CN117639452B - Voltage compensation method, device and equipment of inverter and storage medium - Google Patents
Voltage compensation method, device and equipment of inverter and storage medium Download PDFInfo
- Publication number
- CN117639452B CN117639452B CN202410090315.3A CN202410090315A CN117639452B CN 117639452 B CN117639452 B CN 117639452B CN 202410090315 A CN202410090315 A CN 202410090315A CN 117639452 B CN117639452 B CN 117639452B
- Authority
- CN
- China
- Prior art keywords
- voltage
- inverter
- compensation
- parameter
- parallel
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 63
- 238000004458 analytical method Methods 0.000 claims abstract description 65
- 238000005457 optimization Methods 0.000 claims abstract description 35
- 239000003795 chemical substances by application Substances 0.000 claims description 110
- 239000002245 particle Substances 0.000 claims description 60
- 230000006870 function Effects 0.000 claims description 52
- 238000004364 calculation method Methods 0.000 claims description 45
- 239000011159 matrix material Substances 0.000 claims description 28
- 238000004422 calculation algorithm Methods 0.000 claims description 16
- 238000012544 monitoring process Methods 0.000 claims description 15
- 238000001514 detection method Methods 0.000 claims description 9
- 230000009977 dual effect Effects 0.000 claims description 9
- 230000002457 bidirectional effect Effects 0.000 claims description 8
- 230000004927 fusion Effects 0.000 claims description 8
- 238000013506 data mapping Methods 0.000 claims description 6
- 238000000354 decomposition reaction Methods 0.000 claims description 5
- 230000009466 transformation Effects 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 abstract description 3
- 230000008569 process Effects 0.000 description 25
- 238000012545 processing Methods 0.000 description 16
- 230000008859 change Effects 0.000 description 8
- 238000000638 solvent extraction Methods 0.000 description 7
- 238000010801 machine learning Methods 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 238000006243 chemical reaction Methods 0.000 description 4
- 238000009826 distribution Methods 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000007499 fusion processing Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 239000003990 capacitor Substances 0.000 description 1
- 238000011217 control strategy Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000004146 energy storage Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 238000010921 in-depth analysis Methods 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012067 mathematical method Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012731 temporal analysis Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000000700 time series analysis Methods 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
Abstract
The application relates to the technical field of deep learning, and discloses a voltage compensation method, device and equipment of an inverter and a storage medium. The method comprises the following steps: constructing a double-inverter control system and detecting first voltage waveform data of a first inverter and second voltage waveform data of a second inverter; performing voltage distortion identification and compensation prediction to obtain a first voltage distortion compensation parameter and a second voltage distortion compensation parameter; creating a parallel intelligent agent combination; performing parallel voltage execution compensation strategy analysis to obtain a first parallel voltage execution compensation strategy; performing voltage compensation strategy optimization to obtain a second parallel voltage execution compensation strategy; the application realizes intelligent voltage compensation of the inverter and improves the accuracy of the voltage compensation of the inverter.
Description
Technical Field
The present application relates to the field of deep learning technologies, and in particular, to a voltage compensation method, apparatus, device and storage medium for an inverter.
Background
In today's power systems, inverters are widely used in energy conversion and regulation, in particular in the fields of renewable energy and electric automobiles. However, the inverter is often affected by voltage distortion and real-time load variation during operation, resulting in reduced system performance and reduced energy conversion efficiency. In order to improve the stability and efficiency of inverter systems, researchers have come to focus on voltage compensation methods, especially dual inverter control systems, which can achieve bi-directional energy flow in different modes of operation.
However, problems remain in the current research, such as challenges in accurately identifying and compensating for voltage distortions, dynamic response to real-time load fluctuations, and the like. In addition, the performance of the current method under complex actual conditions needs to be further improved.
Disclosure of Invention
The application provides a voltage compensation method, device and equipment for an inverter and a storage medium.
In a first aspect, the present application provides a voltage compensation method of an inverter, the voltage compensation method of the inverter including:
Constructing a double-inverter control system through a first inverter and a second inverter, and detecting voltage waveforms of the double-inverter control system to obtain first voltage waveform data of the first inverter and second voltage waveform data of the second inverter;
Respectively carrying out voltage distortion identification and compensation prediction on the first voltage waveform data and the second voltage waveform data through a preset convolution long-short time model to obtain a first voltage distortion compensation parameter of the first inverter and a second voltage distortion compensation parameter of the second inverter;
creating a first intelligent agent of the first inverter and a second intelligent agent of the second inverter, and combining the first intelligent agent and the second intelligent agent in parallel connection relationship to obtain a parallel intelligent agent combination;
Acquiring target control parameter data of the double-inverter control system, and performing parallel voltage execution compensation strategy analysis on the parallel intelligent agent combination through the target control parameter data to obtain a first parallel voltage execution compensation strategy;
performing voltage compensation strategy optimization on the first parallel voltage execution compensation strategy according to the first voltage distortion compensation parameter and the second voltage distortion compensation parameter to obtain a second parallel voltage execution compensation strategy;
And carrying out real-time load monitoring on the double-inverter control system to obtain real-time system load data, and carrying out dynamic load compensation parameter analysis on the real-time system load data according to the second parallel voltage execution compensation strategy to obtain a dynamic load compensation parameter combination.
In a second aspect, the present application provides a voltage compensation device of an inverter, the voltage compensation device of the inverter comprising:
the detection module is used for constructing a double-inverter control system through the first inverter and the second inverter, and detecting voltage waveforms of the double-inverter control system to obtain first voltage waveform data of the first inverter and second voltage waveform data of the second inverter;
The prediction module is used for respectively carrying out voltage distortion identification and compensation prediction on the first voltage waveform data and the second voltage waveform data through a preset convolution long-short time model to obtain a first voltage distortion compensation parameter of the first inverter and a second voltage distortion compensation parameter of the second inverter;
The creation module is used for creating a first intelligent agent of the first inverter and a second intelligent agent of the second inverter, and carrying out parallel relation combination on the first intelligent agent and the second intelligent agent to obtain a parallel intelligent agent combination;
The acquisition module is used for acquiring target control parameter data of the double-inverter control system, and performing parallel voltage compensation strategy analysis on the parallel intelligent agent combination through the target control parameter data to obtain a first parallel voltage compensation strategy;
The optimization module is used for performing voltage compensation strategy optimization on the first parallel voltage execution compensation strategy according to the first voltage distortion compensation parameter and the second voltage distortion compensation parameter to obtain a second parallel voltage execution compensation strategy;
And the analysis module is used for carrying out real-time load monitoring on the double-inverter control system to obtain real-time system load data, and carrying out dynamic load compensation parameter analysis on the real-time system load data according to the second parallel voltage execution compensation strategy to obtain a dynamic load compensation parameter combination.
A third aspect of the present application provides a computer apparatus comprising: a memory and at least one processor, the memory having instructions stored therein; the at least one processor invokes the instructions in the memory to cause the computer device to perform the voltage compensation method of the inverter described above.
A fourth aspect of the present application provides a computer readable storage medium having instructions stored therein which, when run on a computer, cause the computer to perform the voltage compensation method of an inverter described above.
According to the technical scheme provided by the application, by the preset convolution long-short time model, the method can accurately identify the distortion characteristics in the voltage waveform, and predict and compensate the distortions in real time, so that the quality and stability of the output voltage of the inverter are improved. By creating a parallel agent combination, a cooperative operation of the two inverters is achieved. By means of parallel analysis, the voltage execution compensation strategy can be optimized, and the efficiency and performance of the whole double-inverter system are improved. Real-time load monitoring and dynamic load compensation parameter analysis are introduced, so that the system can respond to load fluctuation in time and maintain stable output voltage. This is important to cope with the case where the load change is large in practical use. By optimizing the voltage execution compensation strategy, particularly adopting strategy fusion and particle association generation, the system can select the optimal voltage compensation strategy under different operation conditions, and the self-adaptability and the robustness of the system are improved. The system can know the current working state in real time by introducing real-time system load monitoring and dynamic load compensation parameter analysis, so that the compensation parameters can be adjusted more flexibly, the stability and performance of the system under different load conditions are ensured, the intelligent voltage compensation of the inverter is realized, and the accuracy of the voltage compensation of the inverter is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings may be obtained based on these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic diagram of an embodiment of a voltage compensation method of an inverter according to an embodiment of the present application;
Fig. 2 is a schematic diagram of an embodiment of a voltage compensation device of an inverter according to an embodiment of the application.
Detailed Description
The embodiment of the application provides a voltage compensation method, device and equipment of an inverter and a storage medium. The terms "first," "second," "third," "fourth" and the like in the description and in the claims and in the above drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments described herein may be implemented in other sequences than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus.
For easy understanding, the following describes a specific flow of an embodiment of the present application, referring to fig. 1, and one embodiment of a voltage compensation method of an inverter in an embodiment of the present application includes:
Step 101, constructing a double-inverter control system through a first inverter and a second inverter, and detecting voltage waveforms of the double-inverter control system to obtain first voltage waveform data of the first inverter and second voltage waveform data of the second inverter;
It is to be understood that the execution body of the present application may be a voltage compensation device of an inverter, and may also be a terminal or a server, which is not limited herein. The embodiment of the application is described by taking a server as an execution main body as an example.
Specifically, a double-inverter control system corresponding to the first inverter and the second inverter is constructed in a parallel connection mode, and the stability and the output power of the system can be effectively improved in the parallel connection mode, wherein the first inverter is responsible for converting direct current into alternating current so as to meet the requirements of a power grid or a load, and the second inverter is responsible for converting the alternating current back into direct current so as to be used for energy storage or other direct current application. After the construction is completed, the voltage waveform detection is carried out on the double-inverter control system, the voltage signals at the output end of the inverter are collected and converted into a voltage waveform data set which can be digitally processed, and the data set contains important information such as amplitude, frequency and phase of voltage, which is the basis for the establishment of a subsequent compensation strategy. In order to accurately identify and classify various characteristics and anomalies in the voltage waveforms, a preset threshold clustering algorithm is adopted to process an original voltage waveform data set, classification threshold calculation is carried out on the data set through an initial threshold calculation function, and classification thresholds can be dynamically determined according to the characteristics of the voltage waveforms. The obtained initial classification threshold is the initial recognition of the voltage waveform characteristics, and in order to further improve the accuracy and the efficiency of classification, the original voltage waveform data set is subjected to data set division according to the initial result, so that an initial data set division result is obtained. And carrying out parameter adjustment on the initial threshold calculation function according to the initial data set dividing result to obtain a target threshold calculation function, so that the algorithm is better adapted to the actual voltage waveform characteristics. And carrying out classification threshold calculation on the original voltage waveform data set through a target threshold calculation function to obtain a corresponding target classification threshold. And carrying out final data division on the original voltage waveform data set through the target classification threshold value to obtain a target data set division result. This result reflects the classification state of the voltage waveform data after the fine processing, thereby determining the first voltage waveform data of the first inverter and the second voltage waveform data of the second inverter.
102, Respectively carrying out voltage distortion identification and compensation prediction on the first voltage waveform data and the second voltage waveform data through a preset convolution long-short time model to obtain a first voltage distortion compensation parameter of a first inverter and a second voltage distortion compensation parameter of a second inverter;
Specifically, a preset convolution long-short time model is called, and the model is a core of voltage distortion identification and compensation parameter prediction and comprises three main parts of a convolution long-short time network, a parameter prediction network and a parameter discrimination network. The convolution long-short time network is specially used for processing time series data, and combines the space feature extraction capability of the convolution neural network and the time series analysis capability of the long-short time memory network, so that the model can accurately extract the features of voltage distortion from the voltage waveform data. The voltage waveform data output by the first inverter and the second inverter are processed through a convolution long-short time network, and a plurality of first voltage distortion characteristics and second voltage distortion characteristics are respectively extracted, wherein the characteristics comprise amplitude, frequency, phase, harmonic components and the like of the voltage waveform. The extracted voltage distortion characteristics are sent to a parameter prediction network, and the network uses an advanced machine learning algorithm to calculate corresponding original distortion compensation parameters according to the input distortion characteristics. For the first inverter, the network will output a series of first raw distortion compensation parameters; for the second inverter, a series of second original distortion compensation parameters are output. These parameters are preliminary results of voltage compensation strategy formulation that reflect the adjustments that need to be made to counteract the voltage distortion. To ensure the accuracy and validity of these compensation parameters, the compensation parameters are checked and adjusted. And checking the first original distortion compensation parameter and the second original distortion compensation parameter output by the parameter prediction network through the parameter discrimination network. The parameter discrimination network typically uses a set of predefined criteria or empirical rules to evaluate whether the compensation parameters are reasonable and effective to correct the voltage distortion. After the checksum adjustment of the parameter discrimination network, the first voltage distortion compensation parameter of the first inverter and the second voltage distortion compensation parameter of the second inverter are finally obtained, and the parameters are subjected to multiple verification, so that compensation measures required for achieving an ideal voltage waveform can be more accurately reflected.
Step 103, creating a first intelligent agent of the first inverter and a second intelligent agent of the second inverter, and combining the first intelligent agent and the second intelligent agent in parallel connection to obtain a parallel intelligent agent combination;
Specifically, a first agent of the first inverter and a second agent of the second inverter are created, respectively. The first agent consists of three key parts: a first input layer, a first execution compensation policy analysis network, and a first output layer. The first input layer acts as a single hot vector encoding layer, its primary function being to convert the input data into a format that can be more efficiently processed and understood by the subsequent network. Single-hot encoding is a common data processing mode, and can ensure the definition and consistency of model input. The first implementation compensation policy analysis network includes a bi-directional threshold cycle network and a uni-directional threshold cycle network. The bi-directional threshold cycle network is capable of processing both forward and reverse data streams, which makes it more efficient in understanding and analyzing time series data, as it can consider both the context of the current data point. The unidirectional threshold cycle network is more focused on the front-to-back data flow, which is also an efficient way to process time series data. The two networks are combined, so that the voltage compensation strategy can be more comprehensively analyzed and processed, and the generated strategy is ensured to be comprehensive and accurate. The first output layer employs a softmax function. The softmax function is a function that normalizes a vector to a probability distribution, the output of which can be interpreted as a series of probability values, each value corresponding to the degree of confidence of an agent in a particular decision or classification. The second agent is similar in structure to the first agent and includes a second input layer (single thermal vector encoding layer), a second implementation compensation strategy analysis network (including a bi-directional threshold cycle network and a uni-directional threshold cycle network), and a second output layer (softmax function). This design ensures that the two agents are consistent and comparable in processing their respective data and tasks. And combining the first intelligent agent and the second intelligent agent in parallel connection to obtain a parallel intelligent agent combination. In the parallel combination process, the outputs of the two agents are comprehensively considered to ensure that the compensation strategy is not only applicable to a single inverter, but also can exert maximum efficiency when the two inverters work cooperatively. Through the parallel relation combination, the advantages of the two intelligent bodies can be fully utilized, the defects of the two intelligent bodies are overcome, the finally generated voltage compensation strategy is more comprehensive and effective, and the stable operation and the excellent performance of the inverter system are ensured.
104, Acquiring target control parameter data of a double-inverter control system, and performing parallel voltage compensation strategy analysis on the parallel intelligent agent combination through the target control parameter data to obtain a first parallel voltage compensation strategy;
Specifically, target control parameter data of a dual inverter control system is obtained, and the data includes switching time of the inverter and various operation parameters such as frequency, phase and amplitude, and the like, which directly affect the output voltage characteristics and performance of the inverter. After these data are acquired, they are input into a parallel combination of agents, which includes two agents, each responsible for analyzing and predicting the voltage to implement the compensation strategy. In the parallel intelligent agent combination, voltage execution compensation strategy analysis is performed on target control parameter data through a first intelligent agent, and complex characteristics of input data can be understood and interpreted by the first intelligent agent through a neural network structure in the first intelligent agent, such as a bidirectional threshold circulation network and a unidirectional threshold circulation network and other machine learning models, so that a set of voltage execution compensation strategy suitable for a first inverter is predicted, and the set of strategy is called first prediction voltage execution compensation strategy. Similarly, the second agent also analyzes the same target control parameter data to obtain a second predicted voltage to implement a compensation strategy. The two prediction strategies are fused. By comparing, analyzing and integrating the contents and effects of the two sets of prediction strategies, the optimal combination point and complementarity between the two sets of strategies are discovered, and a set of comprehensive voltage executing compensation strategy which can meet the requirements of the first inverter and adapt to the characteristics of the second inverter, namely the first parallel voltage executing compensation strategy, is generated. The comprehensive strategy considers the overall performance and efficiency of the double-inverter system, and ensures the coordination and consistency of each inverter during parallel operation while ensuring the independent operation of each inverter.
Step 105, performing voltage compensation strategy optimization on the first parallel voltage execution compensation strategy according to the first voltage distortion compensation parameter and the second voltage distortion compensation parameter to obtain a second parallel voltage execution compensation strategy;
Specifically, the first and second voltage distortion compensation parameters are subjected to voltage distortion compensation interval prediction, and the compensation range required by each inverter is determined by analyzing and processing the voltage distortion compensation parameters, so that the first and second voltage distortion compensation interval ranges are obtained. These interval ranges provide the basis and constraints for subsequent policy optimization, ensuring that the optimization process proceeds within a viable and efficient range. And setting a strategy optimization function for the two voltage distortion compensation interval ranges. The policy optimization function is a mathematical model specifically designed to precisely adjust the parallel voltage execution compensation policy, which takes into account various factors and constraints of voltage distortion compensation, such as voltage stability, system efficiency, safety, and the like. Through the function, the voltage distortion compensation interval range is converted into a specific compensation strategy adjustment instruction, and criteria and directions are provided for generating the target particle population. And executing a compensation strategy on the first parallel voltage through a target strategy optimization function to perform particle association generation to obtain a target particle population. In this process, a particle swarm optimization algorithm is used to generate and optimize a series of compensation strategies that are considered as particles and are searched and updated in the solution space. The target particle population includes a plurality of sub-particle populations, each sub-particle population representing a particular combination of compensation strategies. And carrying out fitness calculation on each sub-particle population to obtain a particle fitness set. The fitness calculation evaluates the performance and effectiveness of each sub-particle population according to preset evaluation criteria, ensuring that the selected compensation strategy not only can effectively reduce voltage distortion, but also meets other system requirements such as stability and efficiency. By fitness calculation, it can be determined which sub-particle populations are more optimal. And carrying out iterative computation on the particle fitness set until a preset condition is met, so as to generate an optimal solution corresponding to the target particle population. This process is dynamic and adaptive and the system will continually adjust and optimize the position and velocity of the particles until the best compensation strategy combination is found. And further optimizing the first parallel voltage execution compensation strategy through the obtained optimal solution to obtain a second parallel voltage execution compensation strategy. The optimized strategy is based on the results of in-depth analysis and accurate adjustment of the first and second voltage distortion compensation parameters, so that voltage distortion can be more effectively compensated, and the overall performance and stability of the inverter system can be improved.
And 106, carrying out real-time load monitoring on the double-inverter control system to obtain real-time system load data, and carrying out dynamic load compensation parameter analysis on the real-time system load data according to the second parallel voltage execution compensation strategy to obtain a dynamic load compensation parameter combination.
Specifically, the dual inverter control system is monitored in real time, and data about current, voltage, power, etc. are collected in real time using sensors and a data acquisition system. And carrying out linear transformation and covariance matrix calculation on the real-time system load data, and extracting deeper statistical features from the original data to obtain a target covariance matrix. The covariance matrix reflects the relation and the change modes among different load parameters, and is a key for understanding the load dynamics of the system. And carrying out eigenvalue decomposition on the target covariance matrix, and extracting key information from the covariance matrix to obtain a plurality of eigenvalues and corresponding eigenvectors. These eigenvalues and eigenvectors describe the primary direction and pattern of change of the data, helping to understand the complexity and diversity of the system load. On the basis, feature screening and projection matrix construction are carried out on the feature values, the dimension of data is reduced by selecting the most representative feature values, and the most critical information is reserved, so that a data projection matrix for data mapping is obtained. And carrying out data mapping on the real-time system load data according to the data projection matrix, and converting the original high-dimensional data into a lower-dimensional space to obtain standardized system load data. This normalized data is more convenient to analyze and process. And executing a compensation strategy according to the second parallel voltage, and carrying out dynamic load compensation parameter analysis on the standardized system load data. The most suitable load compensation parameters are dynamically calculated from the current load conditions and compensation strategies by algorithms and models, such as machine learning or optimization algorithms. A dynamic load compensation parameter combination is obtained.
In the embodiment of the application, the distortion characteristics in the voltage waveform can be accurately identified through the preset convolution long-short time model, and the distortions are predicted and compensated in real time, so that the quality and stability of the output voltage of the inverter are improved. By creating a parallel agent combination, a cooperative operation of the two inverters is achieved. By means of parallel analysis, the voltage execution compensation strategy can be optimized, and the efficiency and performance of the whole double-inverter system are improved. Real-time load monitoring and dynamic load compensation parameter analysis are introduced, so that the system can respond to load fluctuation in time and maintain stable output voltage. This is important to cope with the case where the load change is large in practical use. By optimizing the voltage execution compensation strategy, particularly adopting strategy fusion and particle association generation, the system can select the optimal voltage compensation strategy under different operation conditions, and the self-adaptability and the robustness of the system are improved. The system can know the current working state in real time by introducing real-time system load monitoring and dynamic load compensation parameter analysis, so that the compensation parameters can be adjusted more flexibly, the stability and performance of the system under different load conditions are ensured, the intelligent voltage compensation of the inverter is realized, and the accuracy of the voltage compensation of the inverter is improved.
In a specific embodiment, the process of executing step 101 may specifically include the following steps:
(1) A double-inverter control system corresponding to a first inverter and a second inverter is constructed in a parallel connection mode, wherein the first inverter is used for converting direct current into alternating current, and the second inverter is used for converting the alternating current into direct current;
(2) Detecting voltage waveforms of the double-inverter control system to obtain an original voltage waveform data set;
(3) Inputting the original voltage waveform data set into a preset threshold clustering algorithm, and performing classification threshold calculation on the original voltage waveform data set through an initial threshold calculation function in the threshold clustering algorithm to obtain a corresponding initial classification threshold;
(4) Carrying out data set division on the original voltage waveform data set and the initial classification threshold value to obtain an initial data set division result;
(5) Parameter adjustment is carried out on the initial threshold calculation function according to the initial data set dividing result, and a target threshold calculation function is obtained;
(6) Classifying threshold calculation is carried out on the original voltage waveform data set through a target threshold calculation function to obtain a corresponding target classifying threshold, and data division is carried out on the original voltage waveform data set through the target classifying threshold to obtain a target data set division result;
(7) And determining first voltage waveform data of the first inverter and second voltage waveform data of the second inverter according to the target data set division result.
Specifically, a parallel connection mode is adopted to construct a double-inverter control system corresponding to the first inverter and the second inverter, the first inverter is designed to convert direct current into alternating current, and the second inverter is responsible for converting the alternating current into direct current, so that the design can meet diversified electric energy conversion requirements, and the flexibility and the adaptability of the system are improved. With this parallel configuration, the two inverters can work together to handle the power conversion task in a more efficient and stable manner. And voltage waveform detection is carried out on the double-inverter control system through a voltage sensor and a data acquisition module which are arranged at the output end of the inverter so as to acquire an original voltage waveform data set. Voltage waveform detection is the basis for the overall voltage compensation process and provides detailed information about the inverter output voltage, including amplitude, frequency, phase, waveform distortion, etc. of the voltage. For example, if the inverter is connected to the power grid, voltage waveform detection may help identify voltage quality issues with the power grid, such as voltage fluctuations, flicker, or harmonic distortion. The original voltage waveform dataset is input into a preset threshold clustering algorithm. And performing classification threshold calculation on the original voltage waveform data set through a predefined initial threshold calculation function to obtain a corresponding initial classification threshold. The threshold clustering algorithm is an unsupervised learning algorithm that can automatically classify data into different categories based on its characteristics. Such an algorithm may help the system distinguish between normal and distorted voltage waveforms, as well as different types of distortion. For example, voltage waveforms are classified into sine wave, saw tooth wave, square wave, etc., each corresponding to a different processing and compensation strategy. And carrying out data set division on the original voltage waveform data set and the initial classification threshold value to obtain an initial data set division result. The raw data is divided into different categories according to an initial classification threshold, each category containing voltage waveform data having similar characteristics. For example, voltage waveforms with all amplitudes within a certain range may be divided into one category, while voltage waveforms with frequencies within another certain range are divided into another category. This partitioning of the data facilitates subsequent analysis and processing, reduces the complexity of the data, and makes the problem more definite. And according to the initial data set dividing result, carrying out parameter adjustment on the initial threshold calculation function to obtain a target threshold calculation function, so that the actual distribution and characteristics of the data can be reflected more accurately. For example, if the initial data partitioning result shows that a certain class of voltage waveforms is too concentrated or scattered, the parameters of the threshold calculation function may be adjusted so that the class of voltage waveforms is more evenly distributed in the future data partitioning. This parameter adjustment is an iterative process that requires constant adjustment and testing until a threshold calculation function is found that yields satisfactory data partitioning results. And carrying out classification threshold calculation on the original voltage waveform data set through a target threshold calculation function to obtain a corresponding target classification threshold, and carrying out data division on the original voltage waveform data set through the target classification threshold to obtain a target data set division result. The resulting target data set partitioning results will reflect the latest and most accurate classification status of the voltage waveforms. And determining first voltage waveform data of the first inverter and second voltage waveform data of the second inverter according to the target data set division result. The target data set partitioning result is applied to the actual inverter control system to select the appropriate voltage compensation strategy and parameters. For example, if the target data set partitioning results show that there is severe harmonic distortion in the output voltage of a certain inverter, the system may select a compensation strategy for the harmonic distortion and calculate the corresponding compensation parameters.
In a specific embodiment, the process of executing step 102 may specifically include the following steps:
(1) And calling a preset convolution long-short time model, wherein the convolution long-short time model comprises the following steps: a convolution long-short time network, a parameter prediction network and a parameter discrimination network;
(2) Extracting voltage distortion characteristics of the first voltage waveform data through a convolution long-short time network to obtain a plurality of first voltage distortion characteristics of the first voltage waveform data, and extracting voltage distortion characteristics of the second voltage waveform data through the convolution long-short time network to obtain a plurality of second voltage distortion characteristics of the second voltage waveform data;
(3) Performing compensation parameter calculation on a plurality of first voltage distortion characteristics of the first voltage waveform data through a parameter prediction network to obtain first original distortion compensation parameters of the first inverter, and performing compensation parameter calculation on a plurality of second voltage distortion characteristics of the second voltage waveform data through the parameter prediction network to obtain second original distortion compensation parameters of the second inverter;
(4) And respectively carrying out compensation parameter verification on the first original distortion compensation parameter of the first inverter and the second original distortion compensation parameter of the second inverter through a parameter discrimination network to obtain a first voltage distortion compensation parameter of the first inverter and a second voltage distortion compensation parameter of the second inverter.
Specifically, three main components of a convolution long-short time model are constructed: a convolution long-short time network, a parameter prediction network and a parameter discrimination network. A convolutional long-short time network is a hybrid network that combines a Convolutional Neural Network (CNN) and a long-short time memory network (LSTM) to efficiently process and analyze time-series data, particularly data having complex time-dependence and spatial characteristics, such as voltage waveforms. In this network, CNN is responsible for extracting spatial features of the voltage waveform, such as amplitude, frequency and phase, while LSTM handles the long-term dependence of time series data, ensuring that the model can understand and remember the changes in the voltage waveform over time. The first voltage waveform data output by the first inverter is processed through a convolution long-short time network, and a series of first voltage distortion characteristics are extracted from the first voltage waveform data. These features include basic features of the voltage waveform such as peaks, averages and standard deviations, as well as more complex features such as harmonic components, flicker and asymmetry. And similarly, processing the second voltage waveform data output by the second inverter, and extracting corresponding second voltage distortion characteristics. And analyzing and processing the voltage distortion characteristics through a parameter prediction network, and calculating corresponding compensation parameters. The parameter prediction network is a neural network specifically designed to predict output parameters based on input characteristics, and may be a fully connected network, a convolutional network, or other type of network. In this network, the first voltage distortion characteristic will be used to calculate a first original distortion compensation parameter for the first inverter, while the second voltage distortion characteristic is used to calculate a second original distortion compensation parameter for the second inverter. These raw distortion compensation parameters are calculated from the voltage distortion characteristics and the operating conditions of the grid and represent the measures that need to be taken to counteract the voltage distortion, such as adjusting the switching time of the inverter, changing the amplitude or phase of the output voltage, etc. However, these raw distortion compensation parameters are not optimal nor meet other requirements of the system, such as stability, efficiency, safety, and the like. Therefore, further checksum optimization of these parameters is required. The parameter discrimination network is a neural network that evaluates and verifies output parameters, including a classification network, a regression network, or other type of network. The first and second original distortion compensation parameters are input into a parameter discrimination network, respectively, and these parameters are evaluated according to predefined criteria and rules. If the parameters meet all the requirements, they will be accepted as final compensation parameters; if not, the system may need to adjust the parameters, or recalculate. In this way, it is ensured that the resulting compensation parameters are not only effective in counteracting voltage distortions, but also meet other requirements.
In a specific embodiment, the process of executing step 103 may specifically include the following steps:
(1) Creating a first intelligent agent of a first inverter, wherein the first intelligent agent comprises a first input layer, a first execution compensation strategy analysis network and a first output layer, the first input layer is a single-heat vector coding layer, the first execution compensation strategy analysis network comprises a bidirectional threshold circulation network and a unidirectional threshold circulation network, and the first output layer is a softmax function;
(2) Creating a second intelligent agent of a second inverter, wherein the second intelligent agent comprises a second input layer, a second execution compensation strategy analysis network and a second output layer, the second input layer is a single-heat vector coding layer, the second execution compensation strategy analysis network comprises a bidirectional threshold circulation network and a unidirectional threshold circulation network, and the second output layer is a softmax function;
(3) And carrying out parallel relation combination on the first intelligent agent and the second intelligent agent to obtain a parallel intelligent agent combination.
Specifically, the first agent is composed of three main parts: a first input layer, a first execution compensation policy analysis network, and a first output layer. The first input layer acts as a single thermal vector coding layer and functions to convert raw input data, such as voltage distortion characteristics or operating parameters, into a form that can be efficiently processed by subsequent networks. Single-hot encoding is a common data processing approach, often used in machine learning and pattern recognition, that can convert classification variables into a format more suitable for algorithmic model processing. For example, different operating states of the inverter or different conditions of the grid may be converted into a single-heat encoded form for further analysis. The first implementation compensation policy analysis network includes a bi-directional threshold cycle network and a uni-directional threshold cycle network. Both networks are tools for processing time series data in deep learning. The bi-directional threshold cycle network is capable of processing both past and future information, and is suitable for voltage compensation tasks where context information needs to be considered. For example, in analyzing voltage waveforms, the bi-directional network may take into account both previous and subsequent voltage changes, thereby more accurately predicting future voltage waveforms and determining compensation strategies. The unidirectional threshold cycle network focuses on the past to future information flow, and is suitable for gradually analyzing time series data, such as gradually analyzing each period of voltage waveform and adjusting the compensation strategy in real time. The first output layer uses a softmax function, which is a commonly used activation function that converts a real vector into a probability distribution. In this embodiment, the softmax function may convert the output of the executing compensation strategy analysis network into a series of probability values, each value representing the trustworthiness of the corresponding compensation strategy. The second agent is similar in structure to the first agent and also includes a second input layer, a second implementation compensation strategy analysis network, and a second output layer. The second input layer adopts a single thermal vector coding layer to convert the data related to the second inverter into a proper form. The second implementation compensation strategy analysis network comprises a bidirectional threshold cycle network and a unidirectional threshold cycle network, and analyzes and processes voltage waveform data and distortion characteristics related to the second inverter. The second output layer converts the analysis results into probability distributions for different compensation strategies using a softmax function. In order to realize the cooperative work of the first inverter and the second inverter and optimize the voltage compensation effect, the first intelligent agent and the second intelligent agent are combined in parallel connection. For example, the outputs of two agents are combined by a decision fusion strategy, such as voting, weighted averaging, or decision-level fusion, and a unified compensation strategy is ultimately determined. In this fusion process, various factors, such as the reliability of each agent, the working condition of the inverter, and the requirements of the power grid, can be considered to ensure that the final voltage compensation strategy is not only applicable to a single inverter, but also can perform the best effect when the two inverters work cooperatively.
In a specific embodiment, the process of executing step 104 may specifically include the following steps:
(1) Acquiring target control parameter data of a double-inverter control system, wherein the target control parameter data comprises switching time and operation parameters;
(2) Inputting target control parameter data into a parallel intelligent agent combination;
(3) Performing voltage execution compensation strategy analysis on the target control parameter data through a first intelligent agent in the parallel intelligent agent combination to obtain a first predicted voltage execution compensation strategy;
(4) Performing voltage execution compensation strategy analysis on the target control parameter data through a second intelligent agent in the parallel intelligent agent combination to obtain a second predicted voltage execution compensation strategy;
(5) And performing strategy fusion on the first predicted voltage executing compensation strategy and the second predicted voltage executing compensation strategy to obtain a first parallel voltage executing compensation strategy.
Specifically, target control parameter data of a dual inverter control system is acquired. The target control parameters typically include switching times of the inverter and operating parameters such as frequency, phase, and amplitude. The switching time determines when the inverter is started and shut down, which directly affects the operating efficiency of the inverter and the stability of the output voltage. The operating parameters determine the specific characteristics of the inverter output voltage, such as waveform, magnitude, frequency, etc. And inputting the target control parameter data into the parallel intelligent agent combination. The parallel agent combination is made up of two agents, each responsible for analyzing a portion of the data and proposing a compensation strategy. The parallel structure can not only improve the processing speed and efficiency, but also improve the reliability and the robustness of the system. In the parallel intelligent agent combination, the first intelligent agent performs voltage execution compensation strategy analysis on the target control parameter data to obtain a first predicted voltage execution compensation strategy. The first agent needs to understand and interpret various characteristics of the input data and predict the most appropriate compensation strategy based on these characteristics. The first predicted voltage execution compensation strategy includes adjusting a switching time of the inverter, changing an amplitude or phase of the output voltage, adding a filter to remove noise in the voltage waveform, and the like. For example, if a large voltage fluctuation of the power grid is detected, the first intelligent agent may recommend adjusting the switching time of the inverter so that it can respond to the change of the power grid more quickly and maintain the stability of the output voltage. Similarly, the second agent analyzes the same target control parameter data to obtain a second predicted voltage and execute a compensation strategy. The second agent's analysis process is similar to the first agent, but the resulting compensation strategy is different because it uses a different algorithm or model. The second predicted voltage executing compensation strategy also includes adjusting the switching time of the inverter, changing the amplitude or phase of the output voltage, etc., but the specific adjustment manner and parameters are different from those proposed by the first agent. For example, the second agent may suggest adding a large-capacity capacitor to the output of the inverter to smooth out the fluctuation of the output voltage and improve the stability of the system. And fusing the two strategies to obtain a final first parallel voltage execution compensation strategy. All suggestions made by the two agents are taken into account in combination and the most appropriate strategy is selected from them. Policy fusion may take a variety of approaches such as weighted averaging, voting, decision trees or neural networks, etc. Various factors, such as the reliability of each strategy, the operating conditions of the inverter, the requirements of the grid, and other constraints of the system, etc., need to be considered in selecting the fusion method. For example, if the policies proposed by two agents are consistent in some way, then these parts may be given higher weights; if there is a conflict in some way, further analysis and adjustment is required to ensure that the final strategy is both effective and viable. Through the fusion process, a voltage execution compensation strategy comprehensively considering all factors is obtained, the strategy can not only effectively offset voltage distortion, but also adapt to different working conditions and requirements, and the optimal performance of an inverter system is ensured.
In a specific embodiment, the process of executing step 105 may specifically include the following steps:
(1) Performing voltage distortion compensation interval prediction on the first voltage distortion compensation parameter to obtain a first voltage distortion compensation interval range, and performing voltage distortion compensation interval prediction on the second voltage distortion compensation parameter to obtain a second voltage distortion compensation interval range;
(2) Performing strategy optimization function setting on the first voltage distortion compensation interval range and the second voltage distortion compensation interval range to obtain a target strategy optimization function;
(3) Performing a compensation strategy on the first parallel voltage through a target strategy optimization function to perform particle association generation to obtain a target particle population, wherein the target particle population comprises a plurality of sub-particle populations;
(4) Respectively carrying out fitness calculation on each sub-particle population to obtain a particle fitness set;
(5) Performing iterative computation on the particle fitness set until a preset condition is met, and generating an optimal solution corresponding to the target particle population;
(6) And performing voltage compensation strategy optimization on the first parallel voltage execution compensation strategy through the optimal solution to obtain a second parallel voltage execution compensation strategy.
Specifically, the first voltage distortion compensation parameter is subjected to voltage distortion compensation interval prediction, and a first voltage distortion compensation interval range is obtained. Historical voltage data and compensation parameters are analyzed using statistical or machine learning methods to predict the compensation range required under different conditions. For example, if the historical data shows that voltage distortion increases as the load suddenly increases, the system predicts that a greater range of compensation will be required under similar conditions. And similarly, carrying out the same prediction process on the second voltage distortion compensation parameter to obtain a second voltage distortion compensation interval range. And performing policy optimization function setting on the first and second voltage distortion compensation interval ranges. This policy optimization function is a rule or model that determines how to select the optimal compensation policy from the compensation intervals. It includes a variety of considerations such as efficiency of compensation, cost, impact on system stability, etc. For example, the policy optimization function may favor selecting a less costly but reasonably effective compensation policy, or selecting the most effective compensation policy with the system stability guaranteed. After determining the policy optimization function, executing a compensation policy on the first parallel voltage through the function to perform particle association generation, which is an optimization method for searching in a solution space through simulating particles. In this process, each particle represents a compensation strategy, and all particles in a particle population together explore the optimal solution. Thereby generating a target particle population comprising a plurality of sub-particle populations, each sub-particle population representing a specific set of compensation strategy combinations. An fitness calculation is performed on each population of sub-particles to determine the effect of each particle (i.e., each compensation strategy). Fitness calculations are typically based on predefined performance metrics such as the degree of reduction in voltage distortion after compensation, impact on system stability, cost, etc. These metrics collectively reflect the merits and merits of each strategy. For example, a strategy that has lower cost and significantly reduces voltage distortion would achieve a higher fitness score. After the fitness calculation is completed, a particle fitness set is obtained, which contains fitness scores of all particles. And (3) carrying out iterative calculation on the particle fitness set, wherein each particle in the particle swarm continuously adjusts its own position according to the fitness of the particle swarm and other particles, namely adjusts a represented compensation strategy to explore a better solution. This process continues until a predetermined condition is met, such as reaching a predetermined number of iterations or the fitness is no longer significantly improved. At the end of the iteration, an optimal solution for the target particle population is generated, which represents the most effective one of all the explored compensation strategies under the current conditions. And optimizing the first parallel voltage execution compensation strategy through the optimal solution to obtain a second parallel voltage execution compensation strategy. The optimal solution is converted into a practically executable compensation strategy and applied to an inverter control system.
In a specific embodiment, the process of executing step 106 may specifically include the following steps:
(1) Carrying out real-time load monitoring on the double-inverter control system to obtain real-time system load data, and carrying out linear transformation and covariance matrix calculation on the real-time system load data to obtain a target covariance matrix;
(2) Performing eigenvalue decomposition on the target covariance matrix to obtain a plurality of eigenvalues and eigenvectors, and performing eigenvalue screening and projection matrix construction on the plurality of eigenvalues to obtain a data projection matrix;
(3) Carrying out data mapping on the real-time system load data according to the data projection matrix to obtain standard system load data;
(4) And performing dynamic load compensation parameter analysis on the standard system load data according to the second parallel voltage execution compensation strategy to obtain a dynamic load compensation parameter combination.
Specifically, the dual inverter control system is monitored in real time, and various sensors and data acquisition systems are used for collecting key parameters such as current, voltage, power and the like. And performing linear transformation and covariance matrix calculation on the collected real-time system load data. Linear transformations are typically used to normalize the data making it more suitable for further analysis and processing. The covariance matrix calculation is to understand the relationship and interdependence between different load parameters. By calculating the covariance matrix, a comprehensive load signature description is obtained that reflects how the various parameters change over time and interact with each other. For example, if the system detects a large covariance between current and voltage, which means that the load is fluctuating more, a finer control strategy is needed to maintain the voltage stable. After the covariance matrix is obtained, eigenvalue decomposition is carried out on the covariance matrix. Eigenvalue decomposition is a mathematical method that extracts eigenvalues and eigenvectors from the covariance matrix that represent the direction of the dominant change of the data. These eigenvalues and eigenvectors provide an effective means for understanding and compressing the data. These feature values are screened. The more eigenvalues that remain, the closer the reconstructed data is to the original data, but the more computationally intensive. Therefore, a balance needs to be found between the accuracy of the data and the computational efficiency. A data projection matrix is constructed based on the eigenvalues and corresponding eigenvectors. This matrix will be used to convert the original high-dimensional data into a lower-dimensional space, a process called data mapping. By data mapping, a more simplified standard system load dataset is obtained, while retaining the main features. And performing a compensation strategy according to the second parallel voltage to perform dynamic load compensation parameter analysis on the data. A predefined rule or model is used to determine how to adjust the compensation parameters based on the real-time load data to achieve an optimal compensation effect. These rules or models are based on historical data, theoretical analysis, or machine learning algorithms. By this analysis, a set of load compensation parameters is dynamically generated that vary with load to ensure that the voltage remains at an ideal state throughout. For example, if the system detects a sudden increase in load resulting in a voltage drop, it may increase the output voltage of the inverter or adjust other relevant parameters to compensate for such drop.
The voltage compensation method of the inverter in the embodiment of the present application is described above, and the voltage compensation device of the inverter in the embodiment of the present application is described below, referring to fig. 2, and one embodiment of the voltage compensation device of the inverter in the embodiment of the present application includes:
The detection module 201 is configured to construct a dual-inverter control system through a first inverter and a second inverter, and perform voltage waveform detection on the dual-inverter control system to obtain first voltage waveform data of the first inverter and second voltage waveform data of the second inverter;
The prediction module 202 is configured to perform voltage distortion recognition and compensation prediction on the first voltage waveform data and the second voltage waveform data through a preset convolution long-short time model, so as to obtain a first voltage distortion compensation parameter of the first inverter and a second voltage distortion compensation parameter of the second inverter;
The creation module 203 is configured to create a first agent of the first inverter and a second agent of the second inverter, and perform a parallel relationship combination on the first agent and the second agent to obtain a parallel agent combination;
The obtaining module 204 is configured to obtain target control parameter data of the dual inverter control system, perform compensation policy analysis on the parallel voltage performed by the parallel agent combination through the target control parameter data, and obtain a first parallel voltage performed compensation policy;
An optimization module 205, configured to perform voltage compensation policy optimization on the first parallel voltage execution compensation policy according to the first voltage distortion compensation parameter and the second voltage distortion compensation parameter, so as to obtain a second parallel voltage execution compensation policy;
and the analysis module 206 is configured to perform real-time load monitoring on the dual inverter control system to obtain real-time system load data, and perform dynamic load compensation parameter analysis on the real-time system load data according to the second parallel voltage execution compensation strategy to obtain a dynamic load compensation parameter combination.
Through the cooperative cooperation of the components, the method can accurately identify distortion characteristics in the voltage waveform and predict and compensate the distortions in real time through a preset convolution long-short time model, so that the quality and stability of the output voltage of the inverter are improved. By creating a parallel agent combination, a cooperative operation of the two inverters is achieved. By means of parallel analysis, the voltage execution compensation strategy can be optimized, and the efficiency and performance of the whole double-inverter system are improved. Real-time load monitoring and dynamic load compensation parameter analysis are introduced, so that the system can respond to load fluctuation in time and maintain stable output voltage. This is important to cope with the case where the load change is large in practical use. By optimizing the voltage execution compensation strategy, particularly adopting strategy fusion and particle association generation, the system can select the optimal voltage compensation strategy under different operation conditions, and the self-adaptability and the robustness of the system are improved. The system can know the current working state in real time by introducing real-time system load monitoring and dynamic load compensation parameter analysis, so that the compensation parameters can be adjusted more flexibly, the stability and performance of the system under different load conditions are ensured, the intelligent voltage compensation of the inverter is realized, and the accuracy of the voltage compensation of the inverter is improved.
The present application also provides a computer device including a memory and a processor, the memory storing computer readable instructions that, when executed by the processor, cause the processor to perform the steps of the voltage compensation method of the inverter in the above embodiments.
The present application also provides a computer readable storage medium, which may be a non-volatile computer readable storage medium, and which may also be a volatile computer readable storage medium, the computer readable storage medium having instructions stored therein which, when executed on a computer, cause the computer to perform the steps of the voltage compensation method of the inverter.
It will be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, systems and units may refer to the corresponding processes in the foregoing method embodiments, which are not repeated herein.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a read-only memory (ROM), a random access memory (random access memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
The above embodiments are only for illustrating the technical solution of the present application, and not for limiting the same; although the application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims (7)
1. A voltage compensation method of an inverter, the voltage compensation method of the inverter comprising:
Constructing a double-inverter control system through a first inverter and a second inverter, and detecting voltage waveforms of the double-inverter control system to obtain first voltage waveform data of the first inverter and second voltage waveform data of the second inverter;
Respectively carrying out voltage distortion identification and compensation prediction on the first voltage waveform data and the second voltage waveform data through a preset convolution long-short time model to obtain a first voltage distortion compensation parameter of the first inverter and a second voltage distortion compensation parameter of the second inverter; the method specifically comprises the following steps: and calling a preset convolution long-short time model, wherein the convolution long-short time model comprises the following steps: a convolution long-short time network, a parameter prediction network and a parameter discrimination network; extracting voltage distortion characteristics of the first voltage waveform data through the convolution long-short time network to obtain a plurality of first voltage distortion characteristics of the first voltage waveform data, and extracting voltage distortion characteristics of the second voltage waveform data through the convolution long-short time network to obtain a plurality of second voltage distortion characteristics of the second voltage waveform data; performing compensation parameter calculation on a plurality of first voltage distortion characteristics of the first voltage waveform data through the parameter prediction network to obtain first original distortion compensation parameters of the first inverter, and performing compensation parameter calculation on a plurality of second voltage distortion characteristics of the second voltage waveform data through the parameter prediction network to obtain second original distortion compensation parameters of the second inverter; respectively checking compensation parameters of a first original distortion compensation parameter of the first inverter and a second original distortion compensation parameter of the second inverter through the parameter discrimination network to obtain a first voltage distortion compensation parameter of the first inverter and a second voltage distortion compensation parameter of the second inverter;
creating a first intelligent agent of the first inverter and a second intelligent agent of the second inverter, and combining the first intelligent agent and the second intelligent agent in parallel connection relationship to obtain a parallel intelligent agent combination; the method specifically comprises the following steps: creating a first intelligent agent of the first inverter, wherein the first intelligent agent comprises a first input layer, a first execution compensation strategy analysis network and a first output layer, the first input layer is a single-hot vector coding layer, the first execution compensation strategy analysis network comprises a bidirectional threshold circulation network and a unidirectional threshold circulation network, and the first output layer is a softmax function; creating a second intelligent agent of the second inverter, wherein the second intelligent agent comprises a second input layer, a second execution compensation strategy analysis network and a second output layer, the second input layer is a single-heat vector coding layer, the second execution compensation strategy analysis network comprises a bidirectional threshold circulation network and a unidirectional threshold circulation network, and the second output layer is a softmax function; carrying out parallel relation combination on the first intelligent agent and the second intelligent agent to obtain a parallel intelligent agent combination;
Acquiring target control parameter data of the double-inverter control system, and performing parallel voltage execution compensation strategy analysis on the parallel intelligent agent combination through the target control parameter data to obtain a first parallel voltage execution compensation strategy;
Performing voltage compensation strategy optimization on the first parallel voltage execution compensation strategy according to the first voltage distortion compensation parameter and the second voltage distortion compensation parameter to obtain a second parallel voltage execution compensation strategy; the method specifically comprises the following steps: performing voltage distortion compensation interval prediction on the first voltage distortion compensation parameter to obtain a first voltage distortion compensation interval range, and performing voltage distortion compensation interval prediction on the second voltage distortion compensation parameter to obtain a second voltage distortion compensation interval range; performing policy optimization function setting on the first voltage distortion compensation interval range and the second voltage distortion compensation interval range to obtain a target policy optimization function; performing a compensation strategy on the first parallel voltage through the target strategy optimization function to perform particle association generation to obtain a target particle population, wherein the target particle population comprises a plurality of sub-particle populations; respectively carrying out fitness calculation on each sub-particle population to obtain a particle fitness set; performing iterative computation on the particle fitness set until a preset condition is met, and generating an optimal solution corresponding to the target particle population; performing voltage compensation strategy optimization on the first parallel voltage execution compensation strategy through the optimal solution to obtain a second parallel voltage execution compensation strategy;
And carrying out real-time load monitoring on the double-inverter control system to obtain real-time system load data, and carrying out dynamic load compensation parameter analysis on the real-time system load data according to the second parallel voltage execution compensation strategy to obtain a dynamic load compensation parameter combination.
2. The method for voltage compensation of an inverter according to claim 1, wherein constructing a dual-inverter control system by a first inverter and a second inverter, and performing voltage waveform detection on the dual-inverter control system, obtaining first voltage waveform data of the first inverter and second voltage waveform data of the second inverter, comprises:
Constructing a double-inverter control system corresponding to a first inverter and a second inverter in a parallel connection mode, wherein the first inverter is used for converting direct current into alternating current, and the second inverter is used for converting the alternating current into direct current;
Detecting voltage waveforms of the double-inverter control system to obtain an original voltage waveform data set;
Inputting the original voltage waveform data set into a preset threshold clustering algorithm, and performing classification threshold calculation on the original voltage waveform data set through an initial threshold calculation function in the threshold clustering algorithm to obtain a corresponding initial classification threshold;
carrying out data set division on the original voltage waveform data set and the initial classification threshold value to obtain an initial data set division result;
parameter adjustment is carried out on the initial threshold computing function according to the initial data set dividing result, and a target threshold computing function is obtained;
Classifying threshold calculation is carried out on the original voltage waveform data set through the target threshold calculation function to obtain a corresponding target classifying threshold, and data division is carried out on the original voltage waveform data set through the target classifying threshold to obtain a target data set division result;
And determining first voltage waveform data of the first inverter and second voltage waveform data of the second inverter according to the target data set division result.
3. The method for voltage compensation of an inverter according to claim 1, wherein the obtaining target control parameter data of the dual inverter control system, performing a compensation strategy analysis on the parallel voltage by the parallel agent combination through the target control parameter data, obtaining a first parallel voltage execution compensation strategy, includes:
Acquiring target control parameter data of the double-inverter control system, wherein the target control parameter data comprises switching time and operation parameters;
inputting the target control parameter data into the parallel intelligent agent combination;
performing voltage execution compensation strategy analysis on the target control parameter data through a first intelligent agent in the parallel intelligent agent combination to obtain a first predicted voltage execution compensation strategy;
performing voltage execution compensation strategy analysis on the target control parameter data through a second intelligent agent in the parallel intelligent agent combination to obtain a second predicted voltage execution compensation strategy;
And performing strategy fusion on the first predicted voltage executing compensation strategy and the second predicted voltage executing compensation strategy to obtain a first parallel voltage executing compensation strategy.
4. The method for voltage compensation of an inverter according to claim 1, wherein the performing real-time load monitoring on the dual inverter control system to obtain real-time system load data, and performing dynamic load compensation parameter analysis on the real-time system load data according to the second parallel voltage execution compensation strategy to obtain a dynamic load compensation parameter combination includes:
Carrying out real-time load monitoring on the double-inverter control system to obtain real-time system load data, and carrying out linear transformation and covariance matrix calculation on the real-time system load data to obtain a target covariance matrix;
Performing eigenvalue decomposition on the target covariance matrix to obtain a plurality of eigenvalues and eigenvectors, and performing eigenvalue screening and projection matrix construction on the eigenvalues to obtain a data projection matrix;
performing data mapping on the real-time system load data according to the data projection matrix to obtain standard system load data;
And performing dynamic load compensation parameter analysis on the standard system load data according to the second parallel voltage execution compensation strategy to obtain a dynamic load compensation parameter combination.
5. A voltage compensation device of an inverter, characterized in that the voltage compensation device of the inverter comprises:
the detection module is used for constructing a double-inverter control system through the first inverter and the second inverter, and detecting voltage waveforms of the double-inverter control system to obtain first voltage waveform data of the first inverter and second voltage waveform data of the second inverter;
The prediction module is used for respectively carrying out voltage distortion identification and compensation prediction on the first voltage waveform data and the second voltage waveform data through a preset convolution long-short time model to obtain a first voltage distortion compensation parameter of the first inverter and a second voltage distortion compensation parameter of the second inverter; the method specifically comprises the following steps: and calling a preset convolution long-short time model, wherein the convolution long-short time model comprises the following steps: a convolution long-short time network, a parameter prediction network and a parameter discrimination network; extracting voltage distortion characteristics of the first voltage waveform data through the convolution long-short time network to obtain a plurality of first voltage distortion characteristics of the first voltage waveform data, and extracting voltage distortion characteristics of the second voltage waveform data through the convolution long-short time network to obtain a plurality of second voltage distortion characteristics of the second voltage waveform data; performing compensation parameter calculation on a plurality of first voltage distortion characteristics of the first voltage waveform data through the parameter prediction network to obtain first original distortion compensation parameters of the first inverter, and performing compensation parameter calculation on a plurality of second voltage distortion characteristics of the second voltage waveform data through the parameter prediction network to obtain second original distortion compensation parameters of the second inverter; respectively checking compensation parameters of a first original distortion compensation parameter of the first inverter and a second original distortion compensation parameter of the second inverter through the parameter discrimination network to obtain a first voltage distortion compensation parameter of the first inverter and a second voltage distortion compensation parameter of the second inverter;
The creation module is used for creating a first intelligent agent of the first inverter and a second intelligent agent of the second inverter, and carrying out parallel relation combination on the first intelligent agent and the second intelligent agent to obtain a parallel intelligent agent combination; the method specifically comprises the following steps: creating a first intelligent agent of the first inverter, wherein the first intelligent agent comprises a first input layer, a first execution compensation strategy analysis network and a first output layer, the first input layer is a single-hot vector coding layer, the first execution compensation strategy analysis network comprises a bidirectional threshold circulation network and a unidirectional threshold circulation network, and the first output layer is a softmax function; creating a second intelligent agent of the second inverter, wherein the second intelligent agent comprises a second input layer, a second execution compensation strategy analysis network and a second output layer, the second input layer is a single-heat vector coding layer, the second execution compensation strategy analysis network comprises a bidirectional threshold circulation network and a unidirectional threshold circulation network, and the second output layer is a softmax function; carrying out parallel relation combination on the first intelligent agent and the second intelligent agent to obtain a parallel intelligent agent combination;
The acquisition module is used for acquiring target control parameter data of the double-inverter control system, and performing parallel voltage compensation strategy analysis on the parallel intelligent agent combination through the target control parameter data to obtain a first parallel voltage compensation strategy;
The optimization module is used for performing voltage compensation strategy optimization on the first parallel voltage execution compensation strategy according to the first voltage distortion compensation parameter and the second voltage distortion compensation parameter to obtain a second parallel voltage execution compensation strategy; the method specifically comprises the following steps: performing voltage distortion compensation interval prediction on the first voltage distortion compensation parameter to obtain a first voltage distortion compensation interval range, and performing voltage distortion compensation interval prediction on the second voltage distortion compensation parameter to obtain a second voltage distortion compensation interval range; performing policy optimization function setting on the first voltage distortion compensation interval range and the second voltage distortion compensation interval range to obtain a target policy optimization function; performing a compensation strategy on the first parallel voltage through the target strategy optimization function to perform particle association generation to obtain a target particle population, wherein the target particle population comprises a plurality of sub-particle populations; respectively carrying out fitness calculation on each sub-particle population to obtain a particle fitness set; performing iterative computation on the particle fitness set until a preset condition is met, and generating an optimal solution corresponding to the target particle population; performing voltage compensation strategy optimization on the first parallel voltage execution compensation strategy through the optimal solution to obtain a second parallel voltage execution compensation strategy;
And the analysis module is used for carrying out real-time load monitoring on the double-inverter control system to obtain real-time system load data, and carrying out dynamic load compensation parameter analysis on the real-time system load data according to the second parallel voltage execution compensation strategy to obtain a dynamic load compensation parameter combination.
6. A computer device, the computer device comprising: a memory and at least one processor, the memory having instructions stored therein;
The at least one processor invoking the instructions in the memory to cause the computer device to perform the voltage compensation method of the inverter of any of claims 1-4.
7. A computer readable storage medium having instructions stored thereon, which when executed by a processor, implement the voltage compensation method of an inverter according to any one of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410090315.3A CN117639452B (en) | 2024-01-23 | 2024-01-23 | Voltage compensation method, device and equipment of inverter and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410090315.3A CN117639452B (en) | 2024-01-23 | 2024-01-23 | Voltage compensation method, device and equipment of inverter and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117639452A CN117639452A (en) | 2024-03-01 |
| CN117639452B true CN117639452B (en) | 2024-04-23 |
Family
ID=90030718
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410090315.3A Active CN117639452B (en) | 2024-01-23 | 2024-01-23 | Voltage compensation method, device and equipment of inverter and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117639452B (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2017103973A (en) * | 2015-12-04 | 2017-06-08 | 株式会社東芝 | Voltage compensation device |
| CN107332261A (en) * | 2017-08-11 | 2017-11-07 | 兰州理工大学 | A kind of micro-capacitance sensor quality of power supply distributed coordination administering method |
| CN111272277A (en) * | 2020-01-21 | 2020-06-12 | 中国工程物理研究院激光聚变研究中心 | Laser pulse waveform measurement distortion correction method and system based on neural network |
| CN111682557A (en) * | 2020-06-18 | 2020-09-18 | 四川大学 | Optimization method for reactive equipment configuration site and optimal compensation capacity of power system |
| CN113391329A (en) * | 2021-06-07 | 2021-09-14 | 中国科学院国家授时中心 | Beidou satellite navigation signal distortion adaptive compensation method |
| CN114492530A (en) * | 2022-01-27 | 2022-05-13 | 中国科学院声学研究所 | Active sonar detection waveform identification method and device based on generation type countermeasure network |
| WO2023012194A1 (en) * | 2021-08-06 | 2023-02-09 | Ecole Nationale Superieure De L'electronique Et De Ses Applications | Active electric compensation device with fast-switching structure |
| CN115933159A (en) * | 2022-10-31 | 2023-04-07 | 燕山大学 | Real-time high-precision wavefront distortion phase compensation system |
| CN116822920A (en) * | 2023-05-23 | 2023-09-29 | 北京杰成合力科技有限公司 | Flow prediction method based on cyclic neural network |
| CN116882301A (en) * | 2023-09-04 | 2023-10-13 | 联和存储科技(江苏)有限公司 | Resistance compensation method, device and equipment of memory chip and storage medium |
| CN117236485A (en) * | 2023-08-15 | 2023-12-15 | 湖北华中电力科技开发有限责任公司 | Method for reducing and predicting power supply quantity errors of power grid based on CNN-BILSTM algorithm |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200265211A1 (en) * | 2019-02-14 | 2020-08-20 | West Virginia University | Fingerprint distortion rectification using deep convolutional neural networks |
-
2024
- 2024-01-23 CN CN202410090315.3A patent/CN117639452B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2017103973A (en) * | 2015-12-04 | 2017-06-08 | 株式会社東芝 | Voltage compensation device |
| CN107332261A (en) * | 2017-08-11 | 2017-11-07 | 兰州理工大学 | A kind of micro-capacitance sensor quality of power supply distributed coordination administering method |
| CN111272277A (en) * | 2020-01-21 | 2020-06-12 | 中国工程物理研究院激光聚变研究中心 | Laser pulse waveform measurement distortion correction method and system based on neural network |
| CN111682557A (en) * | 2020-06-18 | 2020-09-18 | 四川大学 | Optimization method for reactive equipment configuration site and optimal compensation capacity of power system |
| CN113391329A (en) * | 2021-06-07 | 2021-09-14 | 中国科学院国家授时中心 | Beidou satellite navigation signal distortion adaptive compensation method |
| WO2023012194A1 (en) * | 2021-08-06 | 2023-02-09 | Ecole Nationale Superieure De L'electronique Et De Ses Applications | Active electric compensation device with fast-switching structure |
| CN114492530A (en) * | 2022-01-27 | 2022-05-13 | 中国科学院声学研究所 | Active sonar detection waveform identification method and device based on generation type countermeasure network |
| CN115933159A (en) * | 2022-10-31 | 2023-04-07 | 燕山大学 | Real-time high-precision wavefront distortion phase compensation system |
| CN116822920A (en) * | 2023-05-23 | 2023-09-29 | 北京杰成合力科技有限公司 | Flow prediction method based on cyclic neural network |
| CN117236485A (en) * | 2023-08-15 | 2023-12-15 | 湖北华中电力科技开发有限责任公司 | Method for reducing and predicting power supply quantity errors of power grid based on CNN-BILSTM algorithm |
| CN116882301A (en) * | 2023-09-04 | 2023-10-13 | 联和存储科技(江苏)有限公司 | Resistance compensation method, device and equipment of memory chip and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117639452A (en) | 2024-03-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sibalija et al. | An integrated approach to optimise parameter design of multi-response processes based on Taguchi method and artificial intelligence | |
| CN111382906B (en) | Power load prediction method, system, equipment and computer readable storage medium | |
| CN110544177A (en) | Load identification method based on power fingerprint and computer readable storage medium | |
| CN112987666B (en) | Power plant unit operation optimization regulation and control method and system | |
| Qiao et al. | Renewable scenario generation using controllable generative adversarial networks with transparent latent space | |
| CN109902340B (en) | Multi-source-load combined scene generation method considering complex meteorological coupling characteristics | |
| Natarajan et al. | New SVM kernel soft computing models for wind speed prediction in renewable energy applications | |
| CN111489188A (en) | Resident adjustable load potential mining method and system | |
| CN115409649A (en) | Wind power plant energy management system and method thereof | |
| CN110717610A (en) | Wind power prediction method based on data mining | |
| Zhang et al. | Intelligent systems for power system dynamic security assessment: Review and classification | |
| CN110852344A (en) | Intelligent substation network fault classification based method | |
| CN117639452B (en) | Voltage compensation method, device and equipment of inverter and storage medium | |
| CN116701919B (en) | Optimization monitoring method and system for gyro-type hydrogen fuel generator | |
| CN111858343A (en) | Countermeasure sample generation method based on attack capability | |
| CN116681186A (en) | Power quality analysis method and device based on intelligent terminal | |
| CN117010519A (en) | Data processing method of quantum support vector machine, storage medium and related equipment | |
| CN111967919A (en) | System and method for analyzing electricity consumption behavior of residents based on autoregressive and adaptive boosting algorithm | |
| CN111026741A (en) | Data cleaning method and device based on time series similarity | |
| Povkhan et al. | The algorithmic classification trees | |
| CN115883182A (en) | Method and system for improving network security situation element identification efficiency | |
| CN115660038A (en) | Multi-stage integrated short-term load prediction based on error factors and improved MOEA/D-SAS | |
| CN115936303A (en) | Transient voltage safety analysis method based on machine learning model | |
| CN114662576A (en) | Non-invasive storage battery car charging detection method and system based on supervised classification | |
| CN117094478B (en) | Energy scheduling management method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |