Intelligent combustion regulation method and system based on online optical fiber detection of flame signal

Abstract

This invention discloses an intelligent combustion control method and system based on online fiber optic detection of flame signals. The method monitors the flame light signal emitted by the burner via fiber optic cable, analyzes the signal to predict pollutant concentrations, and performs fuzzy adjustment and defuzzification on the burner's feedback signal based on the predicted values, thereby achieving intelligent and adaptive control of the combustion process. This method can quickly respond to changes in combustion conditions, optimize control strategies, ensure the stability and environmental friendliness of the combustion process, overcome the technical bottlenecks of high cost and poor real-time performance, effectively reduce system hardware costs, and improve the intelligence level of the control process.

Description

Technical Field

[0001] This invention relates to the field of flame signal detection technology, and in particular to an intelligent combustion control method and system based on online fiber optic detection of flame signals. Background Technology

[0002] With increasingly stringent requirements for energy conservation and emission reduction, various industrial combustion devices have an urgent need for clean and efficient combustion technologies. Precise control and optimization of the combustion process are crucial for achieving clean and efficient combustion. Flame characteristics directly reflect the combustion state of fuel. Online and precise monitoring of key parameters in the flame and exhaust emissions allows for real-time assessment of the fuel combustion state. Optimizing and adjusting combustion condition parameters based on the combustion state enables proactive control of the combustion process, thereby achieving the goal of clean and efficient combustion.

[0003] The monitoring of the existing combustion process is mainly achieved through traditional flame image processing technology, online detection of combustion device temperature and pressure, and multi-component exhaust gas, among other methods. Flame image monitoring, in particular, uses cameras to capture and analyze flame images to monitor the combustion state and further enable combustion control.

[0004] While existing flame image processing technology is mature, it still faces challenges such as high-cost equipment, unclear signal mapping, large data volume, and high computational complexity, which make real-time control difficult. At the same time, traditional temperature, pressure, and exhaust gas composition measurements cannot directly reflect flame characteristics and require the combination of flame optical signals to accurately determine the combustion status. Traditional control methods based on physical models lack intelligent features. Summary of the Invention

[0005] To address the aforementioned problems, this invention proposes an intelligent combustion control method and system based on online fiber optic detection of flame signals, aiming to improve the real-time performance and intelligence level of burner control at a low cost.

[0006] To address the aforementioned technical problems, the first aspect of this invention proposes an intelligent combustion control method based on online fiber optic detection of flame signals, comprising the following steps:

[0007] The optical fiber acquires the operating parameters of the burner and the emitted flame light signal, converts the free radical signal in the flame light signal into an analog electrical signal, and then converts the analog electrical signal into a digital electrical signal, while simultaneously acquiring the flame temperature and actual pollutant concentration output by the burner.

[0008] The time-domain and frequency-domain features of digital electrical signals are extracted, and the ratio of free radical signals in the digital electrical signals is calculated. A multi-dimensional feature dataset is generated using operating parameters, time-domain features, frequency-domain features, ratios, and flame temperature, and a regular relationship is established with flame stability.

[0009] The feature dataset is used as the input variable of the neural network, and the actual pollutant concentration generated during combustion is used as the output variable to predict the pollutant concentration under different combustion conditions.

[0010] Error analysis is performed on the predicted pollutant concentration within a preset threshold range, and the feedback signal is fuzzified by combining the fuzzy rule base;

[0011] The defuzzified feedback signal is used to dynamically adjust the fuel and air flow of the burner.

[0012] In some implementations, time-domain features include mean, root mean variance, and peak amplitude, while frequency-domain features include dominant frequency components and energy spectrum.

[0013] In some implementations, operating parameters include fuel ratio, equivalence ratio, and combustion power.

[0014] In some implementations, after the feature dataset is generated, the digital electrical signal is denoised based on moving average filtering, and the mean of the denoised feature dataset is normalized according to preset conditions.

[0015] In some implementations, the fuzzy control rule base determines the control strategy based on empirical rules. When the predicted pollutant concentration exceeds the limit, the system adjusts the flow rate of fuel and air according to the control strategy.

[0016] In some implementation methods, the predicted pollutant concentration is calculated as follows:

[0017]

[0018] in, ω represents the predicted pollutant concentration. jk h represents the weights from the hidden layer to the output layer of the neural network. j Forward propagation formula, b k For bias.

[0019] In some implementations, error analysis includes: comparing the predicted pollutant concentration with the actual pollutant concentration within the preset threshold range to obtain an error signal and an error rate of change; converting the error signal and error rate of change into fuzzy linguistic variables through a fuzzy membership function; and defuzzifying the feedback signal based on the fuzzy linguistic variables.

[0020] In some implementations, the fuzzy membership function is:

[0021]

[0022] Where e is the error signal, and a and b are the thresholds of the membership function.

[0023] In some implementations, the defuzzified feedback signal is:

[0024]

[0025] Where e(t) is the error signal, K p K i K d These are the proportional, integral, and derivative coefficients of the PID controller after adjustment using the fuzzy rule base.

[0026] A second aspect of this invention proposes an intelligent combustion control system based on online fiber optic detection of flame signals, comprising:

[0027] The data acquisition module uses optical fibers to collect the operating parameters of the burner and the emitted flame light signal. It converts the free radical signal in the flame light signal into an analog electrical signal, and then converts the analog electrical signal into a digital electrical signal. At the same time, it acquires the flame temperature and actual pollutant concentration output by the burner.

[0028] The data processing module is used to extract the time-domain and frequency-domain features of digital electrical signals, and simultaneously calculate the ratio of free radical signals in the digital electrical signals. It generates a multi-dimensional feature dataset based on operating parameters, time-domain features, frequency-domain features, ratios, and flame temperature, and establishes a regular relationship with flame stability.

[0029] The data analysis module is used to use the feature dataset as the input variable of the neural network, and the neural network uses the actual pollutant concentration generated during the combustion process as the output variable to predict the pollutant concentration under different combustion conditions.

[0030] The fuzzy PID control module is used to perform error analysis on the predicted pollutant concentration within a preset threshold range, and to fuzzify the feedback signal by combining it with a fuzzy rule base;

[0031] The execution module uses defuzzified feedback signals to dynamically adjust the fuel and air flow of the burner.

[0032] The beneficial effects of this invention are as follows: By monitoring the flame light signal emitted by the burner through optical fiber, analyzing the flame light signal to predict pollutant concentration, and performing fuzzy adjustment and defuzzification on the burner's feedback signal based on the predicted value, intelligent and adaptive control of the combustion process is achieved. This method can quickly respond to changes in combustion conditions, optimize control strategies, ensure the stability and environmental friendliness of the combustion process, solve the technical bottlenecks of high cost and poor real-time performance, effectively reduce system hardware costs, and improve the intelligence level of the control process. Attached Figure Description

[0033] Figure 1This is a flowchart of the intelligent combustion control method based on online fiber optic detection of flame signals disclosed in Embodiment 1 of the present invention;

[0034] Figure 2 This is a flowchart of the intelligent combustion control system based on online fiber optic detection of flame signals disclosed in Embodiment 3 of the present invention. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the content of this invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to this invention are shown in the accompanying drawings, not all of them.

[0036] Example 1

[0037] This embodiment proposes an intelligent combustion control method based on online fiber optic detection of flame signals, such as... Figure 1 As shown, it includes the following steps:

[0038] Step 1: The optical fiber acquires the operating parameters of the burner and the emitted flame light signal, converts the free radical signal in the flame light signal into an analog electrical signal, and then converts the analog electrical signal into a digital electrical signal. At the same time, it acquires the flame temperature and actual pollutant concentration (i.e., exhaust gas composition) output by the burner.

[0039] Free radicals are highly reactive chemical groups that can react with other free radicals and molecules, causing combustion to extend in a chain reaction. In this embodiment, by extracting free radical signals from the flame light signal, the fuel combustion process can be analyzed and diagnosed through the distribution characteristics and evolution patterns of these groups, providing a theoretical basis for improving combustion conditions and increasing combustion efficiency. Furthermore, the aforementioned operating parameters include, but are not limited to, fuel ratio, equivalence ratio, and combustion power.

[0040] Step 2: Extract the time-domain and frequency-domain features of the digital electrical signal, and calculate the ratio of free radical signals in the digital electrical signal. Generate a multi-dimensional feature dataset based on operating parameters, time-domain features, frequency-domain features, ratios, and flame temperature, and establish a regular relationship with flame stability.

[0041] The aforementioned time-domain features include mean, root mean variance, and peak amplitude, while the frequency-domain features include the main frequency component and energy spectrum. These are all important features of free radical signals and, together with other real-time information, serve as the input information for step 3.

[0042] Optionally, by extracting the mean of time-domain features i and root mean square and its peak amplitude, where N is the number of sampling points of the signal, x i It is the i-th data point in the signal, from which frequency domain features such as the main frequency component and energy spectrum are extracted.

[0043] Even better, after the feature dataset is generated, the digital electrical signal is denoised using a moving average filter to remove signal noise and improve data reliability. Then, the denoised feature dataset is normalized according to preset conditions, such as Z-score normalization, to unify the data scale. In one example, such as moving average filtering... Denoising is performed, where M is the width of the sliding window and k is the number of points on the left and right sides of the window; then, Z-score normalization is applied. Where μ is the mean of the signal and σ is the standard deviation of the signal.

[0044] Step 3: Use the feature dataset (i.e., containing operating parameters, time-domain features, frequency-domain features, ratios, and flame temperature) as the input variable of the neural network. Use the actual pollutant concentration generated during combustion as the output variable to predict the pollutant concentration under different combustion conditions.

[0045] In this embodiment, the neural network model used is particularly important. This model receives a processed feature dataset and employs a multi-layer feedforward neural network, including an input layer, hidden layers, and an output layer. The input layer receives flame feature data (such as the ratio of free radicals and other combustion parameters). The hidden layer performs a non-linear mapping on the input features. The output layer represents the actual pollutant concentration, and the predicted value is the pollutant concentration. The predicted feature is a vector X = [x1, x2, ..., x...]. n ], where x i Let represent the i-th input feature, and n be the total number of input features.

[0046] The forward propagation formula is as follows:

[0047]

[0048] Where, x i For the input features, ω ij As the weight, b j σ is the bias, and σ is the activation function.

[0049] The method for calculating the predicted values, i.e., the predicted pollutant concentration values, in the output layer is as follows:

[0050]

[0051] in, ω represents the predicted pollutant concentration. jk h represents the weights from the hidden layer to the output layer of the neural network. j Forward propagation formula, bk For bias.

[0052] Then, a loss function is used to measure the difference between the predicted value and the actual value under the current operating conditions:

[0053]

[0054] Among them, y k This represents the actual pollutant concentration. θ represents the predicted value, m represents the number of samples, and θ represents the network weights and biases.

[0055] Then, backpropagation and parameter updates are performed. The backpropagation of the neural network calculates the gradient of the loss function with respect to each weight and bias using the chain rule to update these parameters and minimize the loss function.

[0056] Weight gradient: For the weights from the hidden layer to the output layer, the gradient is:

[0057]

[0058] Among them, h j This is the output of the hidden layer neurons.

[0059] Bias gradient: for the bias b of the output layer k Its gradient is:

[0060]

[0061] Weight Update: The weights and biases are updated using gradient descent, with each iteration updating as follows:

[0062]

[0063]

[0064] Here, α is the learning rate, which controls the magnitude of weight updates.

[0065] Reinforcement learning algorithms continuously optimize the weights and biases of the neural network model using historical data and real-time feedback, enabling the system to cope with dynamic changes during the combustion process and improve the accuracy of predictions.

[0066] Step 4: Perform error analysis on the predicted pollutant concentration within the preset threshold range, and fuzzify the feedback signal using a fuzzy rule base.

[0067] The fuzzy control rule base determines the control strategy based on empirical rules. When the predicted pollutant concentration exceeds the standard, the system adjusts the flow rate of fuel and air according to the control strategy.

[0068] The aforementioned error calculation refers to the process where the input to the fuzzy PID controller is the error signal of the combustion process, and the calculation is performed on the difference between the actual pollutant concentration within a preset low threshold range and the ideal concentration predicted by the neural network. In one example, the error analysis includes: comparing the predicted pollutant concentration with the actual pollutant concentration within the low threshold range to obtain the error signal and the error rate of change; converting the error signal and the error rate of change into fuzzy linguistic variables using a fuzzy membership function; and defuzzifying the feedback signal based on the fuzzy linguistic variables.

[0069] To handle the nonlinearity and uncertainty of the combustion process, the fuzzy PID controller fuzzifies the error signal and the rate of change of error. Through fuzzy membership functions, the error and the rate of change of error are converted into fuzzy linguistic variables, such as "low," "medium," and "high." Membership functions are typically defined using triangular or trapezoidal functions. A fuzzy membership function (which can also be understood as the membership function for an error of "low") can be expressed as:

[0070]

[0071] Where e is the error signal, and a and b are the thresholds of the membership function.

[0072] The system infers the error and the rate of change of error based on a predefined fuzzy rule base. For example, if the error is "high" and the rate of change of error is "increasing", the coefficient of the feedback signal needs to be increased significantly; if the error is "medium" and the rate of change of error is "stable", the feedback signal is fine-tuned.

[0073] Step 5: The defuzzified feedback signal is used to dynamically adjust the fuel and air flow of the burner.

[0074] Defuzzification refers to the process by which the feedback signals obtained from fuzzy inference (such as adjustment values ​​for proportional, integral, and derivative coefficients) are converted into precise numerical control quantities through a defuzzification step. In this embodiment, the defuzzified value is calculated using the centroid method, and the defuzzified feedback signal is as follows:

[0075]

[0076] Where e(t) is the error signal, K p K i K d These are the proportional, integral, and derivative coefficients of the PID controller after adjustment using the fuzzy rule base.

[0077] The fuel and air supply are adjusted based on the defuzzified feedback signal to achieve closed-loop regulation of the combustion process.

[0078] In summary, this method monitors the flame light signal emitted by the burner, predicts the pollutant concentration based on the flame light signal, and performs fuzzy adjustment and defuzzification on the burner's feedback signal based on the predicted value. This enables the system to quickly adapt to constantly changing combustion conditions, gradually optimize the control strategy, and ensure the stability and environmental friendliness of the combustion process.

[0079] Furthermore, the flame light signal in this method can be acquired by an optical fiber sensor, which is significantly different from traditional flame detection methods. It eliminates the need to purchase expensive high-resolution cameras, saving equipment costs and enabling high-precision detection in high-temperature and complex environments.

[0080] Furthermore, by integrating neural network models with reinforcement learning algorithms and fuzzy PID control, real-time combustion optimization and regulation are achieved. It can not only monitor flame conditions but also generate predictions of combustion stability and pollutant concentrations, enabling broader optimization of combustion conditions and making it applicable to various combustion devices.

[0081] Example 2

[0082] This embodiment proposes an intelligent combustion control system based on online fiber optic detection of flame signals, including:

[0083] The data acquisition module uses optical fibers to collect the operating parameters of the burner and the emitted flame light signal. It converts the free radical signal in the flame light signal into an analog electrical signal, and then converts the analog electrical signal into a digital electrical signal. At the same time, it acquires the flame temperature and actual pollutant concentration output by the burner.

[0084] The data processing module is used to extract the time-domain and frequency-domain features of digital electrical signals, and simultaneously calculate the ratio of free radical signals in the digital electrical signals. It generates a multi-dimensional feature dataset based on operating parameters, time-domain features, frequency-domain features, ratios, and flame temperature, and establishes a regular relationship with flame stability.

[0085] The data analysis module is used to use the feature dataset as the input variable of the neural network, and the neural network uses the actual pollutant concentration generated during the combustion process as the output variable to predict the pollutant concentration under different combustion conditions.

[0086] The fuzzy PID control module is used to perform error analysis on the predicted pollutant concentration within a preset threshold range, and to fuzzify the feedback signal by combining it with a fuzzy rule base;

[0087] The execution module uses defuzzified feedback signals to dynamically adjust the fuel and air flow of the burner.

[0088] In this embodiment, the specific functions implemented by each module can be found in steps 1-5 of Embodiment 1, and will not be repeated here.

[0089] Example 3

[0090] This embodiment further proposes an intelligent combustion control system based on online fiber optic detection of flame signals, such as... Figure 2 As shown, the device includes a burner 1, with a filter 2 and an analyzer 13 mounted at the nozzle end of the burner 1. The analyzer 13 consists of a temperature sensor and a gas analyzer. The filter 2 has, in sequence, a lens 3, an optical fiber 4, a photoelectric sensor 5, an amplifier 6, a data acquisition unit 8, a data processing module 9, a control module 10, an execution module 11, and a mass flow meter 12. The output of the mass flow meter 12 is connected to the input of the burner 1. The amplifier 6 is powered by an AD-DC linear low-ripple power supply 7 with an output voltage of ±5V. The current signal in the flame signal is amplified by a two-stage amplifier 6: the first stage is a transimpedance amplifier, converting the current into voltage and performing initial amplification; the second stage is a proportional amplifier, further amplifying the smaller voltage signal. The formula is:

[0091] V out =-I·R f

[0092] Where I is the current signal of the photoelectric sensor, and R f This is the feedback resistor.

[0093] The amplified analog electrical signal is collected and saved by the data acquisition unit 8. The sampling frequency (e.g., 200kHz) and sampling time (e.g., 1s) are determined, and the reading length (e.g., 2000) is used. The signal is then converted into a digital electrical signal by the AD converter in the data acquisition unit and transmitted to the data processing module 9. At the same time, the flame temperature and the actual pollutant concentration are collected by the temperature sensor and the gas analyzer.

[0094] The data processing module 9 preprocesses the acquired signal and performs signal processing according to step 2 of embodiment 1. The processed feature dataset is input to the control module 10. The control module 10 outputs a fuzzy feedback signal. The feedback signal passes through the execution module 11, where the DA converter reconverts the feedback signal into an analog signal and transmits it to the STM32 microcontroller. The STM32 adjusts the fuel and air supply according to the control signal, and the mass flow meter 12 controls the fuel and air flow to achieve closed-loop regulation of the combustion process.

[0095] The above embodiments are merely illustrative of the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All equivalent changes or modifications made based on the essence of the content of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A smart combustion control method based on online fiber optic detection of flame signals, characterized in that, Includes the following steps: The optical fiber acquires the operating parameters of the burner and the emitted flame light signal, converts the free radical signal in the flame light signal into an analog electrical signal, and then converts the analog electrical signal into a digital electrical signal, while simultaneously acquiring the flame temperature and actual pollutant concentration output by the burner. The time-domain and frequency-domain features of the digital electrical signal are extracted, and the ratio of free radical signals in the digital electrical signal is calculated. A multidimensional feature dataset is generated using the operating parameters, the time-domain features, the frequency-domain features, the ratio, and the flame temperature, and a regular relationship is established with the flame stability. The feature dataset is used as the input variable of the neural network, and the actual pollutant concentration generated during combustion is used as the output variable to predict the pollutant concentration under different combustion conditions. Error analysis is performed on the predicted pollutant concentration within a preset threshold range, and the feedback signal is fuzzified using a fuzzy rule base. The defuzzified feedback signal is used to dynamically adjust the fuel and air flow of the burner.

2. The intelligent combustion control method based on online fiber optic detection of flame signals as described in claim 1, characterized in that, The time-domain features include the mean, root mean variance, and peak amplitude, while the frequency-domain features include the dominant frequency component and the energy spectrum.

3. The intelligent combustion control method based on online fiber optic detection of flame signals as described in claim 1, characterized in that, The operating parameters include fuel ratio, equivalence ratio, and combustion power.

4. The intelligent combustion control method based on online fiber optic detection of flame signals as described in claim 1, characterized in that, After the feature dataset is generated, the digital electrical signal is denoised based on moving average filtering, and the mean of the denoised feature dataset is normalized according to preset conditions.

5. The intelligent combustion control method based on online fiber optic detection of flame signals as described in claim 1, characterized in that, The fuzzy control rule base determines the control strategy based on empirical rules. When the predicted value of the pollutant concentration exceeds the standard, the system adjusts the flow rate of fuel and air according to the control strategy.

6. The intelligent combustion control method based on online fiber optic detection of flame signals as described in claim 1, characterized in that, The method for calculating the predicted pollutant concentration is as follows: in, ω represents the predicted pollutant concentration. jk h represents the weights from the hidden layer to the output layer of the neural network. j Forward propagation formula, b k For bias.

7. The intelligent combustion control method based on online fiber optic detection of flame signals as described in claim 1, characterized in that, The error analysis includes: comparing the predicted pollutant concentration with the actual pollutant concentration within the preset threshold range to obtain an error signal and an error rate of change; converting the error signal and the error rate of change into fuzzy linguistic variables using a fuzzy membership function; and defuzzifying the feedback signal based on the fuzzy linguistic variables.

8. The intelligent combustion control method based on online fiber optic detection of flame signals as described in claim 7, characterized in that, The fuzzy membership function is: Where e is the error signal, and a and b are the thresholds of the membership function.

9. The intelligent combustion control method based on online fiber optic detection of flame signals as described in claim 8, characterized in that, The feedback signal after defuzzification is: Where e(t) is the error signal, K p K i K d These are the proportional, integral, and derivative coefficients of the PID controller after adjustment using the fuzzy rule base.

10. An intelligent combustion control system based on online fiber optic detection of flame signals, characterized in that, include: The data acquisition module uses optical fiber to collect the operating parameters of the burner and the emitted flame light signal, converts the free radical signal in the flame light signal into an analog electrical signal, and then converts the analog electrical signal into a digital electrical signal, while simultaneously acquiring the flame temperature and actual pollutant concentration output by the burner. The data processing module is used to extract the time-domain and frequency-domain features of the digital electrical signal, and simultaneously calculate the ratio of free radical signals in the digital electrical signal. It generates a multi-dimensional feature dataset based on the operating parameters, the time-domain features, the frequency-domain features, the ratio, and the flame temperature, and establishes a regular relationship with the flame stability. The data analysis module is used to use the feature dataset as the input variable of the neural network, and the neural network uses the actual pollutant concentration generated during combustion as the output variable to predict the pollutant concentration under different combustion conditions. The fuzzy PID control module is used to perform error analysis on the predicted value of pollutant concentration within a preset threshold range, and to fuzzify the feedback signal in combination with a fuzzy rule base. The execution module uses the defuzzified feedback signal to dynamically adjust the fuel and air flow of the burner.

Images