Adaptive control system for new energy power generation equipment based on multi-scale data fusion

Abstract

This invention discloses an adaptive control system for new energy power generation equipment based on multi-scale data fusion, belonging to the field of power generation control technology. The system includes: acquiring multi-source power datasets corresponding to each composite fuel type; performing multi-scale feature convolution on the multi-source power datasets to output multi-source high-dimensional datasets; fusing the multi-source high-dimensional datasets to generate a fused high-dimensional dataset; receiving equipment fault information and generating equipment control parameter items; performing feature dimensionality reduction on the fused high-dimensional dataset to output a fused low-dimensional dataset; and training an adaptive control model for the equipment to adaptively control the new energy power generation equipment. This invention solves the technical problems of poor adaptability of new energy power generation equipment and unstable power output in existing technologies, achieving a significant improvement in equipment adaptability and effectively reducing power output instability.

Description

Technical Field

[0001] This invention relates to the field of power generation control technology, and more specifically to an adaptive control system for new energy power generation equipment based on multi-scale data fusion. Background Technology

[0002] In the field of new energy power generation, with the rapid development of various new energy power generation technologies, new energy sources such as wind power, solar power, and biomass energy are gradually being widely used. However, because these new energy power generation devices typically operate under different environmental conditions and loads, their operating states have significant uncertainties and complexities. This is especially true when equipment malfunctions or switches operating modes, which can easily lead to instability in power output. Traditional control systems often struggle to cope with such variable operating environments, lacking sufficient adaptability and precision, thus affecting the stability of the power system.

[0003] Existing technologies suffer from poor adaptability to new energy power generation equipment and unstable power output. Summary of the Invention

[0004] This application provides an adaptive control system for new energy power generation equipment based on multi-scale data fusion, which is used to address the technical problems of poor adaptability and unstable power output of new energy power generation equipment in the prior art.

[0005] In view of the above problems, this application provides an adaptive control system for new energy power generation equipment based on multi-scale data fusion, the system comprising:

[0006] The system includes the following modules: a multi-source power dataset acquisition module for acquiring new energy fuel information from new energy power generation equipment, determining whether the new energy power generation equipment is a hybrid fuel power generation mode based on the new energy fuel information, and acquiring multi-source power datasets corresponding to each hybrid fuel type if the new energy power generation equipment is a hybrid fuel power generation mode; a multi-source high-dimensional dataset output module for performing multi-scale feature convolution on the multi-source power datasets through multi-scale convolutional network layers to output multi-source high-dimensional datasets, wherein the multi-scale feature convolution includes temporal feature convolution, spatial feature convolution, and type feature convolution; a fused high-dimensional dataset generation module for fusing the multi-source high-dimensional datasets to generate a fused high-dimensional dataset; an equipment control parameter generation module for receiving equipment fault information from the new energy power generation equipment and generating equipment control parameter items based on the equipment fault information; a fused low-dimensional dataset generation module for performing feature dimensionality reduction on the fused high-dimensional dataset using the t-SNE algorithm to output a fused low-dimensional dataset; and an adaptive control module for training an equipment adaptive control model using the fused low-dimensional dataset and the equipment control parameter items, and performing adaptive control on the new energy power generation equipment based on the equipment adaptive control model.

[0007] A second aspect of this application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program and the processor is used to execute the adaptive control system for new energy power generation equipment based on multi-scale data fusion provided in this application.

[0008] A third aspect of this application provides a computer-readable storage medium having a computer program stored thereon, the computer program being used to execute the adaptive control system for new energy power generation equipment based on multi-scale data fusion provided in this application.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] The system includes a multi-source power dataset acquisition module for acquiring multi-source power datasets corresponding to each composite fuel type; a multi-source high-dimensional dataset output module for performing multi-scale feature convolution on the multi-source power datasets using a multi-scale convolutional network layer to output multi-source high-dimensional datasets; a fused high-dimensional dataset generation module for fusing the multi-source high-dimensional datasets to generate a fused high-dimensional dataset; a device control parameter generation module for generating device control parameters based on device fault information received by the new energy power generation equipment; a fused low-dimensional dataset generation module for performing feature dimensionality reduction on the fused high-dimensional dataset using the t-SNE algorithm to output a fused low-dimensional dataset; and an adaptive control module for training an adaptive control model for the equipment using the fused low-dimensional dataset and the device control parameters to perform adaptive control. This significantly improves the adaptability of the equipment and effectively reduces power output instability. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A schematic diagram of the structure of an adaptive control system for new energy power generation equipment based on multi-scale data fusion, provided in an embodiment of this application;

[0013] Figure 2 This is a schematic diagram illustrating the process of outputting a multi-source high-dimensional dataset and fusing a low-dimensional dataset in an adaptive control system for new energy power generation equipment based on multi-scale data fusion, as provided in an embodiment of this application.

[0014] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0015] Explanation of reference numerals in the attached figures: 10 for multi-source power data acquisition module, 20 for multi-source high-dimensional data output module, 30 for fused high-dimensional data generation module, 40 for equipment control parameter generation module, 50 for fused low-dimensional data generation module, 60 for adaptive control module, 21 for processor, 22 for memory, 23 for input device, and 24 for output device. Detailed Implementation

[0016] This application provides an adaptive control system for new energy power generation equipment based on multi-scale data fusion, which is used to address the technical problems of poor adaptability and unstable power output of new energy power generation equipment in the prior art.

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0018] Examples, such as Figure 1 As shown, this application provides an adaptive control system for new energy power generation equipment based on multi-scale data fusion, the system comprising:

[0019] The multi-source power data set acquisition module 10 is used to acquire new energy fuel information of new energy power generation equipment, determine whether the new energy power generation equipment is a composite fuel power generation mode based on the new energy fuel information, and if the new energy power generation equipment is a composite fuel power generation mode, acquire the multi-source power data set corresponding to each composite fuel type.

[0020] Specifically, the first step is to obtain information on the new energy fuels used in new energy power generation equipment. These equipment can use various types of new energy fuels, such as solar, wind, hydro, and biomass energy, or combinations thereof. Through various sensors and monitoring systems, detailed information about the fuels used in the equipment can be obtained. This information includes the type of fuel, its source, current reserves, and supply stability. For example, for solar power generation equipment, information such as current sunlight intensity and the operational status of the solar panels will be obtained; for wind power generation equipment, information such as current wind speed and direction will be obtained; and for biomass power generation equipment, information such as the supply of biomass fuel and combustion efficiency will be obtained.

[0021] Based on the acquired new energy fuel information, it can be determined whether the new energy power generation equipment operates in a hybrid fuel power generation mode. A hybrid fuel power generation mode means that the equipment simultaneously uses two or more different types of new energy fuels to generate electricity. For example, a power generation device might utilize both solar and wind power, or combine biomass energy with hydropower. Analysis of the fuel information can determine whether the equipment is operating in this hybrid fuel power generation mode.

[0022] If a renewable energy power generation device is identified as operating in a hybrid fuel power generation mode, then it is necessary to obtain multi-source power datasets for each hybrid fuel type. This is because different fuel types have different impacts on the device's operation, necessitating the acquisition of data related to each fuel type separately. These multi-source power datasets may include electrical parameters, environmental data, and device status information related to specific fuel types. For example, if the device uses both solar and wind power generation, then separate multi-source power datasets for the solar and wind power generation components are required. These datasets will provide crucial information for subsequent analysis and control, enabling better management and optimization of the hybrid fuel power generation device's operation.

[0023] By acquiring fuel information from new energy power generation equipment, it is determined whether it is a hybrid fuel power generation mode and the corresponding multi-source power dataset is obtained, providing a data foundation for optimizing the operation and management of hybrid fuel power generation equipment and achieving more efficient and stable power generation control.

[0024] The multi-source high-dimensional dataset output module 20 is used to perform multi-scale feature convolution on the multi-source power dataset through a multi-scale convolutional network layer to output a multi-source high-dimensional dataset. The multi-scale feature convolution includes temporal feature convolution, spatial feature convolution, and type feature convolution.

[0025] Specifically, firstly, multi-scale convolutional network layers are neural network structures capable of processing features at different scales. They can simultaneously analyze multiple dimensions such as time, space, and data type, thereby extracting richer and more comprehensive feature information.

[0026] Temporal feature convolution is primarily used to operate on time-series data in multi-source power datasets. Time-series data includes variations in equipment output power over time, voltage and current fluctuations, etc. Through temporal feature convolution, periodic and trend features are extracted. For example, it can reveal that the equipment output power exhibits a periodic variation pattern within a day, or shows an upward or downward trend over a period of time. These temporal features are crucial for predicting the future state of the equipment and implementing adaptive control.

[0027] Spatial feature convolution focuses on spatial data within multi-source power datasets. In renewable energy power plants, different devices are distributed in different locations, and they interact with each other. Spatial feature convolution can capture these spatial relationships, such as power transmission between different devices and the interaction of electric and magnetic fields. In addition, spatial feature convolution can also consider the differences in the operating status of devices in different geographical locations, such as changes in environmental factors like light intensity and wind speed.

[0028] Type-feature convolution performs convolution operations on different types of data within a multi-source power dataset. Multi-source power datasets for new energy power generation equipment contain various data types, such as electrical parameter data, environmental factor data, and equipment status data. Type-feature convolution can extract correlation features between different types of data. For example, it may discover the relationship between light intensity and the output power of solar power generation equipment, or the relationship between wind speed and the output power of wind power generation equipment. These correlation features can help to better understand the operating mechanisms and performance characteristics of the equipment.

[0029] By performing temporal feature convolution, spatial feature convolution, and typological feature convolution on multi-source power datasets, a multi-source high-dimensional dataset can be obtained. This dataset contains richer and more abstract feature information, providing a stronger foundation for subsequent data analysis and adaptive control.

[0030] The high-dimensional dataset generation module 30 is used to fuse the multi-source high-dimensional datasets to generate a fused high-dimensional dataset.

[0031] Specifically, the purpose of fusion is to integrate multi-source high-dimensional data obtained from feature convolutions at different scales. These data have extracted feature information of different dimensions through temporal feature convolution, spatial feature convolution, and typological feature convolution, respectively. The fusion process is to integrate this scattered information to form a more comprehensive and representative dataset.

[0032] To achieve fusion, a weighted fusion approach is used. Weights are assigned to each high-dimensional feature from different sources based on their importance, and then the weighted features are summed. This highlights the contribution of important features while reducing the influence of less important features. During the fusion process, data consistency and compatibility must be considered. Because different scales of feature convolutions use different processing systems and parameter settings, the resulting high-dimensional data differ. Before fusion, these data need to be preprocessed, such as normalized and standardized, to ensure the data is at the same scale and within the same range. After fusion, the generated fused high-dimensional dataset contains comprehensive feature information across multiple dimensions, including time, space, and type. This dataset better reflects the operating status and characteristics of new energy power generation equipment, providing stronger support for subsequent equipment control and optimization. By fusing multi-source high-dimensional datasets to generate a fused high-dimensional dataset, integrating feature information from different dimensions, the ability to reflect the operating status and characteristics of new energy power generation equipment is improved, providing stronger support for equipment control and optimization.

[0033] The equipment control parameter generation module 40 is used to receive equipment fault information from the new energy power generation equipment and generate equipment control parameter items based on the equipment fault information.

[0034] Specifically, when a fault occurs during the operation of new energy power generation equipment, the fault detection system will promptly detect the fault and generate corresponding fault information. This fault information includes aspects such as the fault type, fault location, and fault severity. For example, it could indicate damage to a critical component, a short circuit, or a sensor malfunction.

[0035] Upon receiving these fault messages, the new energy power generation equipment will activate its fault response mechanism. First, the equipment's control system will analyze and process the fault information, adopting different response strategies based on the type and severity of the fault. For example, in the case of a minor fault, the equipment can continue operating by adjusting its operating parameters, while maintenance personnel are dispatched to inspect and repair it; in the case of a serious fault, the equipment must be stopped immediately to prevent further damage.

[0036] To generate equipment control parameters, the control system calculates parameters by combining fault information, the equipment's current operating status, and pre-set fault handling strategies. For example, if the fault is due to excessive output power, the control system generates control parameters to reduce the output power; if a sensor malfunctions, it adjusts the control algorithm dependent on that sensor or activates a backup sensor. The generated control parameters are sent to the equipment's actuators, such as motors, valves, and switches, to adjust the equipment's operating status. These parameters are also recorded for subsequent fault analysis and processing. In this way, effective control measures can be taken promptly when equipment malfunctions, reducing the impact of faults on equipment operation and improving equipment reliability and stability.

[0037] The low-dimensional dataset generation module 50 is used to perform feature dimensionality reduction on the fused high-dimensional dataset using the t-SNE algorithm and output the fused low-dimensional dataset.

[0038] Specifically, the t-SNE algorithm is used to perform feature dimensionality reduction on the fused high-dimensional dataset to output a fused low-dimensional dataset. First, fused high-dimensional datasets typically contain a large amount of feature information, but high-dimensional data faces problems such as high computational complexity and data sparsity during processing and analysis. Therefore, feature dimensionality reduction is needed to simplify the data while retaining important feature information. The t-SNE algorithm is an effective nonlinear dimensionality reduction system. Its main idea is to construct probability distributions between data points in both high-dimensional and low-dimensional spaces, and then find the representation in the low-dimensional space by minimizing the difference between these two probability distributions. The t-SNE algorithm works as follows: In the high-dimensional space, a Gaussian distribution is calculated for each data point in the fused high-dimensional dataset. This means that for each data point, the distance between it and other data points is calculated, and a probability value is determined based on the distance, representing the similarity between that point and other points. This yields the high-dimensional probability distribution matrix. In the low-dimensional space, a t-distribution is calculated for each data point in the fused low-dimensional dataset. Similarly, the distance between each point and other points is calculated, and the probability value is determined, yielding the low-dimensional probability distribution matrix. Next, the KL divergence between the high-dimensional and low-dimensional probability distribution matrices is calculated. KL divergence represents the degree of difference between the two probability distributions. An objective function is constructed using the KL divergence, aiming to minimize this difference, making the distribution of data points in the low-dimensional space as close as possible to the distribution of data points in the high-dimensional space. To minimize the objective function, gradient descent is used for optimization. By continuously adjusting the positions of data points in the low-dimensional space, the objective function gradually decreases. In each iteration, the direction and distance of movement for each data point are calculated based on the gradient of the objective function. As gradient descent progresses, the distribution of data points in the low-dimensional space is gradually adjusted until the objective function converges. When the objective function converges, a fused low-dimensional dataset is obtained. This low-dimensional dataset retains important feature information from the fused high-dimensional dataset while significantly reducing the data dimensionality. This facilitates subsequent data analysis and processing, such as visualization, clustering, and classification, while also reducing computational resource consumption and improving algorithm efficiency.

[0039] The adaptive control module 60 is used to train an adaptive control model for the device using the fused low-dimensional dataset and the device control parameter items, and to perform adaptive control on the new energy power generation equipment based on the device adaptive control model.

[0040] Specifically, the Support Vector Machine (SVM) algorithm is used to train and apply the adaptive control model for the equipment. For the processing of the fused low-dimensional dataset and equipment control parameters, the fused low-dimensional dataset is used as the input feature vector, containing key information about the new energy power generation equipment in different aspects. The dimensionality-reduced data is more refined and highlights important features. The equipment control parameters can be used as additional features or as part of the label to guide model training. During the training phase, the SVM algorithm aims to find an optimal hyperplane that separates data points of different categories as much as possible while maximizing the classification margin. For the adaptive control problem of new energy power generation equipment, different equipment operating states or control strategies are treated as different categories. By learning from a large amount of fused low-dimensional dataset and corresponding equipment control parameters, the SVM model determines the optimal classification hyperplane, i.e., the best control strategy for the new energy power generation equipment under different conditions. The SVM algorithm determines the position and orientation of the hyperplane by finding support vectors, i.e., the data points closest to the hyperplane. These support vectors are crucial to the model's performance and determine the classification boundary. During training, the SVM algorithm determines the optimal hyperplane parameters by optimizing an objective function to maximize the classification margin while minimizing the classification error. Once training is complete, when a new fused low-dimensional dataset is input, the SVM model quickly performs classification and prediction to determine the corresponding equipment control strategy. For adaptive control of new energy power generation equipment, the most suitable control parameters are automatically selected based on the real-time acquired equipment operation data, i.e., the new fused low-dimensional dataset, to achieve optimal control of the equipment.

[0041] In one possible implementation, such as Figure 2 As shown, the low-dimensional dataset generation module further includes:

[0042] Gaussian distribution calculation is performed on each data point in the fused high-dimensional dataset in a high-dimensional space to output a high-dimensional probability distribution matrix.

[0043] Perform t-distribution calculations on each data point in the fused low-dimensional dataset in a low-dimensional space to output a low-dimensional probability distribution matrix.

[0044] Calculate the KL divergence between the high-dimensional probability distribution matrix and the low-dimensional probability distribution matrix, and construct an objective function using the KL divergence, where the KL divergence represents the difference in probability distribution between the high-dimensional probability distribution matrix and the low-dimensional probability distribution matrix.

[0045] When the objective function converges, the output is a fused low-dimensional dataset.

[0046] Specifically, firstly, the fused high-dimensional dataset contains high-dimensional data information obtained after multi-scale feature extraction and fusion. In the high-dimensional space, for each data point in the dataset, it is necessary to calculate its relationship with all other data points. The purpose of using Gaussian distribution calculation is to determine the probability distribution of each data point in the high-dimensional space. For a specific data point, the distance between it and other data points is calculated using the common Euclidean distance. Based on these distance values, the Gaussian distribution function is used to calculate the similarity probability between the data point and other data points. The Gaussian distribution function has good mathematical properties and can effectively describe the distribution of data points in high-dimensional space. It takes the distance between data points as input and outputs a probability value between 0 and 1, representing the similarity between two data points. The closer the distance, the higher the probability value, indicating that the two data points are more similar; the farther the distance, the lower the probability value, indicating that the two data points are less similar. For each data point in the fused high-dimensional dataset, the above calculation process is repeated: calculate its distance to all other data points, and calculate the similarity probability based on the distance. This yields a high-dimensional probability distribution matrix where the number of rows and columns equals the number of data points in the dataset. Each element in the matrix represents the similarity probability between two data points in the high-dimensional space.

[0047] In a low-dimensional space, the t-distribution is calculated for each data point in the fused low-dimensional dataset to output a low-dimensional probability distribution matrix. The fused low-dimensional dataset is a dataset that has undergone feature dimensionality reduction processing, and its dimensionality is significantly reduced compared to the original fused high-dimensional dataset. In the low-dimensional space, the t-distribution needs to be calculated for each data point in the dataset. The t-distribution is a probability distribution with specific mathematical forms and properties. For each data point, the distance between it and all other data points is calculated. Similar to the high-dimensional space, but in the low-dimensional space, the t-distribution function is used to determine the similarity probability between data points based on these distances. The t-distribution has several advantages when processing low-dimensional data; it can better capture the distribution characteristics of data points in the low-dimensional space, especially for data with complex shapes or uneven distribution. The t-distribution calculation yields a low-dimensional probability distribution matrix. The number of rows and columns in this matrix is ​​equal to the number of data points in the dataset. Each element in the matrix represents the similarity probability of two data points in the low-dimensional space. By calculating the t-distribution for each data point in the fused low-dimensional dataset and outputting the low-dimensional probability distribution matrix, we can understand the distribution and similarity relationships of the data in the low-dimensional space. This provides an important basis for comparing the probability distribution differences between the high-dimensional and low-dimensional spaces and for further optimizing the representation of the low-dimensional space.

[0048] First, KL divergence is an indicator used to measure the difference between two probability distributions. In this step, the high-dimensional probability distribution matrix represents the probability distribution of the data in the high-dimensional space, while the low-dimensional probability distribution matrix reflects the probability distribution of the data in the low-dimensional space after dimensionality reduction. Calculating KL divergence involves comparing and operating on corresponding elements in the two probability distribution matrices. For each element in the high-dimensional probability distribution matrix and its corresponding element in the low-dimensional probability distribution matrix, the KL divergence is calculated according to the formula, measuring the amount of information required to transform from one probability distribution to another. By calculating the KL divergence between the high-dimensional and low-dimensional probability distribution matrices, the degree of difference between the two probability distributions can be quantitatively determined. This difference reflects the changes in the distribution of data in the high-dimensional and low-dimensional spaces. If the KL divergence is large, it indicates a significant difference in the probability distributions between the high-dimensional and low-dimensional spaces, meaning that the dimensionality reduction process has lost a considerable amount of important information; if the KL divergence is small, it indicates that the low-dimensional space has better preserved the probability distribution characteristics of the high-dimensional space. The objective function is constructed using the Karl von Lee (KL) divergence, with the goal of minimizing this divergence. An optimization algorithm continuously adjusts the data representation in the low-dimensional space, making the low-dimensional probability distribution as close as possible to the high-dimensional probability distribution while preserving the important features of the data. This objective function guides the dimensionality reduction process towards a better outcome, ensuring that the fused low-dimensional dataset, after dimensionality reduction, reduces the data dimensionality while maintaining its essential characteristics, thus facilitating subsequent data analysis and processing.

[0049] The objective function is constructed by calculating the KL divergence between the high-dimensional and low-dimensional probability distribution matrices. Its purpose is to make the data distribution in the low-dimensional space as close as possible to the data distribution in the high-dimensional space, while simultaneously reducing the dimensionality of the data. During optimization, the positions of data points in the low-dimensional space are continuously adjusted to reduce the value of the objective function. As optimization progresses, the low-dimensional probability distribution matrix gradually approximates the high-dimensional probability distribution matrix. When the objective function converges, it means that the data distribution in the low-dimensional space has sufficiently approximated the data distribution in the high-dimensional space and has reached a relatively stable state. At this point, the output fused low-dimensional dataset has the following characteristics: First, its dimensionality is significantly reduced, making it easier to perform subsequent data analysis and processing, such as visualization, clustering, and classification, compared to the original fused high-dimensional dataset. Second, although the dimensionality is reduced, due to the convergence of the objective function, this fused low-dimensional dataset still retains most of the important feature information and can better reflect the essential characteristics of the original data. The fused low-dimensional dataset output when the objective function converges achieves both dimensionality reduction and, to a certain extent, retains the key information of the original data, providing strong support for subsequent data analysis and applications.

[0050] In one possible implementation, the fused low-dimensional dataset generation module further includes:

[0051] The objective function is minimized using gradient descent to achieve convergence. The gradient calculation formula is as follows:

[0052] ;

[0053] in, Let be the objective function. Let i be the position of the i-th data point in the low-dimensional space, as gradient descent proceeds. The position of the j-th data point in the low-dimensional space. Data points in high-dimensional space and Similarity between them Data points in low-dimensional space and Similarity between them Data points in low-dimensional space and The Euclidean distance between them is used to gradually adjust the distribution of data points in the low-dimensional space based on gradient descent until the objective function converges.

[0054] Specifically, the objective function is minimized by using gradient descent to achieve convergence of the objective function, thereby obtaining the optimized low-dimensional data distribution. The objective function reflects the degree of difference between the data distributions in high-dimensional and low-dimensional spaces. In the t-SNE algorithm, the core objective is to improve the similarity between point pairs in the high-dimensional space. Similarity between points and pairs in low-dimensional space To be as close as possible means to allow the lower-dimensional space to better preserve the feature information of the higher-dimensional space.

[0055] For the position of the i-th data point in the low-dimensional space It is a key variable in the algorithm. It is continuously adjusted as gradient descent progresses. The position of the j-th data point is adjusted so that it moves in a direction that reduces the objective function value. Simultaneously, the position of the j-th data point in the low-dimensional space... This process also influences the change of the objective function.

[0056] Similarity between data points in high-dimensional space It is calculated based on the distribution of data points in high-dimensional space, reflecting the strength of the relationship between pairs of data points in high-dimensional space. The similarity between data points in low-dimensional space... It is calculated based on the positional relationship of data points in a low-dimensional space.

[0057] Data points in low-dimensional space and Euclidean distance between It measures their actual distance in space. This distance can further influence the similarity score. The calculation.

[0058] In a gradient descent-based system, the gradient of the objective function with respect to each data point in the low-dimensional space is calculated at each step. This gradient indicates the direction in which the objective function value increases the most at the current position. The task is to adjust the data point positions in the direction that the objective function value decreases for each data point in the low-dimensional space. The gradient direction and magnitude at the current position are calculated according to the gradient calculation formula. If the objective function value is large, it means that the current low-dimensional space data distribution is significantly different from the high-dimensional space data distribution. In this case, the gradient will guide the data points to move in the direction that can reduce this difference.

[0059] As gradient descent is continuously performed, the distribution of data points in the low-dimensional space gradually adjusts, and the objective function value changes in each iteration. If the objective function value gradually decreases, it indicates that the data distribution in the low-dimensional space is becoming increasingly similar to that in the high-dimensional space. When the objective function converges, it means that the data distribution in the low-dimensional space has reached a stable state, at which point the difference between the data distributions in the high-dimensional and low-dimensional spaces is minimized, resulting in an optimized fused low-dimensional dataset. This dataset retains important features of the high-dimensional space while reducing the data dimensionality, facilitating subsequent data analysis and processing. By minimizing the objective function using gradient descent and adjusting the positions of data points in the low-dimensional space to minimize the difference in data distribution between the high-dimensional and low-dimensional spaces, an optimized fused low-dimensional dataset is obtained, which is convenient for subsequent data analysis and processing while preserving important features of the high-dimensional space.

[0060] In one possible implementation, the fused low-dimensional dataset generation module further includes:

[0061] Correlation analysis is performed on the fused high-dimensional dataset using the device control parameter items to determine the relevant high-dimensional datasets in the fused high-dimensional dataset.

[0062] The t-SNE algorithm is used to retain the relevant high-dimensional dataset and perform feature dimensionality reduction on the remaining fused high-dimensional dataset, outputting the updated fused low-dimensional dataset.

[0063] Specifically, correlation analysis is performed on the fused high-dimensional dataset using equipment control parameters. These parameters are generated based on fault information from new energy power generation equipment and are closely related to the equipment's operating status and control. The purpose of correlation analysis is to determine which data points in the fused high-dimensional dataset have a strong correlation with the equipment control parameters. Various correlation analysis systems, such as Pearson correlation coefficient and Spearman correlation coefficient, can be used to measure the linear or nonlinear relationship between various features in the fused high-dimensional dataset and the equipment control parameters. If a feature has a high correlation coefficient with a control parameter, it indicates that this feature significantly influences the equipment's control decisions, and the data point containing this feature is considered part of the relevant high-dimensional dataset. After correlation analysis, the relevant high-dimensional dataset within the fused high-dimensional dataset is determined. This relevant high-dimensional dataset contains data points closely related to the equipment control parameters, which have significant reference value for the equipment's adaptive control.

[0064] The t-SNE algorithm is used to retain relevant high-dimensional datasets, and feature reduction is performed on the remaining fused high-dimensional datasets to output an updated fused low-dimensional dataset. The t-SNE algorithm has already been used in previous steps for feature reduction of the fused high-dimensional dataset; its re-application here aims to further reduce the data dimensionality while preserving important information, retaining only the relevant high-dimensional datasets, as these data are closely related to device control and their importance should not be lost during dimensionality reduction. For the remaining fused high-dimensional dataset (excluding the relevant high-dimensional datasets), t-SNE is applied again for feature reduction. The t-SNE algorithm constructs probability distributions between data points in both high-dimensional and low-dimensional spaces and minimizes the difference between these two distributions to find the representation in the low-dimensional space. This process effectively removes noise and redundant information while preserving the main features of the data. After processing with the t-SNE algorithm, an updated fused low-dimensional dataset is obtained. This dataset includes the retained relevant high-dimensional datasets and, through dimensionality reduction of the remaining data, further reduces its dimensionality, making the data easier to analyze and process subsequently. The updated fused low-dimensional dataset can better serve the adaptive control of new energy power generation equipment, providing more accurate information support for the efficient operation and optimization of the equipment.

[0065] In one possible implementation, the multi-source high-dimensional dataset output module further includes:

[0066] The time feature convolution includes performing convolution operations on the time series data of the multi-source power dataset to extract periodic features and trend features.

[0067] The spatial feature convolution includes performing convolution operations on the spatial data of the multi-source power dataset, including the correlation between different devices and the operating status of devices at different locations.

[0068] The type feature convolution involves performing convolution operations on different types of data in the multi-source power dataset to extract the correlation features between different types of data.

[0069] Specifically, temporal feature convolution primarily operates on time-series data in multi-source power datasets. Time-series data includes variations in equipment output power over time, voltage and current fluctuations, etc. By performing convolution operations on this time-series data, periodic and trend features can be extracted. Periodic features refer to recurring patterns in the data over time. For example, the output power of new energy power generation equipment exhibits periodic changes within a day, with higher power during the day and lower power at night. Temporal feature convolution can capture this periodic pattern, providing important reference information for equipment control and optimization. Trend features reflect the changing trends of the data over a longer time range. For example, the output power of equipment may show an increasing or decreasing trend due to equipment aging or changes in environmental factors. Temporal feature convolution can extract these trend features, helping to predict the future state of the equipment and take corresponding maintenance measures in advance.

[0070] Spatial feature convolution focuses on the spatial data of multi-source power datasets. In renewable energy power plants, different devices are distributed in different locations, and they interact with each other. Spatial feature convolution can capture these spatial relationships and extract useful feature information. On one hand, spatial feature convolution can consider the correlation between different devices. For example, there are airflow interactions between multiple wind power generation devices, or power transmission relationships between solar power generation devices and energy storage devices. By performing convolution operations on spatial data, the correlation features between these devices can be extracted, providing a basis for coordinated device control. On the other hand, spatial feature convolution can also consider the differences in the operating status of devices at different locations. For example, environmental factors such as light intensity and wind speed may vary in different geographical locations, which will lead to differences in the output power and operating efficiency of renewable energy power generation devices at different locations. Spatial feature convolution can capture these spatial differences, providing a reference for device layout optimization and operation management.

[0071] Type-based convolution operates on different data types within a multi-source power dataset. Multi-source power datasets for new energy power generation equipment contain various data types, such as electrical parameter data, environmental factor data, and equipment status data. By performing convolution operations on different data types, correlation features between them can be extracted. For example, the relationship between light intensity and the output power of solar power generation equipment, or the relationship between wind speed and the output power of wind power generation equipment, can be discovered. These correlation features help to better understand the operating mechanisms and performance characteristics of the equipment, providing more comprehensive information support for equipment control and optimization. In summary, through temporal feature convolution, spatial feature convolution, and type-based feature convolution, rich feature information can be extracted from multi-source power datasets, providing strong support for subsequent data analysis and adaptive equipment control.

[0072] In one possible implementation, the adaptive control module further includes:

[0073] The new energy power generation equipment is used to acquire multiple sets of multi-source power datasets under multiple operating modes, including load mode, non-load mode, energy storage mode and fault mode.

[0074] Obtain multiple sets of fused low-dimensional datasets corresponding to the multiple sets of multi-source power datasets.

[0075] A multi-mode-device adaptive control model is trained based on the multiple sets of fused low-dimensional datasets, and the new energy power generation equipment is subjected to mode adaptive control using the multi-mode-device adaptive control model.

[0076] Specifically, the first step is to acquire multiple sets of multi-source power datasets for new energy power generation equipment under various operating modes. New energy power generation equipment typically operates in different modes, reflecting its status under different operating conditions. These multiple operating modes include load mode, off-load mode, energy storage mode, and fault mode. In load mode, the new energy power generation equipment is supplying power to an external load, and its output power and voltage parameters are adjusted according to the load demand. In off-load mode, the equipment is idle or only supplies power to the internal system; its operating parameters differ from load mode. Energy storage mode typically involves storing excess electrical energy for later use; the equipment's operation in this mode also has its characteristics. Fault mode is a special state when the equipment malfunctions, and the data in this mode contains fault characteristic information. By collecting multi-source power datasets from these different operating modes, a comprehensive understanding of the equipment's operating status and data characteristics under various conditions can be achieved. Multi-source power datasets include data from multiple sources, such as the equipment's electrical parameters, environmental data, and equipment status information.

[0077] Multiple sets of multi-source power datasets were obtained, each corresponding to a fused low-dimensional dataset. For each acquired dataset, the preceding steps were performed, including multi-scale feature convolution, fusion, and t-SNE dimensionality reduction, ultimately yielding the corresponding fused low-dimensional dataset. These fused low-dimensional datasets retain important feature information from the original multi-source power datasets, but with significantly reduced dimensionality, facilitating subsequent analysis and processing. Each fused low-dimensional dataset reflects the operational characteristics of new energy power generation equipment under a specific operating mode.

[0078] A multi-mode-equipment adaptive control model is trained using the support vector machine algorithm based on multiple sets of fused low-dimensional datasets, and this model is used to perform mode adaptive control on new energy power generation equipment.

[0079] The multi-set fused low-dimensional dataset is obtained by multi-scale feature extraction, fusion and dimensionality reduction of multiple sets of multi-source power datasets of new energy power generation equipment under different working modes. These datasets retain key feature information and reduce data dimensionality, providing an efficient data foundation for model training.

[0080] Dividing multiple fused low-dimensional datasets into training and test sets is a common practice, typically using a specific ratio, such as 70% for training and 30% for testing. This ensures the model has sufficient data to learn patterns during training, while also allowing for validation on the test set to evaluate its performance and generalization ability.

[0081] The kernel function for the support vector machine (SVM) is determined first. Common kernel functions include linear kernel functions, polynomial kernel functions, and radial basis function (RBF) kernel functions. The RBF kernel function performs well in handling nonlinear problems, mapping data to a high-dimensional space to better capture complex relationships within the data. Choosing appropriate kernel function parameters, such as the bandwidth parameter of the RBF kernel function, is crucial to model performance. Optimal parameter values ​​can be found through systems like cross-validation. Cross-validation further divides the training set into multiple subsets, allowing training and validation on different subset combinations to find the best parameter settings.

[0082] The support vector machine model is trained using the training set data, with samples from the fused low-dimensional dataset as input and the corresponding device control parameters as the output target.

[0083] Support Vector Machines (SVMs) find a hyperplane that makes most of the sample points in the training dataset as close as possible to this hyperplane, while also ensuring that the hyperplane has a certain degree of generalization ability. During training, the model searches for support vectors that play a key role in determining the hyperplane. These support vectors are a subset of sample points in the training dataset that lie on or near the boundary of the hyperplane.

[0084] By optimizing the sequence minimum optimization algorithm to solve for the parameters of the support vector machine model, the parameters of the support vector machine can be solved quickly, thus improving training efficiency.

[0085] The trained support vector machine model is evaluated using test set data. Metrics such as mean squared error and mean absolute error are calculated to assess the model's predictive accuracy and generalization ability. If the model performs poorly on the test set, adjusting the kernel function parameters and increasing training data can improve the model. When the renewable energy power generation equipment is running, its operating mode is monitored in real time. Based on the current operating mode, a corresponding fused low-dimensional dataset is selected. This fused low-dimensional dataset is input into the trained multi-mode-device adaptive control model. The model predicts suitable control parameters based on the input data and the current operating mode. These control parameters may include adjustments to electrical parameters such as output power, voltage, and current, or changes in the equipment's operating mode and control strategy. Finally, the predicted control parameters are sent to the renewable energy power generation equipment's control system, which executes the corresponding control operations. In this way, the multi-mode-device adaptive control model can automatically adjust control parameters according to the equipment's real-time operating mode, achieving mode-adaptive control of the renewable energy power generation equipment. The support vector machine algorithm can be used to effectively train a multi-mode-device adaptive control model, and this model can be used to perform mode-adaptive control on new energy power generation equipment. This system can improve the efficiency, stability and reliability of the equipment in different operating modes, and provide strong support for the intelligent control of new energy power generation equipment.

[0086] In one possible implementation, the adaptive control module further includes:

[0087] Obtain the real-time multi-source power dataset of the new energy power generation equipment.

[0088] Based on the real-time multi-source power dataset, output a real-time fused low-dimensional dataset.

[0089] The real-time fused low-dimensional dataset is input into the device adaptive control model to obtain adaptive control parameters that meet the preset power demand, and the new energy power generation equipment is adaptively controlled using the adaptive control parameters.

[0090] Specifically, it involves acquiring real-time multi-source power datasets of new energy power generation equipment. This dataset contains various information about the equipment at the current moment, such as electrical parameters (voltage, current, power, etc.), environmental data (light intensity, wind speed, temperature, etc.), and equipment status information. These data can be collected in real time through sensors and monitoring equipment to understand the equipment's operating status promptly.

[0091] Based on real-time multi-source power datasets, a real-time fused low-dimensional dataset is output. This process is similar to the previous processing of multiple sets of multi-source power datasets, but it is for real-time data. Through multi-scale feature extraction, fusion, and dimensionality reduction, the high-dimensional real-time multi-source power dataset is transformed into a low-dimensional fused low-dimensional dataset. The advantage of this is that it reduces the complexity of the data, facilitates subsequent analysis and processing, and also preserves the key feature information in the data.

[0092] The real-time fused low-dimensional dataset is input into the device's adaptive control model. This model, trained using multiple sets of fused low-dimensional datasets, can predict adaptive control parameters to meet preset power demands based on the input data. For example, if the preset power demand is to maintain a stable output power under certain load conditions, the model will predict appropriate control parameters based on information from the real-time fused low-dimensional dataset. This could include adjusting electrical parameters such as the device's output voltage and current, or changing the device's operating mode. Once the adaptive control parameters are obtained, they can be used to adaptively control the renewable energy power generation equipment. The control system will adjust the equipment's operating state according to these parameters to achieve the goal of meeting the preset power demand. This real-time adaptive control improves the efficiency, stability, and reliability of renewable energy power generation equipment, better adapting to different operating conditions and power demands. This constitutes a complete real-time adaptive control process, ensuring efficient and stable operation of the equipment and meeting various power demands through real-time monitoring and control.

[0093] In one possible implementation, the adaptive control module further includes:

[0094] If the new energy power generation equipment is a hybrid fuel power generation mode, analyze the data correlation of each mode in the hybrid fuel power generation mode and output the correlation matrix.

[0095] The model parameters of the multi-scale convolutional network layer are learned based on the correlation matrix.

[0096] Specifically, when new energy power generation equipment operates in a hybrid fuel power generation mode, the Pearson correlation coefficient algorithm is used to analyze the data correlation of each mode and output a correlation matrix. The Pearson correlation coefficient is an indicator that measures the degree of linear correlation between two variables. For different modes under the hybrid fuel power generation mode, the multi-source power datasets for each mode are first organized. For each pair of different modes, the Pearson correlation coefficient between each characteristic variable in its dataset is calculated. For example, for the power output time series data of two modes, they are treated as two sets of variables. By calculating the Pearson correlation coefficient, the degree of linear correlation between these two time series is determined. If the correlation coefficient is close to 1, it indicates a strong positive correlation between the two modes in terms of power output over time, meaning their trends are very similar. If the correlation coefficient is close to -1, it indicates a strong negative correlation, meaning that when the power output of one mode increases, the power output of the other mode tends to decrease. If it is close to 0, it indicates a weak linear correlation, and the relationship between the two modes in the power output time series is not obvious. For other characteristic variables, such as fuel consumption rate, equipment temperature, voltage, and current, the Pearson correlation coefficient is also calculated one by one. By comprehensively analyzing the correlation coefficients of various feature variables under different modes, a correlation matrix is ​​constructed. Each element in the matrix represents the Pearson correlation coefficient value of two specific modes on a certain feature variable. Using the Pearson correlation coefficient algorithm, the data correlation between different composite fuel power generation modes is accurately quantified, providing an important basis for subsequent model parameter learning of multi-scale convolutional network layers and adaptive control of new energy power generation equipment.

[0097] The correlation matrix provides crucial guidance for parameter learning in multi-scale convolutional network layers. This matrix reflects the data correlations between various modes in a multi-fuel power generation model. By analyzing the values ​​in the matrix, we can understand the similarity and association between different modes on different features. For learning convolution weights, the correlation matrix is ​​used to adjust the weight allocation on different feature channels. If two modes have a high correlation on a certain feature, then the convolution weights on the corresponding feature channel are appropriately increased to better capture and utilize this correlation. The correlation matrix also provides valuable reference for learning the feature convolution scale. Different modes may exhibit correlations at different temporal, spatial, and typological feature scales. If the correlation matrix indicates that some modes have a high correlation at a specific time scale, then the scale parameters of the temporal feature convolutions in the multi-scale convolutional network layers are adjusted to better suit the data of these related modes. Similarly, the scale parameters of spatial and typological feature convolutions can also be adjusted accordingly based on the correlation matrix. The specific learning process employs optimization algorithms such as gradient descent. By calculating the gradient of the loss function with respect to the convolution weights and feature convolution scales, the parameters are updated based on the direction and magnitude of the gradient. When calculating the loss function, the correlation information between modes reflected by the correlation matrix is ​​considered, so that the network can better adapt to the characteristics of the composite fuel power generation mode during the learning process, and improve the processing ability and accuracy of different mode data.

[0098] Example 2: Based on the same inventive concept as the adaptive control system for new energy power generation equipment based on multi-scale data fusion in the previous examples, this example provides an electronic device. Figure 3 This is a schematic diagram of the structure of an electronic device provided in Embodiment 2 of the present invention, showing a block diagram of an exemplary electronic device suitable for implementing the embodiments of the present invention. Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality or scope of the embodiments of the present invention. Figure 3 As shown, the electronic device includes a processor 21, a memory 22, an input device 23, and an output device 24; the number of processors 21 in the electronic device can be one or more. Figure 3 Taking a processor 21 as an example, the processor 21, memory 22, input device 23, and output device 24 in an electronic device can be connected via a bus or other means. Figure 3 Taking the example of a connection between China and Israel via a bus.

[0099] Example 3: Based on the same inventive concept as the adaptive control system for new energy power generation equipment based on multi-scale data fusion in the previous examples, this example provides a computer-readable storage medium for storing software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the adaptive control system for new energy power generation equipment based on multi-scale data fusion in this application. The processor executes various functional applications and data processing of the computer device by running the software programs, instructions, and modules stored in the memory, thereby realizing the aforementioned adaptive control system for new energy power generation equipment based on multi-scale data fusion.

[0100] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0101] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0102] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. An adaptive control system for new energy power generation equipment based on multi-scale data fusion, characterized in that, The system includes: The multi-source power dataset acquisition module is used to acquire new energy fuel information of new energy power generation equipment, determine whether the new energy power generation equipment is a composite fuel power generation mode based on the new energy fuel information, and if the new energy power generation equipment is a composite fuel power generation mode, acquire the multi-source power dataset corresponding to each composite fuel type. The multi-source high-dimensional dataset output module is used to perform multi-scale feature convolution on the multi-source power dataset through multi-scale convolutional network layers to output the multi-source high-dimensional dataset. The multi-scale feature convolution includes temporal feature convolution, spatial feature convolution and type feature convolution. A high-dimensional dataset generation module is used to fuse the multi-source high-dimensional datasets to generate a fused high-dimensional dataset. The equipment control parameter item generation module is used to receive equipment fault information from the new energy power generation equipment and generate equipment control parameter items based on the equipment fault information. A low-dimensional dataset generation module is used to perform feature dimensionality reduction on the fused high-dimensional dataset using the t-SNE algorithm, and output a fused low-dimensional dataset. An adaptive control module is used to train an adaptive control model for the device using the fused low-dimensional dataset and the device control parameter items, and to perform adaptive control on the new energy power generation equipment based on the device adaptive control model.

2. The system as described in claim 1, characterized in that, The low-dimensional dataset generation module further includes: Perform Gaussian distribution calculation on each data point in the fused high-dimensional dataset in high-dimensional space, and output a high-dimensional probability distribution matrix; Perform t-distribution calculation on each data point in the fused low-dimensional dataset in low-dimensional space, and output a low-dimensional probability distribution matrix; Calculate the KL divergence between the high-dimensional probability distribution matrix and the low-dimensional probability distribution matrix, and construct an objective function using the KL divergence, wherein the KL divergence represents the difference in probability distribution between the high-dimensional probability distribution matrix and the low-dimensional probability distribution matrix; When the objective function converges, the output is a fused low-dimensional dataset.

3. The system as described in claim 2, characterized in that, The low-dimensional dataset generation module further includes: ； in, Let be the objective function. Let be the position of the i-th data point in the low-dimensional space, as gradient descent proceeds. The position of the j-th data point in the low-dimensional space. Data points in high-dimensional space and Similarity between them Data points in low-dimensional space and Similarity between them Data points in low-dimensional space and The Euclidean distance between them is used to gradually adjust the distribution of data points in the low-dimensional space based on gradient descent until the objective function converges.

4. The system as described in claim 1, characterized in that, The low-dimensional dataset generation module further includes: Correlation analysis is performed on the fused high-dimensional dataset using the device control parameter items to determine the relevant high-dimensional datasets in the fused high-dimensional dataset; The t-SNE algorithm is used to retain the relevant high-dimensional dataset and perform feature dimensionality reduction on the remaining fused high-dimensional dataset, outputting the updated fused low-dimensional dataset.

5. The system as described in claim 1, characterized in that, The multi-source high-dimensional dataset output module also include: The time feature convolution includes performing convolution operations on the time series data of the multi-source power dataset to extract periodic features and trend features; The spatial feature convolution includes performing convolution operations on the spatial data of the multi-source power dataset, including the correlation between different devices and the operating status of devices at different locations; The type feature convolution involves performing convolution operations on different types of data in the multi-source power dataset to extract the correlation features between different types of data.

6. The system as described in claim 1, characterized in that, The adaptive control module also includes: Obtain multiple sets of multi-source power datasets of the new energy power generation equipment under multiple operating modes, wherein the multiple operating modes include load mode, non-load mode, energy storage mode and fault mode; Obtain multiple sets of fused low-dimensional datasets corresponding to the multiple sets of multi-source power datasets; A multi-mode-device adaptive control model is trained based on the multiple sets of fused low-dimensional datasets, and the new energy power generation equipment is subjected to mode adaptive control using the multi-mode-device adaptive control model.

7. The system as described in claim 1, characterized in that, The adaptive control module also includes: Obtain the real-time multi-source power dataset of the new energy power generation equipment; Based on the real-time multi-source power dataset, output a real-time fused low-dimensional dataset; The real-time fused low-dimensional dataset is input into the device adaptive control model to obtain adaptive control parameters that meet the preset power demand, and the new energy power generation equipment is adaptively controlled using the adaptive control parameters.

8. The system as described in claim 1, characterized in that, The adaptive control module also includes: If the new energy power generation equipment is a hybrid fuel power generation mode, analyze the data correlation of each mode in the hybrid fuel power generation mode and output the correlation matrix; The model parameters of the multi-scale convolutional network layer are learned based on the correlation matrix.

9. An electronic device, characterized in that, The electronic device includes: Memory, used to store executable instructions; The processor, when executing executable instructions stored in the memory, implements the adaptive control system for new energy power generation equipment based on multi-scale data fusion as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the adaptive control system for new energy power generation equipment based on multi-scale data fusion as described in any one of claims 1-8.

Images