[0114] wherein, represents the kth principal component selection weight, represents the eigenvalue of the kth principal component, represents the original variable index set contained in the kth principal component, is the causal influence degree of the jth original variable on the target power quality index.

[0115] S32: Input the reduced dimension features into the pre-trained Bayesian network, quantify the influence degree of each feature on the target power quality index, and generate a factor contribution rate matrix;

[0116] It should be noted that the reduced dimension features refer to a number of low-dimensional features obtained after dimension reduction processing, which can represent the main information of the original multi-dimensional data.

[0117] This step inputs the reduced dimension features into the pre-trained Bayesian network. Since the Bayesian network has learned the probability relationship between the features and the target power quality index, it can use these known relationships to calculate the influence degree of each feature on the target power quality index through probability reasoning. This quantification of influence degree is based on the conditional probability distribution between nodes in the network, and finally forms a factor contribution rate matrix, which clearly shows the contribution of each feature to the target index.

[0118] S33: Based on the factor contribution rate matrix, arrange the influencing factors in descending order of contribution degree, and select the top m factors with cumulative contribution degree exceeding the set threshold as the dominant factors, m is a set value.

[0119] It should be noted that after obtaining the factor contribution rate matrix, the contribution degrees of various influencing factors are sorted from high to low. Then, the sorted factor contribution rates are sequentially accumulated, and when the accumulated contribution degree exceeds a set threshold, the first m factors contained in the accumulated contribution degree at this time are selected as the dominant factors, wherein m is a value set according to actual conditions. In this way, the factors that mainly affect the target power quality index can be accurately screened from numerous influencing factors, and a clear basis for subsequent development of a targeted control strategy is provided.

[0120] The improved PCA method can effectively solve the problems of nonlinearity, dynamic disturbance and data heterogeneity of the inverter grid-connected power quality data, and can also consider data compression efficiency and system causal explanation. Meanwhile, the input quality of the subsequent Bayesian causal analysis model is improved, and the interference of redundant misleading factors is reduced.

[0121] In an embodiment of the present application, the neural network is a long short-term memory neural network, and the input of the power quality dynamic prediction model is data in a multi-dimensional data matrix, and the output is a predicted value of the target power quality index.

[0122] Specifically, in order to realize the advance judgment of power quality change, the embodiment introduces a LSTM (Long Short-Term Memory) neural network model to predict the power quality state in the future 1-5 seconds, including: total harmonic distortion (THD), voltage amplitude stability, frequency fluctuation range, and three-phase voltage symmetry. The prediction result will be used to trigger the adjustment mechanism in advance to avoid lag response. The power quality dynamic prediction model is introduced as follows:

[0123] (1) Input parameter and front-end data docking method.

[0124] The input sequence of the LSTM model comes from the preprocessed data of steps S11 and S21, and specifically includes:

[0125]

[0126] All input data are provided in the form of a feature matrix aligned with a unified timestamp, with a length of a sliding window W, which can be 30-50 sampling points, covering the last 1-5 seconds of historical data.

[0127] (2) LSTM model structure and parameter setting.

[0128] The model adopts a typical multi-layer LSTM architecture, and the structure configuration is as follows:

[0129]

[0130] (3) Model training parameter setting.

[0131]

[0132] The training data is derived from a sample library constructed based on 1000+ hours of grid-connected operation data; the verification accuracy requires an average prediction error of between 3% and 5%.

[0133] In this embodiment, the power quality dynamic prediction model supports 1-5 second rolling prediction in the future, and triggers the power quality adjustment strategy in advance according to the set threshold; for example, when it is predicted that the future THD may exceed the set upper limit (such as 3.0%), the regulator is awakened to perform filtering or voltage control, thereby avoiding the deterioration of power quality caused by control lag.

[0134] In one embodiment of the present application, the control parameters and control strategies of the grid-connected inverter are determined according to the dominant factor, including:

[0135] S41: Determine the controller parameters that need to be adjusted according to the dominant factor.

[0136] It should be noted that the controller parameters of the inverter (such as PI coefficients, current loop bandwidth, PWM modulation mode, etc.) are dynamically adjusted according to the currently identified dominant factor, to improve the system response speed and stability.

[0137] S42: Determine the adjustment value of the controller parameters that need to be adjusted using the pre-constructed fuzzy control rules.

[0138] It should be noted that the key parameters of the controller (such as the proportional coefficient , integral coefficient , current loop bandwidth , PWM modulation mode, etc.) will be dynamically adjusted online according to the real-time identified power quality dominant factor.

[0139] The fuzzy control rule library design (taking PI coefficients as an example) is as follows:

[0140] (1) Input:

[0141] Error signal: (such as THD target and current value); represents the reference value of the control target, represents the current actual measurement value, represents the error at time t;

[0142] Error change rate: ; represents the error at the previous time , and represents the error change value.

[0143] (2) Fuzzy language variable division:

[0144]

[0145] (3) Fuzzy rules (part):

[0146]

[0147] (4) Output defuzzification method: Center of Gravity.

[0148] Formula updating mechanism:

[0149]

[0150]

[0151] wherein, represents the updated proportional coefficient new value of the PI controller, represents the current proportional coefficient old value of the PI controller, represents the proportional coefficient correction amount determined by the center of gravity method; represents the updated integral coefficient new value of the PI controller, represents the current integral coefficient old value of the PI controller, represents the integral coefficient correction amount determined by the center of gravity method.

[0152] Through the above operations, the controller parameter adaptive setting for different disturbance amplitudes and change trends can be realized, and the dynamic stability under the harmonic disturbance and fluctuating load scene can be improved.

[0153] S43: According to the dominant factor, it is identified whether the power grid is in a special scene other than the normal working condition. If yes, it is switched to a more robust control strategy based on fuzzy rules.

[0154] It should be noted that when the power grid is in a special scene such as strong disturbance, high harmonic, high fluctuation, etc., a more robust control strategy such as sliding mode control (SMC) or model predictive control (MPC) mode is automatically determined and switched based on fuzzy rules. The switching mechanism is as follows:

[0155] (1) Judgment variable:

[0156]

[0157] (2) Strategy switching logic:

[0158] If and the high frequency harmonics are concentrated in the 6~10 frequency range: then switch to SVPWM+filter optimized carrier frequency (12~15kHz) modulation strategy;

[0159] If the voltage fluctuation rate ≥ 3%: then start the sliding mode current loop controller (SMC) to replace the PI control; If the grid impedance rises and the response is delayed, it is judged as a weak grid scenario: then start the MPC controller (model predictive control) to predict the target trajectory in advance.

[0160] (3) Fuzzy rule example (segment):

[0161]

[0162] All strategy switching is run through the embedded RTOS control system without human intervention.

[0163] In this embodiment, by self-adaptive setting of the control parameters, the real-time response ability of the control parameters to the disturbance is improved, and the under / over compensation caused by the fixed parameters is avoided; and the above control strategy switching mechanism enhances the robustness in the extreme grid environment, especially suitable for complex power distribution scenarios containing distributed energy and pulsating load. According to the experimental test, after the above switching mechanism is started, the system response time is shortened by 15%, the overmodulation amplitude is reduced by 30%, and the THD is reduced by 1.8% in total.

[0164] In an embodiment of the present application, it further comprises:

[0165] S51: Real-time monitoring of the change trend of the target power quality index (such as THD reduction, voltage recovery speed, etc.);

[0166] S52: If the change trend indicates that the control effect is not up to standard, trigger the control parameter re-tuning and control strategy re-selection;

[0167] S53: If the change trend indicates that the system reaches steady state, keep the current control parameters and control strategy running, and record the data of the current scene as training samples for each model.

[0168] In an embodiment of the present application, it further comprises: using a visual interface to display the current power quality state, control action history, prediction and warning information and equipment running state in real time. At the same time, all data support local storage and remote upload, which is convenient for cloud optimization and algorithm iteration.

[0169] The present application can not only be applied to photovoltaic, wind power and other new energy grid-connected systems, but also can be extended to electric vehicle charging stations, microgrids and other scenes with high requirements for power quality.

[0170] The present application can not only be applied to photovoltaic, wind power and other new energy grid-connected systems, but also can be extended to electric vehicle charging stations, microgrids and other scenes with high requirements for power quality.

[0171] Based on the same inventive concept, the embodiments of the present application also provide an analysis and suppression device for analyzing and suppressing the influencing factors of the power quality of the grid-connected inverter, which is used to implement the above-mentioned method for analyzing and suppressing the influencing factors of the power quality of the grid-connected inverter. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above-mentioned method, and therefore the specific limitations in the following analysis and suppression device embodiments for analyzing and suppressing the influencing factors of the power quality of the grid-connected inverter can be referred to the limitations of the analysis and suppression method for analyzing and suppressing the influencing factors of the power quality of the grid-connected inverter in the above, which will not be described here again.

[0172] Please refer to Figure 2 The embodiments of the present application also provide an analysis and suppression device for analyzing and suppressing the influencing factors of the power quality of the grid-connected inverter, which comprises:

[0173] A data acquisition module is configured to acquire electrical quantity data on the grid-connected side of the inverter and external disturbance data that will affect the power quality, and to calculate power quality characteristic parameters for measuring the power quality based on the electrical quantity data, and to form a multi-dimensional data matrix according to the acquired and calculated data;

[0174] A factor analysis module is configured to identify the dominant factor that reduces the power quality by using a pre-trained power quality multi-factor analysis model according to the multi-dimensional data matrix; the power quality multi-factor analysis model is based on a Bayesian network structure and is used to identify the factor that causes the reduction of the power quality by the causal relationship of the Bayesian network;

[0175] A quality prediction module is configured to predict the target power quality index value by using a pre-trained power quality dynamic prediction model according to the multi-dimensional data matrix; the power quality dynamic prediction model is based on a neural network structure and is used to mine the time series characteristics in the multi-dimensional data matrix and predict the target power quality index value;

[0176] A suppression module is configured to determine the control parameters and control strategies of the grid-connected inverter according to the dominant factor, to determine the control adjustment time of the grid-connected inverter according to the target power quality index value, and to control the grid-connected inverter based on the control adjustment time and the control parameters and control strategies, so as to suppress the reduction of the power quality.

[0177] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above functional units and modules is exemplified, and in actual application, the above functions can be completed by different functional units and modules according to needs, that is, the internal structure of the system is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit or module in the embodiment can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit. In addition, the specific names of each functional unit or module are only for the convenience of mutual distinction, and do not limit the protection scope of the present application. The specific working process of the unit or module in the system can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

[0178] With reference to Figure 3 The embodiment of the present application also provides a computer device, comprising a memory and a processor and a computer program stored in the memory, when the computer program is executed on the processor, the inverter grid-connected device power quality influencing factor analysis and suppression method is realized.

[0179] The computer device can be a desktop computer, a notebook computer, a palm computer and a cloud server and the like. The computer device can include, but is not limited to, a processor, a memory. Those skilled in the art can understand that, Figure 3 The computer device is only an example and does not constitute a limitation on the computer device, and can include more or fewer components than shown, or combine certain components, or different components, for example, can also include an input / output device, a network access device and the like.

[0180] The processor can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0181] The memory can be an internal storage unit of the computer device in some embodiments, such as a hard disk or a memory of the computer device. The memory can also be an external storage device of the computer device in other embodiments, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, and the like. Further, the memory can include both an internal storage unit and an external storage device of the computer device. The memory is used to store an operating system, an application program, a boot loader, data, and other programs, such as program codes of the computer program. The memory can also be used to temporarily store data that has been output or will be output.

[0182] The embodiment of the present application further provides a computer readable storage medium, which stores a computer program. The computer program is run by a processor to implement the influence factor analysis and suppression method of power quality of an inverter grid-connected device according to any one of the above methods.

[0183] In the embodiment, the integrated unit, if implemented in the form of a software functional unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the present application can implement all or part of the processes of the above-mentioned embodiment methods by a computer program to instruct related hardware, and the computer program can be stored in a computer readable storage medium. The computer program is executed by a processor to implement the steps of the above-mentioned various method embodiments. The computer program includes computer program codes, which can be in the form of source code, object code, an executable file, or some intermediate form, etc. The computer readable medium at least includes any entity or device capable of carrying the computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunications signal, and a software distribution medium. For example, a U disk, a mobile hard disk, a magnetic disk or an optical disk, etc. In some jurisdictions, according to legislation and patent practice, the computer readable medium can not be an electrical carrier signal and a telecommunications signal.

[0184] The embodiment of the present application provides a computer program product, which includes a computer program. The computer program is executed by a processor to implement the influence factor analysis and suppression method of power quality of an inverter grid-connected device according to any one of the above methods.

[0185] In the above embodiments, the description of each embodiment is focused on, and the part not described or recorded in a certain embodiment can be referred to the relevant description of other embodiments.

[0186] Those skilled in the art can understand that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software manner depends on the specific application and design constraints of the technical solutions. The skilled person 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.

[0187] In the embodiments disclosed in the present application, it should be understood that the disclosed apparatus / terminal device and method can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely schematic, for example, the division of the modules or units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed each other can be indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0188] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit the present application; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalent; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for analyzing and suppressing the influencing factors of power quality in inverter grid-connected equipment, characterized in that, Includes the following steps: Collect electrical quantity data from the grid-connected side of the inverter, as well as external disturbance data that may affect power quality, and calculate power quality characteristic parameters for measuring power quality based on the electrical quantity data. A multi-dimensional data matrix is ​​formed based on the collected and calculated data. Based on the multidimensional data matrix, the dominant factors that reduce power quality are identified using a pre-trained power quality multi-factor analysis model. The power quality multi-factor analysis model is based on a Bayesian network structure and is used to derive and identify factors that lead to a decrease in power quality through causal relationship deduction of the Bayesian network. Based on the multidimensional data matrix, the target power quality index value is predicted using a pre-trained power quality dynamic prediction model. The power quality dynamic prediction model is based on a neural network structure and is used to mine the time-series features in the multi-dimensional data matrix and predict the target power quality index value. The control parameters and control strategies of the grid-connected inverter are determined based on the dominant factors, and the control adjustment timing of the grid-connected inverter is determined based on the target power quality index value. The grid-connected inverter is controlled based on the control adjustment timing and the control parameters and control strategies to suppress the degradation of power quality. The pre-trained multi-factor power quality analysis model was used to identify the dominant factors that reduce power quality, including: An improved principal component analysis (PCA) method is used to reduce the dimensionality of a multidimensional data matrix, yielding several features after dimensionality reduction. During the dimensionality reduction process, the improved PCA method standardizes the data in the multidimensional data matrix according to the following formula: ； ； In the formula, This represents the result of standardizing the i-th sample and the j-th feature in the multidimensional data matrix. This represents the original value of the i-th sample and the j-th feature. This represents the mean of the original data for the j-th feature. This represents the weight of the j-th feature. This represents the standard deviation of the j-th feature; This represents the coefficient of variation for the j-th feature dimension. This represents the correlation coefficient between the j-th feature dimension and the target variable; The covariance matrix is ​​also updated according to the following formula: ； In the formula, Let represent the covariance matrix at time t. This represents the sliding window size for selecting data from the multidimensional data matrix. Standardized data representing time n; This represents the mean of the standardized data within the sliding window, and T represents the transpose operation; The principal component selection weights are also determined according to the following formula: ； In the formula, This represents the selection weight of the k-th principal component. This represents the eigenvalue of the k-th principal component. This represents the set of original variable indices contained in the k-th principal component. The degree of causal influence of the j-th original variable on the target power quality index; The reduced-dimensional features are input into a pre-trained Bayesian network to quantify the influence of each feature on the target power quality index and generate a factor contribution rate matrix. Based on the factor contribution rate matrix, the influencing factors are arranged in descending order of contribution, and the top m factors whose cumulative contribution exceeds a set threshold are selected as the dominant factors, where m is a set value.

2. The method for analyzing and suppressing the influencing factors of power quality in inverter grid-connected equipment according to claim 1, characterized in that, A multidimensional data matrix is ​​formed based on the collected and calculated data, including: The collected electrical quantity data, external disturbance data, and calculated power quality characteristic parameters are preprocessed to remove noise interference and abnormal measurement errors in the data. All preprocessed data are interpolated and aligned based on a unified time reference to obtain multi-type parameter data corresponding to the same time point. The various parameter data are stored in a data matrix indexed by timestamps, thereby forming a multidimensional data matrix.

3. The method for analyzing and suppressing the influencing factors of power quality in inverter grid-connected equipment according to claim 1, characterized in that, The neural network is a long short-term memory neural network. The input of the power quality dynamic prediction model is the data in the multidimensional data matrix, and the output is the predicted value of the target power quality index.

4. The method for analyzing and suppressing the influencing factors of power quality in inverter grid-connected equipment according to claim 1, characterized in that, The control parameters and control strategies of the grid-connected inverter are determined based on the aforementioned dominant factors, including: The controller parameters that need to be adjusted are determined based on the aforementioned dominant factors; The adjustment values ​​of the controller parameters that need to be adjusted are determined using pre-constructed fuzzy control rules; Based on the dominant factors, identify whether the power grid is in a special scenario outside of normal operating conditions. If so, switch to a more robust control strategy based on fuzzy rules.

5. The method for analyzing and suppressing the influencing factors of power quality in inverter grid-connected equipment according to claim 4, characterized in that, Also includes: Real-time monitoring of the changing trends of target power quality indicators; If the changing trend indicates that the control effect is not up to standard, then the control parameters will be retuned and the control strategy will be reselected. If the changing trend indicates that the system has reached a steady state, then the current control parameters and control strategy are maintained, and the data of the current scenario is recorded as training samples for each model.

6. The method for analyzing and suppressing the influencing factors of power quality in inverter grid-connected equipment according to claim 5, characterized in that, Also includes: It adopts a visual interface to display the current power quality status, control action history, prediction and early warning information, and equipment operating status in real time.

7. A device for analyzing and suppressing the influencing factors of power quality in inverter grid-connected equipment, characterized in that, include: The data acquisition module is used to collect electrical quantity data on the grid-connected side of the inverter, as well as external disturbance data that may affect power quality, and calculate power quality characteristic parameters for measuring power quality based on the electrical quantity data, and form a multi-dimensional data matrix based on the collected and calculated data. The factor analysis module is used to identify the dominant factors that reduce power quality based on the multidimensional data matrix and a pre-trained power quality multi-factor analysis model. The power quality multi-factor analysis model is based on a Bayesian network structure and is used to derive and identify factors that lead to a decrease in power quality through causal relationship deduction of the Bayesian network. The power quality prediction module is used to predict the target power quality index value based on the multidimensional data matrix and a pre-trained power quality dynamic prediction model. The power quality dynamic prediction model is based on a neural network structure and is used to mine the time-series features in the multi-dimensional data matrix and predict the target power quality index value. The suppression module is used to determine the control parameters and control strategy of the grid-connected inverter based on the dominant factors, and to determine the control adjustment timing of the grid-connected inverter based on the changes in the power quality status. Based on the control adjustment timing and the control parameters and control strategy, the grid-connected inverter is controlled to suppress the degradation of power quality. The pre-trained multi-factor power quality analysis model was used to identify the dominant factors that reduce power quality, including: An improved principal component analysis (PCA) method is used to reduce the dimensionality of a multidimensional data matrix, yielding several features after dimensionality reduction. During the dimensionality reduction process, the improved PCA method standardizes the data in the multidimensional data matrix according to the following formula: ； ； In the formula, This represents the result of standardizing the i-th sample and the j-th feature in the multidimensional data matrix. This represents the original value of the i-th sample and the j-th feature. This represents the mean of the original data for the j-th feature. This represents the weight of the j-th feature. This represents the standard deviation of the j-th feature; This represents the coefficient of variation for the j-th feature dimension. This represents the correlation coefficient between the j-th feature dimension and the target variable; The covariance matrix is ​​also updated according to the following formula: ； In the formula, Let represent the covariance matrix at time t. This represents the sliding window size for selecting data from the multidimensional data matrix. Standardized data representing time n; This represents the mean of the standardized data within the sliding window, and T represents the transpose operation; The principal component selection weights are also determined according to the following formula: ； In the formula, This represents the selection weight of the k-th principal component. This represents the eigenvalue of the k-th principal component. This represents the set of original variable indices contained in the k-th principal component. The degree of causal influence of the j-th original variable on the target power quality index; The reduced-dimensional features are input into a pre-trained Bayesian network to quantify the influence of each feature on the target power quality index and generate a factor contribution rate matrix. Based on the factor contribution rate matrix, the influencing factors are arranged in descending order of contribution, and the top m factors whose cumulative contribution exceeds a set threshold are selected as the dominant factors, where m is a set value.

8. A computer device, characterized in that, The device includes a processor and a memory: The memory is used to store computer programs and send the instructions of the computer programs to the processor; The processor executes, according to the instructions of the computer program, a method for analyzing and suppressing the influencing factors of power quality of inverter grid-connected equipment as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements a method for analyzing and suppressing the influencing factors of power quality of inverter grid-connected equipment as described in any one of claims 1-6.

Images