A waste incinerator temperature monitoring control method

Abstract

The application discloses a garbage incinerator temperature monitoring control method, relates to the technical field of electric digital data processing, and aims to solve the problem that the prior art cannot adapt to array hardware and is difficult to realize high-precision dynamic temperature control. Specifically, array-form temperature monitoring data is collected, and two-dimensional Fourier analysis is performed to obtain a frequency spectrum array; a Prewitt operator convolution is used to extract temperature field characteristics; a three-full-connection-layer neural network is activated by using the convolution result to dynamically assign weights, so as to identify a temperature state; a hidden Markov model probability classification is performed; a control mode is adaptively adjusted according to a classification result, an operating condition is matched according to an integer identifier, a control value is stepwise adjusted, and the control value is cached; and an output result is output and cyclically monitored. The application realizes global temperature accurate perception and control adaptation, guarantees that the furnace temperature is stably in an environmental protection interval, and improves incineration environmental protection and equipment intelligence.

Description

Technical Field

[0001] This invention relates to the field of electronic digital data processing technology, and in particular to a temperature monitoring and adaptive control method for waste incinerators that meet the requirements of rapid heating and precise temperature control. Background Technology

[0002] In the field of waste treatment and resource recycling, the technology of converting municipal waste or sludge into sintered bricks through high-temperature incineration has been put into practical use. The core environmental requirement of this technology lies in strictly controlling temperature changes within the incinerator: to avoid the generation of large amounts of dioxins during the heating process, a "rapid heating" mode is required. Specifically, the furnace temperature must be below 300℃ during the feeding and discharging stages to ensure operational safety, and after incineration starts, the furnace temperature must be rapidly increased to above 950℃. This dynamic temperature requirement of "low-temperature standby - rapid heating - high-temperature incineration" places stringent demands on the accuracy of temperature monitoring and the speed of control response in the incinerator.

[0003] To meet the aforementioned temperature control objectives, existing technologies have undergone hardware improvements: on the one hand, by employing arrayed flame nozzles, the overall heating rate within the furnace is increased through multi-nozzle coordinated heating to satisfy the core requirement of "rapid temperature attainment"; on the other hand, to achieve more comprehensive temperature monitoring, arrayed thermometers are introduced to replace traditional single-point monitoring devices (such as thermocouples and infrared point thermometers). Theoretically, arrayed thermometers can collect spatial distribution data of the temperature field within the furnace, providing a richer data foundation for subsequent precise temperature control. Compared to single-point monitoring methods that can only acquire localized temperatures, this hardware configuration represents a revolutionary change.

[0004] However, existing technologies have gaps in key areas, which prevent the full utilization of the hardware advantages of "array-type flame nozzle + array-type thermometer," making it difficult to achieve the expected high-precision and high-adaptability temperature control. 1. Lack of effective processing solutions for array-type temperature data: The data collected by array thermometers contains the spatial distribution characteristics of furnace temperature (such as local temperature gradients and regional hot spot differences), but existing technologies only treat it as independent single-point temperature data and simply summarize it without mining the overall temperature field state information hidden in the data. This cannot support the fine control needs such as "judging whether the heating is uniform" and "identifying local temperature anomalies", which is equivalent to wasting the core value of array monitoring. 2. Control methods without compatible array hardware: Existing control strategies still rely on static parameter modes adapted to "single-point temperature measurement" (such as fixed fuel supply and preset heating rate), which cannot adjust the control logic based on dynamic changes in the temperature field reflected by the array monitoring data. For example, when the array data shows that "the heating rate of a certain area in the furnace lags behind the overall rate" or "there is a risk of local temperature overshoot," the existing control methods cannot specifically adjust the operating parameters of the flame nozzles in the corresponding area. They can only respond by adjusting the overall parameters, which can easily lead to local temperature failures or overall temperature fluctuations, failing to meet the dual requirements of "rapid heating + stable temperature control" for waste incinerators.

[0005] In summary, while existing technologies have provided the basic conditions for precise temperature control of waste incinerators through hardware upgrades, the lack of a complete chain of technical solutions for array temperature data processing and dynamic control adaptation still prevents the achievement of the expected temperature control effect. There is an urgent need for a temperature monitoring and control method that can fully utilize the characteristics of array monitoring and adapt to dynamic temperature control requirements. Summary of the Invention

[0006] In view of this, the present invention provides a method for temperature monitoring and control of a waste incinerator. This method first performs a two-dimensional Fourier spectrum transformation on the array temperature data, then uses the Prewitt operator to extract convolutional features, and then achieves dynamic weighted activation based on a three-fully connected layer neural network. Finally, it combines a hidden Markov model to complete probabilistic classification, thus forming a deep integration technical solution of spatial temperature field analysis, dynamic weighted neural network and probabilistic decision control. This solves the problems in the prior art where the hardware configuration of array flame nozzles and array thermometers lacks an adapted control scheme and it is difficult to achieve high-precision monitoring and dynamic control using array temperature data. This method significantly improves the temperature monitoring accuracy of the waste incinerator, enables adaptive adjustment of the control mode, and provides stable adaptation to all operating conditions, including low-temperature standby, rapid heating, and high-temperature incineration. This ensures that the furnace temperature can quickly reach above 950°C during incineration to suppress dioxin formation, while avoiding equipment damage and environmental risks caused by temperature fluctuations.

[0007] The specific technical solution is a method for monitoring and controlling the temperature of a waste incinerator, including the following steps: S1. Array Data Acquisition: Acquire array-formatted temperature monitoring data from the temperature monitoring module of the waste incinerator to provide basic data for subsequent temperature field analysis; S2. Temperature field feature calculation: The temperature monitoring data is convolved using a preset convolution kernel to obtain the calculation results characterizing the spatial distribution features of the temperature field. S3. Neural Network Adaptation Activation: Based on the calculation results, the preset neural network model is weighted and activated to form a computational model that adapts to the current temperature field characteristics. S4. Temperature State Recognition: The temperature monitoring data is calculated using a weighted and activated neural network model. The spatial distribution law and gradient change characteristics of the temperature field are integrated, and a quantified temperature field feature expression is output as the recognition result. S5. Probabilistic classification decision: The identification results are classified using a preset probabilistic decision model to determine the corresponding temperature classification result; S6. Dynamic adjustment of control mode: Adaptively adjust the control mode according to the matching result between the classification result and the current control mode to match the classification result. S7. Result Output and Cyclic Monitoring: Output the adjustment result to the preset output terminal and return to step S1 to build a continuous temperature monitoring and control cycle. Furthermore, the array-type temperature monitoring data is a spectral array formed by performing two-dimensional Fourier analysis on the temperature readings of the waste incinerator furnace collected by the array thermometer, which is used to characterize the spatial distribution law of the temperature field.

[0008] Furthermore, the preset convolution kernel mentioned in step S2 adopts the Prewitt operator form, and the specific value of the convolution kernel is preset by the user according to the actual temperature monitoring requirements of the waste incinerator.

[0009] Furthermore, the neural network model is a deep feedforward neural network structure with three fully connected layers, and the specifications of the three fully connected layers in the neural network recognition model are all no more than four times the specifications of the temperature monitoring data in the array form.

[0010] Furthermore, the weighted activation of the neural network model by the convolution result in step S3 is specifically as follows: the convolution result output by S2 is augmented with regularized data without changing the proportion of the temperature field spectral characteristics. The augmented convolution result is adjusted to be completely consistent with the number of neurons in each of the three fully connected layers. Then, the adjusted convolution result is multiplied with each neuron in the corresponding fully connected layer one by one. Through this operation, the weights of the neurons are dynamically adjusted, so that the neural network recognition model can respond to the morphological changes of the temperature field of the waste incinerator in real time, thus completing the weighted activation of the model.

[0011] Furthermore, the probabilistic decision-making model described herein employs a Hidden Markov Model.

[0012] Furthermore, both the classification results and the current control mode correspond to three operating conditions sorted by temperature, and both are identified using a unified integer quantization rule. Specifically, the three operating conditions are: the first is shutdown, corresponding to the equipment not operating when there is no temperature inside the furnace; the second is low-temperature standby, corresponding to the feeding and discharging state when the furnace temperature is below 300°C, used to avoid the generation of dioxins during the feeding and discharging process; and the third is high-temperature incineration, corresponding to the waste incineration state when the furnace temperature is not lower than 950°C, used to suppress the generation of dioxins. The integer quantization rule is: different integer values ​​are sequentially assigned according to the operating temperature from low to high to ensure that the classification results and the values ​​of the current control mode are comparable.

[0013] Furthermore, the method for determining the matching between the classification result and the current control mode, and the adaptive adjustment direction, is as follows: based on the unified integer quantization rule, compare the integer value corresponding to the classification result with the integer value corresponding to the current control mode; if the two values ​​are equal, the classification result is determined to be consistent with the current control mode, and no adjustment is required; if the integer value of the classification result is less than the integer value of the current control mode, the two are determined to be inconsistent, and the control value of the current control mode needs to be increased; if the integer value of the classification result is greater than the integer value of the current control mode, the two are determined to be inconsistent, and the control value of the current control mode needs to be decreased.

[0014] Furthermore, the adjustment rule for the control value of the current control mode is as follows: the increase or decrease each time is a preset ratio of the default value of the control value; the adjusted control value of the current control mode is stored in the cache, and when the current control mode is entered again in the future, the control value corresponding to the current control mode is directly read from the cache without recalculation and adjustment.

[0015] Furthermore, the preset output terminal is a software output interface set by the user, which is respectively connected to the screen display output or the log output.

[0016] The beneficial effects achieved by this invention include: 1. Improve temperature monitoring accuracy: By performing two-dimensional Fourier analysis on temperature monitoring data to obtain a spectrum array, and combining it with the Prewitt operator convolution calculation to locate the temperature gradient change region, the spatial blind zone problem of traditional point temperature measurement is solved, and the accurate perception of the temperature field of the entire incinerator is realized. 2. Enhance the adaptability of neural networks: Dynamically weight the convolution results to activate the three fully connected layers of the neural network, enabling the model to adapt to the current temperature field data characteristics in real time, avoiding the lag in the response of static models to nonlinear working conditions, and improving the accuracy of temperature identification; 3. Ensure control stability: Rely on the Hidden Markov Model to realize probabilistic classification decision-making, combined with step-by-step control value adjustment and buffer reuse, to suppress drastic fluctuations in control parameters, avoid furnace temperature overshoot and oscillation, and ensure that the temperature is stable within the environmental protection range (feed and discharge <300℃, incineration ≥950℃). 4. Fully leverage the value of array devices: Design technical solutions based on the characteristics of array thermometers, characterize temperature distribution patterns through spectral arrays, and combine convolutional and neural network fusion analysis to transform the full-domain monitoring advantages of array devices into monitoring and control efficiency; 5. Balancing accuracy and efficiency: The spectrum array simplifies data dimensions and convolution weighting focuses on key features. While reducing the amount of computation, it relies on probabilistic decision-making to quickly output adjustment instructions, meeting the timeliness requirements of control in conditions such as rapid heating of incinerators. 6. Enhance environmental protection and intelligence: By deeply coupling temperature field analysis and probabilistic control, ensure that the furnace temperature is stable and meets the dioxin control requirements, reducing pollutant generation; the software output interface supports multiple forms of result output, improving the visualization and traceability of the control process and enhancing the intelligence of the equipment. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating at least one embodiment of the present invention. Detailed Implementation

[0018] The technical solution of the present invention is further described below, but the scope of protection is not limited to the content described.

[0019] First Implementation Method The first embodiment of the present invention specifically includes the following steps: S1. Acquire array data: Collect array-format temperature monitoring data from the temperature monitoring module of the waste incinerator; S2. Convolution Calculation: Perform convolution calculation on the above temperature monitoring data using a preset convolution kernel to obtain the corresponding convolution result; S3. Activate the neural network: Use the convolution result to weight and activate the neural network recognition model; S4. Neural Network Calculation: The activated neural network recognition model processes the temperature monitoring data and outputs the recognition results. S5. Classification Calculation: The probabilistic graphical model is used to classify the recognition results and generate classification results; S6. Deviation Calculation: Determine whether the classification result is consistent with the current control mode, and adjust the current control mode according to the result of whether they are consistent. S7. Output the result: Output the above judgment result to a specific output terminal, and then return to step S1 to enter the next round of monitoring and control loop.

[0020] Second Implementation Method The second embodiment of the present invention has roughly the same steps and flow as the first embodiment, the core difference being the limitation of key technical parameters, as follows: Temperature monitoring data processing: In this embodiment, the temperature monitoring data is a spectrum array obtained by performing two-dimensional Fourier analysis on the temperature readings collected by the array thermometer. By converting the array temperature data into a spectrum form, the sensitivity to temperature distribution patterns can be effectively enhanced, solving the "spatial blind spot" problem of traditional point measurement and achieving accurate perception of the temperature distribution across the entire incinerator. Convolution kernel selection: The preset convolution kernel adopts the Prewitt operator form, and the specific value of the operator is preset by the user according to the actual incinerator specifications and monitoring accuracy requirements. With the edge detection logic of the Prewitt operator, gradient change regions in temperature data can be directly identified, significantly improving the accuracy of locating local temperature anomalies; Neural Network Structure and Activation Method: The neural network recognition model adopts a deep feedforward neural network structure with three fully connected layers. The weighted activation in step S3 specifically involves: first, proportionally scaling the value distribution of the convolution result to the specifications of the three fully connected layers in the neural network recognition model; then, multiplying the scaled value distribution with the corresponding fully connected layer neurons term by term to dynamically adjust the model weights. This method allows the spatial feature distribution of the convolution output to be proportionally scaled to each layer of the neural network, enabling the model to respond in real-time to changes in the temperature field within the incinerator and improving its adaptability to nonlinear temperature conditions. Neural network specification constraint: In the neural network recognition model, the specification (number of neurons) of the three fully connected layers should not exceed four times the specification (data dimension) of the array-type temperature monitoring data. This constraint simplifies network complexity, balances computational efficiency while ensuring feature extraction capabilities, and avoids overfitting issues in the model.

[0021] Third Implementation Method The third embodiment of the present invention is basically the same as the first embodiment in terms of steps and flow, mainly supplementing and defining the technical details of steps S5, S6, and S7, as follows: Step S5. Selection of probabilistic graphical model: In this embodiment, the probabilistic graphical model adopts the Hidden Markov Model (HMM). By leveraging the state transition characteristics of the Hidden Markov Model, the classification accuracy of the temperature monitoring data identification results is improved, which is especially suitable for scenarios with dynamic temperature changes in incinerators. Step S6. Control mode adjustment logic: The classification result and the current control mode both correspond to "multiple integer values ​​sorted by temperature". Specifically, both are divided into three operating conditions: "shutdown", "low temperature standby", and "high temperature incineration". The three operating conditions are identified by integers: "shutdown" corresponds to integer 1, "low temperature standby" corresponds to integer 2, and "high temperature incineration" corresponds to integer 3. Adjustment rules: If the integer value of the classification result is less than the integer value of the current control mode, increase the control value of the current control mode; if the integer value of the classification result is greater than the integer value of the current control mode, decrease the control value of the current control mode. Adjustment range: Each time the control value of the current control mode is increased or decreased, the adjustment range is 5% of the default value of the control value. This incremental adjustment can effectively avoid drastic fluctuations in flame nozzle airflow and fuel supply, and suppress furnace temperature overshoot oscillations; if the control adaptation cannot be completed even with a 5% adjustment range, it usually means that an emergency shutdown is required, which is not within the scope of consideration for general control scenarios; Cache reuse: The adjusted control value of the current control mode is stored in the system cache. When the control mode is entered again, the control value is read directly from the cache without recalculation, thus improving the control response speed. Step S7. Specific output terminal configuration: The specific output terminal in step S7 is "a software output interface that can be connected to the screen for display output or log output through user settings". Users can choose to display the judgment results on the screen in a visual form or store them in the form of log files according to their actual needs, so as to facilitate subsequent traceability and operation and maintenance analysis.

[0022] The core technical advantage of this invention lies in using the convolution result as the input weights of the neural network, enabling the model to dynamically adapt to the distribution characteristics of the current temperature data and significantly improving the response speed to nonlinear operating conditions. Simultaneously, the input data uses a spectral array derived from two-dimensional Fourier analysis (rather than the original temperature distribution data), ensuring that the weighting and activation of the neural network recognition model is more targeted. The spectral array represents the data distribution pattern of the array thermometer's temperature readings, easily exhibiting single lines, single points, or local high values, inherently indicating a specific temperature distribution pattern. Convolving this type of data representing patterns and using it for neural network weighting and activation fully utilizes the characteristic that different neurons in the neural network correspond to different characteristic temperature distribution patterns, ensuring the accuracy of the calculation results while significantly reducing the computational load.

[0023] In summary, this invention, through a triple technological innovation of "spectral spatial perception - convolutional weighted dynamic network - hidden Markov model decision chain", breaks through the limitations of traditional single-point temperature monitoring and static control. It deeply couples temperature field morphological analysis with probabilistic control decision-making, achieving significant progress in environmental compliance (avoiding dioxin generation), equipment protection (preventing abnormal temperature damage), and intelligence (automated cyclic control and traceable results) of waste incinerator temperature control.

Claims

1. A method for monitoring and controlling the temperature of a waste incinerator, characterized in that, S1. Array Data Acquisition: Acquire array-formatted temperature monitoring data from the temperature monitoring module of the waste incinerator to provide basic data for subsequent temperature field analysis; S2. Temperature field feature calculation: The temperature monitoring data is convolved using a preset convolution kernel to obtain the calculation results characterizing the spatial distribution features of the temperature field. S3. Neural Network Adaptation Activation: Based on the calculation results, the preset neural network model is weighted and activated to form a computational model that adapts to the current temperature field characteristics. S4. Temperature State Recognition: The temperature monitoring data is calculated using a weighted and activated neural network model. The spatial distribution law and gradient change characteristics of the temperature field are integrated, and a quantified temperature field feature expression is output as the recognition result. S5. Probabilistic classification decision: The identification results are classified using a preset probabilistic decision model to determine the corresponding temperature classification result; S6. Dynamic adjustment of control mode: Adaptively adjust the control mode according to the matching result between the classification result and the current control mode to match the classification result. S7. Result Output and Cyclic Monitoring: Output the adjustment result to the preset output terminal and return to S1 to build a continuous temperature monitoring and control cycle.

2. The method as described in claim 1, characterized in that, The array-type temperature monitoring data is a spectrum array formed by performing two-dimensional Fourier analysis on the temperature readings of the waste incinerator furnace collected by the array thermometers, which is used to characterize the spatial distribution law of the temperature field.

3. The method as described in claim 1, characterized in that, The preset convolution kernel mentioned in step S2 adopts the Prewitt operator form, and the specific value of the convolution kernel is preset by the user according to the actual temperature monitoring requirements of the waste incinerator.

4. The method as described in claim 1, characterized in that, The neural network model is a deep feedforward neural network structure with three fully connected layers, and the size of each of the three fully connected layers in the neural network recognition model is no more than four times the size of the temperature monitoring data in the array form.

5. The method as described in claim 1, characterized in that, In step S3, the convolution result is used to weight and activate the neural network model. Specifically, the convolution result output from S2 is augmented with regularized data without changing the proportion of the temperature field spectral characteristics. The augmented convolution result is then adjusted to be exactly the same as the number of neurons in each of the three fully connected layers. The adjusted convolution result is then multiplied one by one with the neurons in the corresponding fully connected layer. This operation enables the dynamic adjustment of the neuron weights, allowing the neural network to recognize the model in real time in response to the morphological changes of the temperature field in the waste incinerator, thus completing the weighting and activation of the model.

6. The method as described in claim 1, characterized in that, The probabilistic decision model described above employs a hidden Markov model.

7. The method as described in claim 1, characterized in that, Both the classification results and the current control mode correspond to three operating conditions sorted by temperature, and both are identified using a unified integer quantization rule. Specifically, the three operating conditions are: first, shutdown, corresponding to equipment not operating when there is no temperature inside the furnace; second, low-temperature standby, corresponding to the feeding and discharging state when the furnace temperature is below 300℃, used to avoid dioxin generation during feeding and discharging; and third, high-temperature incineration, corresponding to the waste incineration state when the furnace temperature is not lower than 950℃, used to suppress dioxin generation. The integer quantization rule is: different integer values ​​are sequentially assigned according to the operating temperature from low to high to ensure the comparability of the classification results and the values ​​of the current control mode.

8. The method as described in claim 7, characterized in that, The method for determining the matching between the classification result and the current control mode and the adaptive adjustment direction is as follows: Based on the unified integer quantization rule, compare the integer value corresponding to the classification result with the integer value corresponding to the current control mode; if the two values ​​are equal, it is determined that the classification result is consistent with the current control mode and no adjustment is needed; if the integer value of the classification result is less than the integer value of the current control mode, it is determined that the two are inconsistent and the control value of the current control mode needs to be increased. If the integer value of the classification result is greater than the integer value of the current control mode, it is determined that the two are inconsistent, and the control value of the current control mode needs to be reduced.

9. The method as described in claim 8, characterized in that, The adjustment rule for the control value of the current control mode is as follows: the increase or decrease each time is a preset ratio of the default value of the control value; the adjusted control value of the current control mode is stored in the cache, and when the current control mode is entered again, the control value corresponding to the current control mode is directly read from the cache without recalculation and adjustment.

10. The method according to any one of claims 1-9, characterized in that, The preset output terminal is a software output interface set by the user, which is respectively connected to the screen display output or the log output.

Images