CN115329659A - Method and system for real-time early warning and intelligent control of dioxin emission in waste incinerator - Google Patents
Method and system for real-time early warning and intelligent control of dioxin emission in waste incinerator Download PDFInfo
- Publication number
- CN115329659A CN115329659A CN202210845636.0A CN202210845636A CN115329659A CN 115329659 A CN115329659 A CN 115329659A CN 202210845636 A CN202210845636 A CN 202210845636A CN 115329659 A CN115329659 A CN 115329659A
- Authority
- CN
- China
- Prior art keywords
- dioxin
- prediction
- control system
- model
- flue gas
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F23—COMBUSTION APPARATUS; COMBUSTION PROCESSES
- F23G—CREMATION FURNACES; CONSUMING WASTE PRODUCTS BY COMBUSTION
- F23G5/00—Incineration of waste; Incinerator constructions; Details, accessories or control therefor
- F23G5/50—Control or safety arrangements
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F23—COMBUSTION APPARATUS; COMBUSTION PROCESSES
- F23N—REGULATING OR CONTROLLING COMBUSTION
- F23N3/00—Regulating air supply or draught
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
- G08B21/02—Alarms for ensuring the safety of persons
- G08B21/12—Alarms for ensuring the safety of persons responsive to undesired emission of substances, e.g. pollution alarms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/08—Fluids
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/08—Thermal analysis or thermal optimisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
Abstract
The invention relates to a dioxin control technology, and aims to provide a method and a system for real-time early warning and intelligent control of dioxin emission in a waste incinerator. The method comprises the following steps: constructing an LSTM-ARIMA fusion prediction model, and using historical data of dioxin in the flue gas of the waste incinerator, which is acquired by a dioxin on-line detection device, as the input of the fusion prediction model; taking the emission data of the waste incinerator obtained by the online dioxin detection device in real time as the input of a fusion prediction model, and outputting a prediction result after fusion calculation, wherein the prediction result is the emission data of the dioxin converted into trichlorobenzene concentration after a set time period; and comparing the prediction result with a preset index, and if the prediction result exceeds the preset index range, sending out a dioxin exceeding early warning. The method realizes the warning of the excessive dioxin emission based on the advanced prediction of the dioxin concentration, solves the problem of hysteresis of the detection of the dioxin generated by burning the wastes, and effectively prevents the phenomenon of the excessive dioxin emission; it is possible to realize the associated control with the waste incinerator operation control system.
Description
Technical Field
The application relates to a dioxin control technology, in particular to a method and a system for real-time early warning and intelligent control of dioxin emission in a waste incinerator.
Background
The waste incineration treatment technology is the most effective disposal method due to the advantages of volume reduction, rapid reduction and the like. However, dioxin persistent organic pollutants are easily generated by waste incineration, and dioxin has strong carcinogenic and teratogenic effects and can poison reproductive systems, immune systems and endocrine systems.
The existing dioxin control method has serious hysteresis, dioxin emission detection is mainly 'off-line sampling-laboratory analysis detection', the detection period is at least one week long, the time domain of detection data and operation parameters cannot be matched, the real-time early warning value is lacked, and quick feedback and adjustment are difficult under abnormal working conditions. Zhejiang university has developed a method for on-line rapid detection of dioxin in incineration flue gas (patent number: ZL 200510023260.1), and the dioxin emission data in flue gas can be obtained in about 15 minutes, which is the fastest dioxin detection method in the world at present, but a plurality of processes and flue gas treatment equipment exist from the waste incineration front end to the flue gas emission tail end, and a period of control vacuum period still exists from the acquisition of detection data to the feedback control, and the dioxin emission cannot be effectively controlled in the period exceeding the standard. If a real-time dioxin early warning technology can be developed based on the online rapid dioxin detection data, the excessive dioxin emission is early warned in real time, the occurrence of a vacuum period is effectively prevented and controlled, and the excessive dioxin phenomenon is reduced.
In addition, the generation and emission of dioxin are influenced by various factors, including material types, combustion abundance, flue dust deposition, oxygen amount, activated carbon adsorption capacity, dust particles, hydrogen chloride concentration and the like, and the low-emission control of dioxin requires the mutual coupling implementation of multiple links such as combustion, ash removal, dust removal, online detection and the like. However, in the conventional control technology, operators can only adjust abnormal working conditions such as excessive pollutants according to operation experience, the operation level of the operators is excessively depended on, the operation levels of different operators are uneven, the fluctuation of the operation working conditions is large, and the stability of the working conditions, the emission of pollutants and the operation efficiency are influenced. Meanwhile, due to a plurality of factors influencing the generation of dioxin, the control targets of the traditional control technology are all aimed at controlling pollutants, and other targets such as economy, stability and the like are difficult to consider.
If a multi-objective optimization control technology can be developed based on dioxin forecast data and operation condition monitoring data, indexes such as economy and environmental assessment benefits are integrated into a control process, and the operation control level of an enterprise is effectively improved.
Disclosure of Invention
The invention aims to solve the technical problem of overcoming the defects in the prior art and provides a real-time early warning and intelligent control system for dioxin discharge of a waste incinerator
In order to solve the technical problem, the solution of the invention is as follows:
the method for giving real-time warning on dioxin emission in the waste incinerator comprises the following steps:
(1) Constructing an LSTM-ARIMA fusion prediction model based on an LSTM (long-short term memory neural network) model and an ARIMA (autoregressive moving average) model; wherein, the output results of the LSTM model and the ARIMA model are dioxin emission data converted into trichlorobenzene concentration; the two models are fused in a way that the prediction results of the two models are respectively added by taking a weight coefficient of 0.5;
(2) Historical data of dioxin in flue gas of a waste incinerator, which is acquired by a dioxin on-line detection device, is used as input of a fusion prediction model; wherein, the input variables of the LSTM model are as follows: the average temperature of the flue gas at the outlet of the first channel, the total air quantity of secondary air, the total air quantity of primary air, the pressure of primary air, the flow rate of the flue gas and dioxin emission data converted into trichlorobenzene concentration in the first 5 hours; input variables for the ARIMA model include: the average temperature of the flue gas at the outlet of the first channel, the total air quantity of secondary air, the total air quantity of primary air, the pressure of primary air and the flow rate of the flue gas in the first 5 hours;
(3) Taking a dioxin detection value and smoke average temperature, air volume, air pressure and smoke flow data in smoke discharged by a waste incinerator, which are acquired by a dioxin online detection device in real time, as input of an LSTM-ARIMA fusion prediction model, outputting a prediction result after fusion calculation, wherein the prediction result is dioxin discharge data converted into trichlorobenzene concentration after a set time period (such as 1 hour);
(4) And comparing the prediction result with a preset index, and if the prediction result exceeds the preset index range, sending out a dioxin exceeding early warning.
As a preferred scheme of the invention, the LSTM model adopts 1 LSTM +1 Dense architecture, the number of LSTM layer neurons is 150, and the Relu function is selected as an activation function; the number of neurons in a Dense full junction layer is 1; initializing parameters of each layer by adopting a random normal function, wherein the seed of a random seed is 6; the input time step of the input layer is 5, and the prediction step of the output layer is 1; the number p of autoregressive terms, the difference times d and the number q of moving average terms in the ARIMA model parameters are respectively 1,2 and 1, and the prediction step length is 1.
The invention further provides an intelligent control method for dioxin emission in a waste incinerator, which comprises the following steps:
(1) Based on an economic objective calculated by the running cost of the incinerator and an environmental benefit objective calculated by predicting emission data based on dioxin, a multi-objective optimization function F (x) is constructed according to the following formula, and the optimization objective is a minimization function F (x):
min F(x)＝[f 1 (x)，f 2 (x)]
in the formula, f 1 (x) For the economic objective function calculated from the running cost of the incinerator, f 2 (x) Calculating an environmental benefit objective function according to dioxin emission prediction data; the dioxin emission prediction data refer to a prediction result output by the fusion prediction model in the step (3) of the method;
(2) Setting constraint conditions of all control variables in the multi-objective optimization function, and then solving the multi-objective optimization function to obtain an optimal control state point of the operation parameters for intelligently controlling the waste incinerator;
(3) And adjusting the working conditions of a pretreatment control system, an incineration control system and a flue gas purification control system of the waste incinerator by using the data of the optimal control state point of the operation parameter through an automatic control mode or an auxiliary control mode, so that the working conditions are kept within a specified range while the pretreatment control system, the incineration control system and the flue gas purification control system operate in the optimal working conditions.
As a preferable aspect of the present invention, in the step (1), the operation cost of the incinerator is calculated according to the following formula:
in the formula: n is the total number of controllable variables, and the controllable variables comprise the dosage of feeding materials, primary air, secondary air, water supply and active carbon and the pressure difference of the cloth bag;
θ i is the unit cost of the ith controlled variable; wherein, the unit cost (yuan/kg) of the feeding amount is calculated according to the power/electricity price of the crane grab, and the unit cost (yuan/m) of the air volume 3 ) Calculated for the real-time power of the fan multiplied by the specification of the fan/electricity price, the unit cost of water amount (yuan/m) 3 ) The unit cost of the activated carbon is obtained by calculating (unit price of water plus real-time power of a water supply pump multiplied by specification of the water pump/electricity price), the unit cost of the activated carbon is obtained by purchasing price (yuan/kg), and the unit cost of pressure difference of the cloth bag (yuan/pa) is the cost of replacing the cloth bag/(allowance of the cloth bag)Maximum differential pressure-design differential pressure);
x i real-time monitoring values for the ith controllable variable, including feeding amount (kg/min) and primary air volume (m) 3 H) secondary air flow (m) 3 H), water supply quantity (m) 3 H), the usage amount (kg) of the active carbon and the pressure difference pa of the cloth bag.
As a preferable aspect of the present invention, in the step (1), the predicted emission amount of dioxin is calculated according to the following formula:
in the formula:
k i is the influence coefficient of the ith controlled variable on the concentration of dioxin and is used for representing the controlled variable x i Ability to regulate dioxins; the numerical value is an empirical coefficient and ranges from 0 to 1;
lambda is dioxin exceeding early warning input, is 1 when exceeding early warning is carried out, and is 0 otherwise;
As a preferred scheme of the invention, in the step (2), when the multi-objective optimization function is solved, an NSGA or PSO intelligent algorithm is applied to solve multiple Pareto solution sets of the function.
As a preferred embodiment of the invention, the control variables x i The constraint conditions of (1) include: the lowest feeding amount; the maximum air quantity and the minimum air quantity of the primary air and the secondary air; the ratio of primary air to secondary air; the maximum input amount of the activated carbon.
The invention also provides a real-time early warning and intelligent control system for dioxin emission of the waste incinerator, which comprises a pretreatment control system, an incineration control system and a flue gas purification control system, wherein the pretreatment control system is used for controlling the operation of the waste incinerator; further comprising:
the online dioxin detection device is arranged at the extreme end position (such as a chimney inlet) of a waste incinerator flue and is used for acquiring the content of dioxin in the flue gas discharged by the waste incinerator in real time;
a dioxin emission value prediction module for performing fusion calculation of the LSTM-ARIMA fusion prediction model of claim 1 and outputting a prediction result;
the dioxin emission early warning module is used for comparing the prediction result of the fusion prediction model with a preset index and sending out a dioxin exceeding early warning according to the comparison result;
an intelligent control module for dioxin discharge, which is used for executing the intelligent control method for dioxin discharge in the waste incinerator of claim 3 and sending control signals or operation action prompts to the pretreatment control system, the incineration control system and the flue gas purification control system.
Compared with the prior art, the invention has the beneficial effects that:
1. according to the method, the dioxin emission standard exceeding early warning is realized by utilizing the dioxin on-line monitoring device and the LSTM-ARIMA fusion prediction model based on the advanced prediction of the dioxin concentration, the problem of hysteresis of waste incineration dioxin detection is solved, and the phenomenon that the dioxin emission standard exceeds is effectively prevented.
2. The intelligent control system realizes the correlation control of intelligent control of dioxin emission and a pretreatment control system, an incineration control system and a flue gas purification control system for controlling the operation of a waste incinerator based on the early warning of the excessive dioxin emission. Factors such as economy, environmental benefit and the like are considered while stable operation is guaranteed, technical propagability is high, dependence of incineration control on operators is reduced, and enterprise economy is improved.
Drawings
Fig. 1 is a block diagram of a real-time warning and intelligent control system for dioxin discharge from a waste incinerator according to the present invention.
Fig. 2 is a block diagram of a real-time dioxin warning module in the embodiment of the present invention;
fig. 3 is a block diagram of a dioxin intelligent control module in the embodiment of the present invention;
FIG. 4 is a logic diagram of an intelligent control module for dioxin in the embodiment of the invention;
FIG. 5 is a diagram of the LSTM1+ ARIMA2 prediction model in the embodiment of the present invention;
FIG. 6 is a diagram illustrating the variation of the loss function when the LSTM1 model and the LSTM2 model are trained in the embodiment of the present invention;
FIG. 7 is a predicted fit curve of the LSTM1 model, the ARIMA2 model and the LSTM1+ ARIMA2 model on the test set in the embodiment of the present invention.
FIG. 8 is a diagram of the structure of input data of the LSTM model during model training according to an embodiment of the present invention.
FIG. 9 is a diagram of the structure of input data of the ARIMA model during model training according to the embodiment of the invention.
Detailed Description
The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. It is obvious that the described embodiments are only the most basic embodiments of the invention, not all embodiments. Other embodiments based on the invention belong to the protection scope of the invention.
The invention provides a real-time warning method for dioxin discharge in a waste incinerator, which comprises the following steps:
(1) And constructing the LSTM-ARIMA fusion prediction model based on the LSTM model and the ARIMA model. Wherein, the output results of the LSTM model and the ARIMA model are dioxin emission data converted into trichlorobenzene concentration; the two models are fused in a way that the prediction results of the two models are respectively added by taking a weight coefficient of 0.5;
the LSTM model adopts 1 LSTM +1 Dense architecture, the number of LSTM layer neurons is 150, and a Relu function is selected as an activation function; the number of neurons in a Dense full junction layer is 1; initializing parameters of each layer by adopting a random normal function, wherein the seed of a random seed is 6; the input time step of the input layer is 5, and the prediction step of the output layer is 1; the number p of autoregressive terms, the difference times d and the number q of moving average terms in the ARIMA model parameters are respectively 1,2 and 1, and the prediction step length is 1.
The ARIMA model input contains trichlorobenzene concentration as input target. And respectively taking the weight coefficients as the output of the final model, namely the final predicted trichlorobenzene concentration. Before data is input, the data is standardized, and corresponding inverse standardization processing is generally performed to restore the true value of the data, and the standardization is only the change of the data dimension and does not change the information of the data, so that the result after the addition is considered as the final prediction result.
(2) Historical data of dioxin in flue gas of a waste incinerator, which is acquired by a dioxin on-line detection device, is used as input of a fusion prediction model; wherein, the input variables of the LSTM model are as follows: the average temperature of the flue gas at the outlet of the first channel, the total air quantity of secondary air, the total air quantity of primary air, the pressure of primary air, the flow rate of the flue gas and dioxin emission data converted into trichlorobenzene concentration in the first 5 hours; input variables for the ARIMA model include: average temperature of flue gas at the outlet of the first channel, total air quantity of secondary air, total air quantity of primary air, pressure of primary air and flow rate of flue gas in the first 5 hours;
(3) Using a dioxin detection value in flue gas discharged by a waste incinerator and flue gas average temperature, air volume, air pressure and flue gas flow data which are obtained by a dioxin on-line detection device in real time as input of an LSTM-ARIMA fusion prediction model, outputting a prediction result after fusion calculation, wherein the prediction result is the dioxin discharge data which is converted into trichlorobenzene concentration after a set time period (such as 1 hour);
(4) And comparing the prediction result with a preset index, and if the prediction result exceeds the preset index range, sending out a dioxin exceeding early warning.
The invention provides an intelligent control method for dioxin emission of a waste incinerator, which comprises the following steps:
(1) Based on an economic objective calculated by the running cost of the incinerator and an environmental benefit objective calculated by predicting emission data based on dioxin, a multi-objective optimization function F (x) is constructed according to the following formula, and the optimization objective is a minimization function F (x):
min F(x)＝[f 1 (x)，f 2 (x)]
in the formula (f) 1 (x) For an economic objective function calculated from the running cost of the incinerator, f 2 (x) For environments calculated from predicted emission data of dioxinsA benefit objective function; the dioxin emission prediction data are prediction results output by the fusion prediction model in the step (3) of the method in claim 1;
(1.1) calculating the running cost of the incinerator according to the following formula:
in the formula: n is the total number of controllable variables, and the controllable variables comprise the dosage of feeding, primary air, secondary air, water supply and active carbon and the pressure difference of a cloth bag;
θ i is the unit cost of the ith controlled variable; wherein, the unit cost (yuan/kg) of the feeding amount is calculated according to the power/electricity price of the crane grab, and the unit cost (yuan/m) of the air volume 3 ) Calculated for the real-time power of the fan multiplied by the specification of the fan/electricity price, the unit cost of water volume (yuan/m) 3 ) The unit cost of the activated carbon is obtained by calculating (unit price of water consumption + real-time power of a water supply pump multiplied by specification of the water pump/electricity price), the unit cost of the pressure difference of the cloth bag (unit/pa) is the cost of replacing the cloth bag/(the maximum pressure difference allowed by the cloth bag-design pressure difference);
x i real-time monitoring values of the ith controllable variable, including feeding amount (kg/min) and primary air volume (m) 3 H) secondary air volume (m) 3 H), water supply (m) 3 H), the usage amount (kg) of the active carbon and the pressure difference pa of the cloth bag.
(1.2) calculating the predicted emission of dioxin according to the following formula:
in the formula:
k i is the influence coefficient of the ith controlled variable on the concentration of dioxin and is used for representing the controlled variable x i Ability to modulate dioxins; the numerical value is an empirical coefficient and ranges from 0 to 1;
lambda is dioxin exceeding warning input (namely an input value when a prediction result is compared with a preset index), and is 1 when exceeding warning is performed, or is 0;
In steps (1.1) and (1.2), the respective control variable x i The constraint conditions of (1) include: the lowest feeding amount; the maximum air quantity and the minimum air quantity of the primary air and the secondary air; the ratio of primary air to secondary air; the maximum input amount of the activated carbon.
(2) Setting constraint conditions of all control variables in the multi-objective optimization function, and then solving the multi-objective optimization function to obtain an optimal control state point of an operation parameter for intelligently controlling the waste incinerator; when the multi-objective optimization function is solved, a multi-objective genetic algorithm (NSGA) or a Particle Swarm Optimization (PSO) algorithm is applied to solve multiple groups of effective solution (Pareto solution) sets of the function.
(3) And adjusting the working conditions of a pretreatment control system, an incineration control system and a flue gas purification control system of the waste incinerator by using the data of the optimal control state points of the operation parameters through an automatic control mode or an auxiliary control mode, so that the working conditions are kept within a specified range while the pretreatment control system, the incineration control system and the flue gas purification control system are operated in the optimal working conditions.
A more detailed description follows:
as shown in fig. 1, the real-time early warning and intelligent control system for dioxin discharge in a waste incinerator comprises a pretreatment control system, an incineration control system and a flue gas purification control system for controlling the operation of the waste incinerator; further comprising: the dioxin on-line detection device is arranged at the tail end of a flue of the waste incinerator (such as at a chimney inlet) and is used for acquiring the content of dioxin in the exhaust smoke of the waste incinerator in real time. The dioxin on-line detection device can detect dioxin on line by combining laser ionization with time-of-flight mass spectrometry, and reference can be made to the prior patent technology (a method for monitoring dioxin on line by combining tunable laser spectrum with time-of-flight mass spectrometry, patent number ZL 200510023260.1). And the dioxin emission value prediction module is used for training an LSTM-ARIMA fusion prediction model according to the historical dioxin data, carrying out fusion calculation based on real-time monitoring data and outputting a prediction result to realize advanced prediction of dioxin emission, and outputting the result to the dioxin emission early warning module. The dioxin emission early warning module is used for comparing the prediction result of the fusion prediction model with a preset index and sending out a dioxin exceeding early warning according to the comparison result; the intelligent control module for dioxin emission constructs a multi-objective optimization function fused with economy and environmental benefits according to the early warning information, the equipment operation data, the combustion detection data and the dioxin emission data, intelligently adjusts according to the function optimal solution set, and optimally controls the operation of the incinerator and the dioxin emission; and sending control signals or operation action prompts to the pretreatment control system, the incineration control system and the flue gas purification control system.
The dioxin emission value prediction module utilizes an LSTM-ARIMA fusion prediction model based on the dioxin variable optimization processing, can predict the dioxin emission concentration converted from trichlorobenzene concentration within set time (such as 1 hour in the future), and then compares real-time dioxin emission data with the prediction data to check the prediction capability of the optimization model. The prediction model is characterized in that dioxin data are used as variables to train an LSTM (long short term memory neural network) model, while the ARIMA (autoregressive moving average) model is trained by only using the dioxin data as a target value, and meanwhile, the results of the LSTM and the ARIMA model are finally fused, so that the prediction precision is improved.
An example of the implementation steps of the LSTM-ARIMA fusion prediction model is:
(1) And respectively constructing an ARIMA prediction model and an LSTM prediction model.
The LSTM-ARIMA fusion mode is various, the invention modifies and creates the model adaptively, combines the characteristics of waste incineration and screens and optimizes the data in the incineration process; and the data and the model are matched with each other through feature screening, so that a better effect is achieved. In addition, the invention performs fusion by distributing the weight to the model result, wherein the weight is 0.5 in the embodiment, the LSTM model has better time sequence information capture capability to the time sequence, while the ARIMA model is a traditional statistical model and is commonly used for trend prediction of the time sequence, and the ARIMA model and the time sequence have a little complementarity, so the prediction effect after fusion is better.
(2) And (3) respectively converting the historical data into input forms required by an ARIMA model and an LSTM model, and training the models. Particularly, based on a comparison experiment, when dioxin is found to be an input variable, the prediction capability of the LSTM model can be enhanced, but the prediction effect of the ARIMA model can be weakened, so that the input variable of the LSTM model contains historical dioxin data, the ARIMA model input variable does not contain historical dioxin data, and the specific input variable of the LSTM model is as follows: the average temperature of the flue gas at the outlet of the first channel in the first 5 hours, the total air quantity of secondary air, the total air quantity of primary air, the pressure of primary air, the flow rate of flue gas and the converted concentration of trichlorobenzene, wherein input variables of an ARIMA model are as follows: average temperature of flue gas at the outlet of the first channel, total air quantity of secondary air, total air quantity of primary air, pressure of primary air and flue gas flow in the first 5 hours. The predicted target of the model was trichlorobenzene reduced concentration (dioxin data) after 1 hour.
As shown in fig. 8 and 9, the input data includes two contents, i.e., an input variable x and an input target y. And the historical data of the dioxin in the ARIMA model is only used as an input target y and is used for verifying and optimizing the prediction capability of the model. The input contains the target y when the model is trained, but the input target does not contain y when the model is applied for prediction, but the predicted y value is output by the model.
(3) Inputting real-time data into an ARIMA model and an LSTM model which are trained, fusing prediction results of the two models respectively with a weight of 0.5 (the weight can be finely adjusted according to respective prediction effects of the models, and the sum of the weights of the two models is equal to 1) to obtain final prediction data, and judging the prediction effects according to MSE and MAE indexes. Experiments show that compared with the non-fusion, the MSE (mean square error) and MAE (mean absolute error) of the prediction result after fusion are obviously reduced.
And the dioxin emission early warning module compares the prediction result of the fusion prediction model with a preset index and sends out the dioxin exceeding early warning according to the comparison result. The key indexes of dioxin generation comprise residence time of smoke in an area above 850 ℃, terminal CO concentration, active carbon usage amount and the like, the monitoring values comprise predicted dioxin concentration, dioxin concentration change rate, delay time and the like, and when the key indexes of dioxin generation are abnormal or the monitoring values exceed standards, the module sends out early warning that the dioxin exceeds the standards. The delay time t is predicted that the concentration of dioxin will exceed a standard value after t min.
The key index of dioxin generation refers to a factor generally recognized in the field and having a great influence or correlation on the generation of dioxin, and can be directly obtained through a monitoring system of an incineration plant, a monitoring value is obtained by predicting and calculating the dioxin based on a prediction model, and the key index of dioxin generation is irrelevant to the monitoring value. The prediction model is only responsible for prediction, early warning is generated by the early warning module, the early warning module monitors key indexes generated by dioxin, calculates monitoring values, and generates early warning when corresponding values exceed set values.
The intelligent control module for dioxin emission sends out control signals or operation action prompts to the pretreatment control system, the incineration control system and the flue gas purification control system. The pretreatment control system is used for monitoring and controlling the normal operation of incinerator feeding equipment, including crushing, grabbing and hanging, conveying, feeding and the like; the incineration control system is used for monitoring and controlling the normal operation of the incinerator, and comprises incinerator temperature, incinerator pressure, flue gas flow rate, primary and secondary air proportion, tail end CO concentration and the like; the flue gas purification control system is used for monitoring and controlling the normal operation of tail flue gas purification equipment, and comprises cloth bag dust removal, activated carbon adsorption and the like.
The intelligent control comprises an automatic control mode and an auxiliary control mode at the same time. The automatic control mode is used for automatically controlling the equipment to operate according to the instruction of the intelligent control system, automatically controlling the pretreatment control system, the incineration control system and the flue gas purification control system to operate under the optimal working condition, and keeping the concentration of the dioxin in a required range. The auxiliary control mode is used for selecting a system needing manual control from a pretreatment control system, an incineration control system and a flue gas purification control system, an intelligent control system is used for providing a recommended operation scheme, and other control systems are still automatically controlled by the modules. The invention aims at the difficult problem that the dioxin emission of the waste incinerator is difficult to predict and control, and realizes the real-time early warning and intelligent control of the dioxin concentration.
The intelligent control content comprises abnormal judgment and intelligent analysis decision. The abnormal judgment refers to judging whether crushing, grabbing and hanging, furnace temperature and furnace pressure, ash removal and slag removal and the like are normal through logic operation, and adjusting abnormal equipment according to set logic operation; the intelligent analysis decision-making means that an optimization function (including economy, reliability, risk, environmental evaluation and other types of objective functions) is established for a control target, particularly, the optimization function is fused with dioxin early warning information lambda during construction, and when the time exceeding is predicted, the corresponding function is remarkably increased so as to cope with the hysteresis of the existing control; after the optimization function is determined, the optimal values of the control variables (feeding amount, primary air quantity, secondary air quantity, water supply amount and the like) are solved by using intelligent optimization algorithms such as NSGA (multi-objective genetic algorithm), PSO (particle swarm optimization algorithm) and the like, and are transmitted to a treatment control system, an incineration control system and a flue gas purification control system, so that real-time intelligent control is achieved.
The step of intelligently analyzing the decision comprises:
step one, constructing a multi-objective optimization function, specifically comprising the following objective functions
(1) Economic objective
The economy of the incinerator is related to the running cost, the garbage disposal amount, the incineration power generation amount and the like, and an economical objective function f is calculated according to the running cost of the incinerator 1 ：
In the formula:
n is the total number of controllable variables, wherein the controllable variables comprise feeding, primary air, secondary air, water supply, activated carbon, cloth bag pressure difference and the like;
x i the ith controllable variable real-time monitoring value comprises feeding amount (kg/min) and primary air amount (m) 3 H) secondary air volume (m) 3 H), water supply quantity (m) 3 H), the usage amount (kg) of the active carbon, the pressure difference (pa) of the cloth bag and the like;
θ i unit cost of the ith controlled variable, e.g. unit cost of air volume (yuan/m) 3 ) The real-time power of the fan is multiplied by the specification of the fan/electricity price, the unit cost (yuan/kg) of the feeding amount is calculated according to the power/electricity price of the crane grab, and the unit cost (yuan/m) of the water amount is calculated 3 ) The unit cost of the activated carbon is obtained by calculating (unit price of water consumption + real-time power of a water supply pump multiplied by specification of the water pump/electricity price), the unit cost of the activated carbon is obtained by purchasing price (yuan/kg), and the unit cost of the pressure difference of the cloth bag (yuan/pa) is the cost of replacing the cloth bag/(the maximum allowable pressure difference of the cloth bag-design pressure difference).
(2) Environmental benefit goal
The environmental benefit relates to the discharge amount of various pollutants, and an environmental benefit objective function f is constructed by predicting the discharge amount of dioxin 2 ：
In the formula:
lambda is the exceeding warning input of dioxin, the exceeding warning is 1, otherwise, the exceeding warning is 0, and when the exceeding warning exceeds the standard, the function value is obviously increased;
k i -the influence coefficient of the ith controlled variable on the concentration of dioxin, and the characterization of the controlled variable x i The dioxin regulating capacity and the empirical coefficient are generally 0 to 1;
Step two, setting each control variable x i The constraint conditions of (1) comprise the lowest feeding amount, the maximum air quantity and the minimum air quantity of primary air and secondary air, the proportion of the primary air and the secondary air, the maximum input amount of the active carbon and the like.
Step three, solving the multi-objective optimization function, wherein according to the constructed multi-objective optimization function F (x) and the constraint conditions thereof:
min F(x)＝[f 1 (x)，f 2 (x)]
s.t.x i min ＜x i ＜x i max
0＜x 3 /x 2 ＜0.5
and (4) minimizing F (x) and solving a plurality of Pareto solution sets of the functions by applying intelligent algorithms such as NSGA, PSO and the like.
Taking a Particle Swarm Optimization (PSO) algorithm as an example, the particle swarm optimization is described mathematically as follows: for a multi-objective optimization problem with n arguments, each particle j contains a position vector x that is n-dimensional j And velocity vector v j When the particle j searches the solution space, the optimal experience position pbest searched by the particle j is saved j . When each iteration starts, the particles adjust their own velocity vector according to their own inertia and experience and the optimal experience position of the group gbest to adjust their own position, and the solving process is:
(1) setting the number m of particles, and randomly initializing the speed v and the position x of the particles;
(2) calculating the adaptive value of each particle according to the objective function, and updating the individual optimal position pbest of the jth particle j And global optimal extremum position gbest;
(3) updating the particle speed v and the position x through formulas (1) and (2);
in the formula, c 1 And c 2 All acceleration factors reflect information exchange among particle swarms, and generally all acceleration factors are 2; t is expressed as the current iteration number; r is a radical of hydrogen 1 And r 2 Is in [0,1 ]]A uniform random number that varies over a range; pbest j Representing the optimal extreme position of the jth particle individual; gbest tableShown as the global optimum extremum location.
(4) And (3) judging whether the maximum iteration number is reached or the global optimal solution number reaches a set value, stopping iteration if the maximum iteration number is reached, outputting a final optimal solution set, and repeating the step (2) if the final optimal solution set is not reached.
Different from single-target optimization, the multi-target optimization may obtain more than one solution, and finally, the most appropriate solution set can be screened out according to management requirements (for example, the smaller the number of times of fan operation is, the better the smaller the variation value of fan power is, and the like), or by using methods such as multi-attribute decision making and the like
The optimal state point of the ith variable decided by the intelligent control system.
Specific application examples of the present invention:
description of (A) data set
The content of the data set is 6 months burning data of a certain garbage burning plant, and the data set has 533 samples and 28 characteristics, and the time interval of each sample is 1 hour. The person correlation coefficients of 28 features are calculated, 6 features of 'average temperature of flue gas at the outlet of a first channel, total air volume of secondary air, total air volume of primary air, primary air pressure, flue gas flow and trichlorobenzene reduced concentration (dioxin data)' are obtained after the features of which the correlation coefficients are more than 0.8 are fused, the 6 features are used for training and predicting in the training, a data set is divided into a training set and a testing set according to the proportion of 7: 3, and meanwhile Min-Max standardization is carried out on the data so as to remove unit limitation of the data and convert the data into dimensionless pure numerical values, so that indexes of different units or orders can be compared and weighted conveniently. The on-line dioxin value is converted from the trichlorobenzene converted concentration, that is, the on-line dioxin value = k × chlorobenzene converted concentration, and the correlation coefficient thereof is 1.
(II) introduction of model
The ARIMA model is called as an autoregressive moving average model, and the basic idea of the ARIMA model is as follows: the data sequence formed by a prediction object over time is regarded as a random sequence, a certain mathematical model is used for approximately describing the sequence, and ARIMA is an earlier prediction method and has application in many fields.
The LSTM model is a long-short term memory model and is a variant of a Recurrent Neural Network (RNN), and the LSTM model makes up the problems of gradient disappearance, gradient explosion, insufficient long-term memory capability and the like of the RNN, so that the recurrent neural network can really and effectively utilize long-distance time sequence information. The LSTM model has a few successful application cases in the research of time series data in different fields such as language modeling and voice recognition, but the LSTM model has less application in the field of industrial prediction.
The application of the model relates to the contents of the waste incineration field and the machine learning field, but related cross personnel in the industry are insufficient and are slow to develop. Traditional workers cannot well understand the characteristics of a machine learning model, and machine learning related workers cannot well understand the characteristics of waste incineration field data and the correlation between the data, so that massive operation data cannot be processed. The invention combines the characteristics of LSTM and ARIMA (the effectiveness of time sequence prediction), and can effectively predict dioxin after screening and processing the burning data characteristics according to the characteristics of waste burning, particularly the influence of the dioxin data as the input variable of the LSTM model but not as the input variable of the ARIMA model on the model prediction effect.
The ARIMA prediction model and the LSTM prediction model are applied in the experiment. For comparison, 5 models are constructed, namely LSTM1, LSTM2, ARIMA 1, ARIMA2, LSTM1+ ARIMA2, wherein the architecture and model parameters of LSTM1 and LSTM2 and the architecture and model parameters of ARIMA 1 and ARIMA2 are the same, and the difference is only whether the dioxin data is used as an input variable. The specific parameters of each LSTM model and ARIMA model in the experiment are as follows:
the LSTM model adopts 1 LSTM +1 Dense architecture, the number of LSTM layer neurons is 150, and the Relu function is selected as an activation function; the number of neurons in the Dense (fully-linked) layer is 1; initializing parameters of each layer by adopting a random normal function, wherein the seed of a random seed is 6; the input time step of the input layer is 5 (first 5 hours) and the prediction step of the output layer is 1 (last 1 hour).
The ARIMA model parameters (the number p of autoregressive terms, the difference degree d and the number q of moving average terms) are respectively (1, 2 and 1), and the prediction step length is 1.
1. Input data processing and training
The LSTM1 model input data contains features: the average temperature of the flue gas at the outlet of the first channel, the total air volume of secondary air, the total air volume of primary air, the pressure of the primary air, the flow rate of the flue gas and the converted concentration of trichlorobenzene; each training sample comprises 5 sample points, namely the input time step is 5; during model training, an Adam algorithm (adaptive momentum estimation algorithm) is adopted as an optimization algorithm, the learning rate is set to be 0.004, MSE (mean square error) is used as a loss function, and the iteration number (epochs) of sample training is 50.
The LSTM2 model input data contains features: the average temperature of the flue gas at the outlet of the first channel, the total air quantity of secondary air, the total air quantity of primary air, the pressure of the primary air and the flow of the flue gas, and other parameters are the same as those of the LSTM1 model.
The ARIMA 1 model input data comprise characteristic first channel outlet flue gas average temperature, secondary air total air quantity, primary air pressure, flue gas flow and trichlorobenzene conversion concentration, and all samples of a training set are used for regression analysis.
The ARIMA2 model input data comprises characteristic first channel outlet flue gas average temperature, secondary air total air quantity, primary air pressure and flue gas flow, and all samples of the training set are used for regression analysis.
The LSTM1+ ARIMA2 model is the fusion of the LSTM1 and ARIMA2 models, namely the predicted values of the two models are fused according to the weight, and the weight of the two models in the experiment is 0.5.
2. Results
After training is finished, the verification set data is used for verifying the model effect, the result is shown in table 1, when the dioxin data are used as input variables of the LSTM model, the MSE and MAE indexes of the result are obviously reduced, and the result shows that the dioxin data can well improve the prediction effect of the LSTM model, but the ARIMA model effect is opposite. Particularly, after two models with better effect, namely the LSTM1 and the ARIMA2, are fused, the prediction effect is better than that of the original model, so that the LSTM-ARIMA fusion prediction model based on feature screening is provided as the prediction model.
TABLE 1 results of the experiment
The parameters of each prediction model in the experiment are locally optimized, and the parameters of the prediction model in practical application are not limited to the parameters used in the experiment.
And when the model prediction accuracy does not meet the requirement, the prediction model is retrained until the requirement is met. And then comparing the real-time dioxin emission data with the prediction data, and detecting and optimizing the prediction capability of the model. And meanwhile, the dioxin prediction module transmits the prediction data to the dioxin concentration early warning module, and the module provides a dioxin real-time prediction data interface and can transmit the dioxin concentration prediction data outwards.
As shown in fig. 2, the real-time dioxin early warning module is used for monitoring the concentration of dioxin, and the settable monitoring value includes 70pg of TEQ/Nm of the concentration of dioxin 3 And a rate of change of dioxin concentration of 10pg TEQ/(Nm & lt/m & gt) 3 X min), delay time 10min (the delay time is that the concentration of dioxin is predicted to exceed a standard value after 10 min), and key indexes of dioxin generation are abnormally monitored and started (including residence time of flue gas in an area with the temperature of above 850 ℃, tail end CO concentration, usage amount of activated carbon, ash removal time interval, primary and secondary air volume and air temperature, integrity of a cloth bag, working condition of pretreatment equipment and the like). When any monitoring value exceeds the standard or a key index is abnormal, the real-time early warning module sends out dioxin exceeding early warning information, and the following are three types of early warning conditions for module early warning:
early warning condition 1: abnormal key indexes of dioxin generation, such as the CO concentration at the tail end exceeding 80mg/Nm when being monitored 3 And other indexes and monitoring values are normal, and at the moment, the module generates early warning information and outputs the early warning information to the intelligent control module.
Early warning condition 2: if the monitoring value exceeds the standard, the dioxin concentration exceeds the set 70pg TEQ/N after predicting 8min (less than 10min of delay time)m 3 And other indexes and monitoring values are normal, and at the moment, the module generates early warning information and outputs the early warning information to the intelligent control module.
Early warning condition 3: critical indexes of dioxin generation are abnormal and monitoring values exceed the standard, for example, the retention time of the smoke in a region above 850 ℃ is less than 2.5 seconds, and the change rate of the dioxin concentration is more than 10pg TEQ/(Nm & lt/m & gt) 3 X min), at this moment, the module generates early warning information and outputs the early warning information to the intelligent control module.
3. The intelligent analysis decision-making method comprises the following specific steps:
step one, constructing a multi-objective optimization function, specifically comprising the following objective functions
(1) Economic objective
The economy of the incineration plant relates to the operation cost, the garbage treatment capacity, the incineration power generation capacity and the like, and an economic objective function f can be obtained according to the operation cost of the incinerator 1 ：
In the formula:
n-total number of controllable variables including primary air, secondary air, activated carbon, etc., in this example n =6;
x i the ith real-time controllable variable monitoring value comprises feeding quantity, primary air quantity, secondary air quantity, water feeding quantity, active carbon usage amount and the like 1 ～x 6 Respectively shows the grate speed (m/min), the feeding amount (kg/min), the primary air amount (m) 3 H) secondary air volume (m) 3 H), water supply (m) 3 H), activated carbon input amount (kg);
θ i -unit cost of the ith controlled variable, unit cost of air volume (yuan/m) 3 ) The real-time power of the fan is multiplied by the specification of the fan/electricity price, and the unit cost (yuan/kg) of the feeding amount is calculated according to the power/electricity price of the crane.
(2) Environmental benefit goal
Environmental benefits relate to the emission of various pollutants, including dioxinsMethod for constructing environmental benefit objective function f by measuring discharge amount 2 ：
In the formula:
lambda-dioxin exceeding early warning input, wherein the exceeding early warning is 1, otherwise, the exceeding early warning is 0;
k i -the adjustment coefficient of the ith controllable variable to the concentration of dioxin represents the controllable variable x i The dioxin regulating capacity and the empirical coefficient are generally 0 to 1;
Step two, setting each control variable x i The constraint conditions of the furnace comprise a grate speed range, a lowest feeding amount, air quantity ranges of primary air and secondary air, a primary air and secondary air proportion and the like.
And step three, solving the multi-objective optimization function, wherein according to the constructed multi-objective optimization function and the constraint conditions thereof:
min F(x)＝[f 1 (x)，f 2 (x)]
s.t.x i min ＜x i ＜x i max
0＜x 4 /x 3 ＜0.5
solving by applying a Particle Swarm Optimization (PSO) algorithm, wherein the solving process is as follows:
(1) setting the number m of particles, and randomly initializing the speed v and the position x of the particles;
(2) calculating the adaptive value of each particle according to the objective function, and updating the individual optimal position pbest of the jth particle j And global optimal extremum position gbest;
(3) updating the particle speed v and the position x through formulas (3) and (4);
in the formula, c 1 And c 2 All acceleration factors reflect information exchange among particle swarms, and generally all acceleration factors are 2; t is expressed as the current iteration number; r is 1 And r 2 Is in [0,1 ]]A uniform random number that varies within a range; pbest j Representing the optimal extreme position of the jth particle individual; the gbest is expressed as a global optimum extremum position.
(4) And (3) judging whether the maximum iteration number is reached or the global optimal solution number reaches a set value, stopping iteration if the maximum iteration number is reached, outputting a final optimal solution set, and repeating the step (2) if the final optimal solution set is not reached.
Different from single-target optimization, the multi-target optimization may obtain more than one solution, and finally, the most appropriate solution set can be screened out according to management requirements (for example, the smaller the number of times of fan operation is, the better the smaller the variation value of fan power is, and the like), or by using methods such as multi-attribute decision making and the like
The optimum state point for the ith variable, as determined by the intelligent control system, in this embodiment,
respectively showing the optimal speed, the optimal feeding quantity, the optimal primary air quantity, the optimal secondary air quantity, the optimal water feeding quantity and the optimal input quantity of the active carbon of the current grate.
The pretreatment control system monitors, controls and displays the running condition of the incinerator feeding equipment, and comprises: normal operation of crushing, grabbing and hanging, conveying and feeding equipment and the like;
the incineration control system is used for monitoring, controlling and displaying the operation condition of the incinerator and comprises: the suspension section of the incinerator is in a negative pressure combustion state, the retention time of the flue gas in a high-temperature interval exceeds 2 seconds, and the CO concentration of the tail flue gas is lower than 50mg/Nm 3 The primary air and the secondary air are reasonable in proportion, slag discharging is smooth, no fire phenomenon exists, dust removal of a tail flue and a heating surface is timely and the like;
the flue gas purification control system is used for monitoring, controlling and displaying the operation state of tail flue gas purification equipment, and comprises: the cloth bag is complete, the temperature of the flue gas at the cloth bag inlet is not more than 150 ℃, the concentration of the dust at the cloth bag outlet is lower than the emission value, the quality of the activated carbon is qualified, the blanking of the activated carbon is normal, and the like.
Meanwhile, the intelligent dioxin control module comprises an automatic control mode and an auxiliary control mode. Under the automatic control mode, the module instructs automatic control equipment operation according to the intelligent control system, and automatic control preliminary treatment control system, burning control system, gas cleaning control system operate at the best operating mode, keep dioxin concentration at the required scope. The auxiliary control mode is used for selecting a system needing manual control from pretreatment control, incineration control and flue gas purification control, an intelligent control system is used for providing a recommended operation scheme, and other control systems are still automatically controlled by the modules.
It will be apparent to those skilled in the art that various applications, additions, modifications and variations can be made to the present invention without departing from the spirit or scope of the invention as hereinafter claimed. If various applications, additions, modifications and variations based on the present invention are within the scope of the claims and their equivalents, the present invention is also intended to encompass these applications, additions, modifications and variations.
Claims (8)
1. A method for giving real-time early warning on dioxin discharge in a waste incinerator is characterized by comprising the following steps:
(1) Constructing an LSTM-ARIMA fusion prediction model based on the LSTM model and the ARIMA model; wherein, the output results of the LSTM model and the ARIMA model are dioxin emission data converted into trichlorobenzene concentration; the two models are fused in a way that the prediction results of the two models are respectively added by taking a weight coefficient of 0.5;
(2) Historical data of dioxin in flue gas of a waste incinerator, which is acquired by a dioxin on-line detection device, is used as input of a fusion prediction model; wherein, the input variables of the LSTM model are as follows: average temperature of flue gas at the outlet of the first channel, total air quantity of secondary air, total air quantity of primary air, pressure of primary air, flue gas flow and dioxin emission data converted into trichlorobenzene concentration in the first 5 hours; input variables for the ARIMA model include: average temperature of flue gas at the outlet of the first channel, total air quantity of secondary air, total air quantity of primary air, pressure of primary air and flow rate of flue gas in the first 5 hours;
(3) Taking a dioxin detection value and average temperature, air volume, air pressure and flue gas flow data of flue gas discharged by a waste incinerator, which are acquired by a dioxin online detection device in real time, as input of an LSTM-ARIMA fusion prediction model, outputting a prediction result after fusion calculation, wherein the prediction result is the dioxin discharge data converted into trichlorobenzene concentration after a set time period;
(4) And comparing the prediction result with a preset index, and if the prediction result exceeds the preset index range, sending out a dioxin exceeding early warning.
2. The method of claim 1,
the LSTM model adopts 1 LSTM +1 Dense architecture, the number of LSTM layer neurons is 150, and a Relu function is selected as an activation function; the number of neurons in a Dense full junction layer is 1; initializing parameters of each layer by adopting a random normal function, wherein the seed of a random seed is 6; the input time step of the input layer is 5, and the prediction step of the output layer is 1;
the number p of autoregressive terms, the difference times d and the number q of moving average terms in the ARIMA model parameters are respectively 1,2 and 1, and the prediction step length is 1.
3. An intelligent control method for dioxin emission of a waste incinerator is characterized by comprising the following steps:
(1) Based on an economic objective calculated by the running cost of the incinerator and an environmental benefit objective calculated by predicting emission data based on dioxin, a multi-objective optimization function F (x) is constructed according to the following formula, and the optimization objective is a minimization function F (x):
min F(x)＝[f 1 (x)，f 2 (x)]
in the formula, f 1 (x) For the economic objective function calculated from the running cost of the incinerator, f 2 (x) Calculating an environmental benefit objective function for the emission data predicted by using the dioxin; the dioxin emission prediction data are prediction results output by the fusion prediction model in the step (3) of the method in claim 1;
(2) Setting constraint conditions of all control variables in the multi-objective optimization function, and then solving the multi-objective optimization function to obtain an optimal control state point of the operation parameters for intelligently controlling the waste incinerator;
(3) And adjusting the working conditions of a pretreatment control system, an incineration control system and a flue gas purification control system of the waste incinerator by using the data of the optimal control state point of the operation parameter through an automatic control mode or an auxiliary control mode, so that the working conditions are kept within a specified range while the pretreatment control system, the incineration control system and the flue gas purification control system operate in the optimal working conditions.
4. A method according to claim 3, wherein in step (1), the operating cost of the incinerator is calculated according to the following equation:
in the formula: n is the total number of controllable variables, and the controllable variables comprise the dosage of feeding materials, primary air, secondary air, water supply and active carbon and the pressure difference of the cloth bag;
θ i is the unit cost of the ith controlled variable; wherein the unit cost of the feeding amount is calculated according to the power/electricity price of the crane, the unit cost of the air volume is calculated according to the real-time power of the fan multiplied by the specification/electricity price of the fan, the unit cost of the water volume is calculated according to the unit price of water consumption plus the real-time power of the water supply pump multiplied by the specification/electricity price of the water pump, the unit cost of the activated carbon is obtained according to the purchase price,the unit cost of the bag pressure difference is the cost for replacing the bag/(the maximum allowable pressure difference of the bag-the design pressure difference);
x i the monitoring value is the ith controllable variable real-time monitoring value and comprises feeding quantity, primary air quantity, secondary air quantity, water supply quantity, active carbon usage and cloth bag pressure difference.
5. The method according to claim 3, wherein in the step (1), the predicted emission amount of dioxin is calculated according to the following formula:
in the formula:
k i the influence coefficient of the ith controlled variable on the concentration of dioxin is used for representing the controlled variable x i Ability to modulate dioxins; the numerical value is an empirical coefficient and ranges from 0 to 1;
lambda is dioxin exceeding early warning input, is 1 when exceeding early warning is carried out, and is 0 otherwise;
6. The method according to claim 3, wherein in the step (2), when the multi-objective optimization function is solved, the multi-objective genetic algorithm or the particle swarm optimization algorithm is applied to obtain a plurality of groups of effective solution sets of the function.
7. A method according to claim 4 or 5, characterized in that each control variable x i The constraint conditions of (1) include: the lowest feeding amount; the maximum air quantity and the minimum air quantity of the primary air and the secondary air; the ratio of primary air to secondary air; activated carbon maximumAnd (4) adding amount.
8. A real-time early warning and intelligent control system for dioxin emission of a waste incinerator comprises a pretreatment control system, an incineration control system and a flue gas purification control system, wherein the pretreatment control system is used for controlling the operation of the waste incinerator; it is characterized by also comprising:
the online dioxin detection device is arranged at the position of the tail end of a flue of the waste incinerator and is used for acquiring the content of dioxin in flue gas discharged by the waste incinerator in real time;
a dioxin emission value prediction module for performing fusion calculation of the LSTM-ARIMA fusion prediction model of claim 1 and outputting a prediction result;
the dioxin emission early warning module is used for comparing the prediction result of the fusion prediction model with a preset index and sending out a dioxin exceeding early warning according to the comparison result;
an intelligent control module for dioxin discharge, which is used for executing the intelligent control method for dioxin discharge in the waste incinerator according to claim 3 and sending control signals or operation action prompts to the pretreatment control system, the incineration control system and the flue gas purification control system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210845636.0A CN115329659B (en) | 2022-07-18 | 2022-07-18 | Method and system for real-time early warning and intelligent control of dioxin emission of waste incinerator |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210845636.0A CN115329659B (en) | 2022-07-18 | 2022-07-18 | Method and system for real-time early warning and intelligent control of dioxin emission of waste incinerator |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115329659A true CN115329659A (en) | 2022-11-11 |
| CN115329659B CN115329659B (en) | 2023-07-14 |
Family
ID=83916881
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210845636.0A Active CN115329659B (en) | 2022-07-18 | 2022-07-18 | Method and system for real-time early warning and intelligent control of dioxin emission of waste incinerator |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115329659B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116538521A (en) * | 2023-05-04 | 2023-08-04 | 单县龙宇生物科技有限公司 | Adaptive planning system based on furfural residue boiler flue gas emission |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101357290A (en) * | 2008-09-25 | 2009-02-04 | 浙江大学 | Method and system of controlling dioxin discharge in flue gas by duty-cycle operation of sulphur-containing inhibitor |
| CN101382288A (en) * | 2008-10-16 | 2009-03-11 | 浙江大学 | Method for controlling discharge of dioxin in incineration process of domestic garbage by sulfur composition in coal |
| JP2009108451A (en) * | 2007-10-31 | 2009-05-21 | Japan Atomic Energy Agency | Heat illness-warning device for protective suit-wearing worker |
| CN101485955A (en) * | 2009-01-15 | 2009-07-22 | 浙江大学 | Method for controlling discharge of dioxins in burning process using sulfurous iron ore |
| JP2016059533A (en) * | 2014-09-17 | 2016-04-25 | 大成建設株式会社 | Heat stroke warning device |
| CN110081440A (en) * | 2019-05-05 | 2019-08-02 | 绿色动力环保集团股份有限公司 | The on-line prediction method of dioxin and early warning and control system in consumer waste incineration |
| CN112183709A (en) * | 2020-09-22 | 2021-01-05 | 生态环境部华南环境科学研究所 | Method for predicting and early warning dioxin exceeding in waste incineration waste gas |
| CN112884203A (en) * | 2021-01-22 | 2021-06-01 | 浙江大学 | Online prediction method for dioxin emission based on Bayesian support vector regression |
| CN113705915A (en) * | 2021-09-01 | 2021-11-26 | 沈阳建筑大学 | CNN-LSTM-ARIMA-based combined short-term power load prediction method |
| CN114266461A (en) * | 2021-12-15 | 2022-04-01 | 北京工业大学 | MSWI process dioxin emission risk early warning method based on visual distribution GAN |
| CN114546797A (en) * | 2022-01-10 | 2022-05-27 | 西安理工大学 | Cloud virtual machine aging prediction method based on time sequence decomposition |
-
2022
- 2022-07-18 CN CN202210845636.0A patent/CN115329659B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009108451A (en) * | 2007-10-31 | 2009-05-21 | Japan Atomic Energy Agency | Heat illness-warning device for protective suit-wearing worker |
| CN101357290A (en) * | 2008-09-25 | 2009-02-04 | 浙江大学 | Method and system of controlling dioxin discharge in flue gas by duty-cycle operation of sulphur-containing inhibitor |
| CN101382288A (en) * | 2008-10-16 | 2009-03-11 | 浙江大学 | Method for controlling discharge of dioxin in incineration process of domestic garbage by sulfur composition in coal |
| CN101485955A (en) * | 2009-01-15 | 2009-07-22 | 浙江大学 | Method for controlling discharge of dioxins in burning process using sulfurous iron ore |
| JP2016059533A (en) * | 2014-09-17 | 2016-04-25 | 大成建設株式会社 | Heat stroke warning device |
| CN110081440A (en) * | 2019-05-05 | 2019-08-02 | 绿色动力环保集团股份有限公司 | The on-line prediction method of dioxin and early warning and control system in consumer waste incineration |
| CN112183709A (en) * | 2020-09-22 | 2021-01-05 | 生态环境部华南环境科学研究所 | Method for predicting and early warning dioxin exceeding in waste incineration waste gas |
| CN112884203A (en) * | 2021-01-22 | 2021-06-01 | 浙江大学 | Online prediction method for dioxin emission based on Bayesian support vector regression |
| CN113705915A (en) * | 2021-09-01 | 2021-11-26 | 沈阳建筑大学 | CNN-LSTM-ARIMA-based combined short-term power load prediction method |
| CN114266461A (en) * | 2021-12-15 | 2022-04-01 | 北京工业大学 | MSWI process dioxin emission risk early warning method based on visual distribution GAN |
| CN114546797A (en) * | 2022-01-10 | 2022-05-27 | 西安理工大学 | Cloud virtual machine aging prediction method based on time sequence decomposition |
Non-Patent Citations (2)
| Title |
|---|
| 刘文慧: "基于多源信息融合的高炉煤气流分布识别和建模研究", 《中国优秀硕士学位论文全文数据库工程科技I辑》, no. 03, pages 023 - 124 * |
| 林晓青: "废物焚烧炉二恶英硫氨基复合阻滞的研究", 《中国博士学位论文全文数据库工程科技I辑》, no. 06, pages 027 - 23 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116538521A (en) * | 2023-05-04 | 2023-08-04 | 单县龙宇生物科技有限公司 | Adaptive planning system based on furfural residue boiler flue gas emission |
| CN116538521B (en) * | 2023-05-04 | 2023-10-20 | 单县龙宇生物科技有限公司 | Adaptive planning system based on furfural residue boiler flue gas emission |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115329659B (en) | 2023-07-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mariajayaprakash et al. | Optimization of process parameters through fuzzy logic and genetic algorithm–A case study in a process industry | |
| CN107944173B (en) | Dioxin soft measurement system based on selective integrated least square support vector machine | |
| CA2594323A1 (en) | Process control and optimization technique using immunological concepts | |
| CN112859780B (en) | Thermal power plant intelligent combustion control method based on cloud data and cloud computing | |
| CN115329659B (en) | Method and system for real-time early warning and intelligent control of dioxin emission of waste incinerator | |
| CN110500596B (en) | Automatic control method for hazardous waste incineration | |
| Wang et al. | Fuzzy modeling of boiler efficiency in power plants | |
| CN111457392B (en) | Intelligent setting method for air quantity in urban domestic garbage incineration process | |
| CN110991756A (en) | MSWI furnace temperature prediction method based on TS fuzzy neural network | |
| CN114265885A (en) | Automatic control method and system for moisture of sintering return powder | |
| CN113175678A (en) | Method and device for monitoring garbage incineration | |
| CN115111594A (en) | Intelligent regulation and control system and method for heat accumulating type thermal oxidation furnace | |
| He et al. | A novel self-organizing TS fuzzy neural network for furnace temperature prediction in MSWI process | |
| Anderson et al. | Multi-objective optimization of operational variables in a waste incineration plant | |
| Li et al. | Neural networks and genetic algorithms can support human supervisory control to reduce fossil fuel power plant emissions | |
| JP4188859B2 (en) | Operation control method and operation control apparatus for waste treatment plant equipment | |
| CN109882883B (en) | Method and device for optimizing boiler coal burning efficiency based on artificial intelligence | |
| JP3231745B2 (en) | Adaptive control method and apparatus in thermal power plant | |
| CN116720446A (en) | Method for monitoring thickness of slag layer of water-cooled wall of boiler in real time | |
| Shen et al. | An adaptive fuzzy approach for the incineration temperature control process | |
| CN110684547A (en) | Optimized control method for biomass pyrolysis carbonization kiln | |
| CN115145152A (en) | Boiler combustion and denitration process collaborative optimization control method | |
| CN211526443U (en) | Automatic combustion control system of garbage incinerator | |
| CN114492727A (en) | Intelligent prediction and control method for inlet smoke concentration of dust removal system | |
| CN114779722B (en) | Intelligent combustion optimization control system and method for coal-fired power station boiler |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |