阅读说明：本技术 一种等离子飞灰熔融炉飞灰熔融温度的软测量方法 (Soft measurement method for fly ash melting temperature of plasma fly ash melting furnace ) 是由 叶泽甫 赵志军 张帅 朱竹军 阎高伟 于 2021-09-10 设计创作，主要内容包括：本发明公开了一种等离子飞灰熔融炉飞灰熔融温度的软测量方法,该方法包括：基于低频、高频微波传感器的飞灰介电常数测量；以飞灰介电常数历史数据为输入神经元,飞灰组分历史数据为输出神经元,建立飞灰组分含量的递推随机权神经网络辨识模型；依据飞灰组分含量实验室数据对模型在线校正；基于飞灰组分含量历史数据和飞灰熔融特性温度历史数据建立递推最小二乘辨识模型；依据飞灰组分含量和飞灰熔融特性温度实验室数据对模型进行在线校正；校正后的模型对飞灰熔融特性温度进行在线测量。通过本发明提供的方法可以对熔融温度实时预测,保证飞灰充分熔融的同时,节约了资源,提高了生产效率。(The invention discloses a soft measurement method for the fly ash melting temperature of a plasma fly ash melting furnace, which comprises the following steps: measuring fly ash dielectric constant based on low-frequency and high-frequency microwave sensors; establishing a recursive random weight neural network identification model of the fly ash component content by taking the fly ash dielectric constant historical data as input neurons and the fly ash component historical data as output neurons; online correcting the model according to the laboratory data of the content of the fly ash components; establishing a recursive least square identification model based on the historical data of the content of the components of the fly ash and the historical data of the melting characteristic temperature of the fly ash; carrying out online correction on the model according to the content of the components of the fly ash and the melting characteristic temperature laboratory data of the fly ash; the corrected model carries out on-line measurement on the melting characteristic temperature of the fly ash. The method provided by the invention can predict the melting temperature in real time, and can save resources and improve the production efficiency while ensuring the full melting of the fly ash.)

1. A soft measurement method for the fly ash melting temperature of a plasma fly ash melting furnace is characterized by comprising the following steps:

s1: fly ash dielectric constant measurement based on a low frequency capacitive sensor; measuring the dielectric constant of the fly ash based on a high-frequency microwave sensor;

s2: establishing a recursive random weight neural network identification model of the fly ash component content by taking the low-frequency dielectric constant historical data and the high-frequency dielectric constant historical data as input neurons and the fly ash component historical data as output neurons;

s3: performing online correction on the recursive random weight neural network identification model of the fly ash component content established in the step S2 according to the laboratory data of the fly ash component content;

s4: establishing a recursive least square identification model based on the historical data of the content of the components of the fly ash and the historical data of the melting characteristic temperature of the fly ash;

s5: online correction is carried out on the recursive least square identification model established in the step S4 according to the content of the components of the fly ash and the laboratory data of the melting characteristic temperature of the fly ash;

s6: the fly ash melting characteristic temperature is measured on line according to the model corrected at step S5.

2. The method of soft measurement of fly ash fusion temperature of plasma fly ash melting furnace of claim 1, wherein step S1 comprises:

s11: the dielectric constant of the sample measured by the low-frequency capacitance sensor is epsilon, and the dielectric constant of the sample measured by the high-frequency microwave sensor is epsilon_rThe sampling period of the low-frequency capacitance sensor and the sampling period of the high-frequency microwave sensor are T, the current time T is taken as a reference, and historical data of the low-frequency capacitance sensor and the high-frequency microwave sensor are respectively expressed as epsilon (T-pT) and epsilon_r(t-pT), fly ash component historical data expressed as u₁(t-pT)，u₂(t-pT)，…，u_M(t-pT), p is 1,2,3 … N, N is the number of samples, M is the number of fly ash components;

s12: at the initial moment, take N₀Individual history data, denoted (xi)₀,U₀)， Wherein epsilon_i＝[ε(t-iT),ε_r(t-iT)]∈R²，u_i＝[u₁(t-iT),u₂(t-iT),…,u_M(t-iT)]∈R^M，i＝1,2,3…N₀And remember epsilon_i＝(ε_i1,ε_i2)，u_i＝(u_i1,u_i2,…u_iM)。

3. The method of soft measurement of fly ash fusion temperature of plasma fly ash melting furnace of claim 2, wherein step S2 comprises:

with low frequency dielectric constant history data epsilon_i1High frequency dielectric constant history data ε_i2For input into neurons, fly ash component historical data u_i1,u_i2,…u_iMFor the output neuron, a single hidden layer neural network containing K hidden layer nodes is established:

wherein, a_j＝(a_j1,a_j2)^TFor input of neuron epsilon_iWeight to jth hidden layer, b_jFor the jth hidden layer neuron bias, β_jmFor the weight of the j hidden layer neuron to the M output layer, M is 1,2,3 … M, u_imFor the output of the mth output neuron, the number of hidden layer neurons K is obtained by a cross-validation method, and g (-) is a neuron activation function:

randomly initializing input layer neurons to hidden layer neuron weights a_jAnd offset b_jCalculating to obtain the hidden layer neuron matrix H₀Comprises the following steps:

the K hidden layer node single hidden layer neural networks and the hidden layer neuron matrix can be obtained as follows:

U₀＝H₀β₀

β₀is beta_jmComposed matrix, U₀Is u_imThe matrix of the composition is composed of a plurality of matrixes,

calculating U by least squares₀＝H₀β₀To obtain

Wherein the content of the first and second substances,is H₀Further solving the generalized inverse matrix of (2):

4. the method of claim 3, wherein step S3 includes:

the data of the low-frequency capacitance sensor and the high-frequency microwave sensor measured by a laboratory instrument at the time t + qT are respectively expressed as epsilon (t + qT) by taking the current time t as a reference_r(t + qT), fly ash component data u₁(t+qT)，u₂(t+qT)，…，u_M(t+qT)，q＝1,2,3…N₁，N₁The number of samples; the hidden layer neuron matrix is obtained by calculation as follows:

the K hidden layer node single hidden layer neural networks and the hidden layer neuron matrix can be obtained as follows:

U₁＝H₁β₁

U₁is u_imA matrix of components, i ═ N₀+1,N₀+2,…,N₀+N₁，β₁In order to be the new weight matrix, the weight matrix,

thus:

solving beta through recursive least squares generalized inverse₁：

Wherein the content of the first and second substances,

further, generalizing to the general case, solving the general recursion formula is:

wherein H_kRepresenting the hidden layer neuron matrix at any one time, H_k+1Is the hidden layer neuron matrix at the next time instant,β_krepresenting hidden layer to output at any one timeWeight vector of layer, beta_k+1The weight vector from the hidden layer to the output layer at the next moment;

order toWhen data is one entry, H_k+1Is denoted by h_k+1，U_k+1Is denoted by u_k+1The recurrence formula is further simplified as:

5. the method of claim 4, wherein step S4 includes:

the historical data of the content of the fly ash components is U_kThe temperature history data of the melting characteristics of the fly ash is y_kThe system history parameter is theta_kThen, then

y_k＝U_kθ_k

The generalized inverse solution by the least square method is adopted to obtain:

θ_k＝U_k+y_k＝(U_k ^TU_k)^-1U_k ^Ty_k。

6. the method of claim 5, wherein step S5 includes:

online correction is carried out on the recursive least square identification model established in the step S4 according to the content of the components of the fly ash and the laboratory data of the melting characteristic temperature of the fly ash;

the newly determined fly ash component content data is U_k+1Fly awayAsh fusion characteristic temperature data y_k+1The system parameter is theta_k+1Then, then

Solving for theta by generalized inverse_k+1：

Wherein the content of the first and second substances,

order toWhen the update data is one entry, U_k+1Is denoted by u_k+1And then:

Technical Field

The invention relates to the field of detection of melting temperature of a plasma fly ash melting furnace, in particular to a soft measurement method of the melting temperature of fly ash of the plasma fly ash melting furnace.

Background

At present, a control system of a plasma fly ash melting furnace is designed by mainly referring to a control system of a conventional garbage incinerator, and the main control mode is as follows: the feeding rate of the raw materials is determined according to the compatible heat value and the treatment capacity of the raw materials, the temperature of a hearth is maintained by adjusting the air inlet amount, the oxygen content and the current of a plasma torch, the full gasification of fly ash is ensured, and the outlet flue gas reaches the standard. In the actual production process, the melting points of the fly ash are different due to different components of the fly ash, and the required combustion temperature of a hearth is different. The hearth combustion temperature determined according to the raw material compatibility heat value can cause insufficient melting gasification due to real heat value fluctuation or fluctuation of treatment capacity, so that the emission standard can not be met; on the other hand, the combustion temperature of the hearth is increased as much as possible, so that the fly ash can be fully gasified, the emission requirement is met, the production cost is increased, and the service lives of the melting furnace and the plasma torch electrode are shortened. Therefore, the determination of the melting point of the fly ash is of great significance to the actual operation of the melting furnace.

At present, the measurement of the melting point of the fly ash is mainly carried out in a laboratory according to the national standard coal ash melting property measurement method (GB/T219-2008), the fly ash is made into a triangular cone and heated in a high-temperature furnace to observe the form change of the ash cone, and four melting characteristic temperatures (deformation temperature, softening temperature, hemisphere temperature and flowing temperature) of the ash cone are recorded. The fly ash melting characteristic temperature measured in a laboratory is accurate in measurement, but the measurement period is long, and the measurement can only be performed by sampling, so that the real-time requirement of the operation of the industrial melting furnace cannot be met.

Disclosure of Invention

In order to more accurately detect the melting temperature of the plasma fly ash melting furnace, the invention provides a soft measurement method of the melting temperature of the fly ash of the plasma fly ash melting furnace.

In order to achieve the technical purpose, the invention adopts the following technical scheme:

a soft measurement method for the fly ash fusion temperature of a plasma fly ash fusion furnace comprises the following steps:

s1: fly ash dielectric constant measurement based on a low frequency capacitive sensor; measuring the dielectric constant of the fly ash based on a high-frequency microwave sensor;

s2: establishing a recursive random weight neural network identification model of the fly ash component content by taking the low-frequency dielectric constant historical data and the high-frequency dielectric constant historical data as input neurons and the fly ash component historical data as output neurons;

s3: performing online correction on the recursive random weight neural network identification model of the fly ash component content established in the step S2 according to the laboratory data of the fly ash component content;

s4: establishing a recursive least square identification model based on the historical data of the content of the components of the fly ash and the historical data of the melting characteristic temperature of the fly ash;

s5: online correction is carried out on the recursive least square identification model established in the step S4 according to the content of the components of the fly ash and the laboratory data of the melting characteristic temperature of the fly ash;

s6: the fly ash melting characteristic temperature is measured on line according to the model corrected at step S5.

Preferably, step S1 includes:

s11: the dielectric constant of the sample measured by the low-frequency capacitance sensor is epsilon, and the dielectric constant of the sample measured by the high-frequency microwave sensor is epsilon_rThe sampling period of the low-frequency capacitance sensor and the sampling period of the high-frequency microwave sensor are T, the current time T is taken as a reference, and historical data of the low-frequency capacitance sensor and the high-frequency microwave sensor are respectively expressed as epsilon (T-pT) and epsilon_r(t-pT), fly ash component historical data expressed as u₁(t-pT)，u₂(t-pT)，…，u_M(t-pT), p is 1,2,3 … N, N is the number of samples, M is the number of fly ash components;

s12: at the initial moment, take N₀Individual history data, denoted (xi)₀,U₀)， Wherein epsilon_i＝[ε(t-iT),ε_r(t-iT)]∈R²，u_i＝[u₁(t-iT),u₂(t-iT),…,u_M(t-iT)]∈R^M，i＝1,2,3…N₀And remember epsilon_i＝(ε_i1,ε_i2)，u_i＝(u_i1,u_i2,…u_iM)。

Preferably, step S2 includes:

with low frequency dielectric constant history data epsilon_i1High frequency dielectric constant history data ε_i2For input into neurons, fly ash component historical data u_i1,u_i2,…u_iMFor the output neuron, a single hidden layer neural network containing K hidden layer nodes is established:

wherein, a_j＝(a_j1,a_j2)^TFor input of neuron epsilon_iWeight to jth hidden layer, b_jFor the jth hidden layer neuron bias, β_jmFor the weight of the j hidden layer neuron to the M output layer, M is 1,2,3 … M, u_imFor the output of the mth output neuron, the number of hidden layer neurons K is obtained by a cross-validation method, and g (-) is a neuron activation function:

randomly initializing input layer neurons to hidden layer neuron weights a_jAnd offset b_jCalculating to obtain the hidden layer neuron matrix H₀Comprises the following steps:

the K hidden layer node single hidden layer neural networks and the hidden layer neuron matrix can be obtained as follows:

U₀＝H₀β₀

β₀is beta_jmComposed matrix, U₀Is u_imThe matrix of the composition is composed of a plurality of matrixes,

calculating U by least squares₀＝H₀β₀To obtain

Wherein the content of the first and second substances,is H₀Further solving the generalized inverse matrix of (2):

preferably, step S3 includes:

the data of the low-frequency capacitance sensor and the high-frequency microwave sensor measured by a laboratory instrument at the time t + qT are respectively expressed as epsilon (t + qT) by taking the current time t as a reference_r(t + qT), fly ash component data u₁(t+qT)，u₂(t+qT)，…，u_M(t+qT)，q＝1,2,3…N₁，N₁The number of samples; the hidden layer neuron matrix is obtained by calculation as follows:

the K hidden layer node single hidden layer neural networks and the hidden layer neuron matrix can be obtained as follows:

U₁＝H₁β₁

U₁is u_imA matrix of components, i ═ N₀+1,N₀+2,…,N₀+N₁，β₁In order to be the new weight matrix, the weight matrix,

thus:

solving beta through recursive least squares generalized inverse₁：

Wherein the content of the first and second substances,

further, generalizing to the general case, solving the general recursion formula is:

wherein H_kRepresenting the hidden layer neuron matrix at any one time, H_k+1Is the hidden layer neuron matrix at the next time instant,β_krepresenting the weight vector, β, from the hidden layer to the output layer at any one time_k+1Is hidden by the next momentIncluding layer-to-output layer weight vectors;

order toThe recurrence formula is further simplified as:

preferably, step S4 includes:

the historical data of the content of the fly ash components is U_kThe temperature history data of the melting characteristics of the fly ash is y_kThe system history parameter is theta_kThen, then

y_k＝U_kθ_k

The generalized inverse solution by the least square method is adopted to obtain:

θ_k＝U_k ⁺y_k＝(U_k ^TU_k)^-1U_k ^Ty_k。

preferably, step S5 includes:

online correction is carried out on the recursive least square identification model established in the step S4 according to the content of the components of the fly ash and the laboratory data of the melting characteristic temperature of the fly ash;

the newly determined fly ash component content data is U_k+1The melting characteristic temperature data of the fly ash is y_k+1The system parameter is theta_k+1Then, then

Solving for theta by generalized inverse_k+1：

Wherein the content of the first and second substances,

order toWhen the update data is one entry, then:

compared with the prior art, the invention has the beneficial effects that:

the invention provides a soft measurement method for the fly ash melting temperature of a plasma fly ash melting furnace, which establishes a recursive random weight neural network identification model of the fly ash component content, and carries out online correction on the model according to the fly ash component content laboratory data; and then establishing fly ash component content historical data and fly ash melting characteristic temperature historical data to establish a recursive least square identification model, and performing online correction on the model according to the fly ash component content and fly ash melting characteristic temperature laboratory data. The method provided by the invention can predict the melting temperature in real time, and can save resources and improve the production efficiency while ensuring the full melting of the fly ash.

Drawings

For a clearer explanation of the embodiments or technical solutions of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for a person skilled in the art to obtain other drawings based on these drawings without creative efforts.

FIG. 1 is a schematic view of the process of the present invention;

FIG. 2 is a schematic diagram of a neural network model of fly ash dielectric constant and fly ash composition according to the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

A method for soft measurement of fly ash fusion temperature of a plasma fly ash fusion furnace, as shown in fig. 1, comprising:

s1: fly ash dielectric constant measurement based on a low frequency capacitive sensor; measuring the dielectric constant of the fly ash based on a high-frequency microwave sensor; the method specifically comprises the following steps:

s11: the dielectric constant of the sample measured by the low-frequency capacitance sensor is epsilon, and the dielectric constant of the sample measured by the high-frequency microwave sensor is epsilon_rThe sampling period of the low-frequency capacitance sensor and the sampling period of the high-frequency microwave sensor are T, the current time T is taken as a reference, and historical data of the low-frequency capacitance sensor and the high-frequency microwave sensor are respectively expressed as epsilon (T-pT) and epsilon_r(t-pT), fly ash component historical data expressed as u₁(t-pT)，u₂(t-pT)，…，u_M(t-pT), p is 1,2,3 … N, N is the number of samples, M is the number of fly ash components;

s12: at the initial moment, take N₀Individual history data, denoted (xi)₀,U₀)， Wherein epsilon_i＝[ε(t-iT),ε_r(t-iT)]∈R²，u_i＝[u₁(t-iT),u₂(t-iT),…,u_M(t-iT)]∈R^M，i＝1,2,3…N₀And remember epsilon_i＝(ε_i1,ε_i2)，u_i＝(u_i1,u_i2,...u_iM)。

S2: establishing a recursive random weight neural network identification model of the content of the fly ash components by taking the low-frequency dielectric constant historical data and the high-frequency dielectric constant historical data as input neurons and the fly ash component historical data as output neurons, as shown in FIG. 2; the method specifically comprises the following steps:

with low frequency dielectric constant history data epsilon_i1High frequency dielectric constant history data ε_i2For input into neurons, fly ash component historical data u_i1,u_i2,…u_iMFor the output neuron, a single hidden layer neural network containing K hidden layer nodes is established:

wherein, a_j＝(a_j1,a_j2)^TFor input of neuron epsilon_iWeight to jth hidden layer, b_jFor the jth hidden layer neuron bias, β_jmFor the weight of the j hidden layer neuron to the M output layer, M is 1,2,3 … M, u_imFor the output of the mth output neuron, the number of hidden layer neurons K is obtained by a cross-validation method, and g (-) is a neuron activation function:

randomly initializing input layer neurons to hidden layer neuron weights a_jAnd offset b_jCalculating to obtain the hidden layer neuron matrix H₀Comprises the following steps:

the K hidden layer node single hidden layer neural networks and the hidden layer neuron matrix can be obtained as follows:

U₀＝h₀β₀

β₀is beta_jmComposed matrix, U₀Is u_imThe matrix of the composition is composed of a plurality of matrixes,

calculating U by least squares₀＝H₀β₀To obtain

Wherein the content of the first and second substances,is H₀Further solving the generalized inverse matrix of (2):

s3: performing online correction on the recursive random weight neural network identification model of the fly ash component content established in the step S2 according to the laboratory data of the fly ash component content; the method specifically comprises the following steps:

the data of the low-frequency capacitance sensor and the high-frequency microwave sensor measured by a laboratory instrument at the time t + qT are respectively expressed as epsilon (t + qT) by taking the current time t as a reference_r(t + qT), fly ash component data u₁(t+qT)，u₂(t+qT)，…，u_M(t+qT)，q＝1,2,3…N₁，N₁The number of samples; the hidden layer neuron matrix is obtained by calculation as follows:

the K hidden layer node single hidden layer neural networks and the hidden layer neuron matrix can be obtained as follows:

U₁＝H₁β₁

U₁is u_imA matrix of components, i ═ N₀+1,N₀+2,…,N₀+N₁，β₁In order to be the new weight matrix, the weight matrix,

thus:

solving beta through recursive least squares generalized inverse₁：

Wherein the content of the first and second substances,

further, generalizing to the general case, solving the general recursion formula is:

wherein H_kRepresenting the hidden layer neuron matrix at any one time, H_k+1Is the hidden layer neuron matrix at the next time instant,β_krepresenting the weight vector, β, from the hidden layer to the output layer at any one time_k+1The weight vector from the hidden layer to the output layer at the next moment;

order toThe recurrence formula is further simplified as:

s4: establishing a recursive least square identification model based on the historical data of the content of the components of the fly ash and the historical data of the melting characteristic temperature of the fly ash; the method specifically comprises the following steps:

the historical data of the content of the fly ash components is U_kThe temperature history data of the melting characteristics of the fly ash is y_kThe system history parameter is theta_kThen, then

y_k＝U_kθ_k

The generalized inverse solution by the least square method is adopted to obtain:

θ_k＝U_k ⁺y_k＝(U_k ^TU_k)^-1U_k ^Ty_k。

s5: online correction is carried out on the recursive least square identification model established in the step S4 according to the content of the components of the fly ash and the laboratory data of the melting characteristic temperature of the fly ash; the method specifically comprises the following steps:

online correction is carried out on the recursive least square identification model established in the step S4 according to the content of the components of the fly ash and the laboratory data of the melting characteristic temperature of the fly ash;

the newly determined fly ash component content data is U_k+1The melting characteristic temperature data of the fly ash is y_k+1Root of Chinese character' systematic ginsengNumber theta_k+1Then, then

Solving for theta by generalized inverse_k+1：

Wherein the content of the first and second substances,

order toWhen the update data is one entry, then:

s6: the fly ash melting characteristic temperature is measured on line according to the model corrected at step S5.