A boiler combustion automatic detection control system based on a big data model

Abstract

The present application relates to the technical field of boiler combustion control, in particular to a boiler combustion automatic detection control system based on a big data model. The system comprises a data acquisition module for acquiring various parameter data during the boiler combustion process; a data processing module for preprocessing and fusing the acquired data; a big data model module for constructing a model based on the processed data, learning the combustion process rules and making predictions; an intelligent control module for generating control instructions according to the model prediction results to optimize the control of the boiler combustion equipment; a monitoring and alarm module for real-time monitoring of the combustion state and equipment operation, and for issuing an alarm and taking emergency measures when an anomaly occurs. The present application overcomes the limitations of traditional boiler combustion control methods, realizes comprehensive and accurate monitoring of the combustion process, intelligent and efficient control, early fault prediction and handling, effectively improves the combustion efficiency, reduces energy consumption and pollutant emissions, enhances the safety and reliability of the boiler operation, and is suitable for various industrial boiler application scenarios.

Description

Technical Field

[0001] This invention relates to the field of boiler combustion control technology, specifically to an automatic detection and control system for boiler combustion based on a big data model. Background Technology

[0002] In industrial production and energy supply, boilers are critical heat conversion devices, and their combustion efficiency, stability, and safety are of paramount importance. Traditional boiler combustion detection and control methods have significant limitations. Monitoring relies on a limited number of basic sensors, which cannot comprehensively and accurately reflect the combustion status; control depends on experience-set fixed parameters, making it difficult to adapt to factors such as fuel quality fluctuations and load changes, and its fault prediction and handling capabilities are also insufficient. Although there have been attempts to improve the system by increasing the number of sensors and adopting simple automated control algorithms, these lack effective data analysis and adaptability to complex systems, failing to meet the high demands of modern industry for boiler operation.

[0003] In recent years, the application potential of big data technology in the industrial field has become increasingly apparent. It can analyze massive amounts of data to uncover patterns and provide decision support. Against this backdrop, applying big data models to the field of boiler combustion detection and control is expected to overcome the shortcomings of traditional methods, achieve more comprehensive, accurate, and intelligent monitoring and control, improve combustion efficiency, reduce energy consumption, reduce emissions, and enhance safety. Summary of the Invention

[0004] An automatic detection and control system for boiler combustion based on a big data model, characterized in that it includes:

[0005] The data acquisition module is used to collect various parameter data during the boiler combustion process;

[0006] The data processing module preprocesses and merges the collected data;

[0007] The big data model module builds models based on processed data to learn the patterns of the combustion process and make predictions.

[0008] The intelligent control module generates control commands based on model prediction results to optimize the control of the boiler combustion equipment;

[0009] The monitoring and alarm module monitors the combustion status and equipment operation in real time, and issues an alarm and takes emergency measures when an abnormality occurs.

[0010] Preferably, the data acquisition module includes various types of sensors, specifically including temperature sensors, pressure sensors, gas composition sensors, and flame monitoring sensors;

[0011] The temperature sensors are installed at different locations inside the boiler furnace to monitor the temperature distribution of the combustion flame in real time, which helps to understand the intensity of combustion and whether the heat release is uniform. At the same time, temperature sensors are also installed at key parts of the flue to monitor the exhaust temperature, thereby indirectly judging the combustion efficiency and whether there is any abnormal heat loss.

[0012] The pressure sensors are installed at the air inlets and outlets of the air duct, the furnace, and the flue. The pressure sensor at the air inlet monitors the initial air pressure of primary and secondary air to ensure that the air supplied to the furnace has a suitable pressure, guaranteeing that the air can participate in the combustion process evenly and in sufficient quantity. The pressure sensor in the furnace is used to monitor the pressure changes in the furnace in real time, because the pressure in the furnace should be maintained within a relatively stable and reasonable range during normal combustion. Abnormal pressure fluctuations may indicate unstable combustion conditions, such as deflagration or poor ventilation. The pressure sensor at the flue monitors the negative pressure in the flue, which can help determine the operation status of the entire ventilation system and whether the smoke exhaust is smooth.

[0013] The gas composition sensor is installed inside the flue to detect the concentrations of oxygen, carbon monoxide, carbon dioxide, and nitrogen oxides. Oxygen concentration reflects the suitability of the air supply during combustion. Excessive oxygen indicates excess air, which carries away a large amount of heat and reduces combustion efficiency; insufficient oxygen may lead to incomplete combustion. Carbon monoxide concentration is a key indicator of complete combustion; excessive carbon monoxide indicates incomplete combustion, requiring timely adjustment of combustion parameters. Changes in carbon dioxide concentration also reflect combustion effectiveness, and combined with analysis of other gas components, can more accurately determine the chemical changes during combustion. Monitoring nitrogen oxide concentration is crucial for controlling pollutant emissions. Real-time monitoring allows for timely adjustments to combustion strategies to reduce pollutant formation.

[0014] The acoustic wave measurement sensor is installed inside the furnace. The acoustic wave emitting device periodically emits acoustic wave signals, and the acoustic wave receiving device receives the acoustic wave signals passing through the furnace. By measuring the propagation time of the acoustic waves on different paths, the temperature field distribution inside the furnace is calculated. The combustion state of the flame inside the furnace is determined based on the temperature field distribution. If the overall average temperature of the temperature field is high and within a reasonable range, it indicates that the combustion reaction is intense and the fuel releases more heat, which means that the combustion state is good. If the temperature distribution range is narrow, it indicates that the temperature inside the furnace is relatively uniform and the combustion state is stable.

[0015] Preferably, the preprocessing and fusion of the collected data specifically includes:

[0016] Data format standardization: The system performs unified format standardization processing on the collected data, converting it into a unified data format;

[0017] Feature extraction: Extract feature information from the collected raw data, including temperature change rate, pressure fluctuation amplitude, gas component concentration change trend, flame stability and flame combustion state;

[0018] Data fusion: The weighted average method is used to fuse multiple temperature sensor data to obtain temperature data for the region; the Kalman filter method is used to fuse pressure sensor data for a region, and then the weighted average method is used to fuse multiple pressure sensor data for the region to obtain pressure data for the region; the evidence-based fusion algorithm is used to fuse evidence about the combustion state provided by sensors in different regions.

[0019] Preferably, the step of fusing multiple temperature sensor data using a weighted average method specifically includes:

[0020] Acquire data from multiple temperature sensors and calculate the fused temperature data using the following formula:

[0021]

[0022] In the formula, T is the temperature after fusion, and t i For the temperature data of the i-th sensor, w i Let be the weight of the i-th sensor, and n be the number of sensors.

[0023] Preferably, the step of fusing multiple pressure sensor data using a weighted average method specifically includes:

[0024] After obtaining the pressure sensor data processed using Kalman filtering, the pressure data from multiple pressure sensors is calculated using the following formula:

[0025]

[0026] In the formula, P is the temperature after fusion, p i For the pressure data of the i-th sensor, w i Let be the weight of the i-th sensor, and n be the number of sensors.

[0027] Preferably, the method of fusing evidence about the combustion state from different sensors using a fusion algorithm based on evidence theory specifically includes:

[0028] Determine the identification frame: The identification frame is the set of all possible combustion states. In a boiler, the combustion states include "normal combustion", "incomplete combustion" and "deflagration". Therefore, the identification frame φ = {normal combustion, incomplete combustion, deflagration}.

[0029] Define the basic probability assignment function: For different types of sensors within each region, assign a basic probability value to each subset in the identification frame based on their measurement data and related knowledge. The basic probability assignment function m satisfies the following conditions:

[0030] m() = 0, where m represents the empty set;

[0031] That is, the sum of the basic probabilities of all subsets is 1;

[0032] The DS synthesis rule is used to fuse the basic probability assignment functions of similar sensors in different regions. For two basic probability assignment functions m1 and m2, their synthesis results are as follows: The definition is as follows:

[0033] for and

[0034] Where K = ∑ B∩C＝ m1(B)m2(C) are the conflict coefficients, representing the degree of conflict between two pieces of evidence;

[0035] Based on the fused basic probability allocation function, the subset with the largest basic probability value is selected as the final combustion state judgment result.

[0036] Preferably, the step of building a model based on the processed data specifically includes:

[0037] Split the dataset: Divide the preprocessed data into a training set, a validation set, and a test set;

[0038] Model training: Train the selected model using the training set and adjust the model's parameters to minimize the loss function;

[0039] Model evaluation: The trained model is evaluated using a validation set, and the model with the best performance is selected.

[0040] Preferably, the intelligent control module uses intelligent control algorithms such as model predictive control and fuzzy control to generate control commands.

[0041] Preferably, the monitoring and alarm module visualizes key parameters and control results, and issues alarm signals in a timely manner in abnormal situations.

[0042] Compared with the prior art, the advantages of this invention are:

[0043] Comprehensive and accurate monitoring: Through the multi-parameter sensor monitoring module, various key parameter information of the boiler combustion process can be obtained in real time and comprehensively, overcoming the limitations of traditional monitoring methods and providing rich data support for accurately judging the combustion status;

[0044] Intelligent and efficient control: Based on the analysis and prediction results of big data models, the automatic control and adjustment module can adjust the boiler's combustion parameters in real time and intelligently to achieve the optimal air-fuel ratio, improve combustion efficiency, and reduce energy consumption and pollutant emissions.

[0045] Powerful fault prediction and handling capabilities: Big data models can perform real-time analysis and prediction of boiler operating status, detect potential fault hazards in advance, and take corresponding measures in a timely manner to effectively prevent the occurrence and escalation of faults, thereby enhancing the safety and reliability of boiler operation. Detailed Implementation

[0046] Example: Application of biomass boilers in paper mills

[0047] Scene description:

[0048] A paper mill mainly relies on biomass boilers to provide the steam heat energy required for the paper production process. Biomass fuels (such as sawdust and straw) have complex and variable characteristics, with significant differences in moisture content, particle size, and other factors. This leads to unstable combustion processes in the boilers, frequently resulting in incomplete combustion, low thermal efficiency, and unplanned shutdowns due to coking problems. This not only affects the continuity of paper production but also increases energy and equipment maintenance costs.

[0049] System setup:

[0050] Sensor installation:

[0051] Inside the furnace of the biomass boiler, 18 high-temperature resistant thermocouple temperature sensors are installed in a grid pattern to accurately monitor the combustion temperature in different areas of the furnace. This facilitates the timely detection of local high or low temperature anomalies. These sensors cover the area from the center of the flame to near the furnace wall, and can comprehensively reflect the temperature distribution of the combustion flame.

[0052] Eight pressure sensors were installed at the inlets of the primary and secondary air ducts and the furnace ventilation openings to accurately measure key pressure parameters such as inlet air pressure and furnace air pressure, thereby determining whether the ventilation system is operating normally and ensuring that air can participate in the combustion process evenly and in sufficient quantity.

[0053] Five sets of gas composition sensors were installed in the flue. Each set of sensors can simultaneously detect the concentration of multiple gases such as oxygen, carbon monoxide, carbon dioxide, sulfur dioxide, and nitrogen oxides. Through real-time analysis of the flue gas composition, the degree of combustion completeness and pollutant emissions can be accurately determined.

[0054] Nine acoustic temperature field measurement sensors are arranged in a 3×3 grid on the furnace wall, forming an acoustic transmitting and receiving device. The acoustic transmitting device transmits a coded modulated acoustic signal with a frequency of 3kHz, and the acoustic receiving device receives the signal and measures the sound wave propagation time. The acoustic algorithm is used to calculate the temperature field distribution inside the furnace, providing more accurate information for combustion control.

[0055] Four flame monitoring sensors were installed near the furnace observation port to monitor the flame's brightness, flicker frequency, color, and other characteristics from different angles, in order to determine the stability of the flame and whether the combustion state is normal.

[0056] Data acquisition and transmission:

[0057] By employing a high-precision timed acquisition mechanism, data from each sensor is collected synchronously every 4 seconds to ensure that the acquired different parameter data strictly correspond in time, laying the foundation for accurate subsequent analysis.

[0058] Considering the on-site environment of the paper mill workshop, shielded twisted-pair cables are used for wired transmission of sensors that are close to the control room and whose wiring is relatively convenient; while for some sensors installed in higher positions and whose wiring is difficult, industrial wireless communication modules are used for data transmission to ensure that all data can be efficiently and stably aggregated to the data processing center.

[0059] Big Data Storage and Management:

[0060] A dedicated big data storage server was built, employing a combination of relational and non-relational databases to store massive amounts of boiler operation data. The relational database is used to store structured sensor parameters, equipment operating status, and other information, facilitating complex queries and correlation analysis; the non-relational database is used to store semi-structured and unstructured data such as historical fault records and real-time image data.

[0061] A comprehensive data backup strategy has been established, with full backups of all data performed daily and incremental backups performed hourly. An off-site disaster recovery center has also been set up to ensure data security and integrity and prevent data loss due to unforeseen circumstances.

[0062] Model building and analysis:

[0063] Data preprocessing:

[0064] A designated person is assigned to conduct a preliminary screening of the collected data daily, eliminating obvious abnormal data points caused by temporary sensor malfunctions, communication interference, or other reasons.

[0065] Data cleaning algorithms are used to perform deep cleaning on the remaining data, removing duplicate and redundant data records. At the same time, for parts with missing values, appropriate interpolation algorithms are used to fill in the missing values ​​based on the time series characteristics of the data and the surrounding normal data.

[0066] To address the differences in the range and magnitude of data collected by different sensors, normalization processing is performed, mapping all data to the [0,1] interval, so that each parameter data has equal weight and comparability in subsequent analysis and model training.

[0067] Feature engineering:

[0068] From the preprocessed massive data, key feature information such as the standard deviation of temperature, the rate of change of pressure, the ratio of the concentrations of each gas component, the characteristics of the temperature field, and the fluctuation amplitude of flame brightness are extracted.

[0069] The data analysis algorithm is used to perform correlation analysis on the feature information to find the correlation between temperature, pressure, gas composition concentration, flame stability and flame combustion state, and multiple data fusion algorithms are used to comprehensively process multi-parameter data.

[0070] Model training:

[0071] Considering the time-series characteristics and multi-factor nonlinear coupling inherent in the biomass boiler combustion process, a hybrid model architecture combining a Long Short-Term Memory (LSTM) network and a Convolutional Neural Network (CNN) based on deep learning was selected. First, the convolutional layers of the CNN automatically extract local feature patterns from the data. Then, LSTM layers are used to model the long-term dependencies of the features in the time series.

[0072] The feature-engineered data was divided into training, validation, and test sets in a 7:2:1 ratio. During training, an adaptive learning rate adjustment algorithm was used, combined with early stopping to prevent overfitting. Through multiple rounds of iterative training, the model's weight parameters were continuously adjusted to minimize the loss function on the validation set, ensuring the model has good generalization ability.

[0073] Automatic control and regulation:

[0074] Real-time monitoring and analysis:

[0075] The system receives data from various sensors in real time and inputs it into a trained hybrid model. The model quickly outputs the current combustion status assessment results, including the specific value of combustion efficiency, whether there is a risk of coking, and the predicted values ​​of emission levels of various pollutants. At the same time, it determines whether the current combustion status belongs to normal combustion, slightly incomplete combustion, coking tendency, or other abnormal status categories.

[0076] For example, based on real-time input data such as temperature and gas composition, the model can determine that a certain area in the furnace may experience localized low temperatures and incomplete combustion due to the accumulation of biomass fuel, and predicts that if adjustments are not made in time, there is a risk of coking.

[0077] Control strategy generation and execution:

[0078] Based on the analysis results of the model, the automatic control and adjustment module quickly generates corresponding control commands. Addressing the issues of incomplete combustion and a tendency to coke, the module automatically fine-tunes the primary and secondary air volumes to increase air turbulence, allowing for more thorough mixing of fuel and air. Simultaneously, it appropriately reduces the feed rate of biomass fuel to prevent excessive fuel accumulation and optimize the combustion process.

[0079] If the model predicts that sulfur dioxide emissions will soon exceed the standard, the system will automatically adjust parameters such as combustion temperature and excess air coefficient to minimize pollutant generation while ensuring combustion efficiency, thus keeping emission levels within compliance ranges. Furthermore, when a potential serious malfunction is detected, the system immediately triggers an emergency shutdown procedure and sends detailed fault alarm information to the operators to ensure boiler safety.

[0080] Implementation results:

[0081] By applying this big data model-based automatic combustion detection and control system to biomass boilers in paper mills, after a period of operation and observation, the average combustion efficiency of the boilers has increased by about 12%, the number of unplanned shutdowns due to coking problems has decreased by 40%, and the emissions of major pollutants such as sulfur dioxide and nitrogen oxides have decreased by about 18%. This has effectively ensured the stable and efficient operation of paper production, while reducing energy consumption and environmental pressure, bringing significant economic and environmental benefits to the enterprise.

[0082] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0083] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A boiler combustion automatic detection control system based on a big data model, characterized in that, The application relates to a boiler combustion state intelligent monitoring and control system. The system comprises: a data acquisition module for acquiring various parameter data in the boiler combustion process; a data processing module for pre-processing and fusing the acquired data; the pre-processing and fusing of the acquired data specifically comprises: data format standardization: uniformly standardizing the acquired data and converting the data into a unified data format; feature extraction: extracting feature information from the data after standardization, including temperature data, pressure data and flame combustion state; data fusion: using a weighted average method to fuse multiple temperature sensor data in the same region to obtain temperature data in the region; using a Kalman filtering method to fuse pressure sensor data in a region, and then using a weighted average method to fuse multiple pressure sensor data in the region to obtain pressure data in the region; using a fusion algorithm based on evidence theory to fuse the evidence provided by sensors in different regions about the combustion state; the use of the weighted average method to fuse multiple temperature sensor data specifically comprises: wherein, is the temperature after fusion, is the temperature data of the th sensor, is the weight of the th sensor, is the number of sensors; acquiring multiple temperature sensor data and calculating fused temperature data through the following formula: the use of the weighted average method to fuse multiple pressure sensor data specifically comprises: P is the temperature after fusion, is the pressure data of the first sensor, is the weight of the first sensor, is the number of sensors; acquiring pressure sensor data after Kalman filtering processing and calculating pressure data of multiple pressure sensors through the following formula: Determining a recognition framework: The recognition framework is a set of all possible combustion states, including "normal combustion", "incomplete combustion", "deflagration" in the boiler, then the recognition framework ; Definition of basic probability assignment function: for each kind of sensor in different regions, a basic probability value is assigned to each subset in the identification framework according to its measurement data and related knowledge, and the basic probability assignment function m satisfies the following conditions: Wherein Indicates an empty set; i.e. the sum of the basic probabilities of all subsets is 1 ; The basic probability assignment functions of the same kind of sensors in different areas are fused using D-S combination rule. For two basic probability assignment functions and their combination result is defined as follows: For and , ; where K = 1 - (1 - 1 / n)1 / n is a conflict coefficient, representing the degree of conflict between two evidences; the use of the fusion algorithm based on evidence theory to fuse the evidence provided by different sensors about the combustion state specifically comprises: selecting a subset with the largest basic probability value as the final combustion state judgment result according to the fused basic probability distribution function; a big data model module for constructing a model based on the processed data, learning the combustion process rule and making a prediction; an intelligent control module for generating a control instruction according to the model prediction result and optimizing the control of the boiler combustion equipment; 2. The boiler combustion automatic detection control system based on big data model according to claim 1, characterized in that, a monitoring and alarm module for monitoring the combustion state and equipment operation in real time, issuing an alarm and taking emergency measures when an abnormality occurs. The data acquisition module comprises various types of sensors, specifically including temperature sensors, pressure sensors, gas component sensors and acoustic wave sensors; the temperature sensors are installed at different positions inside the furnace of the boiler to monitor the temperature distribution of the combustion flame in the furnace in real time; meanwhile, temperature sensors are installed at key positions of the flue to monitor the exhaust gas temperature; the pressure sensors are installed at the air inlet, air outlet, furnace and flue of the air duct; the pressure sensors at the air inlet are used for monitoring the initial air pressure of the primary air and the secondary air; the pressure sensors in the furnace are used for monitoring the pressure change in the furnace in real time; the pressure sensors at the flue are used for monitoring the negative pressure in the flue; the gas component sensors are installed in the flue to detect the gas component concentration of oxygen, carbon monoxide, carbon dioxide and nitrogen oxides; the acoustic wave sensors are installed in the furnace; an acoustic wave emitting device periodically emits acoustic wave signals, and an acoustic wave receiving device receives the acoustic wave signals passing through the furnace; the temperature field distribution in the furnace is calculated by measuring the propagation time of the acoustic wave in different paths, and the flame combustion state in the furnace is determined according to the temperature field distribution.

3. The boiler combustion automatic detection control system based on big data model according to claim 1, characterized in that, The model is constructed based on the processed data, specifically comprising: Divide the data set: divide the preprocessed data into training set, validation set and test set; Model training: use the training set to train the selected model, adjust the parameters of the model to minimize the loss function; Model evaluation: use the validation set to evaluate the trained model, select the model with the best performance.

4. The boiler combustion automatic detection control system based on big data model according to claim 3, characterized in that, The loss function includes cross-entropy loss function, mean square error loss function.

5. The boiler combustion automatic detection control system based on big data model according to claim 1, wherein the intelligent control module adopts model predictive control and fuzzy control intelligent control algorithm to generate control instructions.

6. The boiler combustion automatic detection control system based on big data model according to claim 1, wherein the monitoring and alarm module visualizes key parameters and control results, and sends alarm signals in time in abnormal conditions.

Images