Solar photovoltaic system

Abstract

The invention provides a solar photovoltaic system, and belongs to the technical field of photovoltaics. The system comprises a light condensation module which is used for self-adaptively focusing according to the illumination condition and splitting light; the wide-spectrum absorption photovoltaic module is used for receiving the light transmitted by the condensation module and optimizing spectrum absorption; the intelligent control system is used for generating an optimal strategy through a preset fusion model according to the collected various aspects of data of the solar photovoltaic system, and sending a control instruction according to the optimal strategy; the storage battery module is used for storing electric energy from the wide-spectrum absorption photovoltaic module; and the electric energy output and adaptation module is used for receiving the electric energy of the storage battery module. According to the invention, the absorption range of the solar spectrum is widened; the optimal control strategy can be accurately generated in real time, and the whole solar photovoltaic system can quickly adapt to complex working conditions and always keep an efficient and stable operation state through cooperative regulation and control of all the modules.

Description

Technical Field

[0001] The present invention relates to the field of photovoltaic technology, in particular to a solar photovoltaic system. Background Art

[0002] A solar photovoltaic system is a device that converts solar energy directly into electrical energy, achieving efficient light energy capture and energy conversion by integrating multiple technologies.

[0003] The photovoltaic modules (such as crystalline silicon cells) of traditional solar photovoltaic systems absorb only approximately 40% of the visible light band of the solar spectrum. Ultraviolet light (<400nm) and infrared light (>1100nm) are wasted due to energy mismatch or thermal effects. The full-spectrum energy utilization rate is generally less than 25%. This efficiency loss is particularly significant in high-latitude or polluted areas where the spectral distribution fluctuates greatly. Furthermore, traditional solar photovoltaic system control is mostly based on fixed rules or simple feedback mechanisms. These systems are unable to cope with complex and changing environmental factors, changes in module status, and fluctuations in load demand, and are unable to make optimal control decisions in real time. Summary of the Invention

[0004] The present invention provides a solar photovoltaic system to solve the problems existing in the prior art of full spectrum energy utilization and the inability of solar photovoltaic system control to make optimal regulation decisions in real time.

[0005] In order to achieve the above-mentioned objectives, an embodiment of the present invention provides a solar photovoltaic system on the one hand, which includes: a concentrating module for adaptively focusing and splitting light according to lighting conditions; a wide-spectrum absorption photovoltaic component for receiving light transmitted from the concentrating module and optimizing spectral absorption; an intelligent control system including a distributed sensor group and an intelligent decision-making and regulation module for generating an optimal strategy based on various aspects of the solar photovoltaic system data collected by the distributed sensor group through a preset fusion model, and sending control instructions according to the optimal strategy; a battery module for storing electrical energy from the wide-spectrum absorption photovoltaic component, monitoring battery status, feeding back information to the intelligent control system, and managing battery charging and discharging; and an electric energy output and adaptation module for receiving electric energy from the battery module, and also for adapting and outputting electric energy according to load demand and grid conditions.

[0006] Optionally, the focusing module includes: a focusing lens array, each focusing lens unit in the focusing lens array is configured with an independent light sensor and an angle adjustment device, and the angle of the focusing lens unit is adjusted by the angle adjustment device; a spectral splitting and guiding device, used to split the light focused by the focusing lens array according to the wavelength range, and guide it to the corresponding light absorption area of the wide-spectrum absorption photovoltaic component.

[0007] Optionally, each of the focusing lens units includes a square plane mirror and several light-emitting lens units; a rectangular array of the several light-emitting lens units is distributed on the light-emitting surface of the square plane mirror; the surface of the light-emitting lens unit facing away from the square plane mirror is a plane, a central concave surface is provided in the middle position of the surface of the light-emitting lens unit facing the square plane mirror, and several concentric rings are distributed around the central concave surface on the surface of the light-emitting lens unit facing the square plane mirror; each of the rings is composed of several arc-shaped pieces, and the connecting ends of two adjacent arc-shaped pieces are partially offset and overlapped.

[0008] Optionally, the wide-spectrum absorption photovoltaic component includes a quantum well photovoltaic cell layer and a photochromic spectrum adjustment layer located above the quantum well photovoltaic cell layer. The quantum well photovoltaic cell layer adopts a multi-layer nanoscale quantum well structure, each layer absorbs light for a specific spectral range, and adjusts the energy level distribution through the configured temperature-adaptive electronic structure; the photochromic spectrum adjustment layer is made of photochromic material. Under light of different intensities and spectral compositions, the optical properties of the photochromic spectrum adjustment layer will automatically change to change the optical characteristic parameters for light of different wavelengths.

[0009] Optionally, the method generates an optimal strategy based on various aspects of the solar photovoltaic system data collected by the distributed sensor group through a preset fusion model, and sends control instructions based on the optimal strategy, including: preprocessing the collected various aspects of the solar photovoltaic system data; using the preset fusion model to extract key features that are valuable for state judgment and performance analysis of the solar photovoltaic system based on the preprocessed data; and analyzing the current working condition of the solar photovoltaic system based on the extracted key feature data, and searching for all corresponding action combinations in the predefined action space according to the current working condition, estimating the performance changes of the solar photovoltaic system brought about by different action combinations, and estimating the future reward value corresponding to each action combination; and using layered ε-greedy The strategy selects the current optimal action combination, which is the optimal strategy generated for the current state of the solar photovoltaic system to generate corresponding control instructions; the control instructions are sent to each corresponding module of the solar photovoltaic system to execute corresponding control.

[0010] Optionally, the estimating of the future reward value corresponding to each action combination includes: analyzing, for each action combination, its impact mechanism on the state of each component of the solar photovoltaic system and the overall operating state; determining, through a state transition probability matrix, the probability of the solar photovoltaic system transitioning from the current state to each predicted next state when executing different action combinations; based on the above analysis of the state transition of the solar photovoltaic system caused by the action combination, combined with the current state data of the solar photovoltaic system collected in real time by the distributed sensor group, determining the next state that the solar photovoltaic system immediately enters after executing a certain action combination; calculating the immediate reward value for the new state entered by the solar photovoltaic system based on a preset reward indicator system and the corresponding weight distribution; based on the determined discount factor, combined with the state transition situation and the immediate reward calculation method obtained by the previous analysis, calculating the long-term reward cumulative value of each action combination by recursion or iteration; comprehensively calculating the calculated immediate reward value and the long-term reward cumulative value to determine the future reward value corresponding to each action combination, and comparing the future reward values corresponding to different action combinations to obtain the optimal strategy.

[0011] Optionally, the control instructions are sent to the corresponding modules of the photovoltaic system to perform corresponding controls, including: controlling the action of the angle adjustment device to drive the focusing lens unit to rotate to adjust the angle of the focusing lens unit; adjusting the optical parameters of the spectral splitting and guiding device; adjusting the temperature value applied to the photochromic spectrum adjustment layer to change the transmittance and reflectivity of light of different spectra; adjusting the power storage and distribution strategy of the battery management system; adjusting the power output and the output power of the adaptation module.

[0012] Optionally, the battery module includes a solid electrolyte hybrid energy storage battery pack and a battery management system; the solid electrolyte hybrid energy storage battery pack integrates lithium-ion batteries and solid-state supercapacitors, and realizes rapid storage and release of electric energy through superconducting connection technology; the battery management system monitors battery cell status parameters, exchanges data with the intelligent control system through a distributed intelligent communication network, dynamically adjusts the charging and discharging strategy, and supplies power to the power output and adaptation module through the power transmission line.

[0013] Optionally, the power output and adaptation module includes: an adaptive intelligent inverter, which is used to adjust the waveform, frequency and voltage level of the output AC power according to the grid parameter monitoring data of the intelligent control system; a load intelligent matching and optimization unit, which is used to obtain load type and priority data through a distributed intelligent communication network, and cooperate with the intelligent control system module to realize dynamic power distribution and prioritize power supply to critical loads.

[0014] Optionally, an exciton transport interface layer is provided between the quantum well photovoltaic cell layer and the photochromic spectrum adjustment layer to optimize the photogenerated carrier transport efficiency and improve the photoelectric conversion efficiency through interface state engineering.

[0015] The solar photovoltaic system provided by the present invention broadens the absorption range of the solar spectrum through wide-spectrum absorption photovoltaic modules, and all bands from ultraviolet light to infrared light can be more efficiently absorbed and converted into electrical energy; and through the intelligent control system, it can generate the optimal control strategy in real time and accurately. Through the coordinated regulation of each module, the entire photovoltaic system can quickly adapt to complex working conditions such as different weather, time, and load changes, and always maintain an efficient and stable operating state, effectively improving the reliability and overall performance of the system, and reducing the power generation loss and failure risks caused by non-optimal regulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the present invention or the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. In the drawings: Figure 1 It is a structural block diagram of the solar photovoltaic system provided by the present invention; Figure 2 1 is a flow chart of data preprocessing provided by an embodiment of the present invention; Figure 3 is a flow chart of the control instructions provided by an embodiment of the present invention; Figure 4 This is a schematic structural diagram of a focusing lens unit irradiating light provided by an embodiment of the present invention; Figure 5 is a side view schematic diagram of a focusing lens unit irradiating light provided by an embodiment of the present invention; Figure 6 is a front view of a focusing lens unit provided by an embodiment of the present invention; Figure 7 It is a schematic diagram of the three-dimensional structure of the focusing lens unit provided by an embodiment of the present invention.

[0017] Description of Reference Numerals 10. Concentrating module; 20. Broad spectrum absorption photovoltaic module; 30. Intelligent control system; 40. Battery module; 50. Power output and adaptation module. DETAILED DESCRIPTION

[0018] The following describes the specific implementation of the embodiment of the present invention in detail with reference to the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the embodiment of the present invention and is not used to limit the embodiment of the present invention.

[0019] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application are in compliance with the relevant provisions of national laws and regulations. In the embodiments of this application, certain software, components, models, and other existing solutions in the industry may be mentioned. These should be considered as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of this application, but it does not mean that the applicant has or will necessarily use such solutions.

[0020] As mentioned earlier, the photovoltaic modules (such as crystalline silicon cells) of traditional solar photovoltaic systems absorb only approximately 40% of the visible light band of the solar spectrum. Ultraviolet light (<400nm) and infrared light (>1100nm) are wasted due to energy mismatch or thermal effects. The full-spectrum energy utilization rate is generally less than 25%. This efficiency loss is particularly significant in high-latitude or polluted areas with large spectral distribution fluctuations. Furthermore, traditional solar photovoltaic system control is mostly based on fixed rules or simple feedback mechanisms. These systems are unable to cope with complex and changing environmental factors, changes in module status, and fluctuations in load demand, and are unable to make optimal control decisions in real time.

[0021] Therefore, to address this problem, the present invention provides a solar concentrating lens, which broadens the absorption range of the solar spectrum, and all bands from ultraviolet light to infrared light can be more efficiently absorbed and converted into electrical energy; and through the intelligent control system, it can generate the optimal control strategy in real time and accurately. Through the coordinated regulation of each module, the entire photovoltaic system can quickly adapt to complex working conditions such as different weather, time, and load changes, and always maintain an efficient and stable operating state, effectively improving the reliability and overall performance of the system, and reducing the power generation loss and failure risk caused by non-optimal regulation.

[0022] Figure 1 This is a structural diagram of a solar photovoltaic system provided by an embodiment of the present invention. Please refer to Figure 1 The solar photovoltaic system may include a concentrating module 10 , a wide spectrum absorption photovoltaic component 20 , an intelligent control system 30 , a battery module 40 , and an electric energy output and adaptation module 50 .

[0023] The focusing module 10 is used to adaptively focus and split the light according to the lighting conditions; the wide-spectrum absorption photovoltaic component 20 is used to receive the light transmitted by the focusing module and optimize the spectral absorption; the intelligent control system 30 includes a distributed sensor group and an intelligent decision-making and regulation module, which is used to generate an optimal strategy based on various aspects of the solar photovoltaic system data collected by the distributed sensor group through a preset fusion model, and send control instructions according to the optimal strategy; the battery module 40 is used to store the electrical energy from the wide-spectrum absorption photovoltaic component, monitor the battery status, feedback information to the intelligent control system, and manage battery charging and discharging; the power output and adaptation module 50 is used to receive the power of the battery module, and is also used to adapt and output the power according to the load demand and grid conditions.

[0024] Preferably, the focusing module includes: a focusing lens array, each focusing lens unit in the focusing lens array is equipped with an independent light sensor and an angle adjustment device, and the angle of the focusing lens unit is adjusted by the angle adjustment device; a spectral splitting and guiding device, which is used to split the light focused by the focusing lens array according to the wavelength range and guide it to the corresponding light absorption area of the wide spectrum absorption photovoltaic component.

[0025] For example, the focusing lens array is built with micro-electromechanical system (MEMS) technology, and each focusing lens unit is extremely small and can be independently controlled. These focusing lens units are equipped with miniature light sensors and angle adjustment devices, which can sense the incident angle, intensity and spectral distribution of sunlight in real time. Based on the instructions transmitted by the intelligent control system, the MEMS drive device can accurately adjust the angle of each focusing lens unit so that it can automatically focus sunlight to the best area at different times (such as different times of the day, different seasons) and different weather conditions (sunny, cloudy, overcast, etc.), maximize the collection of light, and effectively overcome the problem of limited light collection efficiency of traditional fixed-angle focusing lenses in different lighting scenes. Assume that the incident sunlight intensity is (unit: ), the optical transmittance of the focusing lens unit is T (dimensionless, ranging from 0 to 1), and the focusing ratio is (dimensionless, indicating the multiple of the intensity of the light after convergence relative to the intensity of the incident light), the intensity of the light after focusing by the focusing lens unit Expressed as: ; The spectral beam splitting and guiding device is used in conjunction with the focusing lens array. It uses optical films and refractive elements (the optical films can be multi-layer dielectric film interference filters, photonic crystal films, etc., and the refractive elements can be aspheric beam splitting prisms, diffraction gratings, etc.) to split the light according to the different wavelengths of sunlight. For example, visible light, ultraviolet light, and infrared light are guided to different paths respectively, providing targeted spectral input for subsequent wide-spectrum absorption photovoltaic modules. Assume that the initial light incident angle is (Unit: degree), the angle adjustment device adjusts the angle change of the focusing lens unit to (Unit: degrees), adjusted light incident angle for: Δ θ The intelligent control system will dynamically adjust the optical parameters of the spectral splitting and guiding device according to the real-time monitored energy proportion of each spectral band and the absorption efficiency of the wide-spectrum absorption photovoltaic module to ensure that light in each band can be distributed and utilized in the best way. This is significantly different from the traditional method of uniformly converging light and can better adapt to the needs of wide-spectrum absorption.

[0026] like Figure 4-Figure 7 As shown, preferably, each focusing lens unit includes a square plane mirror and several light-emitting lens units; the several light-emitting lens units are distributed in a rectangular array on the light-emitting surface of the square plane mirror; the surface of the light-emitting lens unit facing away from the square plane mirror is a plane, and a central concave surface is provided in the middle of the surface of the light-emitting lens unit facing the square plane mirror, and several concentric rings are distributed around the central concave surface on the surface of the light-emitting lens unit facing the square plane mirror; each ring is composed of several arc-shaped pieces, and the connecting ends of two adjacent arc-shaped pieces are partially offset and overlapped.

[0027] For example, the concentrating lens unit is made of optical-grade PMMA material and can be sized from 110cm to 200cm. It features an ultra-thin design: only approximately 6mm thick, with a circular focused spot. Conventional plano-convex lenses are too thick, heavy, and large, resulting in reduced manufacturing precision and, in turn, lower solar energy concentration efficiency. The longer the distance sunlight travels within the optical medium of a concentrating lens, the greater the light loss. Conventional Fresnel optical lenses also have low light efficiency and excessive ineffective grooves.

[0028] Compared to conventional plano-convex lenses and Fresnel optical lenses, the present invention's concentrating lens unit has a smooth outer surface, self-cleansing in rainwater, and is less susceptible to dust obstruction. Grooves on the outer surface of the concentrating lens unit can accumulate dust and sand over time, rapidly reducing the unit's light efficiency. Its ultra-thin structure facilitates injection molding. Concentrating lens units are consumable components within a solar photovoltaic system, making the present invention's concentrating lens unit relatively low in manufacturing cost, reducing replacement costs. Each circular ring of the present concentrating lens unit is composed of several curved segments, with adjacent segments partially offset and overlapping at their ends. This linear mirror design reduces energy loss to only 10% of the incident energy, even without thermal insulation or a selective absorber surface. This outstanding performance stems from the linear mirror's ability to concentrate incident light, allowing heat from the device to be transferred away. This heat, even without thermal insulation, remains significantly higher than the heat loss. This high conversion efficiency effectively addresses the issue of insufficient solar cell conversion efficiency. The square design of the concentrating lens unit allows for more efficient space utilization, allowing more square solar lenses to be installed on the same area of land, thereby improving solar energy collection efficiency. Furthermore, the patented self-cleaning, ultra-thin, low-cost solar concentrating lens incorporates the advantages of circular lenses, eliminates the lower light efficiency disadvantages of Fresnel lenses, and offers a relatively low manufacturing cost, facilitating large-scale promotion and application. The smooth outer surface of the lens allows for self-cleaning in the rain and is less susceptible to dust obstruction, facilitating easy maintenance during use and preventing wind and sand from settling on the lens surface.

[0029] Preferably, the wide-spectrum absorption photovoltaic component includes a quantum well photovoltaic cell layer and a photochromic spectrum adjustment layer located above the quantum well photovoltaic cell layer; the quantum well photovoltaic cell layer adopts a multi-layer nanoscale quantum well structure, each layer absorbs light for a specific spectral range, and adjusts the energy level distribution through the configured temperature-adaptive electronic structure; the photochromic spectrum adjustment layer is made of photochromic material, and under light of different intensities and spectral compositions, the optical properties of the photochromic spectrum adjustment layer will automatically change to change the optical characteristic parameters for light of different wavelengths.

[0030] The core photovoltaic cell layer adopts a quantum well structure. By constructing nanoscale quantum wells in semiconductor materials, discrete energy levels can be generated, thereby achieving the precise capture and efficient absorption of photons of different energies (i.e., light of different wavelengths). Quantum wells can be designed into a multi-layer structure, with each layer optimized for a specific spectral range. For example, some layers focus on absorbing ultraviolet light, some focus on specific frequency bands in visible light, and some focus on infrared light, greatly broadening the effective utilization range of the solar spectrum. Assume that for a specific spectral range (unit: nm ), quantum well photovoltaic cell layer The absorption coefficient of the layer is (unit: ), the thickness of this layer is (Unit: cm), then the absorption efficiency of the layer for the light in this spectral range is It is expressed as (based on the simplified form of the Beer-Lambert law): ; These quantum well photovoltaic cells also have unique temperature adaptive characteristics. Their internal electronic structure will automatically adjust the energy level distribution as the temperature changes, so that they can maintain relatively stable and high photoelectric conversion efficiency at different ambient temperatures (whether it is hot in summer or cold in winter), solving the problem that traditional photovoltaic cells are greatly affected by temperature. The temperature adaptive electronic structure introduces components with different thermal expansion coefficients (such as strained layers, phase change material doping) into the quantum well material or designs dynamic defect structures. When the ambient temperature changes, the thermal expansion of the material or the defect state density changes, resulting in the reversible adjustment of the energy band structure of the quantum well (such as the position of the bottom of the conduction band and the top of the valence band) and the electronic energy level spacing, thereby matching the energy distribution of the solar spectrum at different temperatures. Assume that the quantum well photovoltaic cell layer is at a standard temperature (unit: K ) is the photoelectric conversion efficiency under (dimensionless), the temperature coefficient is (Unit: 1 / K ), the actual operating temperature is T (unit: K ), the actual photoelectric conversion efficiency η Expressed as: .

[0031] Located above the quantum well photovoltaic cell layer is a photochromic spectrum adjustment layer, which is made of photochromic materials (which can be silver halide, metal oxide, spiropyran compounds, etc.). Under light of different intensities and spectral compositions, the optical properties of the photochromic spectrum adjustment layer will automatically change, such as changes in transmittance and reflectivity of light of different wavelengths. The intelligent control system controls the weak electric field or temperature adjustment signal applied to the photochromic spectrum adjustment layer by monitoring the overall power generation performance of the photovoltaic module and the absorption of each spectral band in real time, so that it can dynamically coordinate with the quantum well photovoltaic cell layer, further optimize the absorption and utilization of the solar spectrum, and achieve adaptive adjustment to different lighting environments. This is a function that traditional photovoltaic modules do not have. Take transmittance as an example: suppose the photochromic spectrum adjustment layer is initially responsive to a wavelength of λ The light transmittance is (dimensionless), in light intensity I (unit: W / m 2 ) and a specific spectral composition, the transmittance change is Δ T(λ) (dimensionless), the transmittance after the change T (λ) is expressed as: T (λ)=T0(λ)+Δ T (λ).

[0032] The distributed sensor suite within the intelligent control system is deployed throughout the photovoltaic system. These sensors include light intensity sensors (classified into different spectral bands), temperature sensors (monitoring ambient temperature and the temperatures of key components), angle sensors (monitoring changes in the focusing lens angle), humidity sensors, and other sensors. These sensors all feature wireless communication capabilities, transmitting collected data to the intelligent control system in real time. These sensors utilize a low-power, high-precision design, and some utilize micro-nano manufacturing technologies, resulting in compact size and the ability to accurately capture subtle environmental changes and component status changes. This provides a comprehensive and detailed data foundation for the precise control of the solar photovoltaic system. The intelligent decision-making and control module receives massive amounts of data from the distributed sensor array and generates an optimal strategy using a pre-set fusion model. Based on this optimal strategy, it sends commands to the concentrating module, such as angle adjustment and spectral beam parameter changes. This directs the photochromic spectrum modulation layer in the wide-spectrum absorption photovoltaic module to adjust the optical properties and optimize the operating state of the quantum well photovoltaic cell layer. It also collaborates with the battery management system in the battery module to rationally arrange the storage and distribution of electrical energy, achieving real-time, intelligent, and comprehensive precision control of the entire solar photovoltaic system. This distinguishes itself from traditional, simple, rule-based control methods. The pre-set fusion model utilizes advanced artificial intelligence algorithms (such as deep learning convolutional neural networks combined with reinforcement learning algorithms) for data analysis and modeling. It can quickly learn and grasp the complex relationships between various environmental factors, module status, and the overall photovoltaic system power generation efficiency, thereby generating the optimal control strategy.

[0033] like Figure 2-Figure 3As shown, preferably, generating an optimal strategy based on various aspects of the solar photovoltaic system data collected by the distributed sensor group through a preset fusion model, and sending a control instruction based on the optimal strategy may include steps S301 to S305: Step S301, preprocessing the collected various aspects of the solar photovoltaic system data; Step S302, using the preset fusion model, based on the preprocessed data, extracting key features valuable for state judgment and performance analysis of the solar photovoltaic system; and Step S303, analyzing the current working condition of the solar photovoltaic system according to the extracted key feature data, and searching for all corresponding action combinations in a predefined action space according to the current working condition, estimating the performance changes of the solar photovoltaic system brought about by different action combinations, and estimating the future reward value corresponding to each action combination; and Step S304, using a hierarchical ε-greedy strategy to select the current optimal action combination, the optimal action combination being the optimal strategy generated for the current solar photovoltaic system state to generate a corresponding control instruction; and Step S305, sending the control instruction to each corresponding module of the solar photovoltaic system to execute the corresponding control.

[0034] The various data in step S301 contain information of varying types and formats, with diverse temporal and spatial dimensions. These include light intensity (broken down by spectral band), ambient temperature, humidity, component temperatures (such as the focusing lens and photovoltaic cell layer), angular information (variation in the focusing lens angle), and power-related parameters (current, voltage, etc.) for each module. Preprocessing of this raw data involves data cleaning to remove noise (such as outliers caused by temporary sensor failures or electromagnetic interference) and data normalization to unify data of varying magnitudes into specific ranges to facilitate subsequent calculations and comparisons.

[0035] Taking step S302 as an example, the pre-set fusion model uses a convolutional neural network (CNN) in deep learning combined with a reinforcement learning algorithm as the core data analysis and modeling method. CNNs excel at processing spatially structured data. For example, the distribution of light intensity at different locations and spectral bands can be viewed as two-dimensional image data. CNNs can automatically extract implicit characteristic patterns, such as the distribution characteristics of light spectra under different weather conditions and the spatial variation of the focusing effect. Reinforcement learning algorithms are used to optimize decision-making in dynamic environments. They consider the photovoltaic system as an intelligent agent, with its controllable components (such as the angle and spectral splitting parameters of the concentrating module) as the action space, and performance indicators such as the system's power generation efficiency and power stability as reward signals. Through continuous interaction between the agent and the environment (i.e., the actual photovoltaic system operating environment and the changing states of each component), the reinforcement learning algorithm learns how to select the optimal action under different conditions to maximize long-term rewards, ultimately achieving efficient and stable operation of the entire photovoltaic system. Combining the two forms a fusion model architecture. The CNN component extracts features and recognizes patterns from the input sensor data, converting it into a high-level, abstract representation of the system state. This representation is then passed to the reinforcement learning component. Based on this state representation and historical decision-making experience, the reinforcement learning component generates control strategy recommendations for the current system state through the collaborative work of the policy network (for generating action decisions) and the value network (for evaluating state value). A large amount of annotated historical data is prepared as a training set. The annotations can be manually recorded or verified optimal control strategies at corresponding historical moments, as well as corresponding system performance feedback (such as power generation efficiency and power quality). This training data is used to train the constructed fusion model. During training, the model parameters (such as the convolution kernel weights of the CNN and the connection weights of each layer of the reinforcement learning network) are continuously adjusted through the backpropagation algorithm to gradually reduce the error between the control strategy recommendations output by the model and the annotated optimal strategy. This optimizes the model's predictive ability and enables it to accurately generate reasonable control strategy recommendations based on the input sensor data. At the same time, validation and test sets are used to evaluate and validate the trained model to prevent overfitting (i.e., performing well on the training set but poorly on new data). Based on the evaluation results, the model is further optimized and adjusted, such as adjusting model hyperparameters (learning rate, convolution kernel size, etc.) and adding data augmentation (appropriately transforming and expanding the training data). This ensures that the model has good generalization capabilities and can adapt to different actual operating scenarios and environmental changes.

[0036] The preprocessed data is fed into the fusion model. The CNN component first extracts features from this real-time input data. It automatically extracts key features representing the current solar PV system state from multi-dimensional data such as light intensity, temperature, and angle. These features, for example, represent the spectral distribution of the current light intensity and whether the temperature of each component is within the normal operating range. These features are more abstract and representative than the raw data, more accurately reflecting the real-time state of the system and providing a strong basis for decision-making in the subsequent reinforcement learning component. Based on an understanding of the physical principles and operating mechanisms of the PV system, the extracted features are combined and correlated to construct a feature vector with practical physical meaning and interpretability. For example, the proportion of different light intensity bands is correlated with the corresponding PV module power generation efficiency indicators to form a feature that reflects spectral utilization. Alternatively, ambient temperature is combined with battery pack charge and discharge efficiency characteristics to analyze the impact of temperature on energy storage. At the same time, based on different application scenarios and analysis objectives, the most representative and discriminative feature subsets will be screened out, and redundant or overly correlated features will be removed to reduce data dimensions, improve the efficiency and accuracy of subsequent data analysis and modeling, and ensure that the constructed feature vector can accurately characterize the state characteristics of the photovoltaic system at different times and under different working conditions.

[0037] As an example, step S303 analyzes the current operating conditions of the solar photovoltaic system based on the extracted key feature data, such as the current lighting conditions (light angle, spectral distribution, etc.) and the real-time status of each component (temperature, angle, power parameters, etc.). This provides an accurate description of the overall state of the current photovoltaic system, determining whether the system is in a state of sufficient sunlight but high temperature on a sunny day, or unstable sunlight with variable angles on a cloudy day. The reinforcement learning component, based on a predefined action space (including the range of action options such as angle adjustment of the concentrator module, adjustment of spectral beam splitting parameters, adjustment of the optical properties of the photovoltaic modules, and control of battery charge and discharge power), explores the potential changes in solar photovoltaic system performance caused by different action combinations under the current state. Combining past learning experience (i.e., the mapping relationship between different states and optimal actions accumulated during training), it estimates the future reward value corresponding to each action combination. This reward value comprehensively considers multiple system performance indicators, such as improved power generation efficiency, stable power quality, and battery health maintenance.

[0038] Step S304 is described as follows: ε-greedyThe strategy algorithm estimates the potential benefits of different actions (i.e., the effect of improving performance indicators such as system power generation efficiency and power stability), and combines the value network's evaluation of the current state value. The reinforcement learning algorithm selects one or a group of actions that can maximize long-term rewards, that is, generates the control strategy that is considered optimal at the current moment. For example, on a sunny day when the light angle gradually changes, the strategy may be to gradually fine-tune the angle of the focusing lens to maintain the best focusing effect, while adjusting the photochromic spectrum adjustment layer to increase the absorption of higher energy bands in the visible light, and appropriately control the charging power according to the battery power level to ensure stable storage of electrical energy and continuous and efficient power generation of photovoltaic modules. Layering ε-greedy The strategy includes: 1) classifying all possible action combinations according to their characteristics and importance in affecting the system. For example, it can be divided into different levels, such as the focusing module-related action combination layer (involving focusing lens angle adjustment, spectral beam splitting parameter changes, etc.), the photovoltaic component adjustment action combination layer (photochromic spectrum adjustment layer parameter setting, quantum well photovoltaic cell layer working state optimization, etc.), and the battery management action combination layer (charging and discharging power control, battery balancing management, etc.). Each level is further subdivided into specific action combinations. 2) Setting different initial exploration probabilities for each level ε The value is determined according to the sensitivity and frequency of the impact of each layer of action on the system. For example, the action combination layer of the focusing module responds more frequently to changes in light intensity, so the initial ε 1 is 0.15; the photovoltaic module adjustment action combination layer is set to initial ε 2 is 0.1; the battery management action combination layer has a higher relative stability requirement, and the initial ε 3 is 0.05. At the same time, each layer also records the number of times each action combination has been selected and the cumulative reward value and other related statistical information (initialized to 0), which is used for subsequent evaluation and decision-making similar to the previous scheme. 3) At each decision moment, for each layer, a random number between 0 and 1 is generated and compared with the corresponding layer’s ε If the random number is smaller than the value of the layer ε value, it enters the exploration phase and randomly selects one from the corresponding action combination in the layer; if the random number is greater than or equal to the layer εIf the reward value is not reached, the system enters the utilization phase. Based on the past cumulative rewards of each action combination within this layer, the action combination with the highest average reward value is selected. 4) Then, the action combinations selected at each layer are combined to form a final overall action combination for application to the solar photovoltaic system. This layered approach allows for a more refined balance between exploration and utilization across key links, taking into account the varying impacts of each component on the system while flexibly selecting the appropriate action combination based on the overall system state. Ultimately, this achieves efficient and precise control of the photovoltaic system, screening out the truly optimal action combination to ensure smooth operation under various operating conditions.

[0039] Taking step S305 as an example, the generated optimal strategy is sent to each corresponding module in the photovoltaic system to perform corresponding control operations, such as adjusting the focusing lens angle, changing the PV module spectrum adjustment parameters, and controlling battery charging and discharging. After executing these control instructions, the solar system generates actual operational feedback, including information such as actual power generation efficiency, power quality indicators, and actual status changes of each component. This feedback data is then collected by the distributed sensor group, forming a closed-loop data link for evaluating the effectiveness of the strategy. The feedback data after the actual strategy execution is compared and analyzed with the expected results when the strategy was originally generated. If the actual results deviate from the expected results, it indicates that the current fusion model may need further optimization. At this point, the fusion model is retrained and updated using these new data samples with actual feedback, adjusting the model parameters to better adapt to the actual operation of the photovoltaic system and more accurately generate the optimal strategy. Through this continuous feedback, update, and optimization cycle, the fusion model can continuously improve the accuracy and effectiveness of its strategy generation as the photovoltaic system operates, ensuring that the entire photovoltaic system is always in the optimal operational control state.

[0040] Preferably, in step S303, estimating the future reward value corresponding to each action combination includes: step S3031-step S3036; step S3031, for each action combination, analyzing its impact mechanism on the state of each component of the solar photovoltaic system and the overall operating state; step S3032, determining the probability of the solar photovoltaic system transferring from the current state to each predicted next state when executing different action combinations through the state transition probability matrix; step S3033, based on the above analysis of the state transition of the solar photovoltaic system caused by the action combination, combined with the current state data of the solar photovoltaic system collected in real time by the distributed sensor group, determining the execution The next state that the solar photovoltaic system immediately enters after a certain action combination; step S3034, based on the preset reward indicator system and the corresponding weight distribution, calculate the immediate reward value for the new state entered by the solar photovoltaic system; step S3035, based on the determined discount factor, combined with the state transition situation and the immediate reward calculation method obtained by the previous analysis, calculate the long-term reward cumulative value of each action combination by recursion or iteration; step S3036, comprehensively calculate the calculated immediate reward value and long-term reward cumulative value, determine the future reward value corresponding to each action combination, compare the future reward values corresponding to different action combinations, and obtain the optimal strategy.

[0041] For example, adjusting the angle of the concentrating module in step S3031 changes the focusing effect of sunlight, thereby affecting the distribution of light intensity on the photovoltaic module, potentially leading to changes in module temperature, photoelectric conversion efficiency, and other conditions. Another example is adjusting the optical parameters of the photochromic spectrum adjustment layer in the photovoltaic module, which directly changes the absorption and utilization of light from different spectra, affecting power generation efficiency and the temperature characteristics of the module. Another example is controlling the charge and discharge power of the battery, which changes the battery's state of charge and the number of charge and discharge cycles, thereby affecting its health and the overall system's energy reserves. By establishing a specific correlation model between each action and the change in system state, a clear description is provided of how the system state transitions from the current state to the next after executing a certain combination of actions.

[0042] As an example, step S3032 involves determining the probability of the solar photovoltaic system transitioning from its current state to each possible next state when executing different action combinations, based on methods such as historical data statistics, experimental simulations, or probabilistic analysis based on physical principles. For example, given current light intensity, component temperature, and other conditions, adjusting the focusing lens angle to a certain value can result in an increase in light intensity and a moderate rise in component temperature (a possible next state). Quantifying this probabilistic information allows for a more comprehensive understanding of the uncertainty of state changes caused by the action combination, providing a more detailed basis for subsequently calculating the expected reward value in conjunction with the reward indicator.

[0043] Taking step S3033 as an example, if the current lighting conditions are of a certain intensity and spectral distribution, after executing the combined action of adjusting the focusing lens angle and adjusting the PV module spectrum, the specific state parameters such as the light focusing condition of the solar PV system at the next moment, the temperature variation range of the module, and the expected trend of power generation efficiency changes are obtained through the state transition model analysis.

[0044] To illustrate step S3034, the actual values of each reward metric in the new state (such as power generation efficiency and voltage stability indicators, measured in real time or calculated based on a model) are substituted into the corresponding calculation formula, multiplied by their respective weights, and summed to obtain the immediate reward value immediately after executing the action combination. For example, if power generation efficiency improves in the new state, the corresponding reward contribution is calculated according to the weight of the power generation efficiency metric. This is then combined with the corresponding reward contributions of indicators such as power quality and battery health to obtain the total immediate reward value. This immediate reward value reflects the improvement in system performance caused by executing the action combination and is an important component of estimating future reward values. This step, based on the previous state analysis, realizes the transition from state change to specific reward quantification, paving the way for subsequent consideration of long-term reward accumulation.

[0045] To illustrate step S3035, the discount factor indicates the relative importance of future rewards relative to current rewards. Its value typically ranges from 0 to 1, with values closer to 1 indicating a greater emphasis on long-term rewards and values closer to 0 indicating a greater emphasis on immediate rewards. Its specific value can be determined based on factors such as the PV system's operating characteristics, control objectives, and expectations for long-term performance. For example, for applications prioritizing long-term system stability and battery life, a relatively high discount factor might be chosen to encourage action combinations that contribute to long-term performance improvements but may not yield significant immediate rewards. Starting with the immediate reward after executing the current action combination, the rewards at each future moment are gradually discounted and accumulated based on the discount factor, taking into account the subsequent rewards received by the solar PV system in different states. For example, assuming that a certain immediate reward is currently obtained by executing a certain action combination, the next moment in time will enter different possible states based on the state transition probability, each of which corresponds to a different immediate reward. The rewards at these future moments are then multiplied by the corresponding power of the discount factor (the further away from the current moment, the higher the power, the greater the discount), and then accumulated to obtain the cumulative value after accounting for long-term rewards. This cumulative value comprehensively reflects the overall potential contribution of executing this action combination to improving system performance over the next period of time. It is a key part of estimating future reward values. Combined with the immediate reward value, it comprehensively considers the short-term and long-term impact of the action combination on the system. This step expands the immediate reward to a comprehensive estimate of long-term rewards by introducing the time dimension and discount mechanism, thus improving the reward calculation logic.

[0046] As an example, step S3036 combines the calculated immediate reward value and the accumulated long-term reward value to determine the future reward value for each action combination. This value represents the expected overall performance improvement for the solar PV system from executing that action combination, taking into account various practical factors and system operating characteristics. By comparing the future reward values corresponding to different action combinations, a quantitative basis is provided for selecting the optimal action combination in the reinforcement learning process. This allows the solar PV system to optimize its decisions toward maximizing the reward value, achieving efficient and intelligent control of the system.

[0047] Preferably, the corresponding control in step S305 may include: controlling the action of the angle adjustment device to drive the focusing lens unit to rotate to adjust the angle of the focusing lens unit; adjusting the optical parameters of the spectral splitting and guiding device; adjusting the temperature value applied to the photochromic spectrum adjustment layer to change the transmittance and reflectivity of light of different spectra; adjusting the power storage and distribution strategy of the battery management system; adjusting the power output and the output power of the adapter module.

[0048] Preferably, the battery module includes a solid electrolyte hybrid energy storage battery pack and a battery management system; the solid electrolyte hybrid energy storage battery pack integrates lithium-ion batteries and solid-state supercapacitors, and realizes rapid storage and release of electrical energy through superconducting connection technology; the battery management system monitors battery cell status parameters, exchanges data with the intelligent control system through a distributed intelligent communication network, dynamically adjusts the charging and discharging strategy, and supplies power to the power output and adaptation module through the power transmission line.

[0049] This system utilizes a novel solid-state electrolyte hybrid energy storage battery pack, combining the advantages of lithium-ion batteries and solid-state supercapacitors. The lithium-ion battery provides high-energy density storage, meeting long-term power requirements. The solid-state supercapacitor, with its rapid charge and discharge characteristics and high power density, rapidly absorbs and releases energy during transient fluctuations in photovoltaic power generation (such as sudden changes in light intensity caused by rapid cloud cover or dissipation), ensuring the stability of the overall system's power output. The use of solid-state electrolytes enhances battery pack safety, avoiding the safety hazards of traditional liquid electrolyte batteries, such as leakage and fire. It also enhances the battery's adaptability to extreme environments, such as high and low temperatures, enabling reliable operation in a wider range of environmental conditions. The battery management system (BMS) is closely integrated with the battery pack, monitoring key parameters such as voltage, current, internal resistance, and temperature of each battery cell in real time. Using built-in intelligent algorithms, it accurately assesses the battery's state of health (SOH), remaining capacity (SOC), and charge and discharge performance trends. Based on these monitoring data, the intelligent BMS, on the one hand, feeds back the real-time status of the battery to the intelligent control system so that the system can reasonably arrange the power generation power and load power consumption; on the other hand, it automatically manages the battery pack in a balanced manner to avoid overcharging and over-discharging of individual batteries, thereby extending the overall service life of the battery pack. It can also dynamically adjust the charging and discharging strategy according to the aging of the battery to ensure that the battery pack is always in the best working condition. This is more refined and intelligent than the traditional simple battery management method.

[0050] Preferably, the power output and adaptation module includes: an adaptive intelligent inverter, which is used to adjust the waveform, frequency and voltage level of the output AC power according to the grid parameter monitoring data of the intelligent control system; a load intelligent matching and optimization unit, which is used to obtain load type and priority data through a distributed intelligent communication network, and cooperate with the intelligent control system module to realize dynamic power distribution and give priority to power supply to critical loads.

[0051] The power output stage is equipped with an adaptive smart inverter, which features intelligent waveform modulation and power regulation. The inverter can sense load type, power demand, and grid access conditions (such as voltage, frequency, and phase) in real time. Using built-in intelligent algorithms, it automatically adjusts the waveform, frequency, and voltage level of the output AC power to ensure a perfect match between the output power and the load and the grid, avoiding equipment damage or grid fluctuations caused by substandard power quality. Furthermore, the adaptive smart inverter maintains close communication with the intelligent control system, dynamically adjusting output power based on real-time changes in photovoltaic power generation and battery reserve levels. This ensures efficient utilization and optimal distribution of power, ensuring stable power supply across the entire system in a variety of complex power generation and consumption scenarios. This is a significant difference from traditional inverters with fixed parameter settings. The load intelligent matching and optimization unit is connected to the load side of the entire photovoltaic system and provides categorized management and real-time monitoring of all loads. It understands the importance, power usage patterns, and power characteristics of different loads. Combining the PV system's power generation capacity with the battery's charge status, and working in conjunction with an intelligent control system, it prioritizes stable power supply to critical loads (such as emergency lighting and important medical equipment) while optimizing non-critical loads through intelligent time-sharing and power limiting. This improves overall energy efficiency, avoids system overloads and energy waste caused by unreasonable load demand, and enables the entire PV system to better adapt to diverse power demands. This innovation distinguishes it from existing simple load connection methods.

[0052] In short, the above description is only a preferred embodiment of the technical solution of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the scope of protection of the present invention.

Claims

1. A solar photovoltaic system, characterized in that: The solar photovoltaic system comprises: The focusing module is used to adaptively focus and split the light according to the lighting conditions; A wide spectrum absorption photovoltaic component, used to receive light from the concentrating module and optimize spectral absorption; An intelligent control system, comprising a distributed sensor group and an intelligent decision-making and control module, configured to generate an optimal strategy based on various aspects of the solar photovoltaic system data collected by the distributed sensor group through a preset fusion model, and to send control instructions according to the optimal strategy; A battery module for storing electrical energy from the broad spectrum absorption photovoltaic assembly, monitoring battery status, feeding back information to the intelligent control system, and managing battery charging and discharging; The electric energy output and adaptation module is used to receive the electric energy from the battery module and also to adapt and output the electric energy according to the load demand and the grid conditions.

2. The solar photovoltaic system according to claim 1, characterized in that: The light focusing module comprises: A condenser lens array, wherein each condenser lens unit in the condenser lens array is equipped with an independent light sensor and an angle adjustment device, and the angle of the condenser lens unit is adjusted by the angle adjustment device; The spectral beam splitting and guiding device is used to split the light focused by the condensing lens array according to the wavelength range and guide it to the corresponding light absorption area of the wide spectrum absorption photovoltaic component.

3. The photovoltaic system according to claim 2, characterized in that: Each of the condensing lens units includes a square plane mirror and a plurality of light-emitting lens units; A rectangular array of several light-emitting lens units is distributed on the light-emitting surface of the square plane mirror; the surface of the light-emitting lens unit facing away from the square plane mirror is a plane, a central concave surface is provided in the middle of the surface of the light-emitting lens unit facing the square plane mirror, and several concentric rings are distributed around the central concave surface on the surface of the light-emitting lens unit facing the square plane mirror; each of the rings is composed of several arc-shaped pieces, and the connecting ends of two adjacent arc-shaped pieces are partially offset and overlapped.

4. The photovoltaic system according to claim 1, characterized in that: The broad spectrum absorption photovoltaic module comprises a quantum well photovoltaic cell layer and a photochromic spectrum adjustment layer located above the quantum well photovoltaic cell layer. The quantum well photovoltaic cell layer uses a multi-layer nanoscale quantum well structure, each layer absorbs light in a specific spectral range and adjusts the energy level distribution through a configured temperature-adaptive electronic structure; The photochromic spectrum adjustment layer is made of a photochromic material. Under illumination of different intensities and spectral compositions, the optical properties of the photochromic spectrum adjustment layer automatically change to change the optical characteristic parameters for light of different wavelengths.

5. The photovoltaic system according to claim 1, characterized in that: The method of generating an optimal strategy based on various aspects of the solar photovoltaic system data collected by the distributed sensor group through a preset fusion model and sending control instructions according to the optimal strategy includes: Pre-process the collected data of various aspects of the solar photovoltaic system; Using a preset fusion model, based on pre-processed data, extract key features that are valuable for solar photovoltaic system status judgment and performance analysis; and Analyze the current operating condition of the solar photovoltaic system based on the extracted key feature data, search for all corresponding action combinations in a predefined action space based on the current operating condition, and estimate the future reward value corresponding to each action combination by estimating the performance changes of the solar photovoltaic system caused by different action combinations; and Leverage layering ε-greedy The strategy selects the current optimal action combination, where the optimal action combination is the optimal strategy generated for the current solar photovoltaic system state to generate corresponding control instructions; The control instructions are sent to corresponding modules of the solar photovoltaic system to execute corresponding control.

6. The photovoltaic system according to claim 5, characterized in that: The estimating of the future reward value corresponding to each action combination includes: For each action combination, analyze its impact on the status of each component and the overall operating status of the solar photovoltaic system; The state transition probability matrix is used to determine the probability of the solar photovoltaic system transitioning from the current state to each predicted next state when performing different action combinations; Based on the above analysis of the state transition of the solar photovoltaic system caused by the action combination, combined with the current state data of the solar photovoltaic system collected in real time by the distributed sensor group, the next state that the solar photovoltaic system immediately enters after executing a certain action combination is determined; Calculate the instant reward value for the new state entered by the solar photovoltaic system based on the preset reward indicator system and the corresponding weight distribution; Based on the determined discount factor, combined with the state transitions and immediate reward calculation method analyzed previously, the long-term reward cumulative value of each action combination is calculated recursively or iteratively. The calculated immediate reward value and long-term reward cumulative value are comprehensively calculated to determine the future reward value corresponding to each action combination. The future reward values corresponding to different action combinations are compared to obtain the optimal strategy.

7. The photovoltaic system according to claim 5, characterized in that: The step of sending the control instruction to each corresponding module of the photovoltaic system to execute corresponding control includes: Controlling the action of the angle adjustment device to drive the focusing lens unit to rotate so as to adjust the angle of the focusing lens unit; Adjust the optical parameters of the spectral beam splitting and guiding device; Adjusting the temperature applied to the photochromic spectrum adjustment layer to change the transmittance and reflectivity of light of different spectra; Adjust the energy storage and distribution strategy of the battery management system; Adjust the power output and the output power of the adapter module.

8. The photovoltaic system according to claim 1, characterized in that: The battery module includes a solid electrolyte hybrid energy storage battery pack and a battery management system; The solid electrolyte hybrid energy storage battery pack integrates lithium-ion batteries and solid supercapacitors, and realizes rapid storage and release of electrical energy through superconducting connection technology; The battery management system monitors battery cell status parameters, exchanges data with the intelligent control system through a distributed intelligent communication network, dynamically adjusts the charging and discharging strategy, and supplies power to the power output and adaptation module through the power transmission line.

9. The photovoltaic system according to claim 1, characterized in that: The power output and adaptation module includes: Adaptive smart inverter, used to adjust the waveform, frequency and voltage level of output AC power based on grid parameter monitoring data from the intelligent control system; The load intelligent matching and optimization unit is used to obtain load type and priority data through the distributed intelligent communication network, coordinate with the intelligent control system module to achieve dynamic power distribution, and give priority to ensuring power supply to critical loads.

10. The photovoltaic system according to claim 4, characterized in that: An exciton transport interface layer is provided between the quantum well photovoltaic cell layer and the photochromic spectrum adjustment layer, and the photogenerated carrier transport efficiency is optimized through interface state engineering, thereby improving the photoelectric conversion efficiency.

Images