Photovoltaic photo-thermal panel heat dissipation efficiency optimization method based on neural network and cooling liquid dispersion degree optimization
By combining neural networks and optimization algorithms, the distribution of coolant and import and export design of photovoltaic photothermal system is dynamically adjusted, which solves the problem of low heat dissipation efficiency caused by uneven coolant distribution, and realizes efficient heat dissipation and stable operation of the system.
Patent Information
- Application Number
- CN202510068878.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-16
- Publication Date
- 2025-06-27
AI Technical Summary
In the existing photovoltaic photothermal system, the inlet and outlet design and distribution of coolant are uneven, resulting in low heat dissipation efficiency, poor water dispersion, excessive water temperature difference, high dynamic complexity, and difficult to effectively optimize through traditional methods.
Using a method combining neural network and optimization algorithm, the nonlinear relationship between coolant distribution and heat dissipation efficiency is dynamically adjusted to maximize heat dissipation efficiency.
It significantly improves the heat dissipation efficiency of photovoltaic photothermal plates, reduces the risk of local overheating, improves the overall performance and stability of the system, and ensures the efficiency and reliability of the system in long-term operation.
Smart Images
Figure CN120217822A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to new energy system optimization and artificial intelligence technology, and specifically to a method for optimizing the distribution of coolant in a photovoltaic thermal (PV / T) panel and the design of inlet and outlet ports by combining neural network fitting and optimization algorithms to improve the heat dissipation efficiency and overall performance of the system. Technical Background
[0002] A photovoltaic thermal (PV / T) system is a technology that combines photovoltaic power generation and solar thermal utilization. By flowing coolant on the back of the photovoltaic module to recover heat, both the power generation efficiency and waste heat can be recovered. The design of the inlet and outlet of the coolant and the uneven distribution directly affect the heat dissipation efficiency.
[0003] In the prior art, the optimization of coolant distribution usually adopts fixed design or average flow assumption, which is difficult to solve the following problems:
[0004] Inlet discreteness: The flow distribution at the coolant inlet may be uneven, resulting in local overheating of the plate surface; Outlet uniformity: The temperature difference of the outlet coolant is too large, reflecting insufficient heat dissipation; Dynamic complexity: The coolant distribution is affected by multiple factors (temperature, flow rate, radiation intensity, gravity, etc.), and the optimization is difficult.
[0005] Therefore, there is an urgent need for a method to quantify the influence of the discreteness degree of coolant inlet and outlet on the heat dissipation efficiency, and combine intelligent algorithms to optimize the design of inlet and outlet ports and the coolant distribution to dynamically improve the heat dissipation efficiency of the photovoltaic thermal panel. Summary of the Invention
[0006] The present invention proposes an optimization method for the heat dissipation efficiency of a photovoltaic thermal panel based on neural network and coolant discreteness optimization. By fitting the non-linear relationship between heat dissipation efficiency and coolant distribution through a neural network, and combining optimization algorithms to adjust the design of inlet and outlet ports and the coolant flow rate, the maximization of heat dissipation efficiency is achieved.
[0007] 1. Modeling of Coolant Discreteness
[0008] The flow and heat transfer efficiency of the coolant are directly related to the heat dissipation performance of the system. Therefore, accurately modeling the discreteness degree of the coolant in the system is a crucial step. This process can be divided into the following steps:
[0009] Geometric discretization: Fine geometric discretization is performed on the design area of the coolant inlet and outlet to ensure that the flow and distribution of the coolant in the entire system are reasonably described. At this time, the entire fluid channel can be divided into multiple small units (such as grids or blocks), and each unit corresponds to a calculation node.
[0010] Matrix Generation: Based on the discretized geometric model, establish relevant matrix models to describe the coolant flow and heat exchange processes. These matrices may include the flow velocity matrix, temperature field matrix, heat exchange matrix, etc., which can reflect the flow state and heat dissipation effect of the coolant at different positions.
[0011] Thermodynamic Model Establishment: Based on the basic physical laws of heat conduction, convection, and radiation, construct a heat exchange model for the coolant. This model needs to quantitatively correlate parameters such as the temperature distribution, flow velocity, and heat conduction coefficient of the coolant with the heat dissipation efficiency, thereby providing a theoretical basis for subsequent optimization design.
[0012] 2. Neural Network Fitting
[0013] Using a neural network to model the relationship between the discrete distribution of the coolant and the heat dissipation efficiency can significantly improve the prediction accuracy and optimization efficiency. The specific process is as follows:
[0014] Data Collection and Feature Extraction: First, it is necessary to collect data on the flow pattern and heat dissipation effect of the coolant at different discretization levels through simulation or experimental data. These data include, but are not limited to, the inlet and outlet positions of the coolant, flow velocity, temperature change, and heat dissipation of the system.
[0015] Neural Network Model Design: Select an appropriate neural network structure (such as a multi-layer perceptron, convolutional neural network, etc.), input features such as the geometric discretization, flow velocity, and temperature of the coolant, and output the heat dissipation efficiency as the target variable. The training objective of the neural network is to minimize the prediction error, thereby learning the complex non-linear relationship between the discretization level and the heat dissipation efficiency.
[0016] Training and Fitting: By training on a large amount of experimental data or simulation data, the neural network can fit the relationship formula between the coolant distribution and the heat dissipation effect. This network model can be used as a "black box" model in subsequent optimization design to efficiently predict the heat dissipation effect under different coolant distribution conditions.
[0017] 3. Optimization Design
[0018] The goal of optimization design is to improve the heat dissipation efficiency of the system by dynamically adjusting the coolant distribution and related parameters. Combining multi-objective optimization algorithms, the optimal solution can be found in the complex design space. The specific process is as follows:
[0019] Multi-Objective Optimization Problem Modeling: In optimization design, it is often necessary to balance multiple objectives, such as maximizing the heat dissipation efficiency, minimizing the coolant usage, and controlling the equipment cost. Therefore, first, these objectives need to be transformed into a mathematical model and combined with constraint conditions (such as coolant flow rate limits, system structure limits, etc.) to form an optimization problem.
[0020] Dispersion and Parameter Adjustment: The optimization algorithm simulates the effects of different dispersion degrees by dynamically adjusting the design of the inlet and outlet, the coolant flow rate, the thermodynamic parameters of the coolant, etc. Common optimization methods include genetic algorithms, particle swarm optimization (PSO), simulated annealing, etc., which can find the optimal solution in the high-dimensional design space.
[0021] Neural Network-Assisted Optimization: In the optimization process, the neural network model can be used as a predictor to quickly evaluate the heat dissipation effect under different coolant distribution conditions. By combining the optimization algorithm with the neural network, the optimization efficiency and accuracy can be significantly improved, avoiding a large number of experiments or calculation processes.
[0022] Result Analysis and Feedback: Through multiple rounds of optimization iterations, the optimal inlet and outlet positions and coolant flow distribution plan are finally obtained and verified in combination with the actual operating conditions. The optimization results can not only improve the heat dissipation efficiency but also ensure better stability and long life of the system during operation.
[0023] Scope of Application:
[0024] Photovoltaic-Thermal Hybrid System (PV-T):
[0025] This system combines photovoltaic and solar thermal technologies on the same platform to achieve both power generation and heat energy collection simultaneously. Photovoltaic panels convert light energy into electrical energy through the photovoltaic effect, and the solar thermal part absorbs solar energy to produce hot water or steam for industrial, household, or air conditioning applications. Due to the combination of these two energy conversion methods, such systems require optimized heat dissipation designs to manage the temperatures of photovoltaic modules and collectors simultaneously to ensure the efficient operation of the system.
[0026] Concentrating Photovoltaic-Thermal (CPV-T) System:
[0027] The concentrating photovoltaic-thermal system uses optical concentration technology to focus sunlight onto smaller photovoltaic-thermal components, which makes the light intensity received on the component surface much higher than that of conventional photovoltaic systems. Under concentrating conditions, a large amount of heat is generated inside the system, and the optimized distribution of the coolant is crucial. By optimizing the coolant flow and temperature uniformity, the high-temperature area can be effectively reduced, avoiding problems such as a decrease in photovoltaic efficiency or thermal damage caused by overheating, thus significantly improving the overall performance and stability of the system.
[0028] Centralized Photovoltaic-Thermal Power Generation System:
[0029] In large-scale centralized photovoltaic-thermal power plants, multiple photovoltaic-thermal units are combined into a large system, and these systems usually face complex thermal management problems. The system needs to operate under different climate conditions, so the distribution and flow pattern of the coolant must be able to adapt to various environmental changes. By optimizing the flow path of the coolant, the temperature difference in the system can be effectively reduced, the temperature uniformity of each photovoltaic-thermal unit can be maintained, and the overall power generation efficiency can be ensured.
[0030] Household and commercial photovoltaic-thermal energy supply systems:
[0031] For photovoltaic-thermal shared systems for household and commercial applications, the space and resources of the system are limited, and heat dissipation management is particularly important. Especially in areas with high temperatures or strong sunlight, how to effectively manage the temperature distribution of photovoltaic cells and water heaters, avoid local overheating, and improve the energy efficiency of the system is the key to system design. Through this optimized design method, more efficient heat dissipation effects can be achieved, ensuring the stability of the system during long-term operation.
[0032] Photovoltaic-thermal systems in high-temperature environments:
[0033] In deserts or high-temperature regions, the solar radiation intensity is high, and the thermal load of photovoltaic-thermal systems is also large. To avoid performance losses caused by system overheating under high-temperature conditions, optimizing the flow and distribution of the coolant is an effective way to improve system stability. Especially through neural networks and multi-objective optimization algorithms, the cooling system can be automatically adjusted according to different environmental conditions to maintain an ideal temperature distribution, extend the service life of the system, and improve energy efficiency.
[0034] Mobile photovoltaic-thermal systems:
[0035] Mobile photovoltaic-thermal systems, such as photovoltaic-thermal vehicles, mobile solar thermal power generation equipment, etc., also require efficient thermal management technologies. In these systems, due to limited ground space and the system may move frequently, traditional heat dissipation methods may not meet the requirements. By optimizing the coolant flow distribution, the heat dissipation effect of these mobile systems under different working conditions can be effectively improved, ensuring the efficient and stable operation of the system. Description of the Drawings
[0036] To more intuitively display the technical features, core principles, and specific implementation effects of the present invention, the drawings combine key contents such as the modeling of coolant distribution, the optimization process, and the comparison of optimization results. These drawings are intended to assist in understanding the innovation and practical application value of the present invention, and can help relevant personnel in the technical field to more comprehensively master the implementation details and advantages of the present invention. It should be particularly noted that the following drawings are only for illustration and explanation, and should not be regarded as a limitation of the technical solution of the present invention. The actual application of the present invention can be adjusted and optimized according to different requirements.
[0037] 1.Figure 1 Schematic diagram of coolant distribution, showing the distribution state of coolant on the photovoltaic-thermal panel, and marking the "with water" areas (1) and "without water" areas (0) in matrix form.
[0038] 2. Figure 2 Schematic diagram of neural network architecture, showing the neural network structure used to fit the relationship between coolant distribution and heat dissipation efficiency.
[0039] 3. Figure 3 Schematic diagram of the temperature distribution of the battery panel before optimization.
[0040] 4. Figure 4 Schematic diagram of the temperature distribution of the battery panel after optimization.
[0041] 5. Figure 5 Flowchart of coolant distribution optimization, showing the complete process of coolant distribution optimization, including various steps of matrix generation, neural network fitting, and optimization algorithm iteration. Detailed implementation method
[0042] 1. Construction of with-water and without-water matrix
[0043] Generate matrix: Inspired by the diffusion process of rivers flowing into the sea, lakes or plains, and combined with the actual application requirements of the coolant on the photovoltaic-thermal panel, a matrix generation algorithm that simulates random diffusion behavior is designed. Its core is to simulate the natural phenomenon of coolant diffusing downward from the water inlet on the thermal panel, including the preferential guidance of the main flow path, the randomness of branch diffusion, and the hierarchical advancement limited by coverage conditions.
[0044] 1.1 Initial condition setting
[0045] Divide the surface of the photovoltaic-thermal panel into matrix units of m×n, each unit representing the coverage state of the coolant, and the initial value is set to "0" (no water coverage).
[0046] The water inlet is used as the starting point of the matrix (for example, a certain position at the top of the matrix), and the initial assignment is "1" (with water coverage).
[0047] The coolant starts to diffuse from the water inlet, and the diffusion speed, range and direction are determined by momentum, random perturbation and hierarchical advancement rules.
[0048] M is a binary matrix of m×n, representing the coolant distribution state.
[0049] 1.2 Main flow and branch diffusion rules
[0050] Main flow direction guidance: The coolant preferentially diffuses along the main direction of the matrix (usually downward), simulating the characteristic that water flow advances to the lower layer area under the action of gravity. The main flow path forms a relatively coherent main coverage area.
[0051] Branch diffusion: During the mainstream expansion process, the coolant diffuses to the left and right sides with a certain probability, forming an irregular branch coverage area. The branch range is determined by random perturbation, and the formula is: r = v·t·P branch
[0052] Where:
[0053] r is the random range of branch diffusion,
[0054] v is the water flow velocity,
[0055] t is the diffusion time,
[0056] P branch is the branch diffusion probability.
[0057] 1.3 Hierarchical advancement rule
[0058] The hierarchical advancement simulates the characteristic of the coolant diffusing layer by layer, but does not require the current layer to be fully covered. Instead, it advances downward based on dynamic conditions:
[0059] If at least one cell in the current layer is covered by the coolant (i.e., the value is "1"), the coolant can flow to the next layer. If the current layer is completely uncovered (i.e., all cell values are "0"), the diffusion cannot continue to advance, and the matrix generation stops.
[0060] This rule conforms to the actual diffusion behavior law of the coolant, ensuring that the effective flow of the previous layer supports the subsequent diffusion without forcing full coverage.
[0061] 1.4 Introduction of random perturbation
[0062] Simulate the random offset phenomenon of the coolant affected by the environment during actual diffusion:
[0063] Each time the diffusion direction is selected, a random perturbation term Δx, Δy ~ N(0, σ 2 )
[0064] Where Δx and Δy represent the random expansion amounts in the horizontal and vertical directions, and they follow a normal distribution with a mean of 0 and a variance of σ 2 of.
[0065] The introduction of random perturbation introduces uncertainty, making the generated matrix more in line with real natural phenomena.
[0066] 1.5 Termination condition
[0067] Simulate the diffusion boundary of the coolant. The following situations will cause the diffusion to stop:
[0068] All cells of the matrix have been covered (i.e., all are "1");
[0069] Diffuse to the matrix boundary;
[0070] The diffusion probability or momentum of the coolant decays to zero and cannot expand further.
[0071] 1.6 Matrix state update
[0072] In each diffusion step, the cells covered by the coolant are marked as "1", and the uncovered areas remain "0". The final generated matrix presents a coverage state that randomly expands outward from the main flow path.
[0073] 2. Neural network modeling and training
[0074] 2.1 Input and output design
[0075] Input data:
[0076] Coolant distribution matrix M: Represents the spatial distribution of the coolant on the plate surface.
[0077] Environmental parameters:
[0078] Coolant inlet temperature: T in
[0079] Coolant flow rate: v
[0080] Radiation intensity: I
[0081] Ambient temperature: T ambient
[0082] Output data: System heat transfer efficiency: η cooling
[0083] 2.2 Neural network architecture
[0084] To achieve the fitting of complex non-linear relationships, the neural network adopts the following architecture:
[0085] Input layer: Expand the M matrix into a one-dimensional vector and concatenate it with other parameters to form an input feature vector.
[0086] Hidden layer:
[0087] Multiple fully connected layers (Dense Layer), with the activation function using the rectified linear unit. The Dropout layer is used to prevent overfitting.
[0088] Output layer:
[0089] Single output node: Indicates that this is the output of a neural network model or a certain computational model, predicting the heat transfer efficiency.
[0090] Heat transfer efficiency: η cooling , usually used to describe the performance of the cooling system.
[0091] 2.3 Loss Function and Optimization Method
[0092] Loss function: Mean Squared Error (MSE) is adopted and defined as:
[0093]
[0094] Where:
[0095] y i is the actual efficiency,
[0096] is the predicted efficiency.
[0097] Optimization method: Adam optimizer with dynamically adjusted learning rate.
[0098] 2.4 Data Generation and Training
[0099] Data generation: Use simulation software (such as CFD tools) to generate heat dissipation performance data under different coolant distributions and record relevant environmental parameters simultaneously.
[0100] Training process: Data normalization; Matching input data with target output; Splitting the dataset into training set, validation set and test set; Training the neural network through multiple rounds of iteration until the loss function converges.
[0101] 3. Heat Transfer Efficiency Research
[0102] Through the fitted neural network, the following aspects are studied:
[0103] Nonlinear relationship between coolant distribution and efficiency: Analyze how the distribution patterns of 1s and 0s in matrix M affect the heat transfer efficiency; Optimize the coolant distribution to enable more effective cooling in key areas.
[0104] Identification of locally overheated areas: Analyze the influence of excessive consecutive 0s in matrix M on the heat transfer efficiency; Determine the areas with insufficient local cooling and propose improvement plans.
[0105] Dynamic optimization design: Adjust the coolant distribution according to different environmental parameters (such as radiation intensity) to form an adaptive design.
[0106] 4. Optimization Design and Improvement
[0107] In the optimization of the coolant distribution of photovoltaic-thermal panels, the improvement of system performance not only depends on the heat dissipation efficiency, but also the key indicators of the temperature distribution of the battery panels need to be comprehensively considered. Using the prediction results of the neural network and combining with optimization algorithms (such as genetic algorithm or particle swarm algorithm), the iteration of the coolant distribution and the inlet and outlet design is realized. The optimization objectives include the following four aspects:
[0108] 4.1 Optimization Objectives:
[0109] 4.1.1 Average temperature of the solar panel (T avg , the lower the better)
[0110] It represents the average temperature of the overall photovoltaic-thermal panel. The lower the temperature, the higher the power generation efficiency of the solar panel, and it is also beneficial to extend the equipment life.
[0111] The optimization objective is to minimize T avg .
[0112] 4.1.2 Temperature uniformity coefficient of the solar panel (C uniform , the higher the better)
[0113] It measures the degree of uniformity of the surface temperature distribution of the photovoltaic-thermal panel and is defined as:
[0114]
[0115] T min is the lowest temperature on the surface,
[0116] T avg is the average temperature of the panel surface.
[0117] The more uniform the temperature distribution, the higher the reliability of the photovoltaic module operation.
[0118] The optimization objective is to maximize C uniform .
[0119] 4.1.3 Maximum temperature of the solar panel (T max , the lower the better)
[0120] It represents the local maximum temperature of the photovoltaic-thermal panel. Excessive temperature may cause local damage to the photovoltaic module. The optimization objective is to minimize T max
[0121] 4.1.4 Pressure loss (ΔP, the lower the better)
[0122] It is the pressure loss when the coolant flows inside the photovoltaic-thermal panel, which directly affects the energy consumption and operation efficiency of the cooling system. The optimization objective is to minimize ΔP.
[0123] 4.2 Mathematical expression of the optimization problem
[0124] Combining the above four optimization objectives, the optimization problem can be defined as a multi-objective optimization problem:
[0125] Minimize: [T avg , -C uniform , T max , ΔP]
[0126] Constraints:
[0127] Dispersion Degree D of Coolant Inlet and Outlet in and D out is less than the set threshold D threshold Parameters such as coolant flow rate and temperature must meet the equipment operation limits.
[0128] In multi-objective optimization problems, there are usually multiple conflicting objectives, and it is impossible to optimize all objectives simultaneously with a single solution. The Pareto front provides a set of solutions that have no obvious disadvantages in each objective, representing a compromise between different objectives. In other words, each Pareto optimal solution cannot improve one objective without sacrificing others.
[0129] In this application, it is often necessary to balance multiple objectives such as heat dissipation efficiency and cost. Through the Pareto front, we can find the balance point between these objectives, providing a set of optional optimal solutions for decision-makers to help them select the most suitable solution for current needs.
[0130] 4.3 Optimization Algorithm Selection and MATLAB Implementation
[0131] Optimization Algorithm Selection
[0132] Genetic Algorithm (GA):
[0133] In this application, by simulating operations such as selection, crossover, and mutation in the biological evolution process, the optimal solution or approximate optimal solution of the problem is found. Genetic algorithms are widely used in optimization problems, especially in cases where the search space is very large and the complexity is relatively high. It is suitable for solving multi-objective optimization problems and finding a better-performing coolant distribution by simulating the natural selection process.
[0134] Particle Swarm Optimization (PSO):
[0135] PSO finds the optimal solution by a group of particles (individuals) searching in the solution space. Each particle represents a candidate solution to the problem. During the search process, the particle updates its position and velocity based on its own historical experience and the overall experience of the group.
[0136] MATLAB Built-in Multi-Objective Optimization Toolbox:
[0137] Use the GMAULITIOBI function to handle multi-objective optimization problems and generate the Pareto front.
[0138] Optimization Process
[0139] Step 1: Define the objective function
[0140] The optimization objective T avg, C uniform , T max , ΔP is written as the objective function in MATLAB:
[0141] Step 2: Set the optimization variables and constraints
[0142] Define the optimization variables related to the coolant distribution (such as the inlet and outlet positions, coolant flow rate, etc.) and the constraint conditions:
[0143] Step 3: Call the multi-objective optimization function
[0144] Run the optimization using the GMAULTIOBI function in MATLAB
[0145] 4.4 Expected results
[0146] Through optimization, the following improvements can be obtained:
[0147] The overall temperature of the solar panel decreases, significantly improving the power generation efficiency; the temperature distribution becomes more uniform, reducing the risk of local overheating; the pressure loss decreases, reducing the energy consumption of the cooling system; the best compromise of performance is achieved among multiple objectives, and the optimization results can be applied to the design of actual photovoltaic-thermal systems.
Claims
1. A photovoltaic thermal panel heat dissipation efficiency optimization system based on neural network, characterized in that: include: Mathematicalization module, neural network fitting module, optimization module and result evaluation module; the mathematicalization module is used to abstract the heat dissipation optimization problem of photovoltaic thermal panels into a mathematical model, determine the design variables (including the inlet and outlet positions of the coolant, flow distribution, etc.) and constraints that affect the heat dissipation performance; define the value range of the design variables according to the experimental data and system parameters, and transfer them to the neural network fitting module; the neural network fitting module uses a multi-layer neural network structure to perform high-precision fitting on the complex relationship between the heat dissipation efficiency and the design parameters by establishing a nonlinear correlation model between the inlet and outlet parameters of the coolant and the heat dissipation efficiency of the photovoltaic thermal panel; the optimization module uses a global optimization algorithm (such as NSGA-II+ARSBX multi-objective optimization algorithm) to dynamically adjust the design variables to achieve multi-objective optimization of the heat dissipation efficiency, temperature uniformity, maximum temperature and pressure loss of the photovoltaic thermal panel; wherein different optimization objectives generate a set of optimal design schemes through the Pareto frontier method; the result evaluation module evaluates the heat dissipation performance and energy consumption level of each design scheme based on the optimized design variables and combined with CFD simulation verification, and records the schemes that meet the design requirements through the cloud database to provide data support for subsequent designs.
2. According to the neural network-based photovoltaic thermal panel heat dissipation efficiency optimization system of claim 1, it is characterized in that: The connection relationship between the modules is specifically set as follows: the mathematization module is connected to the neural network fitting module; the neural network fitting module is connected to the optimization module and the result evaluation module respectively; the optimization module and the result evaluation module are connected to each other to achieve iterative optimization of design parameters.
3. A photovoltaic thermal panel heat dissipation optimization method based on neural network and optimization algorithm, characterized in that: The following steps are involved: (1) The discrete characteristics of the coolant inlet and outlet are analyzed through mathematical modeling, and the coolant distribution of the photovoltaic thermal panel is abstracted into a binary matrix model, where "1" represents the coolant coverage area and "0" represents the uncovered area; combined with external parameters such as radiation intensity, ambient temperature, and coolant flow rate, a photovoltaic thermal panel thermodynamic model including dynamic heat transfer effect is constructed; design variables (such as coolant inlet position and flow rate) and constraints (such as pressure loss, temperature difference, etc.) are defined as optimization inputs; (2) Based on the neural network, nonlinear fitting of heat dissipation efficiency and system characteristics is performed. A multi-layer fully connected neural network is used. The input is the coolant distribution matrix and related environmental parameters, and the output is the heat dissipation efficiency, temperature distribution and pressure loss index. In the network design, RELU is used as the activation function to capture complex nonlinear relationships, and the Dropout mechanism is combined to prevent overfitting. The error between the prediction and the true value is calculated through the loss function MSE, and the learning rate is dynamically adjusted using the Adam optimizer, ultimately achieving a high-precision heat dissipation performance fitting model. (3) Use a multi-objective optimization algorithm to generate an optimized design solution. Based on the fitted heat dissipation efficiency model, with NSGA-II+ARSBX or particle swarm optimization algorithm (PSO) as the core, set a multi-objective optimization function: minimize the average and maximum temperatures of the solar panels, maximize the temperature uniformity coefficient, and minimize the cooling fluid loss. During the optimization process, generate a multi-objective optimal solution set through the Pareto frontier, and screen the solution set based on user needs to ensure that the optimization solution meets the design goals; (4) Perform simulation verification and data recording on the optimization results, and use CFD simulation to analyze the actual heat dissipation effect of the optimization solution, including temperature distribution, pressure loss change, and overall heat dissipation efficiency; the simulation data is stored in the cloud database and associated with the optimization model, providing a reliable basis for subsequent iterative design; (5) Improve optimization efficiency based on intelligent decision support system, by combining Pareto solution set with design weights to calculate solutions that meet specific design priorities; The optimization solution is imported into the machine learning model, and the potential relationship between design variables and performance indicators is analyzed through reverse deduction to further improve design efficiency.
4. The photovoltaic thermal panel heat dissipation optimization method based on neural network and optimization algorithm according to claim 3 is characterized in that: During the optimization process, nonlinear fitting models are used to predict the heat dissipation efficiency and pressure loss, and the optimization performance is calculated through explicit objective functions.
5. The photovoltaic thermal panel heat dissipation optimization method based on neural network and optimization algorithm according to claim 3 is characterized in that: The optimized solutions in the Pareto frontier solution set are body-picked, and the multiple solution sets are screened by evaluating the body-picking distance to determine the optimal combination of design variables, and finally output a set of solutions that meet the balanced design requirements.
6. The photovoltaic thermal panel heat dissipation optimization method based on neural network and optimization algorithm according to claim 3 is characterized in that: By introducing multi-attribute decision-making methods (such as TOPSIS, MOORA or CODAS methods), the design schemes in the Pareto solution set are further sorted in multiple dimensions, the optimal solution is selected and the decision results are output.
7. The photovoltaic thermal panel heat dissipation optimization method based on neural network and optimization algorithm according to claim 3 is characterized in that: In the process of optimization algorithm, a variety of machine learning models (such as BP neural network, RBF network, GPR model) are used to accurately fit and quickly predict the optimization solution to improve the efficiency of the algorithm under complex conditions.
8. The photovoltaic thermal panel heat dissipation optimization method based on neural network and optimization algorithm according to claim 3 is characterized in that: Combining forward design and reverse analysis, the neural network model is trained based on historical optimization data to input design variables to output heat dissipation performance indicators, or to output recommended design parameters through reverse deduction of input performance indicators, thereby achieving dynamic optimization and rapid iteration.