Realization method of intelligent control system of waste incineration feeder based on neural network

Abstract

The invention discloses a waste incineration feeder intelligent control system implementation method based on a neural network, and the method comprises the following steps: S1, obtaining historical data of a waste incineration power plant, carrying out the preprocessing of the historical data, and dividing a preprocessed data set into a training set and a test set; s2, based on a neural network structure, constructing a feeder intelligent control model; s3, training the feeder intelligent control model by using the training set, evaluating the feeder intelligent control model by using the test set, and obtaining an optimal model, an optimal feeder stroke deviation and a trolley speed parameter; s4, the optimal feeder stroke deviation and the trolley speed sum are transmitted to an on-site DCS controller, and the DCS controller analyzes parameters and generates corresponding control instructions; and the feeder control mechanism adjusts the speed and the stroke deviation of the feeder trolley according to the control instruction. According to the method, the optimal debugging parameters can be automatically recommended, and the data analysis workload of debugging personnel is effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of waste incineration power generation, and in particular to a method for realizing an intelligent control system of a waste incineration feeder based on a neural network. Background Art

[0002] The growth in the number of waste-to-energy plants has led to several challenges, most notably insufficient waste supply, which has left many incinerators starving and unable to operate at full capacity. Furthermore, the diverse and complex nature of waste sources, including the co-incineration of food waste, sewage sludge, and general industrial solid waste, has further exacerbated the challenges of capacity utilization for waste-to-energy plants.

[0003] Before the intelligent feeder control system could be officially implemented at a waste-to-energy plant, commissioning engineers had to undertake a crucial preliminary task: they collected comprehensive historical data, meticulously screened it manually to eliminate outliers, and thoroughly analyzed the data, plotting multiple operating curves for comparative analysis. Drawing on their extensive commissioning experience, the engineers were able to initially determine the appropriate range for the commissioning parameters. Based on this, they manually evaluated key indicators such as boiler load status and furnace temperature to set initial values for the pusher trolley's speed and stroke deviation parameters. These parameters were then input into the pusher DCS control system, and multiple manual adjustments were performed to achieve the goal of intelligent feeder control. However, this over-reliance on manual trial and error and empirical adjustments was not only inefficient and labor-intensive, but also often struggled to achieve the desired control results due to its subjectivity and empirical limitations.

[0004] Over time, fluctuations in core operating parameters such as waste calorific value, composition, and moisture content can cause the original control parameters and logic to gradually deviate from on-site operating conditions, thereby weakening the performance of the feeder's intelligent control system. To address this issue, commissioning personnel must revisit the site and conduct an in-depth analysis of the incinerator's operating data from the past two weeks. Drawing on their experience, they analyze the deviations between the original parameters and logic and the current combustion conditions, allowing them to make targeted adjustments to the commissioning parameter ranges. Only after multiple manual adjustments and revisions can intelligent control of the feeder be restored. Therefore, regular parameter calibration and optimization of the intelligent feeder control system is crucial. Summary of the Invention

[0005] Purpose of the invention: The purpose of the present invention is to provide a waste incineration feeder intelligent control system based on neural network, which is used to optimize the feeding control during the waste incineration process, reduce the workload of debugging personnel, and improve combustion efficiency and stability.

[0006] Technical solution: A method for implementing an intelligent control system for a waste incineration feeder based on a neural network, comprising the following steps:

[0007] S1, obtain historical data of the waste incineration power plant and preprocess the historical data, and divide the preprocessed data set into a training set and a test set;

[0008] S2, based on the neural network structure, builds the feeder intelligent control model;

[0009] S3, use the training set to train the feeder intelligent control model, and use the test set to evaluate the feeder intelligent control model to obtain the optimal model, optimal feeder stroke deviation and trolley speed parameters;

[0010] S4, transmits the optimal feeder stroke deviation and trolley speed to the on-site DCS controller, and the DCS controller analyzes the parameters and generates corresponding control instructions; the feeder control mechanism adjusts the speed and stroke deviation of the feeder trolley according to the control instructions.

[0011] Furthermore, the historical data includes boiler load, furnace temperature, material layer thickness setting value, left material layer thickness measurement value, middle material layer thickness measurement value, and right material layer thickness measurement value data;

[0012] The acquired historical data is cleaned, and the regression filling method is used to fill in the missing data caused by the interruption of data collection.

[0013] Furthermore, the average value of the historical data of the garbage layer thickness in the previous week is taken as the set value of the material layer thickness. The speed range of the pushing trolley is 0.2~2.0mm / s. The difference between the measured value of the material layer thickness and the set value of the material layer thickness is -0.05~0.05m. The basic value of the trolley stroke is 800mm.

[0014] Furthermore, in the feeder intelligent control model, the input parameters are: boiler load, furnace temperature, material layer thickness set value, left material layer thickness measurement value, middle material layer thickness measurement value, right material layer thickness measurement value, feeding speed, combustion speed, boiler oxygen content, and air bag pressure; the output parameters are: feeder stroke deviation and feeder trolley speed parameter.

[0015] Furthermore, in step S4, when there is a deviation between the measured material layer thickness and the set material layer thickness, the feeder intelligent control model is optimized.

[0016] Compared with the prior art, the present invention has the following significant effects:

[0017] 1. This invention builds an intelligent feeding control model based on the incinerator's operating data over the past two weeks, enabling automatic recommendation of optimal commissioning parameters. This effectively reduces the data analysis workload for commissioning personnel, significantly improves the accuracy of recommended parameters, and significantly accelerates the commissioning process at the project site.

[0018] 2. The present invention optimizes the parameter configuration of the intelligent feeding control system based on historical operating condition data, and automatically recommends the best debugging parameters. By regularly adjusting the parameters of the intelligent control system of the waste incineration feeder, it not only significantly reduces the workload of the debugging personnel, but also greatly improves the combustion efficiency and effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is the structure diagram of the intelligent control system for waste incineration feeder based on neural network;

[0020] Figure 2 is a flow chart of the present invention;

[0021] Figure 3 This is the modeling flow chart of the feeder intelligent control model;

[0022] Figure 4 Schematic diagram of the neural network model structure. DETAILED DESCRIPTION

[0023] The present invention will be described in further detail below with reference to the accompanying drawings and specific implementations.

[0024] like Figure 1 The figure shows the structure of the intelligent control system for waste incineration feeders based on neural networks, including feeder intelligent control model, underlying control system and field instruments. The intelligent control system for waste incineration feeders based on neural networks is based on the combustion principle of waste and the process of incinerators. It uses advanced detection technology and model control technology to train the neural network model through the data set of the project's nearly two weeks of operation. Based on the obtained neural network model, the recommended debugging parameters are obtained: feeder stroke deviation and feeder trolley speed parameters, which monitor the material layer thickness and combustion status in the incinerator in real time to achieve continuous and stable operation of waste incineration, make up for the lag caused by manual operation, and reduce the labor intensity of operators. The flow chart of the implementation method of the intelligent control system for waste incineration feeders based on neural networks is shown below. Figure 2 As shown, the detailed implementation process includes the following steps:

[0025] Step 1: Obtain historical data of the waste incineration power plant and preprocess the historical data;

[0026] The boiler load, furnace temperature, material layer thickness setting value, left material layer thickness measurement value, middle material layer thickness measurement value, and right material layer thickness measurement value data of the waste incineration power plant in the past two weeks were collected to obtain the data volume corresponding to the historical time length of each type of data of the waste incinerator.

[0027] The acquired historical data is cleaned. For missing data caused by data collection interruptions, this invention uses a regression-based infill method to restore the missing data with minimal distortion. For data anomalies caused by field instrument failures, system acquisition signal delays, and other reasons, the anomalous data needs to be processed. This invention treats outliers as missing values.

[0028] Step 2: Build a feeder intelligent control model;

[0029] In this embodiment, the feeder intelligent control model adopts a neural network structure, such as Figure 4 The figure shows a schematic diagram of the neural network model structure of the present invention, wherein the input parameters x_1, x_2, ..., x_n are respectively: boiler load, furnace temperature, material layer thickness set value, left material layer thickness measurement value, middle material layer thickness measurement value, right material layer thickness measurement value, feeding speed, incineration speed, boiler oxygen content, gas bag pressure and other parameters; the output parameters z_1 and z_2 are respectively: feeder stroke deviation and feeder trolley speed parameters.

[0030] y_1, y_2, …, y_m represent hidden layers.

[0031] The present invention adopts the gradient descent method to obtain the optimal model, and the formula is as follows:

[0032]

[0033] Among them, α is the step size, θ is the unknown coefficient to be solved, and J(θ) is the loss function with the minimum variance of θ.

[0034] The present invention adopts the activation function Sigmoid, which is a common S-shaped function in biology. The neural network model needs to re-update the model parameters according to the specific process and sampling time to adapt to the changes in working conditions.

[0035]

[0036] Figure 3 This is the modeling flow chart of the feeder intelligent control model. The preprocessed data set is used to train the neural network model. First, the preprocessed data set is divided into a training set and a test set. Then, the training set is used to train the neural network model, and the test set is used to evaluate the established model. The above steps are repeatedly performed to optimize the model parameters and obtain the optimal model and the optimal feeder stroke deviation and feeder trolley speed parameters.

[0037] The average value of the historical data of the garbage layer thickness in the previous week is taken as the set value of the material layer thickness.

[0038] The speed range of the pushing trolley is 0.2~2.0mm / s, the difference between the measured value of the material layer thickness and the set value of the material layer thickness is -0.05~0.05m, and the basic value of the trolley stroke is 800mm.

[0039] The set value of the material layer thickness of the waste incinerator, the measured value of the material layer thickness on the left, the measured value of the material layer thickness in the middle, the measured value of the material layer thickness on the right, the furnace temperature, the boiler load, the gas bag pressure, the boiler oxygen content, the feeding grate speed, and the incineration grate speed are used as input data, and the pusher trolley speed and stroke deviation are used as output. The S-type function is used as the activation function, and the gradient descent method is used in the parameter space to minimize the loss function. The feeding grate model is trained and optimized to obtain the optimal feeder trolley speed and stroke deviation.

[0040] Step 3: Send the obtained optimal feeder trolley speed and stroke deviation to the DCS controller to dynamically adjust the feeder trolley speed and stroke deviation;

[0041] The optimal feeder trolley speed and stroke deviation are transmitted to the on-site DCS (Distributed Control System) controller, which analyzes the parameters and generates corresponding control instructions. The feeder control mechanism adjusts the feeder trolley speed and stroke deviation according to the control instructions. And based on whether the measured material layer thickness is close to the set material layer thickness, it is determined whether the feeder intelligent control model should be optimized to achieve intelligent control of the feeder, reduce manual workload, and improve the load stability of the incinerator. Among them, the set value of the material layer thickness is set by the operator according to the tail flame (combustion) situation.

[0042] When the deviation between the measured material layer thickness and the set material layer thickness is greater than 0.05m, the feeder intelligent control model is optimized.

[0043] Step 4: During the system operation period, the operating parameters and actual control effects are continuously collected, and the control effects are evaluated against the currently preset parameters such as feeding speed and deviation. When the actual operating conditions deviate from the preset parameters, the system automatically alarms and gives new recommended empirical parameters. The technical staff remotely evaluates the operating effects and analyzes the feasibility of the empirical parameters, thereby realizing self-diagnosis of the intelligent feeding control system and self-tuning under human supervision.

Claims

1. A method for realizing an intelligent control system of a waste incineration feeder based on a neural network, characterized in that: The steps are as follows: S1, obtain historical data of the waste incineration power plant and preprocess the historical data, and divide the preprocessed data set into a training set and a test set; S2, based on the neural network structure, builds the feeder intelligent control model; S3, use the training set to train the feeder intelligent control model, and use the test set to evaluate the feeder intelligent control model to obtain the optimal model, optimal feeder stroke deviation and trolley speed parameters; S4, transmits the optimal feeder stroke deviation and trolley speed to the on-site DCS controller, and the DCS controller analyzes the parameters and generates corresponding control instructions; the feeder control mechanism adjusts the speed and stroke deviation of the feeder trolley according to the control instructions.

2. The method for realizing the intelligent control system of the waste incineration feeder based on neural network according to claim 1 is characterized in that: The historical data includes boiler load, furnace temperature, material layer thickness setting value, left material layer thickness measurement value, middle material layer thickness measurement value, and right material layer thickness measurement value data; The acquired historical data is cleaned, and the regression filling method is used to fill in the missing data caused by the interruption of data collection.

3. The method for realizing the intelligent control system of the waste incineration feeder based on neural network according to claim 2 is characterized in that: The average value of the historical data of the garbage layer thickness in the previous week is taken as the set value of the material layer thickness. The speed range of the pushing trolley is 0.2~2.0mm / s. The difference between the measured value of the material layer thickness and the set value of the material layer thickness is -0.05~0.05m. The basic value of the trolley stroke is 800mm.

4. The method for realizing the intelligent control system of the waste incineration feeder based on neural network according to claim 1 is characterized in that: In the feeder intelligent control model, the input parameters are: boiler load, furnace temperature, material layer thickness set value, left material layer thickness measurement value, middle material layer thickness measurement value, right material layer thickness measurement value, feeding speed, combustion speed, boiler oxygen content, and air bag pressure. The output parameters are: feeder stroke deviation and feeder trolley speed parameter.

5. The method for realizing the intelligent control system of the waste incineration feeder based on neural network according to claim 1 is characterized in that: In step S4, when there is a deviation between the measured material layer thickness and the set material layer thickness, the feeder intelligent control model is optimized.

Images