Coal-fired unit nitrogen oxide prediction method, system and equipment based on support vector machine and medium
By using a support vector machine model and the unit's operating parameters to establish a nitrogen oxide concentration prediction model, the problem of inaccurate nitrogen oxide concentration measurement elements at the inlet of the denitrification system was solved, and the accuracy of nitrogen oxide concentration prediction and monitoring reliability was improved.
Patent Information
- Application Number
- CN202510787570.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-13
- Publication Date
- 2025-11-14
AI Technical Summary
The denitrification system operates in a harsh environment, and the inlet nitrogen oxide concentration measuring element is inaccurate.
A support vector machine model was adopted. By acquiring the unit's operating parameters, preprocessing and normalizing them, the relationship between nitrogen oxide generation concentration and nitrogen oxide concentration was determined. An optimization algorithm was used to optimize the kernel function and penalty factor to establish a nitrogen oxide concentration prediction model. The model was then trained and tested.
It enables accurate prediction of nitrogen oxide concentration, improves the reliability of monitoring and the feasibility of the model, and avoids the uncertainties of direct measurement.
Smart Images
Figure CN120954550A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of nitrogen oxide prediction for coal-fired power units, and in particular to a method, system, equipment, and medium for predicting nitrogen oxides in coal-fired power units based on support vector machines. Background Technology
[0002] Predicting the concentration of nitrogen oxides at the denitrification inlet based on unit operating parameters is particularly important, as it lays the foundation for exploring methods to reduce nitrogen oxide concentration through combustion adjustments.
[0003] With the development of statistical theory and the research of advanced algorithms, current nitrogen oxide prediction methods are mainly divided into three types: least squares method, artificial neural network method, and support vector machine method. Support vector machine is a computational method based on the principle of minimizing structural risk. It takes into account both empirical risk and confidence range, and is well-suited for classification and prediction problems. It has excellent prediction performance for small and medium-sized sample prediction problems. Summary of the Invention
[0004] In view of the above-mentioned problems, the present invention is proposed.
[0005] Therefore, the problem to be solved by the present invention is that the inlet nitrogen oxide concentration measuring element of the denitrification system is inaccurate due to the harsh operating environment.
[0006] To address the aforementioned technical problems, this invention provides the following technical solution: a method for predicting nitrogen oxide emissions from coal-fired power units based on support vector machines (SVMs), comprising: acquiring unit operating parameters and preprocessing the parameter data; determining the relationship between the unit operating parameters and the nitrogen oxide generation concentration; inputting the preprocessed parameter data and the determined relationship logic into a support vector machine for training; the training includes: determining the type of kernel function, optimizing the penalty factor and kernel function parameters of the support vector machine prediction model using an optimization algorithm, training based on the input data and logic to obtain a nitrogen oxide concentration prediction model; substituting the input parameters of the test data into the prediction model, comparing the model's predicted output with the actual output of the test data; analyzing and evaluating the accuracy of the prediction model through the comparison results, and optimizing it accordingly.
[0007] As a preferred embodiment of the nitrogen oxide prediction method for coal-fired power units based on support vector machines described in this invention, the unit operating parameters include coal type characteristic parameters, unit load, excess air coefficient, coal mill operating mode, primary air volume, secondary air volume, and combustion damper opening; the preprocessing of the parameter data includes removing data that does not conform to the actual operating conditions or data that shows bad points, and then normalizing the data after removal.
[0008] As a preferred embodiment of the nitrogen oxide prediction method for coal-fired power units based on support vector machines described in this invention, the determination of the relationship between unit operating parameters and nitrogen oxide generation concentration includes obtaining an indirect relationship between unit operating parameters and nitrogen oxide generation concentration using a controlled variable method based on historical operating data. Specifically, this includes indirectly relating the relationship between coal type characteristic parameters and nitrogen oxide generation concentration to the relationship between coal type characteristic parameters and coal mill outlet temperature. That is, for coal with special materials, a fixed outlet temperature is set for each type of coal with special materials based on historical data. For coal with a moisture content exceeding 30% and a volatile matter content exceeding 50%, the outlet temperature is set to 60℃. The outlet temperature for other coal types is expressed as: outlet temperature = (82 - volatile matter) * 5 / 3 + 5. The volatile matter refers to the proportion of volatile gases and liquid products released during the pyrolysis of coal at high temperatures, determined through laboratory testing. The outlet temperature of the coal mill is set according to the different characteristics of the coal type. Different coal type parameters result in different set outlet temperatures. When the type of coal used changes, the outlet temperature of the coal mill needs to be adjusted. In the calculation, the outlet temperature of the coal mill is selected to indirectly represent the characteristic parameters of the coal type.
[0009] As a preferred embodiment of the method for predicting nitrogen oxides in coal-fired power units based on support vector machines as described in this invention, the determination of the relationship between unit operating parameters and nitrogen oxide generation concentration further includes indirectly relating the relationship between unit load and nitrogen oxide generation concentration to the relationship between unit load and oxygen content in the boiler. That is, different unit load ranges correspond to different oxygen content ranges in the boiler, and the oxygen content range in the boiler affects the concentration of nitrogen oxides at the boiler outlet.
[0010] As a preferred embodiment of the nitrogen oxide prediction method for coal-fired power units based on support vector machines described in this invention, the relationship between the excess air coefficient and the nitrogen oxide generation concentration is indirectly the relationship between the excess air coefficient and the combustion state inside the boiler. That is, different fluctuation ranges of the excess air coefficient reflect different combustion states inside the boiler, and different combustion states affect the concentration of nitrogen oxides at the boiler outlet. The excess air coefficient is calculated by comprehensively considering the oxygen content, primary air volume, and secondary air volume inside the boiler.
[0011] As a preferred embodiment of the nitrogen oxide prediction method for coal-fired power units based on support vector machines described in this invention, the relationship between the coal mill operating mode and the nitrogen oxide generation concentration is indirectly the relationship between the coal mill operating mode and the ignition time and residence time of pulverized coal in the furnace. That is, different coal mill operations lead to different flame center positions, and different flame center positions lead to different ignition times and residence times of pulverized coal in the furnace, thereby affecting the concentration of nitrogen oxides at the boiler outlet; the relationship between the burnout damper opening and the nitrogen oxide generation concentration is indirectly the relationship between the burnout damper opening and the flame distribution in the furnace.
[0012] As a preferred embodiment of the nitrogen oxide prediction method for coal-fired power units based on support vector machines described in this invention, the step of evaluating the accuracy of the prediction model by comparing the results includes converting the output predicted by the model into curves of the actual output of the test data, and visually observing the accuracy of the prediction model in the form of images.
[0013] To address the aforementioned technical problems, this invention provides the following technical solution: a system for predicting nitrogen oxide emissions from coal-fired power units based on support vector machines (SVMs), comprising: a preprocessing module, a training module, and a comparison and optimization module; the preprocessing module acquires unit operating parameters and preprocesses the parameter data; determines the relationship between unit operating parameters and nitrogen oxide generation concentration; the training module inputs the preprocessed parameter data and the determined relationship logic into the support vector machine for training; the training includes determining the type of kernel function, optimizing the penalty factor and kernel function parameters of the support vector machine prediction model using an optimization algorithm, and training based on the input data and logic to obtain a nitrogen oxide concentration prediction model; the comparison and optimization module substitutes the input parameters of the test data into the prediction model, compares the model's predicted output with the actual output of the test data, analyzes and evaluates the accuracy of the prediction model through the comparison results, and optimizes it.
[0014] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the support vector machine-based method for predicting nitrogen oxides in coal-fired power units as described above.
[0015] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the support vector machine-based method for predicting nitrogen oxides in coal-fired power units as described above.
[0016] The beneficial effects of this invention are as follows: Since the nitrogen oxide sensor at the inlet of the denitrification system often becomes inaccurate due to the high temperature and high dust environment, this invention achieves accurate prediction of nitrogen oxide concentration through the support vector machine model, effectively avoiding the uncertainty of direct measurement and improving the reliability of monitoring.
[0017] This invention indirectly transforms parameters that are difficult to measure accurately in practice (such as the chemical composition of coal and the position of the flame center in the furnace) into variables that are easy to collect and measure, such as the coal mill outlet temperature, unit load, and damper opening, which significantly improves the feasibility and predictive stability of the model. Attached Figure Description
[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 This is a flowchart of the overall process for predicting nitrogen oxide emissions from coal-fired power units based on support vector machines, as described in Example 1.
[0020] Figure 2 This is a flowchart of the model training process for a nitrogen oxide prediction method for coal-fired power units based on support vector machines, as shown in Example 1.
[0021] Figure 3 This is a comparison chart of actual and predicted values in the training set of a nitrogen oxide prediction method for coal-fired power units based on support vector machines, as shown in Example 2.
[0022] Figure 4 This is a comparison chart of actual and predicted values for the test set of a nitrogen oxide prediction method for coal-fired power units based on support vector machines in Example 2. Detailed Implementation
[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0024] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0025] Example 1, referring to Figure 1 and Figure 2 This is the first embodiment of the present invention, which provides a method for predicting nitrogen oxide emissions from coal-fired power units based on support vector machines, including, as follows: Figure 1 As shown:
[0026] Step 1: Obtain the unit's operating parameters and preprocess the parameter data.
[0027] The unit operating parameters include coal type characteristics, unit load, excess air coefficient, coal mill operating mode, primary air volume, secondary air volume, and combustion damper opening.
[0028] Preprocessing the parameter data includes removing data that does not conform to the actual operating conditions or that shows bad points, and then normalizing the data after removal.
[0029] Specifically, the unit's operating parameters can be obtained using a PI (Plant Information System), and the steps are as follows:
[0030] Connecting to the PI server: Log in to the PI system client (such as PIProcessBook, PIDataLink, or PI Vision), enter a valid account and permissions, and connect to the enterprise-level data server.
[0031] Select target parameter tags: Each unit operating parameter has a unique tag number in the PI system; select according to modeling needs, such as "unit load", "oxygen content", "primary air volume", "coal mill outlet temperature" and other relevant parameters.
[0032] Set the time range: In the "Trend Query" or "Batch Export" interface, set the time range, such as "January 1, 2022 to January 7, 2022"; set the time interval (such as sampling once per minute, or averaging per hour).
[0033] Data cleaning and export: Remove discrete points, missing values, and abnormal mutations (such as negative values and static dead values), export to CSV / Excel format or write directly to the database as training samples for subsequent SVM models.
[0034] In another optional embodiment, the unit operating parameters can also be obtained through the unit's real-time system, with the following steps:
[0035] Accessing the DCS or MIS system interface: Log in using an authorized account through the DCS site or dispatch workstation to access the corresponding monitoring interface (such as boiler monitoring, coal pulverizing system, air volume measurement, etc.).
[0036] The screen for locating the required variables: Real-time systems typically display the collected data in real time through "trend charts" or "parameter browsers"; common parameters include: current burner status, coal mill operating load, damper opening, furnace oxygen content, etc.
[0037] Data Acquisition and Snapshot Recording: Acquisition methods can be selected as real-time snapshot data (current value) or historical curve tracing (change curve within a certain time period). Export formats are usually CSV, Excel, or access to a third-party database via OPC.
[0038] Compare or supplement the collected data: For variables that are missing or have low accuracy in individual data collection devices (such as the actual opening degree of the combustion damper and CO concentration), supplement them through the real-time system.
[0039] Step 2: Determine the relationship between unit operating parameters and nitrogen oxide generation concentration.
[0040] Based on historical operating data, the indirect relationship between unit operating parameters and nitrogen oxide formation concentration was obtained using the controlled variable method, specifically including:
[0041] (1) Coal Type Characteristics: Large coal-fired power units can be classified into several categories based on their volatile matter content, including anthracite, lean coal, bituminous coal, and lignite. The nitrogen content varies significantly among these different coal types, which affects the final nitrogen oxide concentration. Furthermore, the volatile matter content also differs, leading to variations in air distribution. Generally, lignite produces the lowest nitrogen oxide emission concentration. The combustion characteristics of different coal types influence the ignition zone and temperature distribution within the boiler, all of which affect the nitrogen oxide concentration to some extent. Therefore, it is evident that the type of coal has a significant impact on the nitrogen oxide concentration at the boiler outlet.
[0042] Considering the actual situation on site, the different coal types designed for different units, as well as the influence of geographical environment and transportation methods, and taking into account the power plant's daily coal replenishment plan, the coal type changes frequently and the data collected by the chemical department does not include nitrogen. Therefore, the model of this invention does not directly select the relevant parameters of the coal type as input parameters, but instead selects the outlet temperature of the coal mill to indirectly reflect the characteristics of the coal type.
[0043] The setting of the coal mill outlet temperature according to the power plant's operational requirements is determined based on the moisture and volatile matter parameters of the coal used, and the specific rules are as follows:
[0044] That is, for coal with special materials, a fixed outlet temperature is set for each type of coal with special materials based on historical data.
[0045] For imported coal with a moisture content exceeding 30% and a volatile matter content exceeding 50%, the outlet temperature is set at 60℃.
[0046] The outlet temperature for other coal types is expressed as follows:
[0047] Outlet temperature = (82 - volatile matter) * 5 / 3 + 5
[0048] The volatile matter refers to the ratio of volatile gases and liquid products released during the pyrolysis of coal at high temperatures, as determined in the laboratory.
[0049] The outlet temperature of the coal mill is set according to the different characteristics of the coal. Different coal parameters require different set outlet temperatures. When the type of coal used changes, the outlet temperature of the coal mill needs to be adjusted. In the calculation, the outlet temperature of the coal mill is selected to indirectly represent the characteristic parameters of the coal.
[0050] (2) Unit load: The relationship between unit load and nitrogen oxide generation concentration is indirectly the relationship between unit load and oxygen content in the boiler. That is, different unit load ranges correspond to different oxygen content ranges in the boiler, and the oxygen content range in the boiler affects the concentration of nitrogen oxides at the boiler outlet.
[0051] The unit load and most of the unit's operating parameters are interconnected. Under normal circumstances, the higher the unit load, the lower the oxygen content in the boiler. When the load is 300 MW, the oxygen content is about 5%; when the load is 400 MW, the oxygen content is about 3%; and when the load is above 500 MW, the oxygen content is usually less than 2%. Although the oxygen content decreases as the load increases, the amount of coal and the internal temperature of the furnace both increase, resulting in an overall increase in nitrogen oxide production.
[0052] In fact, the oxygen content of the generating unit has a more significant impact on nitrogen oxides than the boiler's internal temperature. Typically, only when oxygen fluctuations are relatively small will the concentration of nitrogen oxides decrease with decreasing load and boiler internal temperature. The unit load is also directly related to the total coal and air volume. Therefore, we determined the unit load as the input parameter for the prediction model.
[0053] (3) Excess air coefficient: The relationship between excess air coefficient and nitrogen oxide generation concentration is indirectly the relationship between excess air coefficient and combustion state inside the boiler. That is, different fluctuation ranges of excess air coefficient reflect different combustion states inside the boiler, and different combustion states affect the concentration of nitrogen oxides at the boiler outlet.
[0054] The excess air coefficient α refers to the ratio of the actual amount of air supplied for fuel combustion to the theoretical amount of air. The excess air coefficient has the greatest impact on nitrogen oxides (NOx), especially fuel-type NOx. Relevant data confirms that NOx concentration decreases as the oxygen content inside the furnace decreases. The excess air coefficient α has a significant impact on both fuel-type and thermal NOx. When the excess air coefficient α fluctuates within the range of 0.8 to 1.1, the following pattern emerges: the larger the excess air coefficient α, the better the combustion inside the boiler, the higher the furnace temperature, and the better the combustion of nitrogen-containing organic matter. In this case, both fuel-type and thermal NOx increase. When α > 1.1, as the excess air coefficient increases, the furnace temperature tends to decrease, leading to a reduction in thermal NOx, but fuel-type NOx continues to increase.
[0055] It is evident that the excess air coefficient has a significant impact on nitrogen oxides, and oxygen content, primary air volume, and secondary air volume are the main parameters reflecting the excess air coefficient. Therefore, we selected the oxygen content, primary air volume, and secondary air volume inside the furnace as the input parameters for the prediction model.
[0056] (4) Coal mill operation mode: The relationship between coal mill operation mode and nitrogen oxide generation concentration is indirectly the relationship between coal mill operation mode and ignition time of pulverized coal and residence time in the furnace. That is, different coal mill operation modes lead to different flame center positions.
[0057] This example demonstrates a unit with burners arranged in a counter-current configuration along the front and rear walls, with three layers of burners on each wall. Under normal circumstances, the coal mills are started sequentially from the lower layer to the upper layer, and shut down in the reverse order. The operation of different coal mills results in different flame center positions. With symmetrical combustion along the front and rear walls, the flame center is located in the center of the furnace. When the number of operating coal mills is odd, the flame center will deviate. Different flame center positions lead to variations in the ignition time and residence time of the pulverized coal in the furnace, thus affecting the concentration of nitrogen oxides at the boiler outlet.
[0058] (5) Burnout damper opening: The relationship between the burnout damper opening and the nitrogen oxide generation concentration is indirectly the relationship between the burnout damper opening and the flame distribution in the furnace.
[0059] Manually adjusting the burnout damper opening can modify the flame distribution within the furnace, improve the combustion environment, optimize combustion, reduce nitrogen oxide formation, and also adjust the temperature deviation between the two sides. Therefore, we selected the burnout damper opening as the input parameter for the prediction model.
[0060] To further explain, the innovation of this invention lies in indirectly transforming difficult-to-measure parameters into easily measurable parameters with high measurement accuracy. In actual production, parameters such as the chemical content of coal, oxygen content in the furnace, combustion state in the furnace, flame center position, and flame distribution are difficult to measure. Furthermore, due to the harsh operating environment inside the furnace, the measuring sensors may malfunction, further reducing the measurement accuracy. Therefore, transforming difficult-to-measure parameters into directly measurable parameters such as the coal mill outlet temperature, unit load, coal mill operating mode, and combustion damper opening is beneficial for predicting nitrogen oxide generation concentration and improving accuracy.
[0061] Step 3: Input the preprocessed parameter data and the determined relational logic into a support vector machine for training. Determine the type of kernel function, and optimize the penalty factor and kernel function parameters of the support vector machine prediction model using an optimization algorithm. Based on the input data and logic, train the model to obtain the nitrogen oxide concentration prediction model, such as... Figure 2 As shown.
[0062] Specifically, the optimization algorithm can optimize the penalty factor and kernel function parameters of the support vector machine prediction model using the grid optimization algorithm. The steps are as follows:
[0063] Constructing the parameter search space: Construct a two-dimensional parameter grid, with the horizontal axis representing the penalty factor C and the vertical axis representing the kernel function parameter γ.
[0064] To cover as many cases as possible, set a set of candidate values for each of the two parameters, for example:
[0065] Penalty factor: Gradually increase from a small value (e.g., 0.1) to a large value (e.g., 100).
[0066] Kernel function parameters: Gradually increase from a small value (e.g., 0.001) to a large value (e.g., 10).
[0067] Set an appropriate step size between each candidate value so that each coordinate point in the entire grid corresponds to a set of parameter combinations.
[0068] Cross-validation is used to evaluate the performance of each parameter combination: For each parameter combination in the grid, cross-validation is used for evaluation. Specifically, the training dataset is divided into several equal parts (e.g., 5 parts). Each time, one part is selected as the validation set, and the rest are used as the training set. The training and prediction process is repeated, and the average prediction error is calculated. This average error represents the performance of that parameter combination.
[0069] Selecting the parameter combination that minimizes the prediction error: Among all parameter combinations, select the set of parameters that minimizes the prediction error as the hyperparameters of the final support vector regression model. This optimization objective can be expressed by the formula:
[0070]
[0071] Among them, C * Let γ be the finally selected optimal penalty factor. * The parameter 'argmin' represents the final selected optimal kernel function parameters, 'prediction error' represents the mean square error or root mean square error obtained from cross-validation, and 'argmin' represents the parameter combination that minimizes the error.
[0072] Train the final model using the optimal parameters: The optimal parameters C selected in the previous step... * and γ * The model is then applied to the entire training dataset to train the final prediction model. This model is then used to predict the test data and compared with the actual values to verify whether the optimized model has good generalization ability and prediction performance.
[0073] In another optional embodiment, the optimization algorithm can also employ a Bayesian optimization algorithm to optimize the penalty factor and kernel function parameters of the support vector machine prediction model, with the following steps:
[0074] Define the objective function: The objective function uses the prediction error (such as cross-validation MSE) as the evaluation index, takes the combination of parameters to be optimized (C, γ) as input, and outputs the corresponding model prediction error.
[0075] The objective function takes the form of: input SVM parameters (C, γ), output the mean prediction error (MSE) of the support vector regression model.
[0076] Initialize sample points: Randomly select several initial points in the parameter space (e.g., 5 different sets of (C, γ)); For each set of parameters, train the SVM model and calculate the corresponding prediction error. These initial data will be used to fit the probabilistic model of the objective function (usually Gaussian process regression GP).
[0077] Construct a probabilistic surrogate model for the objective function: Use a Gaussian process (GP) to model the prediction error function in the parameter space. For any combination of parameters (C, γ), GP can not only predict the error, but also provide an uncertainty estimate.
[0078] The acquisition function is used to select the next attempt point; the acquisition function value is calculated based on the surrogate model, and the commonly used one is "Expected Improvement (EI)". The point with the largest EI value is selected as the parameter combination for the next attempt; this is very effective in balancing "utilizing known optimality" and "exploring untried areas".
[0079] Evaluate new points and update the model: Train the support vector machine model using the newly selected (C, γ) combination, calculate the prediction error, add it to the existing samples, and update the Gaussian process surrogate model.
[0080] Iterate until the termination condition is met: the number of iterations reaches a preset number (e.g., 30 times), the optimal solution tends to stabilize in several iterations, and the EI value drops below the threshold (indicating that the room for improvement is very small).
[0081] Output optimal parameters: The final parameter combination is the one that minimizes the prediction error (C). * γ * ), used to train the final support vector regression model.
[0082] Step 4: Substitute the input parameters of the test data into the prediction model, compare the model's predicted output with the actual output of the test data, analyze and evaluate the accuracy of the prediction model through the comparison results, and optimize it.
[0083] The accuracy of the prediction model is evaluated by comparing the results. This includes plotting the model's predicted output against the actual output of the test data to visually observe the accuracy of the prediction model in a graphical form.
[0084] Example 2, refer to Figure 3 and Figure 4The second embodiment of the present invention differs from the first embodiment in that: a method for predicting nitrogen oxides in coal-fired power units based on support vector machines further includes verifying and explaining the technical effects used in the method, and using scientific demonstration methods to verify the real effect of the method.
[0085] Forty sets of data were selected as training and testing samples. The samples included various boiler operating parameters. The input parameters were: unit load (indirectly oxygen content), coal quantity and outlet temperature of the six coal mills, primary air volume, secondary air volume, carbon monoxide concentration, and combustion damper opening. The output parameter was nitrogen oxide concentration. Among them, 300 sets of data were used as training samples and 100 sets of data were used as testing samples. The data are shown in Table 1.
[0086] Table 1 Operating parameters of some units
[0087]
[0088] Since the parameters selected for the unit, such as load, air volume, coal volume, temperature, oxygen content, and nitrogen oxide concentration, vary greatly in order of magnitude and unit, the mapminmax function is used to normalize the data.
[0089] The kernel function selected in the prediction model is the RBF kernel function, and the grid optimization algorithm is selected to optimize the penalty factor and kernel function parameters of the support vector machine prediction model.
[0090] Both model generation and prediction are implemented using built-in functions of libsvm. The model uses svmtrain and svmpredict for training and prediction respectively, and the resulting C... * and γ * By setting the relevant properties of the kernel function and then substituting the normalized data into the function, training and prediction can be achieved. The obtained data is then denormalized and compared with the true value, and the results are analyzed in the form of images.
[0091] like Figure 3 and Figure 4 As shown in the figure, the red line segment represents the actual value of the nitrogen oxide concentration at the denitrification inlet, and the blue line segment represents the predicted value output by the prediction model. It can be seen that the prediction model can accurately predict the nitrogen oxide concentration. Through error calculation, the support vector machine prediction model has an MSE of 6.12425 and an RMSE of 2.4784.
[0092] This embodiment takes the nitrogen oxide concentration at the denitrification inlet of a 600MW coal-fired power unit as the research object. The nitrogen oxide concentration is predicted by the support vector machine algorithm. The optimal parameters are obtained by grid optimization. The model is trained with 300 sets of data and the accuracy of the model is verified by 100 sets of data. The error is calculated. The results show that the present invention indirectly transforms difficult-to-measure parameters into easy-to-measure parameters, and the support vector machine model can achieve accurate prediction of nitrogen oxide concentration.
[0093] Example 3, the third embodiment of the present invention, differs from the previous two embodiments in that it provides a nitrogen oxide prediction method system for coal-fired power units based on support vector machines (SVMs). The system includes a preprocessing module, a training module, and a comparison and optimization module. The preprocessing module acquires unit operating parameters and preprocesses the parameter data, determining the relationship between the unit operating parameters and the nitrogen oxide concentration. The training module inputs the preprocessed parameter data and the determined relationship logic into the support vector machine for training. The training includes determining the type of kernel function and optimizing the penalty factor and kernel function parameters of the support vector machine prediction model using an optimization algorithm. Based on the input data and logic, a nitrogen oxide concentration prediction model is obtained. The comparison and optimization module substitutes the input parameters of the test data into the prediction model, compares the model's predicted output with the actual output of the test data, analyzes and evaluates the accuracy of the prediction model through the comparison results, and optimizes it.
[0094] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0095] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.
[0096] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.
[0097] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented in combination with any of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.
[0098] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A method for predicting nitrogen oxide emissions from coal-fired power plants based on support vector machines, characterized in that: include, Acquire unit operating parameters and preprocess the parameter data; Determine the relationship between unit operating parameters and nitrogen oxide generation concentration; The preprocessed parameter data and the determined relational logic are input into the support vector machine for training. The training includes determining the type of kernel function, optimizing the penalty factor and kernel function parameters of the support vector machine prediction model through an optimization algorithm, and training based on the input data and logic to obtain a nitrogen oxide concentration prediction model. Substitute the input parameters of the test data into the prediction model, and compare the model's predicted output with the actual output of the test data. The accuracy of the prediction model is evaluated by comparing the results, and then optimized.
2. The method for predicting nitrogen oxide emissions from coal-fired power units based on support vector machines as described in claim 1, characterized in that: The unit operating parameters include coal type characteristic parameters, unit load, excess air coefficient, coal mill operating mode, primary air volume, secondary air volume, and combustion damper opening. The preprocessing of parameter data includes removing data that does not conform to the actual operating conditions or data that shows bad points from the operating parameters, and then normalizing the data after removal.
3. The method for predicting nitrogen oxide emissions from coal-fired power units based on support vector machines as described in claim 2, characterized in that: Determining the relationship between unit operating parameters and nitrogen oxide generation concentration includes obtaining an indirect relationship between unit operating parameters and nitrogen oxide generation concentration based on historical operating data using a controlled variable method. Specifically, this includes... The relationship between the coal type characteristic parameters and the nitrogen oxide generation concentration is indirectly the relationship between the coal type characteristic parameters and the coal mill outlet temperature. That is, for coal with special materials, a fixed outlet temperature is set for each type of coal with special materials based on historical data. For coal with a moisture content exceeding 30% and a volatile matter content exceeding 50%, the outlet temperature is set at 60℃. The outlet temperature for other coal types is expressed as: outlet temperature = (82 - volatile matter) * 5 / 3 + 5. The volatile matter content refers to the ratio of volatile gases and liquid products released during the pyrolysis of coal at high temperature, which is determined in the laboratory. The outlet temperature of the coal mill is set according to the different characteristics of the coal. Different coal parameters require different set outlet temperatures. When the type of coal used changes, the outlet temperature of the coal mill needs to be adjusted. In the calculation, the outlet temperature of the coal mill is selected to indirectly represent the characteristic parameters of the coal.
4. The method for predicting nitrogen oxide emissions from coal-fired power units based on support vector machines as described in claim 3, characterized in that: The determination of the relationship between unit operating parameters and nitrogen oxide generation concentration also includes indirectly relating the relationship between unit load and nitrogen oxide generation concentration to the relationship between unit load and oxygen content in the boiler. That is, different unit load ranges correspond to different oxygen content ranges in the boiler, and the oxygen content range in the boiler affects the nitrogen oxide concentration at the boiler outlet.
5. The method for predicting nitrogen oxide emissions from coal-fired power units based on support vector machines as described in claim 4, characterized in that: The relationship between the excess air coefficient and the concentration of nitrogen oxides is indirectly the relationship between the excess air coefficient and the combustion state inside the boiler. That is, different fluctuation ranges of the excess air coefficient reflect different combustion states inside the boiler, and different combustion states affect the concentration of nitrogen oxides at the boiler outlet. The excess air coefficient is calculated by combining the oxygen content, primary air volume, and secondary air volume in the boiler.
6. The method for predicting nitrogen oxide emissions from coal-fired power units based on support vector machines as described in claim 5, characterized in that: The relationship between the coal mill operation mode and the nitrogen oxide generation concentration is indirectly the relationship between the coal mill operation mode and the ignition time and residence time of pulverized coal in the furnace. That is, different coal mill operations lead to different flame center positions. Different flame center positions lead to different ignition times and residence times of pulverized coal in the furnace, which in turn affect the concentration of nitrogen oxides at the boiler outlet. The relationship between the opening degree of the combustion damper and the concentration of nitrogen oxides generated indirectly represents the relationship between the opening degree of the combustion damper and the flame distribution in the furnace.
7. The method for predicting nitrogen oxide emissions from coal-fired power units based on support vector machines as described in claim 6, characterized in that: The process of evaluating the accuracy of the prediction model by comparing the results includes plotting the model's predicted output against the actual output of the test data to visually observe the accuracy of the prediction model in the form of an image.
8. A method system for predicting nitrogen oxide emissions from coal-fired power units based on support vector machines, employing the nitrogen oxide prediction method for coal-fired power units based on support vector machines as described in any one of claims 1 to 7, characterized in that: It includes a pre-processing module, a training module, and a comparison and optimization module; The preprocessing module acquires the unit's operating parameters and preprocesses the parameter data; Determine the relationship between unit operating parameters and nitrogen oxide generation concentration; The training module inputs the preprocessed parameter data and the determined relational logic into the support vector machine for training; the training includes determining the type of kernel function, optimizing the penalty factor and kernel function parameters of the support vector machine prediction model through an optimization algorithm, and training based on the input data and logic to obtain a nitrogen oxide concentration prediction model; The comparison and optimization module substitutes the input parameters of the test data into the prediction model, compares the output predicted by the model with the actual output of the test data, analyzes and evaluates the accuracy of the prediction model through the comparison results, and optimizes it.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the method for predicting nitrogen oxides in coal-fired power units based on support vector machines, as described in any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the method for predicting nitrogen oxides in coal-fired power units based on support vector machines, as described in any one of claims 1 to 7.