Method and system for predicting ash level of ash hopper of coal electric boiler based on LSTM (Long Short Term Memory)
Through the LSTM-based ash level prediction method for coal-electric boiler ash bucket, the problem of difficult to predict the ash level changes in the ash transmission system is solved, and intelligent monitoring and prediction of the ash level is realized, system operation efficiency is improved and energy consumption is reduced.
Patent Information
- Application Number
- CN202510023849.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-07
- Publication Date
- 2025-05-09
AI Technical Summary
The existing ash delivery system lacks intelligent control and cannot predict changes in the ash level, resulting in the phenomenon of over-full or over-empty ash bucket, affecting the system's operating efficiency and increasing energy consumption.
Using the LSTM-based ash level prediction method for coal-electric boiler ash bucket, we continuously obtain the ash data and boiler operating parameters of the ash bucket, build multiple sample data pairs, train a bidirectional LSTM model, obtain the ash level prediction model, and collect the operation parameters in real time for prediction.
The prediction of the changes in the gray level is achieved, ensuring that the gray level always fluctuates within a reasonable range, avoiding the phenomenon of over-full or over-empty ash bucket, improving system operation efficiency and reducing energy consumption.
Smart Images
Figure CN119961883A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of ash conveying systems in thermal power plants, and in particular to a method and system for predicting ash levels in ash hoppers of coal-fired power boilers based on LSTM. Background Art
[0002] At present, new energy sources such as wind power and solar energy are connected to the power grid for power generation. However, new energy power generation has inherent characteristics such as unstable power generation and frequent frequency changes, which makes it difficult to meet the needs of stable operation of the power grid. Thermal power generation has become a key force to ensure the stable operation of the power grid because of its ability to flexibly change loads and perform frequency peak regulation. In thermal power generation, coal quality issues bring many challenges to thermal power generation. The implementation of coal blending strategies and the complexity and diversity of coal types make the ash volume and ash content generated by coal-fired boilers complex and changeable, and poorly stable.
[0003] In the ash conveying system of a thermal power plant, the monitoring and control of ash level is crucial to system efficiency and operational stability. As the core equipment in a thermal power plant, the combustion process of a coal-fired boiler will produce a large amount of ash, which needs to be effectively discharged through the ash conveying system.
[0004] The existing ash conveying system generally lacks intelligent control and cannot predict the change of ash level, which leads to the ash hopper being overfilled or overempty from time to time, affecting the system operation efficiency and increasing energy consumption. In particular, the ash characteristics and emissions of coal-fired power boilers have certain fluctuations during operation, which brings great challenges to the traditional ash conveying system. Summary of the invention
[0005] The present invention provides a method and system for predicting the ash level of an ash hopper of a coal-fired power boiler based on LSTM, which solves the problem that the existing ash conveying system generally lacks intelligent control and cannot predict the ash level change, resulting in the ash hopper being overfilled or overempty from time to time, affecting the system operation efficiency and increasing energy consumption.
[0006] In a first aspect, the present invention provides a method for predicting ash level in an ash hopper of a coal-fired power boiler based on LSTM, comprising the following steps:
[0007] Continuously obtain multiple pieces of ash level data corresponding to the ash hopper of the coal-fired power boiler at multiple time points, wherein each adjacent time point is separated by a time step, and obtain the boiler operation parameters with time dependence corresponding to each time step;
[0008] The operating parameters corresponding to the time step between the current time point and the previous time point are used as input, and the gray level data corresponding to the current time point is used as output to construct multiple sample data pairs;
[0009] The bidirectional LSTM model is trained through multiple sample data to obtain the gray level prediction model;
[0010] The operating parameters of the coal-fired power boiler are collected in real time and input into the ash level prediction model to obtain the corresponding ash level data.
[0011] Preferably, the time series processing of the plurality of sample data pairs further requires standardization of the operating parameters of each time step in the sample data, and the standardization process includes:
[0012] Get the mean and standard deviation of each running data at each time step;
[0013] The mean of each run data was subtracted and divided by the standard deviation to obtain the standardized run parameters.
[0014] Preferably, the boiler operating parameters include boiler temperature, pressure and flow rate.
[0015] Preferably, the training of the bidirectional LSTM model by using multiple sample data to obtain the gray level prediction model comprises the following steps:
[0016] Receive the operating parameters of multiple time steps through the input layer of the bidirectional LSTM model;
[0017] The LSTM layer of the bidirectional LSTM model is used to extract the time series features of the operating parameters at each time step;
[0018] The extracted time series features are mapped through the fully connected layer of the bidirectional LSTM model to obtain the corresponding multiple gray level data;
[0019] The corresponding gray level data is output through the output layer of the bidirectional LSTM model;
[0020] The bidirectional LSTM model is iteratively trained using the AdamW optimizer.
[0021] Preferably, a plurality of ash level data corresponding to the ash hopper of the coal-fired power boiler at multiple time points are continuously obtained by a nuclear level meter.
[0022] Preferably, after predicting the ash level data corresponding to the boiler operating parameters collected in real time based on the mapping relationship, the ash transport cycle is dynamically adjusted according to the predicted ash level data.
[0023] In a second aspect, the present invention provides a ash level prediction system for an ash hopper of a coal-fired power boiler based on LSTM, comprising:
[0024] An acquisition module is used to continuously acquire multiple pieces of ash level data corresponding to the ash hopper of the coal-fired power boiler at multiple time points, wherein each adjacent time point is separated by a time step, and to acquire the boiler operation parameters with time dependence corresponding to each time step;
[0025] A construction module is used to take the operating parameters corresponding to the time step between the current time point and the previous time point as input, and the gray level data corresponding to the current time point as output, to construct multiple sample data pairs;
[0026] The training module is used to train the bidirectional LSTM model through multiple sample data to obtain a gray level prediction model;
[0027] The prediction module is used to collect the operating parameters of the coal-fired power boiler in real time and input them into the ash level prediction model to obtain the corresponding ash level data.
[0028] In a third aspect, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the above-mentioned LSTM-based coal-fired power boiler ash hopper ash level prediction method is implemented.
[0029] In a fourth aspect, the present invention provides a computer-readable storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned LSTM-based coal-fired power boiler ash hopper ash level prediction method is implemented.
[0030] Compared with the prior art, the present invention has the following beneficial effects:
[0031] The present invention first continuously obtains a plurality of ash level data corresponding to the ash hopper of a coal-fired power boiler at multiple time points and the operating parameters corresponding to each time point, and inputs them into a bidirectional LSTM model to train the model to obtain a ash level prediction model. The present invention obtains the mapping relationship between the ash level data and the corresponding operating parameters based on the bidirectional LSTM model, and collects the boiler operating parameters in real time and inputs them into the ash level prediction model to obtain the corresponding ash level data, thereby realizing the prediction of ash level changes. Based on the predicted ash level, the system can ensure that the ash level always fluctuates within a reasonable range, avoids the phenomenon of the ash hopper being overfull or overempty, and improves the system operation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0033] Figure 1 It is a flow chart of a method for predicting ash level in ash hopper of coal-fired power boiler based on LSTM of the present invention;
[0034] Figure 2 The figure is a schematic diagram of the installation position of the passive nuclear material level meter of the present invention. DETAILED DESCRIPTION
[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0036] In order to solve the shortcomings of the prior art, the present invention provides a method for predicting the ash level in the ash hopper of a coal-fired power boiler based on LSTM. Figure 1 , including the following steps:
[0037] Step 1: Continuously obtain multiple ash level data corresponding to the ash hopper of the coal-fired power boiler at multiple time points, where there is a time step between two adjacent time points, and obtain the corresponding time-dependent operating parameters in each time step.
[0038] The present invention first loads historical data from the storage system and performs timestamp alignment on these data. For each piece of gray level sensor data, multiple pieces of boiler data in the previous 60 minutes are searched as input features, and multiple pieces of data of the gray level sensor in the next hour are used as prediction targets. The time alignment of the data ensures the temporal consistency of the input and output, and ensures that the prediction model can learn the correlation between boiler data and gray level changes.
[0039] The historical data include various boiler operating parameters, such as temperature, pressure and flow rate, etc. These data are used to predict future ash level changes and have obvious time-dependent characteristics.
[0040] The ash level data comes from the sensor data of the passive nuclear material level meter, which reflects the real-time situation of the ash level in the ash hopper.
[0041] Reference Figure 2 , the passive nuclear level meter monitors the ash level changes in the ash hopper and silo pump, realizing non-contact monitoring and avoiding the common dust blockage and electromagnetic interference problems of traditional sensors. The nuclear level meter uses radioactive isotope rays to penetrate the material and uses the difference in dust absorption of the rays to determine the ash level. It has the characteristics of high accuracy and wide application range, and is particularly suitable for high-temperature and high-dust thermal power plant environments.
[0042] The source nuclear level meter is installed at the key position of the ash hopper and the silo pump to continuously monitor the change of the ash level. The level meter penetrates the material in the ash hopper through radioactive rays and detects the intensity of the material absorbing the rays to obtain the specific height of the material. Since the density of dust and air is much smaller than that of solid materials, the change of the ray intensity can accurately reflect the height of the ash level.
[0043] The signal of the passive nuclear level meter is filtered, amplified and digitally processed before being sent to the central control system to ensure the accuracy and real-time performance of the ash level data. The signal processing module has anti-interference capabilities and can operate stably in complex environments.
[0044] Step 2: Take the operating parameters corresponding to the time step between the current time point and the previous time point as input, and the gray level data corresponding to the current time point as output, to construct multiple sample data pairs.
[0045] For each prediction task, the input includes boiler data of multiple time steps, and multiple important features are extracted for each time step. These features cover key parameters of boiler operation, such as combustion temperature, pressure, and flow rate.
[0046] The output is gray level data of multiple time steps, each time step includes the readings of multiple gray level sensors, which respectively reflect the future changes of multiple gray levels.
[0047] Step 3: Perform time series processing on multiple sample data to obtain the mapping relationship between the operating parameters of the boiler at each time step and the current time point corresponding to the time step.
[0048] Since the boiler sensor data and the gray level sensor data have different numerical ranges, directly inputting them into the deep learning model may cause gradient problems during training. Therefore, we standardize the boiler data at each time step, that is, subtract the mean and divide by the standard deviation, so that the data of each feature is distributed in a similar numerical range. Standardization can not only accelerate the convergence speed of the model, but also improve the prediction accuracy.
[0049] The system extracts the correlation between the ash delivery cycle and ash level changes by analyzing historical ash level change data.
[0050] The gray level prediction algorithm is used to perform time series processing on multiple sample data. The present invention combines a time series regression model (LSTM) based on deep learning to perform real-time analysis on the data of the nuclear level meter. The algorithm takes into account the gray level change trend of the ash hopper and silo pump, and combines historical data to predict future gray level changes. Through the gray level prediction algorithm, the system can adjust the ash delivery cycle in advance before the ash hopper is about to reach the critical gray level, avoid the ash hopper being overfilled or overempty, and improve the operational reliability of the system.
[0051] By performing time series processing on boiler data (input data) and ash level sensor data (output data), the algorithm can predict the future ash level based on the boiler's operating status in different time periods. In particular, the algorithm uses boiler sensor data at multiple time steps and uses data preprocessing techniques such as standardization and regularization to improve prediction accuracy and reduce the risk of model overfitting.
[0052] Long Short-Term Memory (LSTM) is a recurrent neural network (RNN) that can effectively process time series data. Compared with traditional RNN, LSTM can better capture the characteristics of long-term dependence and avoid the gradient vanishing problem by introducing memory units and gate mechanisms. Specifically, LSTM contains three core gate structures: forget gate, input gate and output gate, which work together to determine the model's "memory" process of historical data.
[0053] In the gray level prediction algorithm of the present invention, a bidirectional LSTM model is designed, that is, data processing is performed in both the forward and backward directions of the time series to capture the complex dependency relationship between boiler data and gray level data. Bidirectional LSTM can consider both past and future information at the same time, thereby improving the accuracy of prediction.
[0054] The bidirectional LSTM model includes an input layer, an LSTM layer, a fully connected layer, and an output layer. The input layer inputs boiler data of multiple time steps, each of which contains multiple features. The LSTM layer uses a multi-layer bidirectional LSTM network, each of which contains multiple hidden units. By stacking multiple LSTM layers, the model can learn higher-order time-dependent characteristics in boiler data. The fully connected layer maps the output of the LSTM layer to multiple predicted time steps through the fully connected layer, and predicts the readings of multiple gray level sensors at each time step. The output layer outputs gray level data of multiple time steps.
[0055] With this design, the model is able to capture subtle changes in the boiler's operating status and use this to predict future ash levels.
[0056] The AdamW optimizer is used for model training. AdamW is an optimization algorithm based on adaptive gradient descent. It combines the fast convergence of the Adam optimizer and the anti-overfitting ability of L2 regularization, and can perform well on large-scale time series data.
[0057] In order to measure the deviation between the model prediction results and the true value, the mean square error (MSE) is used as the loss function. MSE calculates the square difference between the predicted value and the true value, which can effectively capture the model error and guide the model to update parameters.
[0058] The model parameters are optimized through multiple rounds of iterative training. In order to prevent the model from overfitting, L2 regularization is introduced in the training process, and the early stopping method is adopted, that is, the training is stopped when the loss on the validation set no longer decreases, so as to avoid the model from overfitting on the training set.
[0059] Step 4: Collect boiler operating parameters in real time, and predict the gray level changes corresponding to the boiler operating parameters collected in real time based on the mapping relationship.
[0060] In actual applications, the trained model can process boiler data in real time and predict the future ash level. The prediction results will be fed back to the control module of the conveying system to ensure that the system is always in the best conveying state. Compared with the traditional fixed parameter adjustment method, the ash level prediction algorithm based on deep learning can adaptively adjust the conveying parameters, greatly improving the conveying efficiency and reducing energy consumption.
[0061] Through ash level monitoring and prediction, the system of the present invention can realize comprehensive monitoring of ash level and ensure smooth and efficient transportation process. At the same time, the introduction of deep learning algorithm enables the system to continuously optimize itself and improve the intelligence level of the system.
[0062] According to the ash position prediction results, the system intelligently adjusts the ash conveying cycle to optimize the ash conveying operation and ensure that the ash conveying operation is carried out at the best time to reduce ash conveying energy consumption and equipment wear. This adjustment mechanism makes the ash conveying process more flexible, can adapt to the load changes of thermal power plants, and improve the overall efficiency of the system. This intelligent control method not only improves the conveying efficiency, but also significantly reduces the energy consumption of the system, providing a technical solution with broad application prospects for the industrial field.
[0063] Based on the same concept, the present invention also provides a LSTM-based coal-fired power boiler ash hopper ash level prediction system, including an acquisition module, a construction module, a training module and a prediction module.
[0064] The acquisition module is used to continuously acquire multiple ash level data corresponding to the ash hopper of the coal-fired power boiler at multiple time points, wherein each adjacent time point is separated by a time step, and to acquire the boiler operation parameters with time dependence corresponding to each time step.
[0065] The construction module is used to take the operating parameters corresponding to the time step between the current time point and the previous time point as input, and the gray level data corresponding to the current time point as output, to construct multiple sample data pairs.
[0066] The training module is used to train the bidirectional LSTM model through multiple sample data to obtain a gray level prediction model.
[0067] The prediction module is used to collect the operating parameters of the coal-fired power boiler in real time and input them into the ash level prediction model to obtain the corresponding ash level data.
[0068] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the above-mentioned LSTM-based coal-fired power boiler ash hopper ash level prediction method is implemented.
[0069] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned LSTM-based coal-fired power boiler ash hopper ash level prediction method is implemented.
[0070] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.
[0071] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A method for predicting ash level in ash hopper of coal-fired power boiler based on LSTM, characterized in that: The following steps are involved: Continuously obtain multiple pieces of ash level data corresponding to the ash hopper of the coal-fired power boiler at multiple time points, wherein each adjacent time point is separated by a time step, and obtain the boiler operation parameters with time dependence corresponding to each time step; The operating parameters corresponding to the time step between the current time point and the previous time point are used as input, and the gray level data corresponding to the current time point is used as output to construct multiple sample data pairs; The bidirectional LSTM model is trained through multiple sample data to obtain the gray level prediction model; The operating parameters of coal-fired power boilers are collected in real time and input into the ash level prediction model to obtain the corresponding ash level data.
2. The ash level prediction method of the coal-fired power boiler ash hopper based on LSTM as claimed in claim 1 is characterized in that: Before training the bidirectional LSTM model using multiple sample data, it is necessary to standardize the operating parameters of each time step in the sample data. The standardization process includes: Get the mean and standard deviation of each running data at each time step; The mean of each run data was subtracted and divided by the standard deviation to obtain the standardized run parameters.
3. The ash level prediction method of coal-fired power boiler ash hopper based on LSTM as claimed in claim 1, characterized in that: The boiler operating parameters include boiler temperature, pressure and flow rate.
4. The ash level prediction method for coal-fired power boiler ash hopper based on LSTM as claimed in claim 1, characterized in that: The bidirectional LSTM model is trained by using multiple sample data to obtain a gray level prediction model, including the following steps: Receive the operating parameters of multiple time steps through the input layer of the bidirectional LSTM model; The LSTM layer of the bidirectional LSTM model is used to extract the time series features of the operating parameters at each time step; The extracted time series features are mapped through the fully connected layer of the bidirectional LSTM model to obtain the corresponding multiple gray level data; The corresponding gray level data is output through the output layer of the bidirectional LSTM model; The bidirectional LSTM model is iteratively trained using the AdamW optimizer.
5. The ash level prediction method of coal-fired power boiler ash hopper based on LSTM as claimed in claim 1, characterized in that: The nuclear material level meter is used to continuously obtain multiple ash level data corresponding to the ash hopper of the coal-fired power boiler at multiple time points.
6. The ash level prediction method of coal-fired power boiler ash hopper based on LSTM as claimed in claim 1, characterized in that: After predicting the ash level data corresponding to the boiler operating parameters collected in real time based on the mapping relationship, the ash conveying cycle is dynamically adjusted according to the predicted ash level data.
7. A LSTM-based coal-fired power boiler ash hopper ash level prediction system, characterized in that: include: An acquisition module is used to continuously acquire multiple pieces of ash level data corresponding to the ash hopper of the coal-fired power boiler at multiple time points, wherein each adjacent time point is separated by a time step, and to acquire the boiler operation parameters with time dependence corresponding to each time step; A construction module is used to take the operating parameters corresponding to the time step between the current time point and the previous time point as input, and the gray level data corresponding to the current time point as output, to construct multiple sample data pairs; The training module is used to train the bidirectional LSTM model through multiple sample data to obtain a gray level prediction model; The prediction module is used to collect the operating parameters of the coal-fired power boiler in real time and input them into the ash level prediction model to obtain the corresponding ash level data.
8. A computer device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, the method for predicting the ash level in the ash hopper of a coal-fired power boiler based on LSTM is implemented.
9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the LSTM-based coal-fired power boiler ash hopper ash level prediction method described in any one of claims 1 to 6 is implemented.