A method and system for intelligently controlling generator output

Abstract

The present invention discloses a method and system for intelligent control of generator output, which relates to the field of intelligent power control, including collecting generator operation data, sampling the generator output waveform, and extracting harmonic distortion data using fast Fourier transform; based on the generator operation data, predicting future changes through a long short-term memory network model to obtain load change trends; according to the load change trend, combined with harmonic distortion data, identifying the correlation between harmonic sources and loads, and generating a harmonic compensation scheme; based on the harmonic compensation scheme and the load change trend, generating a dual control strategy to adjust the generator operating parameters. The present invention generates a dual control strategy by combining the harmonic compensation scheme with the load change trend, thereby realizing dynamic adjustment of the generator output power, reactive power, excitation current and speed, and optimizing the generator operating parameters and energy efficiency.

Description

Technical Field

[0001] The present invention relates to the field of electric power intelligent control, and in particular to a method and system for intelligent control of generator output. Background Art

[0002] In recent years, with the continuous growth of global energy demand, generators, as key equipment in power systems, play a vital role in various power applications due to their stability and efficiency. During the operation of generators, problems such as voltage fluctuations, harmonic distortion, and load instability often occur. These problems not only affect the working efficiency of the generator, but may also pose a serious threat to the stability of the entire power system. With the rapid development of artificial intelligence and big data technology, more and more intelligent control methods have been proposed, especially in terms of harmonic distortion and load change trend prediction. New technical applications are gradually replacing traditional control methods. For example, using fast Fourier transform (FFT) to extract harmonic components in power systems, using deep learning models such as long short-term memory networks (LSTM) to predict future load change trends, and generating intelligent control schemes based on the analysis of the correlation between load and harmonic sources have become the mainstream direction of current research on intelligent control of generators.

[0003] However, existing technical solutions still have shortcomings in many aspects. Traditional harmonic compensation methods often rely on static mathematical models or fixed compensation strategies, and lack the ability to adapt to load changes and dynamic changes in harmonic sources in real time. Although modern technologies such as FFT can extract harmonic distortion data, many existing methods fail to fully consider the complex nonlinear relationship between harmonic sources and loads in practical applications, resulting in inaccurate compensation effects. Most of the existing prediction models only focus on a single load change trend, while ignoring the coordinated changes between harmonic distortion and load, and are unable to fully optimize the dynamic control needs of generators. In addition, many studies focus on optimizing a single control target, such as power output stability, and fail to combine harmonic compensation with load prediction to achieve a more comprehensive and efficient control strategy. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a method for intelligently controlling generator output to solve the problem that harmonic compensation and load change trend prediction cannot work together.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides a method for intelligently controlling generator output, which includes collecting generator operation data, sampling the generator output waveform, and extracting harmonic distortion data using fast Fourier transform; based on the generator operation data, predicting future changes through a long short-term memory network model to obtain load change trends; based on the load change trends, combined with harmonic distortion data, identifying the correlation between harmonic sources and loads, and generating a harmonic compensation plan; based on the harmonic compensation plan and the load change trend, generating a dual control strategy to adjust the generator operation parameters; The generator operation data is collected, the generator output waveform is sampled, and the harmonic distortion data is extracted using fast Fourier transform. The specific steps are: Install a sensor at the output end of the generator, set the sensor acquisition frequency, and collect the operation simulation signal; The operation analog signal is digitized through an analog-to-digital converter to obtain an operation digital signal; Use a bandpass filter to denoise the running digital signal; Perform normalization processing on the de-noised operating digital signal to obtain the generator operating data; Perform fast Fourier transform on the generator operation data, extract the amplitude, frequency and phase of the harmonics, and calculate the total harmonic distortion rate; Based on the amplitude, frequency, phase and total harmonic distortion of the harmonics, a harmonic characteristic matrix is ​​constructed to obtain the harmonic distortion data.

[0007] As a preferred solution of the generator output intelligent control method described in the present invention, the generator operation data includes voltage signals, current signals, frequency signals, engine speed signals and environmental operation parameters.

[0008] As a preferred solution of the method for intelligently controlling the output of a generator according to the present invention, the following specific steps are used to predict future changes based on the generator operation data and obtain the load change trend through a long short-term memory network model: The generator operation data is divided into a training set, a validation set and a test set, and a multi-dimensional feature matrix is ​​constructed using the training set; The empirical mode decomposition algorithm is used to decompose the multidimensional feature matrix into time series, extract the trend component, periodic component and random component, and the decomposition expression is: ; in, is a multidimensional feature matrix, is the trend component, is the periodic component, is the random component, is the number of intrinsic mode functions contained in the trend component, is the index of the number of eigenmode functions, For the The intrinsic mode functions, is the maximum independent variable point set, For the The energy of an eigenmode function, is the time variable; Based on the trend component, periodic component and random component, a three-layer long short-term memory network model is constructed to generate the load change trend at the next moment; According to the load change trend at the next moment, the input feature weight is dynamically adjusted, and the prediction performance of the long short-term memory network model is optimized through the feedback mechanism. The dynamic adjustment expression is: ; ; in, is the input feature weight after dynamic adjustment, is the index of the number of input features, For in time Time Input feature weights, is the learning rate, is the loss function Weighting the input features The gradient of is the loss function, is the time window length, For time The true value of For time The predicted value at time ; The adjusted LSTM network model was evaluated by the mean square error algorithm and combined with the validation set to further optimize the LSTM network model. The test set is input into the optimized long short-term memory network model to obtain the load change trend.

[0009] As a preferred solution of the method for intelligently controlling the output of a generator according to the present invention, the method comprises: identifying the correlation between harmonic sources and loads based on the load change trend and combining the harmonic distortion data to generate a harmonic compensation solution, and the specific steps are as follows: Based on the harmonic distortion data, a dynamic Bayesian network is established, and the expectation maximization algorithm is used to train the dynamic Bayesian network model to optimize the grid structure and conditional probability distribution; Based on the conditional probability distribution of the dynamic Bayesian network and combined with the load change trend, the correlation index between the harmonic source and the load is defined, and the expression is: ; in, For the The correlation index between the harmonic source and the load, is the harmonic index, For time Load change trend right The partial derivative of For time Load variation trend at For time Time harmonic eigenvalues, is the category index of the harmonic feature, for right The conditional probability distribution of In dynamic Bayesian networks The parent node set of is the integral identifier; Correlation identification is performed based on the correlation index between the harmonic source and the load, and a harmonic compensation scheme is generated based on the correlation identification result.

[0010] As a preferred solution of the method for intelligently controlling the output of a generator according to the present invention, the steps of performing correlation identification based on the correlation index between the harmonic source and the load and generating a harmonic compensation solution based on the correlation identification result are as follows: By setting the correlation threshold, the harmonics are divided into a set of classification results of high correlation, medium correlation and low correlation; According to the harmonic correlation classification results, harmonic intensity compensation is performed through a hierarchical compensation mechanism. The high correlation compensation expression is: ; in, For the The compensation strength of the high-correlation harmonics, is a high correlation harmonic index, is the second-order partial derivative of the load change trend with respect to the high-correlation harmonic eigenvalue, is the Gaussian kernel function, is the smoothing parameter, is the conditional probability distribution of high correlation harmonics, is the integral identifier; The correlation compensation expression is: ; in, For the The compensation strength of the relevant harmonics, is the medium correlation harmonic index, is the partial derivative of the load variation trend with respect to the harmonic eigenvalue of the correlation, is the hyperbolic tangent function, is the adjustment coefficient, is the conditional probability distribution of the harmonics with medium correlation, is the integral identifier; For low-correlation harmonics, compensation is ignored; The compensation strengths of high, medium and low correlation components are integrated into a harmonic compensation scheme.

[0011] As a preferred solution of the method for intelligently controlling the output of a generator according to the present invention, the dual control strategy is generated based on the harmonic compensation scheme and the load change trend to adjust the operating parameters of the generator. The specific steps are as follows: Integrate harmonic compensation schemes and load change trends into a unified data structure; Preprocess the unified data structure to generate a dual control strategy; According to the dual control strategy, the generator output power and generator reactive power are adjusted; According to the dual control strategy, the generator excitation current is adjusted and the generator speed is optimized.

[0012] In a second aspect, the present invention provides an intelligent control system for generator output, including an operation data extraction module, a load trend prediction module, a harmonic compensation generation module and a parameter dual control module; the operation data extraction module is used to collect generator operation data, sample the generator output waveform, and extract harmonic distortion data using a fast Fourier transform; the load trend prediction module is used to predict future changes based on the generator operation data through a long short-term memory network model to obtain load change trends; the harmonic compensation generation module is used to identify the correlation between harmonic sources and loads based on the load change trend and in combination with harmonic distortion data, and generate a harmonic compensation scheme; the parameter dual control module is used to generate a dual control strategy based on the harmonic compensation scheme and the load change trend to adjust the generator operation parameters.

[0013] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the method for intelligent control of generator output as described in the first aspect of the present invention is implemented.

[0014] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the method for intelligent control of generator output as described in the first aspect of the present invention is implemented.

[0015] The beneficial effects of the present invention are as follows: the present invention collects generator operation data and uses fast Fourier transform to extract harmonic distortion data, and combines the long short-term memory network model to predict the load change trend, thereby realizing high-precision monitoring and prediction of the generator operation status and load demand. Based on the harmonic distortion data and the load change trend, a dynamic Bayesian network is used to identify the correlation between the harmonic source and the load, and a harmonic compensation scheme is generated through a graded compensation mechanism, thereby effectively suppressing the influence of harmonics. Combining the harmonic compensation scheme with the load change trend, a dual control strategy is generated to realize dynamic adjustment of the generator output power, reactive power, excitation current and speed, and optimize the generator operation parameters and energy efficiency. The present invention not only improves the stability and power quality of the generator operation, but also enhances the adaptability to complex load scenarios, providing a new technical path for the realization of intelligent power systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0017] Figure 1 This is a flow chart of the intelligent control method for generator output in Example 1.

[0018] Figure 2 This is a module diagram of the generator output intelligent control system in Example 1. DETAILED DESCRIPTION

[0019] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0020] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0021] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0022] Example 1, reference Figure 1 and Figure 2, which is the first embodiment of the present invention, provides a method for intelligently controlling output of a generator, comprising the following steps: S1: Collect generator operation data, sample the generator output waveform, and extract harmonic distortion data using fast Fourier transform.

[0023] Specifically, the following steps are included: S1.1: Install a sensor at the output end of the generator, set the sensor acquisition frequency, and collect the operation simulation signal.

[0024] Specifically, a precision sensor assembly is installed at the output end of the generator, including a voltage sensor (for collecting output voltage signals), a current transformer (for collecting output current signals), a frequency meter (for monitoring real-time output frequency), a temperature sensor (for monitoring ambient temperature and internal temperature rise of the generator), a humidity sensor (for collecting ambient humidity data), and a vibration sensor (for collecting ambient vibration data). The collected data stream is a continuous analog signal, and the sampling frequency is set to =10kHz to meet the requirements of harmonic distortion analysis (the highest analysis frequency is , i.e. 5kHz).

[0025] Furthermore, the sampling frequency of 10kHz can cover the common harmonic frequency range of power lines (such as 3rd, 5th, 7th harmonics, etc.), and at the same time has sufficient sampling accuracy to ensure the effectiveness of subsequent FFT analysis.

[0026] S1.2: Digitally process the operation analog signal through an analog-to-digital converter to obtain an operation digital signal.

[0027] It should be understood that the collected analog signal is digitized by a high-precision analog-to-digital converter (ADC) to generate a discrete signal. The time interval of the discrete sampling points is .

[0028] S1.3: Use a bandpass filter to denoise the running digital signal.

[0029] Specifically, the passband range of the bandpass filter is set to 0-5 kHz.

[0030] S1.4: Normalize the denoised operating digital signal to obtain generator operating data.

[0031] Specifically, the minimum-maximum normalization algorithm is used to limit the amplitude of the denoised running digital signal to the range of 0 to 1, eliminating the dimensional differences between different signal sources and ensuring the accuracy of subsequent FFT calculations.

[0032] S1.4.1: Generator operating data includes voltage signals, current signals, frequency signals, engine speed signals and environmental operating parameters.

[0033] It should be noted that the environmental operating parameters include ambient temperature, internal operating temperature of the generator, ambient humidity and external vibration data.

[0034] S1.5: Perform fast Fourier transform on the generator operation data, extract the amplitude, frequency and phase of the harmonics, and calculate the total harmonic distortion rate.

[0035] Specifically, the voltage signal and current signal in the generator operation data are subjected to fast Fourier transform (FFT) to extract the frequency, phase, fundamental amplitude and harmonic amplitude of the harmonics. The total harmonic distortion rate is calculated based on the fundamental amplitude and harmonic amplitude, and the expression is: ; in, is the total harmonic distortion rate, is the highest harmonic order, is the harmonic order index coefficient, For the The amplitude of the subharmonics, is the fundamental amplitude ( ).

[0036] Furthermore, The value range is , where 0 means no harmonic distortion and 1 means full harmonic distortion.

[0037] S1.6: Based on the amplitude, frequency, phase and total harmonic distortion of the harmonics, a harmonic characteristic matrix is ​​constructed to obtain harmonic distortion data.

[0038] Specifically, the expression of the harmonic characteristic matrix is: ; in, is the harmonic characteristic matrix. Combined with the harmonic characteristic matrix and total harmonic distortion Generates harmonic distortion data, formatted in JSON format.

[0039] Preferably, the generator operation data is collected through precision sensors, and a 10kHz sampling rate is used to meet the requirements of harmonic analysis. High-precision ADC digital processing and bandpass filtering denoising are combined to ensure the integrity and purity of the collection. The minimum-maximum normalization algorithm is used to eliminate dimensional differences and improve the accuracy of the analysis. The amplitude, frequency and phase of the harmonics are extracted using fast Fourier transform, and the total harmonic distortion (THD) is calculated to quantify the degree of distortion. The constructed harmonic characteristic matrix expresses the characteristics of each harmonic. The harmonic characteristic matrix and THD are output in JSON format for easy storage and application, providing a reliable basis for optimizing generator performance.

[0040] S2: Based on the generator operation data, the long short-term memory network model is used to predict future changes and obtain the load change trend.

[0041] Specifically, the following steps are included: S2.1: Divide the generator operation data into training set, validation set and test set, and use the training set to construct a multi-dimensional feature matrix.

[0042] Specifically, the generator operation data is divided into training set, validation set and test set in a ratio of 7:1:2. The training set is used to train the long short-term memory network model; the validation set is used to adjust the parameters in the long short-term memory network model to prevent overfitting; the test set is used for the final performance evaluation of the long short-term memory network model to measure the generalization ability.

[0043] Each column of the multidimensional feature matrix is ​​a voltage signal, a current signal, a frequency signal, an engine speed signal, and an environmental operating parameter. Each row of the multidimensional feature matrix is ​​the operating data of each group of generators. The rows and columns are combined into a multidimensional feature matrix. .

[0044] S2.2: Using the empirical mode decomposition algorithm, the multidimensional feature matrix is ​​decomposed into time series to extract the trend component, periodic component and random component. The decomposition expression is: ; in, is a multidimensional feature matrix, is the trend component, is the periodic component, is the random component, is the number of intrinsic mode functions contained in the trend component, is the index of the number of eigenmode functions, For the The intrinsic mode functions, is the maximum independent variable point set, For the The energy of an eigenmode function, is the time variable.

[0045] Furthermore, The expression is: ; in, For the The energy of an eigenmode function, is the upper limit of the sampling time for the operating data of this group of generators, The lower bound of the sampling time for the group of generator operation data.

[0046] Preferably, through time series decomposition, complex time series signals are decomposed into different components to capture long-term trends, short-term periodic changes and random disturbances respectively, thereby improving the ability of the long short-term memory network model to capture features.

[0047] S2.3: Based on the trend component, periodic component and random component, a three-layer long short-term memory network model is constructed to generate the load change trend at the next moment.

[0048] It should be understood that a three-layer deep LSTM network is constructed, and each layer processes the trend component, the periodic component, and the random component to obtain multi-level features. Through hierarchical modeling, the characteristics of different components are processed separately, solving the problem that traditional LSTM only processes a single sequence, while improving the modeling ability of complex dynamic characteristics.

[0049] S2.4: According to the load change trend at the next moment, the input feature weight is dynamically adjusted, and the prediction performance of the long short-term memory network model is optimized through the feedback mechanism. The dynamic adjustment expression is: ; ; in, is the input feature weight after dynamic adjustment, is the index of the number of input features, For in time Time Input feature weights, is the learning rate, is the loss function Weighting the input features The gradient of is the loss function, is the time window length, For time The true value of For time The predicted value at time .

[0050] Specifically, based on the back propagation and gradient descent algorithms, through the feedback mechanism, the input feature weights are adjusted according to the prediction error at each moment, and the adjusted input feature weights are used to make predictions at the next moment. Through repeated adjustments and feedback, the long short-term memory network model is gradually optimized to achieve a more accurate load prediction effect.

[0051] S2.5: The adjusted LSTM network model is evaluated by the mean square error algorithm and combined with the validation set to further optimize the LSTM network model.

[0052] Specifically, input the validation set data to the adjusted LSTM network model, compare the output prediction value with the actual value in the validation set, and evaluate by calculating the mean square error (MSE). If the MSE of the LSTM network model is small, it means that the performance on the validation set is good, otherwise it means that the prediction error is large. When the MSE value is large, use grid search or random search method to adjust the hyperparameters of the LSTM network model for optimization. Hyperparameters include learning rate, number of neurons in the hidden layer, batch size, time step, and optimization algorithm. After multiple training and evaluations, select the hyperparameter combination with the optimal MSE to determine the final LSTM network model.

[0053] S2.6: Input the test set into the optimized long short-term memory network model to obtain the load change trend.

[0054] Specifically, the optimized LSTM network model is loaded into the test environment, the test set data is input, the predicted output is recorded, and the load change trend is obtained. The load change trend includes the generator power output and reactive power changes.

[0055] Preferably, the ability of the long short-term memory network model to capture and predict load changes is improved by combining data partitioning, feature matrix construction, empirical mode decomposition (EMD) and multi-layer LSTM networks. By decomposing the multidimensional feature matrix into trend, cycle and random components, it can better handle complex dynamic characteristics. Dynamically adjusting the input feature weights and using the mean square error algorithm for optimization further improves the prediction accuracy and generalization ability. The accuracy and reliability in practical applications are ensured through test set verification.

[0056] S3: Based on the load change trend and combined with the harmonic distortion data, the correlation between the harmonic source and the load is identified to generate a harmonic compensation plan.

[0057] Specifically, the following steps are included: S3.1: Based on the harmonic distortion data, a dynamic Bayesian network is established, and the expectation maximization algorithm is used to train the dynamic Bayesian network model to optimize the grid structure and conditional probability distribution.

[0058] Specifically, a dynamic Bayesian network (DBN) is established based on harmonic distortion data, and the network structure is defined and modeled. Each node of the DBN model represents a certain feature of the harmonic distortion data, and the node set includes total harmonic distortion, harmonic amplitude, frequency and phase. The edges of the DBN model represent the causal relationship between variables, and a directed graph is used to represent causal dependencies. The conditional probability distribution (CPD) of each node defines the probability distribution of the node given the parent node. For each node, its parent node is determined and its conditional probability distribution is established.

[0059] After building the preliminary DBN model, the DBN model is trained using the expectation maximization (EM) algorithm until the parameters converge or the maximum number of iterations is reached. The convergence criterion can be based on the change in log-likelihood values ​​or the prediction error.

[0060] Then, the grid structure is optimized through heuristic algorithms (greedy algorithms and simulated annealing, etc.) and cross-validation methods. The conditional probability distribution of each node is adjusted through the back-propagation algorithm to gradually reduce the prediction error.

[0061] S3.2: Based on the conditional probability distribution of the dynamic Bayesian network and combined with the load change trend, the correlation index between the harmonic source and the load is defined, and the expression is: ; in, For the The correlation index between the harmonic source and the load, is the harmonic index, For time Load change trend right The partial derivative of represents the sensitivity of the load to changes in harmonic characteristics, For time Load variation trend at For time Time harmonic eigenvalues, taken from the harmonic feature matrix in the harmonic distortion data, is the category index of the harmonic feature, for right The conditional probability distribution of In dynamic Bayesian networks The parent node set of Is the score identifier.

[0062] Preferably, combined with the conditional probability distribution and load sensitivity analysis , dynamically describes the causal relationship between the load and the harmonic characteristics. The correlation index can be used to evaluate the correlation between the harmonic source and the load, so that different harmonic compensation schemes can be generated according to the correlation index to deal with various complex situations.

[0063] S3.3: Perform correlation identification based on the correlation index between the harmonic source and the load, and generate a harmonic compensation plan based on the correlation identification result.

[0064] S3.3.1: By setting the correlation threshold, the harmonics are classified into a set of classification results with high correlation, medium correlation and low correlation.

[0065] Specifically, we set the correlation threshold and define three types of correlation levels: High correlation:

[0066] Medium correlation:

[0067] Low correlation:

[0068] in, is the high correlation threshold, is the low correlation threshold, through the correlation index Range , combined with the distribution characteristics to set the threshold, such as setting , .

[0069] S3.3.2: According to the harmonic correlation classification results, harmonic intensity compensation is performed through a graded compensation mechanism. The high correlation compensation expression is: ; in, For the The compensation strength of the high-correlation harmonics, is a high correlation harmonic index, is the second-order partial derivative of the load change trend with respect to the high-correlation harmonic eigenvalue, is the Gaussian kernel function, is the smoothing parameter, is the conditional probability distribution of high correlation harmonics, Is the score identifier.

[0070] Preferably, for high-correlation components, a high-correlation compensation formula is designed to maximize the compensation strength and ensure that the load harmonic distortion is minimized.

[0071] S3.3.3: The correlation compensation expression is: ; in, For the The compensation strength of the relevant harmonics, is the medium correlation harmonic index, is the partial derivative of the load variation trend with respect to the harmonic eigenvalue of the correlation, is the hyperbolic tangent function, is the adjustment coefficient, is the conditional probability distribution of the harmonics with medium correlation, Is the score identifier.

[0072] Preferably, for medium correlation components, the medium correlation compensation formula is designed to adjust the harmonic impact with moderate strength.

[0073] S3.3.4: For low correlation harmonics, ignore compensation.

[0074] S3.3.5: Integrate the compensation strengths of high, medium, and low correlation components into a harmonic compensation scheme.

[0075] Specifically, by weighted summing the compensation intensities of various components, the compensation intensities of high, medium and low correlation components are integrated into a harmonic compensation scheme.

[0076] Preferably, by combining harmonic distortion data with load change trends, a dynamic Bayesian network is used to accurately model the relationship between harmonic sources and loads, and the mutual influence between the two is flexibly analyzed through conditional probability distribution. By setting correlation thresholds, harmonics are divided into high, medium, and low correlation levels, and compensation schemes of different strengths are designed in a targeted manner to ensure the accuracy and effectiveness of harmonic compensation. Through a graded compensation mechanism, the impact of harmonic sources is optimized to ensure the power quality and working stability of the generator, while improving the intelligence and adaptability of the compensation scheme.

[0077] S4: Based on the harmonic compensation scheme and load change trend, a dual control strategy is generated to adjust the generator operating parameters.

[0078] Specifically, the following steps are included: S4.1: Integrate harmonic compensation schemes and load change trends into a unified data structure.

[0079] Specifically, the data in the harmonic compensation scheme and load change trend are formatted into a time series data table, arranged by time steps, and saved in JSON format.

[0080] S4.2: Preprocess the unified data structure and generate a dual control strategy.

[0081] Specifically, the load change trend data is smoothed and filtered to remove high-frequency noise to ensure that the extracted trend data is stable. The compensation intensity value in the harmonic compensation scheme is normalized to unify the value range for subsequent adjustment.

[0082] S4.3: Adjust the generator output power and generator reactive power according to the dual control strategy.

[0083] Specifically, the output power of the generator is adjusted according to the load change trend. By increasing the main valve opening, the output power of the generator is increased to meet the peak load demand. By reducing the main valve opening, the output power of the generator is reduced to avoid overload.

[0084] According to the load change trend, the reactive power output of the generator is adjusted by changing the excitation current of the engine to maintain the generator voltage stability. The reactive power output is increased by increasing the excitation current to support voltage stability. The reactive power output is reduced by reducing the excitation current to prevent overexcitation.

[0085] S4.4: According to the dual control strategy, adjust the generator excitation current and optimize the generator speed.

[0086] Specifically, the excitation current of the generator is adjusted to change the rotor magnetic field strength and adjust the compensation strength of the generator output voltage waveform. The compensation strength is improved by increasing the excitation current to suppress negative sequence harmonics. The compensation strength is reduced by reducing the excitation current or adjusting the grounding method of the neutral point to weaken the zero sequence harmonics. According to the frequency characteristics of the harmonics, the generator speed is adjusted to make the harmonic frequency of the output waveform deviate from the resonant frequency of the sensitive load. The adjustment range is controlled within the allowable speed variation range of the generator (rated speed ±5%).

[0087] Preferably, by combining the harmonic compensation scheme and the load change trend, a dual control strategy is generated to achieve fine-tuning of the generator operation. Unify and integrate the harmonic compensation requirements and load change trends to ensure data consistency and operability. Pre-process and smooth the load data, optimize the harmonic compensation value, and enhance the model's resistance to noise. Adjust the active and reactive power output of the generator to ensure the load demand while avoiding overload, and achieve precise suppression of harmonics by optimizing the excitation current regulation. Adjust the generator speed to avoid resonance of harmonic frequencies with sensitive loads, and improve the stability of the generator and the quality of power.

[0088] The present embodiment also provides a generator output intelligent control system, including: an operation data extraction module, which is used to collect generator operation data, sample the generator output waveform, and extract harmonic distortion data using fast Fourier transform; a load trend prediction module, which is used to predict future changes based on the generator operation data through a long short-term memory network model to obtain load change trends; a harmonic compensation generation module, which is used to identify the correlation between harmonic sources and loads based on the load change trend and in combination with harmonic distortion data, and generate a harmonic compensation plan; a parameter dual control module, which is used to generate a dual control strategy based on the harmonic compensation plan and the load change trend, and adjust the generator operation parameters.

[0089] This embodiment also provides a computer device, which is suitable for the case of the intelligent control method of generator output, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the intelligent control method of generator output proposed in the above embodiment.

[0090] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0091] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for realizing intelligent control of generator output as proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0092] In summary, the present invention achieves high-precision monitoring and prediction of the generator operating status and load demand by: collecting generator operating data and extracting harmonic distortion data using fast Fourier transform, and combining the long short-term memory network model to predict the load change trend. Based on the harmonic distortion data and the load change trend, a dynamic Bayesian network is used to identify the correlation between the harmonic source and the load, and a harmonic compensation scheme is generated through a hierarchical compensation mechanism, thereby effectively suppressing the influence of harmonics. Combining the harmonic compensation scheme with the load change trend, a dual control strategy is generated to achieve dynamic adjustment of the generator output power, reactive power, excitation current and speed, and optimize the generator operating parameters and energy efficiency. The present invention not only improves the stability and power quality of the generator operation, but also enhances the adaptability to complex load scenarios, providing a new technical path for the realization of intelligent power systems.

[0093] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for intelligently controlling generator output, characterized in that: include, Collect generator operation data, sample generator output waveform, and extract harmonic distortion data using fast Fourier transform; Based on the generator operation data, the long short-term memory network model is used to predict future changes and obtain the load change trend; According to the load change trend and combined with the harmonic distortion data, the correlation between the harmonic source and the load is identified to generate a harmonic compensation plan; Generate a dual control strategy based on the harmonic compensation scheme and load change trend to adjust the generator operating parameters; The generator operation data is collected, the generator output waveform is sampled, and the harmonic distortion data is extracted using fast Fourier transform. The specific steps are: Install a sensor at the output end of the generator, set the sensor acquisition frequency, and collect the operation simulation signal; The operation analog signal is digitized through an analog-to-digital converter to obtain an operation digital signal; Use a bandpass filter to denoise the running digital signal; Perform normalization processing on the de-noised operating digital signal to obtain the generator operating data; Perform fast Fourier transform on the generator operation data, extract the amplitude, frequency and phase of the harmonics, and calculate the total harmonic distortion rate; Based on the amplitude, frequency, phase and total harmonic distortion of the harmonics, a harmonic characteristic matrix is ​​constructed to obtain the harmonic distortion data.

2. The method for intelligently controlling generator output according to claim 1, characterized in that: The generator operation data includes a voltage signal, a current signal, a frequency signal, an engine speed signal and environmental operation parameters.

3. The method for intelligently controlling generator output according to claim 2, characterized in that: Based on the generator operation data, the future change prediction is performed through the long short-term memory network model to obtain the load change trend. The specific steps are: The generator operation data is divided into a training set, a validation set and a test set, and a multi-dimensional feature matrix is ​​constructed using the training set; The empirical mode decomposition algorithm is used to decompose the multidimensional feature matrix into time series, extract the trend component, periodic component and random component, and the decomposition expression is: ； in, is a multidimensional feature matrix, is the trend component, is the periodic component, is the random component, is the number of intrinsic mode functions contained in the trend component, is the index of the number of eigenmode functions, For the The intrinsic mode functions, is the maximum independent variable point set, For the The energy of an eigenmode function, is the time variable; Based on the trend component, periodic component and random component, a three-layer long short-term memory network model is constructed to generate the load change trend at the next moment; According to the load change trend at the next moment, the input feature weight is dynamically adjusted, and the prediction performance of the long short-term memory network model is optimized through the feedback mechanism. The dynamic adjustment expression is: ； ； in, is the input feature weight after dynamic adjustment, is the index of the number of input features, For in time Time Input feature weights, is the learning rate, is the loss function Weighting the input features The gradient of is the loss function, is the time window length, For time The true value of For time The predicted value at time ; The adjusted LSTM network model was evaluated by the mean square error algorithm and combined with the validation set to further optimize the LSTM network model. The test set is input into the optimized long short-term memory network model to obtain the load change trend.

4. The method for intelligently controlling generator output according to claim 3, characterized in that: The harmonic source and load correlation are identified based on the load change trend and combined with the harmonic distortion data to generate a harmonic compensation scheme. The specific steps are: Based on the harmonic distortion data, a dynamic Bayesian network is established, and the expectation maximization algorithm is used to train the dynamic Bayesian network model to optimize the grid structure and conditional probability distribution; Based on the conditional probability distribution of the dynamic Bayesian network and combined with the load change trend, the correlation index between the harmonic source and the load is defined, and the expression is: ； in, For the The correlation index between the harmonic source and the load, is the harmonic index, For time Load change trend right The partial derivative of For time Load variation trend at For time Time harmonic eigenvalues, is the category index of the harmonic feature, for right The conditional probability distribution of In dynamic Bayesian networks The parent node set of is the integral identifier; Correlation identification is performed based on the correlation index between the harmonic source and the load, and a harmonic compensation scheme is generated based on the correlation identification result.

5. The method for intelligently controlling generator output according to claim 4, characterized in that: The method of performing correlation identification based on the correlation index between the harmonic source and the load and generating a harmonic compensation scheme based on the correlation identification result is specifically as follows: By setting the correlation threshold, the harmonics are divided into a set of classification results of high correlation, medium correlation and low correlation; According to the harmonic correlation classification results, harmonic intensity compensation is performed through a hierarchical compensation mechanism. The high correlation compensation expression is: ； in, For the The compensation strength of the high-correlation harmonics, is a high correlation harmonic index, is the second-order partial derivative of the load change trend with respect to the high-correlation harmonic eigenvalue, is the Gaussian kernel function, is the smoothing parameter, is the conditional probability distribution of high correlation harmonics, is the integral identifier; The correlation compensation expression is: ； in, For the The compensation strength of the relevant harmonics, is the medium correlation harmonic index, is the partial derivative of the load variation trend with respect to the harmonic eigenvalue of the correlation, is the hyperbolic tangent function, is the adjustment coefficient, is the conditional probability distribution of the harmonics with medium correlation, is the integral identifier; For low-correlation harmonics, compensation is ignored; The compensation strengths of high, medium and low correlation components are integrated into a harmonic compensation scheme.

6. The method for intelligently controlling generator output according to claim 5, characterized in that: The dual control strategy is generated based on the harmonic compensation scheme and the load change trend to adjust the generator operating parameters. The specific steps are: Integrate harmonic compensation schemes and load change trends into a unified data structure; Preprocess the unified data structure to generate a dual control strategy; According to the dual control strategy, the generator output power and generator reactive power are adjusted; According to the dual control strategy, the generator excitation current is adjusted and the generator speed is optimized.

7. A generator output intelligent control system, based on the generator output intelligent control method according to any one of claims 1 to 6, characterized in that: Including operation data extraction module, load trend prediction module, harmonic compensation generation module and parameter dual control module; The operation data extraction module is used to collect the generator operation data, sample the generator output waveform, and extract the harmonic distortion data using fast Fourier transform; The load trend prediction module is used to predict future changes based on the generator operation data through a long short-term memory network model to obtain the load change trend; The harmonic compensation generation module is used to identify the correlation between harmonic sources and loads based on the load change trend and in combination with the harmonic distortion data, and to generate a harmonic compensation solution; The parameter dual control module is used to generate a dual control strategy based on the harmonic compensation scheme and the load change trend to adjust the generator operation parameters.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the generator output intelligent control method described in any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for intelligently controlling the output of a generator as described in any one of claims 1 to 6 are implemented.

Images