Method and device for optimizing glass fiber combustion recovery of decommissioned fan blades

Abstract

The application relates to the technical field of retired wind turbine blade recycling, in particular to a retired wind turbine blade glass fiber combustion recycling optimization method and device. The method comprises the following steps: inputting initial combustion data of a retired wind turbine blade glass fiber combustion recycling system into a preset machine learning model to obtain initial output data of the retired wind turbine blade glass fiber combustion recycling system; substituting the initial output data into a preset multi-objective function to iteratively optimize the initial combustion data and obtain optimized combustion data; and based on the optimized combustion data, the retired wind turbine blade glass fiber combustion recycling system is regulated and controlled; wherein the multi-objective function is a function with the maximum glass fiber recovery rate, the minimum combustion energy consumption and the minimum nitrogen oxide emission as the targets. In this way, the application can improve the glass fiber recovery efficiency under the premise of ensuring the economy and environmental protection.

Description

Technical Field

[0001] This invention relates to the field of decommissioned wind turbine blade recycling technology, and in particular to an optimized method and apparatus for recycling glass fiber from decommissioned wind turbine blades through combustion. Background Technology

[0002] With the rapid development of the wind power industry, a large number of wind turbine blades are gradually reaching the end of their service life and being decommissioned, making decommissioned wind turbine blades a significant type of solid waste. The main components of wind turbine blades include glass fiber and resin matrix. Glass fiber possesses excellent properties such as high strength and corrosion resistance, giving it extremely high resource recycling value. Currently, combustion recovery is one of the mainstream technologies for recycling glass fiber from decommissioned wind turbine blades. Its principle is to remove the resin matrix from the blades through combustion, thereby achieving the separation and recycling of glass fiber.

[0003] In the combustion recovery process, the proper matching of combustion parameters directly determines the glass fiber recovery rate, combustion energy consumption, and pollutant emissions. To optimize combustion parameters, existing technologies mostly employ computational fluid dynamics (CFD) simulation methods. These methods simulate the gas-solid two-phase flow, heat and mass transfer, and chemical reaction processes within the combustion furnace to analyze the impact of different combustion parameters on the recovery effect. However, CFD simulation involves solving complex physicochemical models, resulting in long simulation times and low computational efficiency. A single complete simulation often takes several hours or even days, making it difficult to achieve rapid iterative optimization of combustion parameters.

[0004] Meanwhile, in existing technologies, there is a mutually restrictive relationship between glass fiber recovery rate, combustion energy consumption and pollutant emissions. For example, increasing the combustion temperature may increase the glass fiber recovery rate, but it will lead to increased combustion energy consumption and increased nitrogen oxide emissions. Existing methods are difficult to achieve synergistic optimization of the three factors and often can only focus on optimizing a single indicator, failing to take into account recycling efficiency, economy and environmental protection.

[0005] Based on this, the present invention proposes an optimized method and apparatus for the combustion and recycling of glass fiber from decommissioned wind turbine blades to solve the above-mentioned technical problems. Summary of the Invention

[0006] This invention describes an optimized method and apparatus for the combustion and recycling of glass fiber from decommissioned wind turbine blades, which can improve the recycling efficiency of glass fiber while ensuring both economic efficiency and environmental friendliness.

[0007] According to a first aspect, the present invention provides an optimized method for the combustion and recycling of glass fiber from decommissioned wind turbine blades, comprising: The initial combustion data of the decommissioned wind turbine blade glass fiber combustion and recovery system is input into a preset machine learning model to obtain the initial output data of the decommissioned wind turbine blade glass fiber combustion and recovery system. The initial output data is substituted into a preset multi-objective function to iteratively optimize the initial combustion data, thereby obtaining optimized combustion data; Based on the optimized combustion data, the glass fiber combustion and recovery system for the decommissioned wind turbine blades is adjusted. The multi-objective function is a function that aims to maximize the glass fiber recycling rate, minimize combustion energy consumption, and minimize nitrogen oxide emissions.

[0008] According to a second aspect, the present invention provides an optimized apparatus for the combustion and recycling of glass fiber from decommissioned wind turbine blades, comprising: The first data processing unit is configured to input the initial combustion data of the decommissioned wind turbine blade glass fiber combustion and recovery system into a preset machine learning model to obtain the initial output data of the decommissioned wind turbine blade glass fiber combustion and recovery system. The second data processing unit is configured to substitute the initial output data into a preset multi-objective function to iteratively optimize the initial combustion data and obtain optimized combustion data. The third data processing unit is configured to regulate the glass fiber combustion and recovery system of the decommissioned wind turbine blades based on the optimized combustion data. The multi-objective function is a function that aims to maximize the glass fiber recycling rate, minimize combustion energy consumption, and minimize nitrogen oxide emissions.

[0009] Thirdly, embodiments of this specification also provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the method described in any embodiment of this specification.

[0010] Fourthly, embodiments of this specification also provide a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the methods described in any embodiment of this specification.

[0011] According to the optimization method and apparatus for the combustion and recycling of glass fiber in decommissioned wind turbine blades provided by the present invention, the initial combustion data of the glass fiber combustion and recycling system for decommissioned wind turbine blades is first input into a preset machine learning model. Through the model's rapid mapping and accurate prediction of the combustion process, the initial output data corresponding to the system is obtained. Subsequently, a multi-objective optimization function is constructed with the optimization objectives of maximizing glass fiber recycling rate, minimizing combustion energy consumption, and minimizing nitrogen oxide emissions. Based on this objective function, the initial combustion data is iteratively optimized through multiple rounds, continuously updating and optimizing the combustion data, ultimately obtaining combustion data with optimized comprehensive performance. After determining the optimal parameters, they are sent as control commands to the glass fiber combustion and recycling system for decommissioned wind turbine blades, enabling real-time adjustment and precise control of the combustion data. Through the above method, the present invention can significantly improve the recycling efficiency and quality of glass fiber while considering operational economy, energy consumption cost, and environmental emission requirements, achieving synergistic optimization of efficient resource regeneration, energy consumption reduction, and pollutant emission reduction. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. 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 effort.

[0013] Figure 1 A schematic flowchart of an optimized method for the combustion and recycling of glass fiber from decommissioned wind turbine blades according to one embodiment is shown. Figure 2 A schematic block diagram of an optimized apparatus for the combustion and recycling of glass fiber from decommissioned wind turbine blades according to one embodiment is shown. Detailed Implementation

[0014] The solution provided by the present invention will now be described with reference to the accompanying drawings.

[0015] Figure 1 This diagram illustrates an optimized method for the combustion and recycling of glass fiber from decommissioned wind turbine blades according to one embodiment. It is understood that this method can be executed by any device, equipment, platform, or cluster of devices with computing and processing capabilities. Figure 1 As shown, the method includes: Step 100: Input the initial combustion data of the decommissioned wind turbine blade glass fiber combustion and recovery system into the preset machine learning model to obtain the initial output data of the decommissioned wind turbine blade glass fiber combustion and recovery system. Step 102: Substitute the initial output data into the preset multi-objective function to iteratively optimize the initial combustion data and obtain the optimized combustion data; Step 104: Based on the optimized combustion data, adjust the glass fiber combustion and recovery system for decommissioned wind turbine blades; Among them, the multi-objective function is a function that aims to maximize the glass fiber recycling rate, minimize combustion energy consumption, and minimize nitrogen oxide emissions.

[0016] In this embodiment, the initial combustion data of the decommissioned wind turbine blade glass fiber combustion and recovery system is first input into a preset machine learning model. Through the model's rapid mapping and accurate prediction of the combustion process, the initial output data of the system is obtained. Subsequently, a multi-objective optimization function is constructed with the optimization objectives of maximizing glass fiber recovery rate, minimizing combustion energy consumption, and minimizing nitrogen oxide emissions. Based on this objective function, the initial combustion data is iteratively optimized through multiple rounds, continuously updating and optimizing the combustion data until the combustion data with optimized overall performance is obtained. After determining the optimal parameters, they are sent as control commands to the decommissioned wind turbine blade glass fiber combustion and recovery system for real-time adjustment and precise control of the combustion data. Through this method, the present invention can significantly improve the recycling efficiency and quality of glass fiber while considering operational economy, energy cost, and environmental emission requirements, achieving synergistic optimization of efficient resource regeneration, energy consumption reduction, and pollutant emission reduction.

[0017] In one embodiment of the present invention, the initial combustion data includes the initial combustion temperature, the initial excess air coefficient, the initial material residence time, and the initial feed rate; the initial output data includes the initial glass fiber recovery rate, the initial combustion energy consumption, and the initial nitrogen oxide emissions.

[0018] In this embodiment, the initial combustion data are key operating parameters of the glass fiber combustion and recovery system for decommissioned wind turbine blades, including the initial combustion temperature, initial excess air coefficient, initial material residence time, and initial feed rate, which directly affect the recovery effect, energy consumption level, and pollutant emissions. The initial output data are the system's operational response indicators under the corresponding initial combustion parameters, mainly including the initial glass fiber recovery rate, initial combustion energy consumption, and initial nitrogen oxide emissions, which can intuitively reflect the recovery efficiency, economic performance, and environmental compliance level under the current operating conditions.

[0019] In one embodiment of the present invention, the initial output data is substituted into a preset multi-objective function to iteratively optimize the initial combustion data, thereby obtaining optimized combustion data, including: The particle swarm size is set to a first preset threshold, each particle corresponds to a set of combustion data, a multi-objective function is used as the appropriate function, and particle velocity, individual optimal solution, and global optimal solution are initialized at the same time. Each particle updates its own velocity and position based on its own velocity, its individual optimal solution, and the global optimal solution; Each time the location is updated, the output data corresponding to the combustion data is predicted by a preset machine learning model, and the output data is substituted into the fitness function to calculate the fitness value. Compare the current particle fitness value with its own historical best and global best, update the individual best solution and global best solution, and retain particles that meet the conditions of glass fiber recovery rate greater than the second preset threshold and nitrogen oxide emission less than the third preset threshold. When the particle swarm size reaches the first preset threshold, the iteration stops in order to obtain the Pareto optimal solution set. Based on the Pareto optimal solution set, the optimized combustion data is determined.

[0020] In this embodiment, iterative optimization calculations are performed on the initial combustion data according to a preset multi-objective function to obtain optimized combustion data with the best overall performance. First, the particle swarm size is set to a first preset threshold, mapping each particle to a complete set of combustion data to construct a multi-dimensional optimization space. The aforementioned multi-objective function is used as the fitness function of the algorithm to evaluate the quality of each set of combustion parameters. Before the iteration begins, the particle velocity, individual optimal solution, and global optimal solution are initialized.

[0021] During the iterative optimization process, each particle updates its velocity and position synchronously according to a preset update rule based on its current velocity, the individual optimal solution obtained from historical searches, and the global optimal solution of the entire particle swarm, thereby achieving dynamic adjustment of combustion parameters within the feasible region. Each time a particle completes a position update, a new set of combustion data is generated. A pre-trained machine learning model then rapidly and accurately predicts the output indicators corresponding to this set of combustion data, obtaining output data such as glass fiber recovery rate, combustion energy consumption, and nitrogen oxide emissions. The predicted output data is substituted into the fitness function to calculate the current particle's fitness value, which serves as a quantitative basis for parameter quality. Subsequently, the current particle's fitness value is compared with its own historical optimal fitness value and the global optimal fitness value, and the individual optimal solution and the global optimal solution are updated accordingly, ensuring that optimal information is continuously transmitted and optimized during the iteration process. A constraint screening mechanism is set during the iteration process, retaining only valid particles with a glass fiber recovery rate greater than a second preset threshold and nitrogen oxide emissions less than a third preset threshold, while eliminating invalid solutions that do not meet process requirements and environmental indicators, improving optimization efficiency and result reliability. When the number of iterations reaches the first preset threshold and the convergence condition is met, the entire iteration process stops, and the Pareto optimal solution set obtained by non-dominated sorting is output. Based on this Pareto optimal solution set, a set of optimized combustion data that balances economic efficiency, energy consumption level, and environmental emission indicators is finally determined.

[0022] In this embodiment, the first preset threshold can be set to 100, the second preset threshold can be set to 85%, and the third preset value can be set to 150 milligrams per cubic meter.

[0023] In one embodiment of the present invention, the multi-objective function is constructed using the following formula:

[0024] In the formula, The function value of a multi-objective function. For glass fiber recycling rate, For combustion energy consumption, For nitrogen oxide emissions, The constraint penalty function, It is a dynamic weight matrix. The correction coefficient for the first objective. This is the correction factor for the second objective. This is the correction factor for the third objective. To constrain the penalty coefficient, To ensure the quality of the recycled glass fiber, This refers to the total mass of glass fibers in the blades of a decommissioned wind turbine. This is a measurement error correction term. This represents the total heat of the fuel consumed during combustion. This represents the total power consumption of auxiliary equipment during the combustion process. Waste heat recovered from combustion flue gas, The total feed mass during the combustion process. This refers to the total mass of nitrogen oxides produced during combustion. This refers to the total volume of flue gas produced during combustion. This represents the flue gas dilution coefficient.

[0025] In this embodiment, existing technologies often use fixed weights to optimize recovery rate, energy consumption, and pollutant emissions, which cannot adapt to the optimization needs of different scenarios, such as the difference in weight requirements between environmentally prioritized scenarios and economically prioritized scenarios. The dynamic weight matrix designed in this patent can flexibly adjust the weight ratios of glass fiber recovery rate, combustion energy consumption, and nitrogen oxide emissions according to actual production needs, ensuring that the sum of the weights is 1. This guarantees the rigor of the optimization logic and achieves multi-scenario adaptability, solving the pain point of poor scenario adaptability in existing technologies. Furthermore, this invention introduces a target correction coefficient to eliminate dimensional differences and improve optimization accuracy. Existing technologies often fail to consider the dimensional differences of different optimization objectives and directly perform weighted calculations, leading to significant deviations in optimization results; at the same time, the lack of a target correction mechanism cannot offset minor deviations in the actual measurement and calculation process. Integrating constraint penalty terms achieves deep binding between optimization and process constraints. Existing multi-objective optimization equations often only focus on optimizing the objective function, failing to deeply integrate process constraints with the optimization process, resulting in some optimization results exceeding the feasible range of actual processes and thus being unapplicable. This patent organically combines the constraint penalty function with the overall optimization equation. When the combustion parameters exceed the process constraint range, the overall optimization target value is increased by the penalty coefficient, and invalid solutions are automatically eliminated. This ensures that the optimization process is always carried out within the process feasible domain, realizing the practicality and feasibility of the optimization results.

[0026] In this embodiment, the dynamic weight matrix can be , The correction coefficient for the first objective. This is the correction factor for the second objective. The range of values ​​for the correction coefficient for the third objective is all within .

[0027] In one embodiment of the present invention, the constraint penalty function is constructed by the following formula:

[0028] In the formula, The combustion temperature, The excess air coefficient, For material residence time, For feed rate, For the first The actual values ​​of each constraint parameter For the first The upper limit value of each constraint parameter. For constraint parameter numbers, For the first The lower limit value of each constraint parameter.

[0029] In this embodiment, the constraint penalty function of the present invention adopts a segmented penalty logic to achieve a match between constraint judgment and penalty intensity. Existing penalty functions often employ a single penalty logic, applying the same penalty method regardless of the degree of parameter deviation from the constraint. This fails to accurately quantify the degree of deviation, resulting in either insufficient penalty leading to residual invalid solutions or excessive penalty affecting optimization efficiency. The segmented penalty function designed in this invention clearly distinguishes between three scenarios: "meeting the constraint," "exceeding the upper limit," and "below the lower limit." No penalty is applied to particles that meet the constraint; for particles that exceed the constraint, a penalty value is calculated based on the degree of deviation, with larger deviations resulting in larger penalty values, thus achieving a match between penalty intensity and deviation degree. Second, the constraint parameters are comprehensive and highly targeted, closely aligning with actual process requirements. Existing penalty functions often only constrain combustion parameters, failing to include output indicators within the constraint scope. This leads to some optimization results meeting combustion parameter constraints, but with recovery efficiency and pollutant emissions not meeting expectations, failing to achieve the synergistic goal of "efficiency-energy saving-environmental protection." The penalty function of this patent not only includes combustion temperature, excess air coefficient, material residence time, and feed rate, but also incorporates two major output indicators, glass fiber recovery rate and nitrogen oxide emissions, into the constraint range, comprehensively covering the constraints of process operation and ensuring that the optimization results simultaneously meet the requirements of process feasibility and target achievement.

[0030] In this embodiment, the constraint parameters refer to combustion temperature, excess air coefficient, material residence time, feed rate, glass fiber recovery rate, and nitrogen oxide emissions. If any of these parameters exist... (Only parameters exceeding the upper limit are summed), if there exists any (Only parameters exceeding the lower limit are summed).

[0031] In one embodiment of the present invention, the preset machine learning model includes an input layer, a feature preprocessing layer, a GBDT core training layer, an SVR correction layer, a feature fusion layer, and an output layer connected in sequence. The input layer is used to receive initial combustion data, the feature preprocessing layer is used to purify, standardize, and enhance the features of the combustion data, the GBDT core training layer is used to capture the nonlinear mapping features between the combustion data and the output indicators, the SVR correction layer is used to correct the prediction bias of the GBDT core training layer, the feature fusion layer receives the GBDT nonlinear features and the SVR correction features and outputs a fused feature matrix, and the output layer receives the fused feature matrix and outputs the initial output data.

[0032] In this embodiment, the preset machine learning model adopts a hybrid architecture of "GBDT+SVR", including an input layer, a feature preprocessing layer, a GBDT core training layer, an SVR correction layer, a feature fusion layer, and an output layer connected in sequence. The input layer is specifically used to receive the initial combustion data from the glass fiber combustion and recovery system for decommissioned wind turbine blades, ensuring the comprehensiveness and relevance of the data input. The feature preprocessing layer is responsible for systematically processing the input combustion data, sequentially performing data purification, standardization, and feature enhancement operations. By removing outliers, eliminating dimensional differences, and mining the interaction features between parameters, it effectively improves the quality of the input data. The GBDT core training layer, as the core unit of the model, focuses on capturing the complex nonlinear mapping relationship between the preprocessed combustion data and output indicators, and mining the influence of combustion parameters on glass fiber recovery rate, combustion energy consumption, and nitrogen oxide emissions. The SVR correction layer specifically corrects the prediction bias of the GBDT core training layer, effectively reducing the impact of model overfitting and further improving prediction accuracy. The feature fusion layer receives the nonlinear mapping features output from the GBDT core training layer and the corrected features output from the SVR correction layer. It achieves feature fusion through weighted concatenation, outputting a fused feature matrix that combines accuracy and completeness. Finally, the output layer receives this fused feature matrix, performs regression calculations, and accurately outputs the initial output data, namely the initial glass fiber recovery rate, initial combustion energy consumption, and initial nitrogen oxide emissions.

[0033] In this embodiment, the training data for the preset machine learning model consists of multiple sets of complete operating condition samples. Each set of samples includes four input parameters and three output parameters. The input parameters specifically include: combustion temperature ranging from 600℃ to 1000℃, excess air coefficient ranging from 1.05 to 1.3, material residence time ranging from 30 minutes to 120 minutes, and feed rate ranging from 50 kg / h to 200 kg / h. The output parameters specifically include: glass fiber recovery rate ranging from 85% to 95%, combustion energy consumption ranging from 800 kJ / kg to 1200 kJ / kg, and nitrogen oxide emissions ranging from [missing data]. The training data was obtained through a combination of numerical simulation and experimental testing. The samples include typical and boundary operating points for each parameter. The data is evenly distributed and highly representative, and can fully reflect the nonlinear mapping law between input parameters and output indicators during combustion. This provides a stable and reliable data foundation for model training, ensuring that the model has high prediction accuracy and generalization ability across the entire operating range.

[0034] In one embodiment of the present invention, the machine learning model is one or more combinations of gradient boosting decision tree, support vector regression, or artificial neural network.

[0035] The foregoing has described specific embodiments of the invention. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0036] According to another embodiment, the present invention provides an optimized apparatus for the combustion and recycling of glass fiber from decommissioned wind turbine blades. Figure 2 A schematic block diagram of an optimized apparatus for the combustion and recovery of glass fiber from decommissioned wind turbine blades according to one embodiment is shown. It will be understood that this apparatus can be implemented by any device, equipment, platform, or cluster of devices with computing and processing capabilities. Figure 2 As shown, the device includes: a first data processing unit 200, a second data processing unit 202, and a third data processing unit 204. The main functions of each component are as follows: The first data processing unit 200 is configured to input the initial combustion data of the decommissioned wind turbine blade glass fiber combustion and recovery system into a preset machine learning model to obtain the initial output data of the decommissioned wind turbine blade glass fiber combustion and recovery system. The second data processing unit 202 is configured to substitute the initial output data into a preset multi-objective function to iteratively optimize the initial combustion data and obtain optimized combustion data. The third data processing unit 204 is configured to regulate the glass fiber combustion and recovery system of the decommissioned wind turbine blades based on the optimized combustion data. The multi-objective function is a function that aims to maximize the glass fiber recycling rate, minimize combustion energy consumption, and minimize nitrogen oxide emissions.

[0037] In one embodiment of the present invention, the initial combustion data includes the initial combustion temperature, the initial excess air coefficient, the initial material residence time, and the initial feed rate; the initial output data includes the initial glass fiber recovery rate, the initial combustion energy consumption, and the initial nitrogen oxide emissions.

[0038] In one embodiment of the present invention, the second data processing unit 202 is configured to perform the following operations: The particle swarm size is set to a first preset threshold, each particle corresponds to a set of combustion data, the multi-objective function is used as the appropriate function, and the particle velocity, individual optimal solution, and global optimal solution are initialized at the same time. Each particle updates its own velocity and position based on its own velocity, its individual optimal solution, and the global optimal solution; Each time the location is updated, the output data corresponding to the combustion data is predicted by the preset machine learning model, and the output data is substituted into the fitness function to calculate the fitness value; Compare the current particle fitness value with its own historical best and global best, update the individual best solution and global best solution, and retain particles that meet the conditions of glass fiber recovery rate greater than the second preset threshold and nitrogen oxide emission less than the third preset threshold. When the particle swarm size reaches the first preset threshold, the iteration stops to obtain the Pareto optimal solution set; Based on the Pareto optimal solution set, the optimized combustion data is determined.

[0039] In one embodiment of the present invention, the multi-objective function is constructed by the following formula:

[0040] In the formula, The function value of a multi-objective function. For glass fiber recycling rate, For combustion energy consumption, For nitrogen oxide emissions, The constraint penalty function, It is a dynamic weight matrix. The correction coefficient for the first objective. This is the correction factor for the second objective. This is the correction factor for the third objective. To constrain the penalty coefficient, To ensure the quality of the recycled glass fiber, This refers to the total mass of glass fibers in the blades of a decommissioned wind turbine. This is a measurement error correction term. This represents the total heat of the fuel consumed during combustion. This represents the total power consumption of auxiliary equipment during the combustion process. Waste heat recovered from combustion flue gas, The total feed mass during the combustion process. This refers to the total mass of nitrogen oxides produced during combustion. This refers to the total volume of flue gas produced during combustion. This represents the flue gas dilution coefficient.

[0041] In one embodiment of the present invention, the constraint penalty function is constructed by the following formula:

[0042] In the formula, The combustion temperature, The excess air coefficient, For material residence time, For feed rate, For the first The actual values ​​of each constraint parameter For the first The upper limit value of each constraint parameter. For constraint parameter numbers, For the first The lower limit value of each constraint parameter.

[0043] In one embodiment of the present invention, the preset machine learning model includes an input layer, a feature preprocessing layer, a GBDT core training layer, an SVR correction layer, a feature fusion layer, and an output layer connected in sequence. The input layer is used to receive the initial combustion data. The feature preprocessing layer is used to purify, standardize, and enhance the features of the combustion data. The GBDT core training layer is used to capture the nonlinear mapping features between the combustion data and the output indicators. The SVR correction layer is used to correct the prediction bias of the GBDT core training layer. The feature fusion layer receives the GBDT nonlinear features and the SVR correction features and outputs a fused feature matrix. The output layer receives the fused feature matrix and outputs the initial output data.

[0044] In one embodiment of the present invention, the machine learning model is one or more combinations of gradient boosting decision tree, support vector regression, or artificial neural network.

[0045] According to another embodiment, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed in a computer, causes the computer to perform a combination Figure 1 The method described.

[0046] According to another embodiment, an electronic device is also provided, including a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, it implements a combination... Figure 1 The method described.

[0047] The various embodiments in this invention are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the apparatus embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0048] Those skilled in the art will recognize that, in one or more of the examples above, the functions described in this invention can be implemented using hardware, software, firmware, or any combination thereof. When implemented in software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium.

[0049] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solution of the present invention should be included within the scope of protection of the present invention.

Claims

1. An optimized method for the combustion and recycling of glass fiber from decommissioned wind turbine blades, characterized in that, include: The initial combustion data of the decommissioned wind turbine blade glass fiber combustion and recovery system is input into a preset machine learning model to obtain the initial output data of the decommissioned wind turbine blade glass fiber combustion and recovery system. The initial output data is substituted into a preset multi-objective function to iteratively optimize the initial combustion data, thereby obtaining optimized combustion data; Based on the optimized combustion data, the glass fiber combustion and recovery system for the decommissioned wind turbine blades is adjusted. The multi-objective function is a function that aims to maximize the glass fiber recycling rate, minimize combustion energy consumption, and minimize nitrogen oxide emissions.

2. The method according to claim 1, characterized in that, The initial combustion data includes the initial combustion temperature, the initial excess air coefficient, the initial material residence time, and the initial feed rate; the initial output data includes the initial glass fiber recovery rate, the initial combustion energy consumption, and the initial nitrogen oxide emissions.

3. The method according to claim 2, characterized in that, The step of iteratively optimizing the initial combustion data according to a preset multi-objective function to obtain optimized combustion data includes: The particle swarm size is set to a first preset threshold, each particle corresponds to a set of combustion data, the multi-objective function is used as the appropriate function, and the particle velocity, individual optimal solution, and global optimal solution are initialized at the same time. Each particle updates its own velocity and position based on its own velocity, its individual optimal solution, and the global optimal solution; Each time the location is updated, the output data corresponding to the combustion data is predicted by the preset machine learning model, and the output data is substituted into the fitness function to calculate the fitness value; Compare the current particle fitness value with its own historical best and global best, update the individual best solution and global best solution, and retain particles that meet the conditions of glass fiber recovery rate greater than the second preset threshold and nitrogen oxide emission less than the third preset threshold. When the particle swarm size reaches the first preset threshold, the iteration stops to obtain the Pareto optimal solution set; Based on the Pareto optimal solution set, the optimized combustion data is determined.

4. The method according to claim 2, characterized in that, The multi-objective function is constructed using the following formula: In the formula, The function value of a multi-objective function. For glass fiber recycling rate, For combustion energy consumption, For nitrogen oxide emissions, The constraint penalty function, It is a dynamic weight matrix. The correction coefficient for the first objective. This is the correction factor for the second objective. This is the correction factor for the third objective. To constrain the penalty coefficient, To ensure the quality of the recycled glass fiber, This refers to the total mass of glass fibers in the blades of a decommissioned wind turbine. This is a measurement error correction term. This represents the total heat of the fuel consumed during combustion. This represents the total power consumption of auxiliary equipment during the combustion process. Waste heat recovered from combustion flue gas, The total feed mass during the combustion process. This refers to the total mass of nitrogen oxides produced during combustion. This refers to the total volume of flue gas produced during combustion. This represents the flue gas dilution coefficient.

5. The method according to claim 4, characterized in that, The constraint penalty function is constructed using the following formula: In the formula, The combustion temperature, The excess air coefficient, For material residence time, For feed rate, For the first The actual values ​​of each constraint parameter For the first The upper limit value of each constraint parameter. For constraint parameter numbers, For the first The lower limit value of each constraint parameter.

6. The method according to claim 1, characterized in that, The preset machine learning model includes an input layer, a feature preprocessing layer, a GBDT core training layer, an SVR correction layer, a feature fusion layer, and an output layer connected in sequence. The input layer receives the initial combustion data. The feature preprocessing layer cleans, standardizes, and enhances the combustion data. The GBDT core training layer captures the nonlinear mapping features between the combustion data and the output indicators. The SVR correction layer corrects the prediction bias of the GBDT core training layer. The feature fusion layer receives the GBDT nonlinear features and the SVR correction features and outputs a fused feature matrix. The output layer receives the fused feature matrix and outputs the initial output data.

7. The method according to claim 1, characterized in that, The machine learning model is one or more combinations of gradient boosting decision tree, support vector regression, or artificial neural network.

8. An optimized device for the combustion and recycling of glass fiber from decommissioned wind turbine blades, characterized in that, include: The first data processing unit is configured to input the initial combustion data of the decommissioned wind turbine blade glass fiber combustion and recovery system into a preset machine learning model to obtain the initial output data of the decommissioned wind turbine blade glass fiber combustion and recovery system. The second data processing unit is configured to substitute the initial output data into a preset multi-objective function to iteratively optimize the initial combustion data and obtain optimized combustion data. The third data processing unit is configured to regulate the glass fiber combustion and recovery system of the decommissioned wind turbine blades based on the optimized combustion data. The multi-objective function is a function that aims to maximize the glass fiber recycling rate, minimize combustion energy consumption, and minimize nitrogen oxide emissions.

9. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed in a computer, causes the computer to perform the method described in any one of claims 1-7.

Images