A Method and System for Adaptive Improvement of Photovoltaic Power Generation Performance Based on Multimodal Sensing

Abstract

This application relates to the field of photovoltaic power generation and discloses a method and system for adaptively improving photovoltaic power generation performance based on multimodal sensing. The method includes the following steps: Real-time acquisition of data on light intensity, temperature, wind speed, electrical parameters, and meteorological images through a multimodal sensor network; construction of a dynamic four-dimensional tensor; extraction of spatiotemporal correlation features using an incremental tensor decomposition algorithm to achieve dynamic modeling of the photovoltaic system state; division of the photovoltaic array into multiple sub-regions; collaborative optimization using distributed optimization algorithms and sparse communication protocols to maximize the system's power generation efficiency while ensuring voltage and temperature constraints; mapping the optimized parameters to a nonlinear manifold space; and adjusting the control input using gradient descent based on the Riemannian manifold control law to achieve globally optimal control of the system. This invention can improve the power generation efficiency of photovoltaic systems in dynamic environments, reduce communication load, and enhance system stability.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation technology, specifically to a method and system for adaptively improving photovoltaic power generation performance based on multimodal sensing. Background Technology

[0002] Photovoltaic power generation, as an important form of clean energy, is widely used in residential, industrial, and power grid systems. Its basic system typically consists of photovoltaic arrays, inverters, controllers, and monitoring units, and is a key component of the modern energy structure. However, despite continuous advancements in photovoltaic power generation technology, existing photovoltaic systems still face many technical challenges, particularly in terms of system operating efficiency and stability.

[0003] Currently, most photovoltaic (PV) systems still rely on traditional maximum power point tracking (MPPT) algorithms. These algorithms are primarily based on single sensor data and ignore various environmental factors that dynamically change within the system, such as sunlight intensity, temperature, and wind speed. This makes it difficult for the system to achieve optimal power generation efficiency in complex environments. Furthermore, existing technologies generally suffer from excessive communication overhead, especially in large-scale PV systems. Frequent data exchange between sub-regions often leads to excessive network load, affecting the system's real-time performance and response speed.

[0004] Furthermore, although manifold control theory has been applied to some optimization control problems, its application in photovoltaic systems is still in its early stages. Traditional optimization methods fail to effectively incorporate the nonlinear characteristics of photovoltaic systems, and most methods cannot fully consider the influence of complex constraints such as temperature and voltage, resulting in the inability to achieve globally optimal control performance. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for adaptively improving photovoltaic power generation performance based on multimodal sensing, which improves the power generation efficiency of photovoltaic systems in dynamic environments, while reducing communication load and ensuring system stability.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for adaptively improving photovoltaic power generation performance based on multimodal sensing, comprising the following steps:

[0007] Collect multimodal data of photovoltaic power generation systems, including irradiance, temperature, wind speed, electrical parameters, and meteorological image data;

[0008] The multimodal data is constructed into a dynamic tensor, and spatiotemporal correlation features are extracted based on the incremental tensor decomposition algorithm;

[0009] The distributed optimization algorithm is used to coordinate the control of sub-regions of the photovoltaic array to generate globally optimal parameters.

[0010] The optimized parameters are mapped to a nonlinear manifold space, and the operating state of the photovoltaic system is adjusted by the manifold control law.

[0011] Preferably, the multimodal data includes:

[0012] Spatial dimension: gridded distribution data of the photovoltaic array;

[0013] Time dimension: Time series data within a sliding time window;

[0014] Modal dimensions: illumination intensity matrix, thermal imaging images, and wind speed sensor data;

[0015] Physical quantity dimensions: voltage, current, temperature.

[0016] Preferably, the step of constructing the multimodal data into a dynamic tensor includes:

[0017] The collected data on light intensity, temperature, humidity, wind speed, and electrical parameters are organized into a four-dimensional tensor. Where S represents the spatial dimension of the photovoltaic array, T represents the time dimension, M represents the modal dimension, and P represents the physical quantity dimension;

[0018] The photovoltaic array data at each moment is mapped to each layer of the tensor to generate a spatiotemporal feature matrix while maintaining the dynamic nature of the time information.

[0019] Preferably, the step of extracting spatiotemporal correlation features based on the incremental tensor decomposition algorithm includes:

[0020] The dynamic tensor x is decomposed using Tucker decomposition. t Decomposed into core tensors And factor matrix U (n) The product of (n = 1, 2, 3, 4), where the core tensor Representing spatiotemporal characteristics, factor matrix R represents the features of each dimension. n Let N be the rank of the factor matrix. n The size of the corresponding dimension;

[0021] The incremental tensor decomposition algorithm is used to update the core tensor and factor matrix by minimizing the reconstruction error. The objective function is:

[0022]

[0023] in, The Frobenius norm, α, represents the dynamic tensor reconstruction error. t-k Here, λ is the time decay factor, and λ is the regularization parameter. For the factor matrix U(n) The Frobenius norm;

[0024] Core tensors based on incremental updates And factor matrix U (n) Spatiotemporal features are extracted to construct an environmental adaptability model for photovoltaic arrays.

[0025] Preferably, the step of coordinating the control of sub-regions of the photovoltaic array based on a distributed optimization algorithm to generate globally optimal parameters includes:

[0026] The photovoltaic array is divided into multiple sub-regions, each containing multiple photovoltaic units, and a local optimization objective is assigned to each sub-region.

[0027] The alternating direction multiplier method is used to perform collaborative optimization of sub-regions, and the global optimum is achieved by iteratively updating local control variables.

[0028] Calculate the local optimization function and generate the globally optimal control parameters based on the global optimization constraints.

[0029] Preferably, the objective function for the coordinated control of sub-regions of the photovoltaic array based on the distributed optimization algorithm is:

[0030]

[0031] Among them, f i (x i Let x be the local optimization objective function for the i-th sub-region. i Let A be the control variable for the i-th sub-region. i Let b be the constraint matrix. i Here, K is the constraint vector, and K is the total number of sub-regions.

[0032] In each iteration, the local optimization variables are updated using the following update formula:

[0033]

[0034] in, Let ρ be the control variable for the i-th sub-region in the k-th iteration, and let ρ be the step size factor. Let λ be the gradient of the local objective function. i For Lagrange multipliers, This is the transpose of the constraint matrix of the i-th sub-region;

[0035] Through multiple iterations and updates, the parameters converge to the globally optimal parameters.

[0036] Preferably, the constraints in the collaborative optimization process include:

[0037] Global voltage constraints: Where V i V is the voltage of the i-th sub-region. max This is the maximum voltage of the system;

[0038] Global temperature constraints: Where T i Let T be the temperature of the i-th sub-region. th This represents the system's maximum temperature threshold.

[0039] Communication constraints between sub-regions: Control information can only be exchanged with adjacent sub-regions.

[0040] Preferably, the step of mapping the optimized parameters to the nonlinear manifold space includes:

[0041] The current-voltage characteristics of the photovoltaic system are modeled as a Riemannian manifold. manifold It has a nonlinear geometric structure;

[0042] By calculating the geodesic distance of the system, a control objective function is established and mapped to the manifold space;

[0043] Based on the manifold control law, the operating state of the photovoltaic system is optimized by adjusting the control variables.

[0044] Preferably, the step of adjusting the operating state of the photovoltaic system through the manifold control law includes:

[0045] The control objective function of the photovoltaic system is defined as follows:

[0046]

[0047] Where x(t) is the state of the system at time t, x opt This is the globally optimal state. The distance of the geodesic line on the surface;

[0048] By optimizing the above control objective function, the optimal control input u(t) is obtained and mapped onto the manifold to control the system parameters;

[0049] The control input u(t) is updated using gradient descent and the system is kept in an optimized state.

[0050] This invention also provides a photovoltaic power generation performance adaptive improvement system based on multimodal sensing, comprising:

[0051] A multimodal data acquisition module is used to acquire light intensity, temperature, wind speed, and electrical parameters.

[0052] The dynamic tensor decomposition module is configured to perform incremental tensor decomposition and extract spatiotemporal features.

[0053] The distributed optimization module is used to coordinate the control of sub-regions and generate globally optimal parameters;

[0054] The manifold embedding control module is configured to map parameters to a nonlinear manifold space and adjust the system's operating state.

[0055] This invention provides a method and system for adaptively improving photovoltaic power generation performance based on multimodal sensing.

[0056] It has the following beneficial effects:

[0057] 1. This invention, by modeling the current-voltage characteristics of a photovoltaic array and employing a distributed optimization algorithm, can significantly improve the power generation efficiency of a photovoltaic system while satisfying voltage and temperature constraints, ensuring optimal performance under different environmental conditions.

[0058] 2. This invention employs a sparse communication protocol, which allows only similar sub-regions to exchange information based on the similarity of each sub-region, reducing unnecessary communication overhead and thus effectively reducing the communication load of the system and improving the communication efficiency of the entire system.

[0059] 3. This invention utilizes manifold control laws in Riemannian manifold space to adjust the control input based on the dynamic state of the photovoltaic system, ensuring the system remains in an optimized state. This geodesic distance-based control method enables the system state to more accurately approach the global optimum, thereby achieving more efficient energy conversion.

[0060] 4. This invention can respond to environmental changes in real time and make dynamic adjustments based on multimodal sensing data, enabling the photovoltaic system to adapt to different operating environments, maintain optimal working conditions, and avoid performance degradation due to environmental changes.

[0061] 5. By comprehensively considering various constraints such as temperature and voltage, and using global optimization objectives for coordinated adjustment, this invention ensures that the photovoltaic system can effectively avoid potential risks such as excessive temperature while generating electricity efficiently, thereby enhancing the stability and reliability of the system. Attached Figure Description

[0062] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0063] Figure 2 This is a schematic diagram of the system architecture of the present invention.

[0064] Among them, 10 is the multimodal data acquisition module; 20 is the dynamic tensor decomposition module; 30 is the distributed optimization module; and 40 is the manifold embedding control module. Detailed Implementation

[0065] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] Please see the appendix Figure 1 This invention provides a method for adaptively improving photovoltaic power generation performance based on multimodal sensing. It aims to enhance the performance of photovoltaic power generation systems by integrating various environmental and electrical parameter data and combining them with advanced mathematical and control algorithms. This method fully considers the multimodal data characteristics of the photovoltaic system during operation, and through precise modeling and optimized control, achieves the goals of improving photovoltaic system power generation efficiency and extending system lifespan.

[0067] like Figure 1 As shown, the adaptive improvement method for photovoltaic power generation performance based on multimodal sensing may include the following steps:

[0068] S1. Collect multimodal data from the photovoltaic power generation system;

[0069] S2. Construct multimodal data into dynamic tensors and extract spatiotemporal correlation features based on incremental tensor decomposition algorithm;

[0070] S3. Based on the distributed optimization algorithm, the sub-regions of the photovoltaic array are coordinated and controlled to generate the globally optimal parameters;

[0071] S4. Map the optimized parameters to a nonlinear manifold space and adjust the operating state of the photovoltaic system through the manifold control law.

[0072] The following is a detailed description of each step in the method of the present invention, providing a comprehensive explanation of the specific implementation principles, technical details, and processes of each step.

[0073] Step S1 involves the acquisition of multimodal data from the photovoltaic power generation system, specifically the real-time acquisition and recording of data on irradiance, temperature, wind speed, electrical parameters, and meteorological images. This data provides the foundation for subsequent optimization and control, ensuring that the photovoltaic system's state information is accurately reflected, thereby supporting efficient performance improvements.

[0074] First, the system includes a light intensity sensor, a temperature and humidity sensor, a wind speed sensor, a current and voltage sensor, and a meteorological image acquisition device. The light intensity sensor collects the direct and diffuse light intensity of the area; the temperature and humidity sensor monitors the temperature of the photovoltaic modules and the ambient temperature; and the wind speed sensor measures wind speed data. The current and voltage sensors are used to collect the system's electrical parameters (voltage, current, etc.), and the meteorological image acquisition device records meteorological image data of the environment (e.g., cloud cover changes, weather conditions).

[0075] The light intensity sensor can acquire the real-time light intensity of the environment surrounding the photovoltaic system. The data acquisition time can be set according to actual needs, typically sampling every minute or hour. Specifically, the light intensity data includes solar radiation intensity and diffuse light intensity, reflecting changes in solar energy. The acquisition of light intensity data can be accomplished through an integrated photovoltaic sensor module.

[0076] Temperature and humidity sensors are responsible for measuring the temperature of the photovoltaic modules and the surrounding environment in real time, and monitoring each photovoltaic unit through temperature sensors. The sensor operates based on thermocouple or thermistor principles, and the data acquisition accuracy can be adjusted according to system requirements. Temperature data is a crucial parameter affecting photovoltaic power generation efficiency, as increased temperature leads to a decrease in the conversion efficiency of photovoltaic cells.

[0077] Wind speed sensors measure wind speed based on ultrasonic or mechanical principles, monitoring its impact on heat dissipation and power generation efficiency. Wind speed data provides a valuable reference for changes in ambient temperature, especially important when photovoltaic systems experience significant temperature fluctuations, enabling more accurate and optimized control.

[0078] Current and voltage sensors collect electrical parameters such as current, voltage, and power, and transmit the data to the processing module through a corresponding data acquisition module. These sensors can monitor the operating status of the photovoltaic array in real time, especially under changes in system load or sunlight, helping to assess power generation efficiency.

[0079] Meteorological image acquisition equipment, such as high-resolution cameras or infrared thermal imaging devices, is used to monitor meteorological conditions around photovoltaic systems in real time. Meteorological image data helps determine the impact of weather changes on photovoltaic systems, such as cloud cover or sudden weather events, thereby optimizing the regulation strategies for photovoltaic systems.

[0080] The multimodal data includes sensor data from multiple dimensions. All data is collected synchronously through the data acquisition module and transmitted to the central data processing module. Data transmission employs either wireless or wired communication technology to ensure real-time transmission and processing of the collected data. Through synchronously transmitted multimodal data, the system can comprehensively monitor the status of each photovoltaic module and its operating environment, thereby accurately predicting the system's power generation efficiency.

[0081] Next, the collected multimodal data will undergo denoising and standardization processing through the data preprocessing module to ensure data accuracy and consistency. Denoising includes data filtering and outlier detection. Data standardization involves normalizing or standardizing the data from different modalities to ensure that all parameters have the same dimensions, facilitating subsequent modeling and analysis.

[0082] Ultimately, the processed multimodal data is constructed into a dynamic tensor, which contains multiple dimensions such as time, space, modes, and physical quantities, and can comprehensively describe the operating status of the photovoltaic power generation system.

[0083] By acquiring and processing this multimodal data, the system can gain a comprehensive understanding of the environment and operating status of the photovoltaic array, providing support for subsequent steps, especially in the extraction of spatiotemporal features and system optimization, ensuring the accuracy and real-time nature of decision-making.

[0084] In step S2, this embodiment involves constructing and decomposing the collected multimodal data to extract the spatiotemporal correlation features of the photovoltaic array, thereby providing support for the optimized control of the photovoltaic system. Specifically, this step includes two main sub-processes: data construction and tensor decomposition. Through these two sub-processes, the potential patterns and dynamic changes in the operating state of the photovoltaic array can be effectively mined, thus providing a data foundation for subsequent environmental adaptability modeling.

[0085] First, the collected data on light intensity, temperature, humidity, wind speed, and electrical parameters were organized into a four-dimensional dynamic tensor. The structure of this tensor is defined as follows:

[0086]

[0087] in,

[0088] S represents the spatial dimension of the photovoltaic array, representing different sub-regions or modules;

[0089] T represents the time dimension, indicating the timestamp of data collection, usually in minutes or hours;

[0090] M represents the modal dimension, which includes data types such as light intensity, temperature, and wind speed;

[0091] P represents the dimension of physical quantities, including electrical parameters such as voltage, current, and power.

[0092] Organizing this data into a four-dimensional tensor allows for convenient mapping of the photovoltaic array's data at each moment. The photovoltaic array data at each moment is mapped to form a slice of the tensor, thereby generating a spatiotemporal feature matrix. This spatiotemporal feature matrix not only preserves the dynamic nature of temporal information but also accurately reflects the diversity of spatial dimensions. Therefore, the four-dimensional tensor X... t This provides sufficient structured data for subsequent spatiotemporal feature extraction.

[0093] After the data is constructed, step S2 further extracts spatiotemporal correlation features using an incremental tensor decomposition algorithm.

[0094] Specifically, the Tucker decomposition is used to decompose the dynamic tensor χ. t Decomposition is performed. Tucker decomposition decomposes a multidimensional tensor χ. t Decomposed into core tensors and multiple factor matrices U (n) The product of these can be expressed as:

[0095]

[0096] in,

[0097] It is the core tensor, representing the core information of spatiotemporal characteristics;

[0098] It is a factor matrix, representing the characteristics of each dimension of the tensor.

[0099] R n The rank of the factor matrix determines the dimension of feature extraction;

[0100] N n It represents the size of the corresponding dimension, indicating the number of elements in that dimension.

[0101] Tucker decomposition can effectively reduce the dimensionality of high-dimensional data and extract its spatiotemporal features. These features can reflect the dynamic changes of photovoltaic arrays under different times, spaces, and physical quantities, providing a foundation for subsequent model training and optimization.

[0102] Next, an incremental tensor decomposition algorithm was used to optimize the results of the Tucker decomposition. This algorithm updates the core tensor by minimizing the reconstruction error. And factor matrix U (n) The objective function for optimization is expressed as:

[0103]

[0104] in:

[0105] It is the Frobenius norm of the tensor reconstruction error, representing the difference between the decomposition result and the actual data;

[0106] α t-k It is a time decay factor, used to adjust the weight of the influence of data at different time points on the model, ensuring that the contribution of historical data to the model gradually decreases;

[0107] λ is a regularization parameter that prevents overfitting;

[0108] It is the factor matrix U (n) The Frobenius norm is used to control the complexity of the matrix.

[0109] By minimizing the objective function, the incremental tensor decomposition algorithm can continuously update the core tensor. And factor matrix U (n) The algorithm extracts spatiotemporal correlation features from photovoltaic array data through an optimization process. Its incremental update method allows the model to quickly adapt to new data in a constantly changing environment, optimizing the extraction of spatiotemporal features in real time.

[0110] Ultimately, the core tensor is based on incremental updates. And factor matrix U (n) The extracted spatiotemporal features provide data support for constructing an environmental adaptability model for photovoltaic arrays.

[0111] In step S3, this embodiment uses a distributed optimization algorithm to collaboratively control sub-regions of the photovoltaic array, generating globally optimal parameters. This step involves dividing the photovoltaic array, setting local optimization objectives, and using the Alternating Directional Multiplier Method (ADMM) for iterative optimization to ultimately determine the globally optimal control parameters. This process improves the power generation efficiency of the photovoltaic array and ensures stable system operation.

[0112] First, the photovoltaic array is divided into multiple sub-regions. Each sub-region contains multiple photovoltaic units, which can be grouped based on the similarity of their physical location or electrical parameters. Each sub-region is assigned a local optimization objective to a specific control module to ensure the optimized operation of each sub-region, ultimately achieving overall optimization through global collaborative control.

[0113] In this embodiment, the goal of the distributed optimization algorithm is to achieve global optimum by minimizing the local optimization function of each sub-region. The specific optimization objective function is:

[0114]

[0115] in,

[0116] f i (x i ) is the local optimization objective function of the i-th sub-region, representing the performance optimization objective of that sub-region;

[0117] x i It is the control variable for the i-th sub-region, used to adjust the operating state of the photovoltaic unit;

[0118] A i It is a constraint matrix that describes the constraints of this sub-region;

[0119] b i It is a constraint vector, representing the specific value of each constraint condition;

[0120] K is the total number of sub-regions.

[0121] The objective function aims to minimize the local optimization objectives of all sub-regions while ensuring that the constraints of each sub-region are satisfied. These constraints include voltage, temperature, and communication limitations between sub-regions.

[0122] During the optimization process, local optimization variables are updated in each iteration to drive the optimization of the global system. The specific update formula is as follows:

[0123]

[0124] in,

[0125] It is the control variable of the i-th sub-region in the k-th iteration;

[0126] ρ is the step size factor, which controls the magnitude of each update;

[0127] It is a local objective function The gradient represents the direction and rate of change of the objective function;

[0128] λ i It is the Lagrange multiplier for the i-th sub-region, used to balance the constraints and the optimization objective;

[0129] It is the transpose of the constraint matrix of the i-th subregion, used to adjust the control variables under constraints.

[0130] Through multiple iterations, the local control variables gradually approach the global optimal solution, and finally the optimal control parameters for all sub-regions are obtained.

[0131] In the collaborative optimization process, constraints are crucial to the optimization process and mainly include the following aspects:

[0132] Global voltage constraints:

[0133]

[0134] Among them, V i V is the voltage of the i-th sub-region. max This is the maximum voltage of the system. This constraint ensures that the voltage of the entire photovoltaic system remains within a safe range.

[0135] Global temperature constraints:

[0136]

[0137] Among them, T i Let T be the temperature of the i-th sub-region. th This represents the maximum temperature threshold of the system. This constraint is used to ensure that the system does not overheat, avoiding excessive temperature that could lead to reduced system efficiency or damage.

[0138] Communication constraints between sub-regions:

[0139] This constraint requires that sub-regions can only exchange control information with adjacent sub-regions, ensuring efficient information transmission and reducing communication costs. This constraint helps improve the efficiency of optimization algorithms and the real-time response capability of the system.

[0140] To further reduce communication overhead, a sparse communication protocol is employed. By calculating the similarity of feature matrices between sub-regions (e.g., using cosine similarity), only sub-regions with high similarity are allowed to exchange information. The formula for calculating cosine similarity is:

[0141]

[0142] If the similarity is greater than a certain preset threshold θ, then the two sub-regions are allowed to exchange intermediate variables.

[0143] Through these constraints, the system can coordinate the operating states of each sub-region to achieve global optimization while ensuring safety and stability.

[0144] In practical implementation, the Alternating Direction Multiplier Method (ADMM) is applied to the updating of local optimization variables. The advantage of ADMM lies in its ability to decompose a large-scale optimization problem into multiple smaller subproblems and achieve global optimum through iterative solutions. In each iteration, by updating the control variables of each sub-region, coordination and synchronization among the various parts of the system are ensured, ultimately yielding a globally optimal solution that satisfies all constraints.

[0145] Finally, after multiple iterations, the control variables of all sub-regions converge to the global optimum, thus obtaining the final globally optimal control parameters. This process ensures the optimal performance of the photovoltaic array under different environmental conditions and enables dynamic adjustments based on real-time data, improving the system's adaptability and reliability.

[0146] For step S4, in this embodiment, the optimized parameters are mapped to a nonlinear manifold space, and the operating state of the photovoltaic system is adjusted through a manifold control law to achieve optimized system control. The specific implementation process includes modeling the current-voltage characteristics of the photovoltaic system as a Riemannian manifold, establishing and mapping the control objective function to the manifold space, and adjusting the control input through a manifold control law to optimize the system's operating state.

[0147] First, the current-voltage characteristics of the photovoltaic system are modeled as a Riemannian manifold. The nonlinear geometry of the manifold better reflects the dynamic characteristics of the system. Assume the state space of the photovoltaic system is... Each state This represents the operating status of the photovoltaic system, such as parameters like voltage, current, and temperature. The choice of manifold is based on the physical characteristics of the system; the geometric properties of the manifold can effectively capture the nonlinear relationship between current and voltage characteristics.

[0148] Then, the control objective function is established by calculating the geodesic distance of the system. A geodesic is the shortest path between two points on a manifold, and the differences between system states can be measured using the metric structure of the manifold. Specifically, the geodesic distance of the system is:

[0149]

[0150] in,

[0151] x(t) is the state of the system at time t;

[0152] x opt It is the globally optimal state;

[0153] It is the geodesic distance between two points on the manifold, representing the difference between the system state and the optimal state.

[0154] Based on this geodesic distance, construct the control objective function:

[0155]

[0156] The significance of this control objective function lies in the fact that minimizing it can make the system's state x(t) approach the global optimal state x. opt This allows for the optimization of system performance.

[0157] Next, based on the manifold control law, the operating state of the photovoltaic system is optimized by adjusting the control input. The control objective is to minimize the aforementioned control objective function. To achieve this optimization objective, the control input u(t) is first defined, and the optimal control input u is obtained by optimizing the control objective function. opt In the manifold space, the control input u(t) needs to be mapped onto the manifold so that the system state evolves along the shortest path of the manifold to approach the global optimum.

[0158] The optimization process for the control input is achieved using gradient descent. Specifically, gradient descent minimizes the control objective function J by iteratively updating the control input u(t). The formula for updating the control input is:

[0159]

[0160] in,

[0161] u(t) (k) It is the control input in the k-th iteration;

[0162] α is the learning rate, which controls the magnitude of each update;

[0163] It is the gradient of the control objective function J(u(t)) with respect to the control input u(t), which represents the direction of change of the control input.

[0164] Through continuous iteration, the system's control input gradually approaches the optimal control input u. opt (t), thus making the system state tend towards the global optimal state.

[0165] Furthermore, to ensure the system remains in an optimal state, gradient descent is used to update the control input in each iteration, and the system state is kept close to the global optimum x by adjusting the control input. opt This process can effectively improve system performance, ensure the maximum power generation efficiency of the photovoltaic array, and avoid the adverse effects of factors such as excessively high temperatures.

[0166] In this process, the optimized parameters are mapped to a nonlinear manifold space through a manifold control law, and the control input is adjusted using the manifold control law to ultimately achieve the optimized operating state of the photovoltaic system. This process can effectively improve the system's power generation efficiency while ensuring its stability and adaptability.

[0167] Ultimately, the optimization process achieves the following effect: by optimizing the control input, the system state x(t) gradually approaches the global optimal state x. opt This maximizes the power generation efficiency of the photovoltaic system and ensures that constraints such as voltage and temperature are met.

[0168] In summary, this invention first acquires real-time data on irradiance, temperature, wind speed, electrical parameters, and meteorological images of a photovoltaic system using a multimodal sensor network, constructing a dynamic four-dimensional tensor to fuse spatiotemporal features. Then, an incremental tensor decomposition algorithm is used to extract spatiotemporal correlation features, and the core tensor is dynamically updated through Tucker decomposition and regularization optimization. Furthermore, the photovoltaic array is divided into sub-regions, and the distributed alternating direction multiplier method (ADMM) combined with a sparse communication protocol is used for collaborative optimization to maximize power generation efficiency while satisfying global voltage and temperature constraints. Finally, the optimization parameters are mapped to a nonlinear manifold space, and a control objective function is constructed based on the geometric properties of the Riemannian manifold. The control input is adjusted using gradient descent to allow the system to converge to the optimal state along the geodesic. This method enables photovoltaic systems to achieve high-efficiency power generation, low communication load, and stable operation in dynamic environments.

[0169] The photovoltaic power generation performance adaptive improvement system based on multimodal sensing described below can be referred to in correspondence with the photovoltaic power generation performance adaptive improvement method based on multimodal sensing described above.

[0170] Please see the appendix Figure 2 The present invention also provides a photovoltaic power generation performance adaptive improvement system based on multimodal sensing, comprising:

[0171] The multimodal data acquisition module 10 is used to acquire light intensity, temperature, wind speed and electrical parameters;

[0172] The dynamic tensor decomposition module 20 is configured to perform incremental tensor decomposition and extract spatiotemporal features.

[0173] The distributed optimization module 30 is used to coordinate the control of sub-regions and generate globally optimal parameters;

[0174] The manifold embedding control module 40 is configured to map parameters to a nonlinear manifold space and adjust the system operating state.

[0175] The system in this embodiment can be used to execute the above method embodiments, and its principle and technical effect are similar, so they will not be described again here.

[0176] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A photovoltaic power generation performance self-adaptive improvement method based on multi-modal perception, characterized in that, The method comprises the following steps: Collecting multi-modal data of a photovoltaic power generation system, including light intensity, temperature, wind speed, electrical parameters and meteorological image data; Constructing the multi-modal data into a dynamic tensor and extracting spatio-temporal correlation features based on an incremental tensor decomposition algorithm; Coordinately controlling sub-regions of the photovoltaic array based on a distributed optimization algorithm to generate globally optimal parameters; Mapping the optimized parameters to a nonlinear manifold space and adjusting the operating state of the photovoltaic system through a manifold control law; The step of extracting spatio-temporal correlation features based on the incremental tensor decomposition algorithm comprises: Dynamic tensors are decomposed using Tucker decomposition. Decomposed into core tensors sum factor matrix The product of the core tensor Representing spatiotemporal characteristics, factor matrix Representing the features of each dimension, Let be the rank of the factor matrix. The size of the corresponding dimension; Updating the core tensor and factor matrix by minimizing the reconstruction error using the incremental tensor decomposition algorithm, and the optimization objective function is: wherein, denotes the Frobenius norm of the dynamic tensor reconstruction error, is a time decay factor, is a regularization parameter, is a factor matrix the Frobenius norm of the factor matrix Incremental update based core tensor and factor matrices extract the spatio-temporal features and further construct the environmental adaptability model of the photovoltaic array; The step of coordinately controlling sub-regions of the photovoltaic array based on the distributed optimization algorithm to generate globally optimal parameters comprises: Dividing the photovoltaic array into multiple sub-regions, each containing multiple photovoltaic units, and assigning a local optimization objective to each sub-region; Coordinately optimizing the sub-regions using the alternating direction multiplier method, and updating the local control variables through iteration to achieve global optimization; Calculating the local optimization function and generating globally optimal control parameters according to global optimization constraints; The constraint conditions in the coordinately optimization process comprise: Global voltage constraint: wherein V is the voltage of the th sub-region, Vmaxis the maximum voltage of the system; Global temperature constraint: wherein T is the temperature of the th sub-region, Tmax is the maximum temperature threshold of the system; Communication constraints between sub-regions: only allowing exchange of control information with adjacent sub-regions.

2. The photovoltaic power generation performance self-adaptive improvement method based on multi-modal perception according to claim 1, characterized in that, The multi-modal data comprises: Spatial dimension: grid distribution data of the photovoltaic array; Temporal dimension: time series data within a sliding time window; Modal dimension: light intensity matrix, thermal imaging image, wind speed sensor data; Physical quantity dimension: voltage, current, temperature.

3. The photovoltaic power generation performance self-adaptive improvement method based on multi-modal perception according to claim 1, characterized in that, The step of constructing the multi-modal data into a dynamic tensor comprises: The collected light intensity, temperature, humidity, wind speed and electrical parameter data are organized as a four-dimensional tensor wherein represents the spatial dimension of the photovoltaic array, represents the temporal dimension, represents the modal dimension, represents the physical quantity dimension; Mapping the photovoltaic array data at each time point to each layer of the tensor to generate a spatio-temporal feature matrix, and maintaining the dynamics of the time information.

4. The photovoltaic power generation performance self-adaptive improvement method based on multi-modal perception of claim 1, characterized in that, The optimization objective function for coordinately controlling sub-regions of the photovoltaic array based on the distributed optimization algorithm is: wherein, is a local optimization objective function for the th sub-region, is a control variable for the th sub-region, is a constraint matrix, is a constraint vector, is a total number of sub-regions; In each iteration, the local optimization variable is updated by the following formula: in, For the first The sub-regions in the first Control variables in the next iteration. Step size factor The gradient of the local objective function. For Lagrange multipliers, For the first Transpose of the constraint matrix of each subregion; Through multiple iterations of updating, the globally optimal parameters are converged.

5. The photovoltaic power generation performance self-adaptive improvement method based on multi-modal perception according to claim 1, characterized in that, The step of mapping the optimized parameters to a nonlinear manifold space comprises: Modeling current-voltage characteristics of photovoltaic systems as riemannian manifolds where the manifold has a nonlinear geometric structure; Calculating the geodesic distance of the system, establishing a control objective function and mapping it to the manifold space; Based on the manifold control law, the operating state of the photovoltaic system is optimized by adjusting the control variables.

6. The photovoltaic power generation performance self-adaptive improvement method based on multi-modal perception according to claim 5, characterized in that, The step of adjusting the operating state of the photovoltaic system through the manifold control law comprises: Defining the control objective function of the photovoltaic system as: wherein, is the state of the system at time is the state of the system at time is the globally optimal state, is the geodesic distance on By optimizing the above control objective function, the optimal control input is obtained and map it onto the manifold to control the parameters of the system; updating the control input using a gradient descent method and keeping the system in an optimized state.

7. A photovoltaic power generation performance self-adaptive improvement system based on multi-modal perception, configured to perform the photovoltaic power generation performance self-adaptive improvement method based on multi-modal perception according to any one of claims 1-6, characterized in that, Comprise: A multi-modal data acquisition module for acquiring light, temperature, wind speed and electrical parameters; A dynamic tensor decomposition module configured to perform incremental tensor decomposition and extract spatio-temporal features; A distributed optimization module for coordinately controlling sub-regions and generating globally optimal parameters; A manifold embedding control module configured to map parameters to a nonlinear manifold space and adjust the operating state of the system.

Images