Hearth temperature measuring system under MISO framework

Abstract

The invention discloses a hearth temperature measurement system under an MISO framework, and the system comprises a sensor module which is used for obtaining the operation data of a unit; the data caching module is used for storing input and output data; the preprocessing module is used for preprocessing the unit data; the HSP processing module is used for outputting a hearth temperature measurement value; after the HSP processing module obtains the preprocessed data, a cubic spline function is constructed, a Hammerstein modular model is constructed, PSO optimization and de-normalization processing are carried out, and then a measurement result of the boiler furnace temperature is output; and the data output module is used for outputting a visual image. The hearth temperature soft measurement system provided by the invention can realize accurate measurement of the hearth temperature, and is suitable for further popularization and application.

Description

Technical Field

[0001] The present invention relates to the technical field of furnace temperature soft measurement, and in particular to a furnace temperature measurement system under a MISO framework. Background Art

[0002] The soft measurement method for furnace temperature is a technology that uses existing process parameters, mathematical models, and statistical methods to estimate furnace temperature without direct temperature sensors or data acquisition equipment. The steps include selecting and determining auxiliary variables, acquiring and processing sample data, establishing soft measurement models, and online correction. The core is the establishment of an offline model. There are many methods, such as those based on process mechanism analysis, regression analysis, artificial neural networks, pattern recognition, fuzzy mathematics, state estimation, correlation analysis, and modern nonlinear information processing technology, but they can be roughly divided into the following three categories:

[0003] a) Heat Balance Method: This method calculates the furnace temperature by establishing a heat balance equation, taking into account the heat input and output of the furnace. Based on the principles of thermodynamics, this method can better reflect the actual process and accurately calculate the furnace temperature when the input parameters are accurate.

[0004] b) Empirical model method: Based on historical data, a regression model or curve fitting model is established to correlate temperature with other process parameters (such as fuel flow, air volume, flue gas composition, etc.). The model is simple and can quickly achieve temperature estimation when there is abundant data.

[0005] c) Statistical and data-driven methods: These methods use regression analysis, neural networks (CN118009322A, CN102175345A), and support vector machines to establish a mapping relationship between data parameters and furnace temperature. Each method has its own advantages and disadvantages. For example, regression analysis offers fast calculations but requires high data quality; neural networks and support vector machines can adapt to complex high-dimensional problems but require longer training times.

[0006] Among soft-sensing methods, the heat balance method requires consideration of multiple factors when establishing the heat balance equation, resulting in a highly complex model. Some thermophysical parameters (such as thermal conductivity and heat loss) are difficult to accurately measure in practice. Furthermore, for complex furnace configurations or when burning a mixture of multiple fuels, the model may require recalibration. Empirical model methods can compromise measurement accuracy when data is insufficient, and they generally assume linear input and output, which can be unsatisfactory for complex systems. Summary of the Invention

[0007] Based on the above, the purpose of the present invention is to propose a multiple-input single-output (MISO) furnace temperature soft measurement system based on the HSP (Hammerstein-Spline-PSO) model.

[0008] In order to achieve the above technical objectives, the technical solution adopted by the present invention is: a furnace temperature measurement system under the MISO framework, comprising:

[0009] Sensor module, used to obtain unit operation data;

[0010] Data cache module, used to store input and output data;

[0011] Preprocessing module, used for preprocessing unit data;

[0012] HSP processing module, used to output furnace temperature measurement value;

[0013] After obtaining the pre-processed data, the HSP processing module constructs a cubic spline function, builds a Hammerstein modular model, performs PSO optimization and denormalization processing, and then outputs the measurement results of the boiler furnace temperature;

[0014] Data output module, used to output visualization images.

[0015] Furthermore, the sensor module includes a temperature sensor, a pressure sensor, a flow sensor, an oxygen content sensor, a current sensor, and a weighing sensor;

[0016] The temperature sensors are respectively arranged at the high-temperature superheater outlet, the reheat steam pipe, the air preheater primary air outlet, the air preheater secondary air outlet, the low-temperature economizer outlet, the feed water pump outlet, the fuel pump pipe outlet and the inside of the furnace;

[0017] The pressure sensors are respectively arranged at the main steam valve, the reheat steam main pipe, the primary air outlet of the air preheater and the secondary air outlet of the air preheater;

[0018] The flow sensors are respectively arranged at the feed water pump outlet, the main steam pipe, the primary air inlet, the secondary air inlet and the superheater feed water inlet;

[0019] The oxygen content sensors are respectively arranged at the outlet of the blower and the outlet of the rear wall;

[0020] The current sensors are respectively arranged on the primary fan motor and the blower motor;

[0021] The weighing sensor is arranged on the coal conveyor belt.

[0022] Furthermore, the unit operation data obtained through the sensor module include: main feed water flow, main steam flow, main steam valve front pressure, high temperature superheater outlet steam temperature, reheat steam main pipe pressure, reheat steam main pipe temperature, oxygen content, air preheater B outlet primary air temperature, air preheater B outlet secondary air temperature, air preheater B outlet primary air pressure, air preheater B outlet secondary air pressure, total fuel quantity, total air volume, total primary air volume, total secondary air volume, A low temperature economizer outlet flue gas temperature, main feed water temperature, rear Oxygen content at the wall outlet, ignition oil temperature, superheater B-water reduction flow rate, A primary fan motor current, A supply fan motor current, B primary fan motor current, B supply fan motor current, A primary fan adjustable vane valve position feedback, B primary fan adjustable vane valve position feedback, A supply fan adjustable vane valve position feedback, B supply fan adjustable vane valve position feedback, A steam-driven induced draft fan adjustable vane valve position feedback, B steam-driven induced draft fan adjustable vane valve position feedback, B air preheater flue gas temperature, furnace temperature signal;

[0023] The five variables, namely main feed water flow, high-temperature superheater outlet steam temperature, reheat steam main pipe temperature, A low-temperature economizer outlet flue gas temperature, and furnace temperature, are selected as characteristic variables.

[0024] Furthermore, the sensor transmits data every 20 seconds, continuously collecting 1,350 sets of data as samples, each of which contains the above five variables.

[0025] Furthermore, the data cache module is used to store historical data and real-time data detected by the sensor module, as well as the boiler furnace temperature output by the data output module;

[0026] At the same time, the data cache module can also transmit the data measured by the sensor to the data output module for error analysis and accuracy detection.

[0027] Furthermore, the preprocessing module is used to preprocess the acquired data. The preprocessing steps include:

[0028] Firstly, the data is processed by threshold denoising method, and then the denoised data is normalized and dimensionless using Min-Max method.

[0029] Furthermore, after obtaining the pre-processed data, the HSP processing module constructs a cubic spline function, builds a Hammerstein modular model, performs PSO optimization, and performs denormalization processing to output the measurement results of the boiler furnace temperature. Specifically, the steps include:

[0030] S10: Obtain preprocessed data

[0031] The main feed water flow, high-temperature superheater outlet steam temperature, reheat steam main pipe temperature, and B air preheater outlet primary air temperature processed by the pre-processing module are used as feature input data to construct the input matrix;

[0032] S20: Constructing a cubic spline function

[0033] The cubic spline function expression is as follows:

[0034]

[0035] Among them, (α0,α1,α2,α3,β1,β2...β j ...β n-1 ) is an arbitrary constant and is also the parameter to be identified in the nonlinear part of the Hammerstein model; n is the number of subintervals after the interval [a, b] is divided, x j is the split node;

[0036] Based on the four variable data of the pre-treated main feed water flow, the high-temperature superheater outlet steam temperature, the reheat steam main pipe temperature, and the B air preheater outlet primary air temperature, four cubic spline functions are constructed:

[0037]

[0038] The parameter superscripts (1), (2), (3), and (4) represent four variables respectively; the parameters to be identified in each spline function are independent of each other;

[0039] S30: Constructing Hammerstein module model

[0040] The difference equation of the Hammerstein model is expressed as:

[0041]

[0042] The nonlinear part is expressed as follows:

[0043]

[0044] The linear part expression is as follows:

[0045]

[0046] In the above formula, (u1(k), u2(k), u3(k), u4(k)) are the four input signals at time k, y(k) and e(k) are the output and noise at time k respectively, and x(k) is the output of the nonlinear part and the input of the linear part at time k. The above method uses a cubic spline function to approximate the memoryless nonlinear function f(·), where (γ1, γ2, γ3, γ4) are the weight coefficients corresponding to the four input variables, and the sum is 1. A(q -1 ) and B(q -1 ) are nth and mth order backshift operator polynomials respectively, d is the system delay, assuming zn +a1z n-1 +a2z n-2 +...+a n = 0 are all within the unit circle;

[0047] S40: PSO Optimization

[0048] The particle position X in the PSO algorithm i =[x i1 ,x i2 ,...,x id ,...,x iD ]for: The minimum fitness E of the Hammerstein-Spline subfunction is optimized by PSO A , that is, the error between the actual output of the system and the calculated output; the error is measured by the following criterion function:

[0049]

[0050] Among them, y(k) is the actual output of the system, is the output of the estimated model; the system identification problem is transformed into the value of minimizing formula (16), and then the model parameters are estimated; N represents the total number of outputs;

[0051] The PSO optimization steps are as follows:

[0052] a) Initialize each particle in a population of size Q, including the position and velocity of the particle; the initial optimal position of the particle p i For itself, then, calculate the initial optimal position p of the group according to the fitness function g ;

[0053] b) During each iteration, calculate the p of each particle in the group i and the p of the entire group g , then calculate the update speed of each particle and update the particle position;

[0054] c) The conditions for the PSO algorithm to end are: reaching the maximum number of iterations or the Hammerstein-Spline objective function reaches a certain set value. If either of the two conditions is met, the algorithm ends. Otherwise, the cycle b) is repeated until the algorithm ends.

[0055] Furthermore, the data output module is electrically connected to the HSP processing module, and can output the results of the HSP processing module and form a visual image.

[0056] Furthermore, the data output module is a display, from which the obtained boiler furnace temperature value can be directly read, and the data output module can also apply to the data cache module to obtain the historical furnace temperature;

[0057] The furnace temperature measured by the thermocouple obtained from the data cache module is checked for accuracy against the calculated value and a root mean square error curve is output. If the error is less than the specified accuracy, the calculation is considered reliable. The information of the data output module is also recorded in the data cache module so that it can be directly retrieved later.

[0058] This paper improves the existing single-input single-output (SISO) furnace temperature soft-sensing method and proposes a MISO-based Hammerstein-Spline-PSO furnace temperature measurement system. The improvements and technical problems solved by this system are as follows:

[0059] 1) Traditional temperature calculation models typically use linear or simple nonlinear models. The Hammerstein-Spline model combines the Hammerstein model's ability to handle input nonlinearity with the flexible fitting capabilities of the spline function. This improves the model's accuracy in fitting highly nonlinear system dynamics and more accurately reflects the complex relationship between input variables and output temperature, thereby improving measurement accuracy. This solves the problem of maintaining model simplicity while enhancing the model's adaptability to the nonlinear dynamics of complex systems.

[0060] 2) This paper uses PSO to solve the problem of identifying parameters in the Hammerstein-Spline algorithm. Compared with traditional optimization methods based on gradient descent or other deterministic algorithms, PSO provides global search capabilities, can find the optimal solution in a wider range, reduce the risk of falling into local optimality, and thus optimize the computational performance of the entire model. It solves the problem of optimizing a large number of unknown parameters in the Hammerstein-Spline model.

[0061] 3) Compared to the SISO model, the MISO model can handle multiple input variables, such as feedwater flow, superheater outlet steam temperature, and reheat steam temperature, which are key factors affecting furnace temperature. This model can more comprehensively map furnace temperature changes, improve the responsiveness and efficiency of the control system, and thus optimize combustion efficiency and reduce energy consumption. It also solves the problem of how to integrate multi-source data and jointly evaluate their impact on furnace temperature.

[0062] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art:

[0063] Compared with traditional linear and simple nonlinear models, the present invention combines the nonlinear processing of the Hammerstein model and the high flexibility of Spline, thereby improving the model's adaptability to nonlinear dynamics and measurement accuracy; unlike the gradient descent method that is prone to falling into local optimality, the present invention uses the PSO algorithm to optimize the parameters to be identified in Hammerstein-Spline, thereby improving global search capabilities and optimization efficiency; unlike the simple SISO architecture, the MISO architecture can integrate a variety of influencing factors, enabling the model to effectively process and integrate a variety of input variables under different industrial environments and conditions. The present invention embeds the HSP program as a processing module into the actual system, enhancing its practical application value, namely the accurate measurement of furnace temperature. In addition, if more sensors or actuators need to be added in the future, the system embedded with the HSP module can adapt to new needs through simple software updates or hardware expansions. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0065] Figure 1 This is a simplified structural diagram of the HSP-based soft measurement system for furnace temperature in thermal power plants;

[0066] Figure 2 This is a flow chart of the furnace temperature measurement method;

[0067] Figure 3 This is the structure diagram of the Hammerstein model based on MISO;

[0068] Figure 4 This is a simplified flow chart of the PSO algorithm. DETAILED DESCRIPTION

[0069] The present invention will be described in further detail below with reference to the accompanying drawings and examples. It is particularly noted that the following examples are intended only to illustrate the present invention and are not intended to limit the scope of the present invention. Similarly, the following examples are only some embodiments of the present invention and are not intended to be exhaustive. All other embodiments obtained by those of ordinary skill in the art without creative effort are intended to fall within the scope of protection of the present invention.

[0070] Refer to the attached Figure 1As shown, this embodiment provides a multiple-input single-output (MISO) furnace temperature soft measurement system based on the HSP (Hammerstein-Spline-PSO) model, which includes: a sensor module, a data cache module, a preprocessing module, an HSP processing module, and a data output module.

[0071] The sensor module 100 is used to obtain the unit operation data.

[0072] In this embodiment, the sensor module includes the following sensor types: temperature sensor, pressure sensor, flow sensor, oxygen content sensor, current sensor, and weighing sensor. The temperature sensors are located at the high-temperature superheater outlet, reheat steam pipe, air preheater primary air outlet, air preheater secondary air outlet, low-temperature economizer outlet, feedwater pump outlet, fuel pump pipe outlet, and furnace interior; pressure sensors are located at the main steam valve, reheat steam main pipe, air preheater primary air outlet, and air preheater secondary air outlet; flow sensors are located at the feedwater pump outlet, main steam pipe, primary air inlet, secondary air inlet, and superheater feedwater inlet; oxygen content sensors are located at the blower outlet and rear wall outlet; current sensors are located at the primary fan motor and blower motor; and weighing sensor is located on the coal conveyor belt.

[0073] According to the above-mentioned arrangement of sensors, the following data can be obtained: main feed water flow, main steam flow, main steam valve pressure, high temperature superheater outlet steam temperature, reheat steam main pipe pressure, reheat steam main pipe temperature, oxygen content, air preheater B outlet primary air temperature, air preheater B outlet secondary air temperature, air preheater B outlet primary air pressure, air preheater B outlet secondary air pressure, total fuel quantity, total air volume, total primary air volume, total secondary air volume, A low temperature economizer outlet flue gas temperature, main feed water temperature, rear wall outlet oxygen 32 variables including flow rate, ignition oil temperature, superheater B-water reduction flow, A primary fan motor current, A blower motor current, B primary fan motor current, B blower motor current, A primary fan adjustable vane valve position feedback, B primary fan adjustable vane valve position feedback, A blower adjustable vane valve position feedback, B blower adjustable vane valve position feedback, A steam-driven induced draft fan adjustable vane valve position feedback, B steam-driven induced draft fan adjustable vane valve position feedback, B air preheater flue gas temperature, and furnace temperature signal.

[0074] Considering the model's complexity and speed, we empirically simplified the 32 variables to five: main feedwater flow rate, high-temperature superheater outlet steam temperature, reheat steam main pipe temperature, low-temperature economizer A outlet flue gas temperature, and furnace temperature (for accuracy verification). This simplification helps shorten model training time and improve its training rate. The sensor transmits data every 20 seconds. This example continuously collects 1,350 sets of data as samples, each containing the five variables mentioned above.

[0075] The data cache module 200 is used to store input and output data.

[0076] In this embodiment, the data cache module connects to the sensor module via a twisted-pair cable. Transmission based on the CAN bus protocol ensures reliable transmission even in high-noise environments. As the system's storage unit, the data cache module stores historical and real-time data detected by the sensor module, as well as the boiler furnace temperature output by the data output module. Furthermore, the data cache module transmits sensor data to the data output module for error analysis and accuracy testing.

[0077] The pre-processing module 300 is used to pre-process the unit operation data;

[0078] In this embodiment, the acquired unit operating data is preprocessed. The preprocessing steps include: first, using a threshold denoising method to process the data to improve model performance, enhance model robustness, and reduce its computational complexity. Then, the denoised data is normalized and dimensionless using the Min-Max method to eliminate dimensionality effects and improve algorithm efficiency.

[0079] The HSP processing module 400 is used to output the furnace temperature measurement value.

[0080] Refer to the attached Figure 2 As shown, in this embodiment, after the HSP processing module obtains the pre-processed data, it measures the boiler furnace temperature by constructing a cubic spline function, building a Hammerstein modular model, and performing particle swarm optimization (PSO) optimization and denormalization processing, and outputs the measurement results. The specific steps include:

[0081] S10: Obtain preprocessed data

[0082] The main feed water flow, high-temperature superheater outlet steam temperature, reheat steam main pipe temperature, and B air preheater outlet primary air temperature processed by the pre-processing module are used as feature input data to construct the input matrix;

[0083] S20: Constructing a cubic spline function

[0084] Splines were originally thin wooden strips used by cartographers to draw smooth curves. The purpose of using these strips was to require that the curves pass through some known points and that the connections between each curve segment are smooth.

[0085] Given a partition of the interval [a,b], a=x0 <x1<...<x n =b. In order to construct a specific k-degree polynomial spline function, let

[0086]

[0087] say is the mth-order half-truncated power function, especially when m=0, we have

[0088]

[0089] Formula (2) is called the unit jump function.

[0090] The k-degree spline function is

[0091]

[0092] where {α j}, {β j}, are two sets of arbitrary constants.

[0093] The cubic spline function is the most widely used spline function in engineering. The cubic spline function is:

[0094]

[0095] Among them, (α0,α1,α2,α3,β1,β2...β j ...β n-1 ) is an arbitrary constant and is also the parameter to be identified in the nonlinear part of the Hammerstein model; n is the number of subintervals after the interval [a, b] is divided, x j For splitting nodes.

[0096] In this embodiment, four types of variable data are collected, so four cubic spline functions need to be generated:

[0097]

[0098] The parameter superscripts (1), (2), (3), and (4) represent four variables respectively; the parameters to be identified in each spline function are independent of each other.

[0099] S30: Constructing Hammerstein module model

[0100] In this embodiment, the Hammerstein model of MISO is as shown in the attached figure. Figure 3 As shown. In the case of MISO, the Hammerstein model consists of a memoryless nonlinear module and a memory linear module in series, and can be used to represent most practical systems containing nonlinear characteristics. In general, the Hammerstein model difference equation is described as:

[0101]

[0102] The nonlinear part is expressed as follows:

[0103]

[0104] The linear part expression is as follows:

[0105]

[0106] In the above formula, (u1(k), u2(k), u3(k), u4(k)) are the four input signals at time k; y(k) and e(k) are the output and noise at time k, respectively; x(k) is the output of the nonlinear part and the input of the linear part at time k, which is immeasurable; in this embodiment, the memoryless nonlinear function f(·) is approximated by the cubic spline function; (γ1, γ2, γ3, γ4) are the weight coefficients corresponding to the four input variables, and the sum is 1; A(q -1 ) and B(q -1 ) are nth and mth order backshift operator polynomials respectively, and d is the system delay; n and m in the above formula (12) are the highest orders of the polynomials. The larger n is, the slower the system response is. The two reflect the dynamic characteristics of the system. n Corresponding polynomial coefficients, reflecting the weights of different orders; b0~b m Similarly, assuming z n +a1z n-1 +a2z n-2 +...+a n = 0 are all within the unit circle to ensure the stability of the discrete system; under this assumption, the system dynamic response converges and will not diverge or continue to oscillate, ensuring that the system is controllable.

[0107] S40: PSO Optimization

[0108] The PSO algorithm is a type of swarm intelligence stochastic optimization algorithm that simulates the foraging method of bird flocks. It adopts the two concepts of "swarm" and "evolution" commonly used in intelligent optimization methods. The fitness function is generally related to a certain identification criterion. Compared with other intelligent algorithms, this algorithm is unique in that it borrows the "position" and "speed" in physics to find model parameters. Each solution is abstracted as a massless and volumeless particle in the solution space. It flies in the solution space at a certain speed and determines its next speed and position by learning the optimal flight path of itself and other particles until the optimal solution is found.

[0109] The standard particle swarm optimization algorithm is used to solve continuous optimization problems, which can be described as follows (taking the minimization problem as an example): Let S be R n A bounded subset of There is an n-dimensional real function f:

[0110] S→R n , find X min ∈S, so that Satisfy f(X min )≤f(X).

[0111] Generally, the initialization process of the PSO algorithm is as follows: set the dimension of the solution space to D, the population size to Q, and the position and velocity of the i-th particle to X i =[x i1 ,x i2 ,...,x id ,...,x iD ] and V i =[v i1 ,v i2 ,...,v id ,...,v iD ], the particle position search range is [x min ,x max ], the particle velocity search range is [v min ,v max ]. Finally, we give the Q min ,x max ]The initial individual X obeys the uniform distribution i (i=1,2,...,Q), Q in [v min ,v max ]The initial individual V follows a uniform distribution i (i=1,2,...,Q). In addition, the individual extreme value p i =[p i1 ,p i2 ,...,p id ,...,p iD ], the global extreme value p of the populationg =[p g1 ,p g2 ,...,p gd ,...,p gD ].

[0112] The i-th particle evolves from generation t to generation t+1 according to the following formula:

[0113]

[0114] Where d = 1, 2, ..., D; i = 1, 2, ..., Q. D is the dimension of each particle, indicating the search for the optimal solution in the D-dimensional solution space, that is, the number of parameters to be optimized; Q is the size of the group, that is, the group consisting of Q particles together to seek the optimal solution of the model; w is the inertia weight of the particle velocity change, which is the parameter of the inertial motion of the particle in the search space; c1 is the particle's own learning coefficient, and c2 is the particle's group learning coefficient. In a statistical sense, these two parameters represent the acceleration weight coefficients that push each particle towards the individual optimal value and the global optimal value, respectively; rand() is a random number in the interval (0, 1).

[0115] The value of w is determined by (14):

[0116]

[0117] Where, t max is the maximum number of iterations of the algorithm, t is the number of current cycles, w max is the maximum inertia weight coefficient, w min is the minimum inertia weight coefficient. Generally, w max =0.9, w min =0.4 can meet the requirements. The larger the w value, the stronger the global search ability of the algorithm. The smaller the w value, the stronger the local search ability of the algorithm.

[0118] In the search space, particles update their speed and position in this iteration according to the individual optimal value and group optimal value in the previous learning evolution process until the maximum number of iterations is reached or the model error meets the requirements, and the global optimal solution p in the search space is obtained. g location.

[0119] In this embodiment, the particle position X i =[x i1 ,x i2 ,...,x id ,...,x iD ]for: The minimum fitness E of the Hammerstein-Spline subfunction is optimized by PSOA , that is, the error between the actual output of the system and the calculated output; the error is measured by the following criterion function:

[0120]

[0121] Among them, y(k) is the actual output of the system, is the output of the estimated model; the system identification problem is transformed into minimizing the value of formula (16), and then estimating the model parameters; N is the total number of outputs.

[0122] Refer to the attached Figure 4 As shown, the PSO optimization process is as follows:

[0123] a) Initialize each particle in a population of size Q, including the particle's position and velocity; generally, the particle's initial optimal position p i For itself, then, calculate the initial optimal position p of the group according to the fitness function g ;

[0124] b) During each iteration, calculate the p of each particle in the group i and the p of the entire group g , and then calculate the update speed of each particle according to formula (13), and then update the particle position;

[0125] c) There are two conditions for the termination of the PSO algorithm: reaching the maximum number of iterations or the Hammerstein-Spline objective function reaching a certain set value. If either of the two conditions is met, the algorithm ends. Otherwise, the algorithm repeats b) until the algorithm ends.

[0126] The data output module 500 is used to output a visual image.

[0127] In this embodiment, the data output module is electrically connected to the HSP processing module and can output the HSP processing module's results and create a visual image, namely, the calculated boiler furnace temperature value. This data output module serves as a display, from which the calculated boiler furnace temperature value can be directly read, or historical furnace temperature values ​​can be requested from the data cache module. Furthermore, the furnace temperature measured by the thermocouple obtained from the data cache module is compared with the calculated value for accuracy, and a root mean square error curve is output. If the error is less than the specified accuracy, the calculation is deemed reliable. Information from the data output module is also recorded in the data cache module for direct access later.

[0128] The above descriptions are only some embodiments of the present invention and do not limit the scope of protection of the present invention. Any equivalent device or equivalent process transformation made by using the contents of the description and drawings of the present invention, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A furnace temperature measurement system under the MISO framework, characterized in that: include: Sensor module, used to obtain unit operation data; Data cache module, used to store input and output data; Preprocessing module, used for preprocessing unit data; HSP processing module, used to output furnace temperature measurement value; After obtaining the pre-processed data, the HSP processing module constructs a cubic spline function, builds a Hammerstein modular model, performs PSO optimization and denormalization processing, and then outputs the measurement results of the boiler furnace temperature; Data output module, used to output visualization images.

2. The furnace temperature measurement system under the MISO framework according to claim 1 is characterized in that: The sensor module includes a temperature sensor, a pressure sensor, a flow sensor, an oxygen content sensor, a current sensor, and a weighing sensor; The temperature sensors are respectively arranged at the high-temperature superheater outlet, the reheat steam pipe, the air preheater primary air outlet, the air preheater secondary air outlet, the low-temperature economizer outlet, the feed water pump outlet, the fuel pump pipe outlet and the inside of the furnace; The pressure sensors are respectively arranged at the main steam valve, the reheat steam main pipe, the primary air outlet of the air preheater and the secondary air outlet of the air preheater; The flow sensors are respectively arranged at the feed water pump outlet, the main steam pipe, the primary air inlet, the secondary air inlet and the superheater feed water inlet; The oxygen content sensors are respectively arranged at the outlet of the blower and the outlet of the rear wall; The current sensors are respectively arranged on the primary fan motor and the blower motor; The weighing sensor is arranged on the coal conveyor belt.

3. The furnace temperature measurement system under the MISO framework according to claim 2 is characterized in that: The unit operation data obtained through the sensor module include: main feed water flow, main steam flow, main steam valve pressure, high temperature superheater outlet steam temperature, reheat steam main pipe pressure, reheat steam main pipe temperature, oxygen content, air preheater B outlet primary air temperature, air preheater B outlet secondary air temperature, air preheater B outlet primary air pressure, air preheater B outlet secondary air pressure, total fuel quantity, total air volume, total primary air volume, total secondary air volume, A low temperature economizer outlet flue gas temperature, main feed water temperature, rear wall Outlet oxygen quantity, ignition oil temperature, superheater B-water reduction flow rate, A primary fan motor current, A blower motor current, B primary fan motor current, B blower motor current, A primary fan adjustable vane valve position feedback, B primary fan adjustable vane valve position feedback, A blower adjustable vane valve position feedback, B blower adjustable vane valve position feedback, A steam-driven induced draft fan adjustable vane valve position feedback, B steam-driven induced draft fan adjustable vane valve position feedback, B air preheater flue gas temperature, furnace temperature signal; The five variables, namely main feed water flow, high-temperature superheater outlet steam temperature, reheat steam main pipe temperature, A low-temperature economizer outlet flue gas temperature, and furnace temperature, are selected as characteristic variables.

4. The furnace temperature measurement system under the MISO framework according to claim 3 is characterized in that: The sensor transmits data every 20 seconds, continuously collecting 1,350 sets of data as samples, each of which contains the above five variables.

5. The furnace temperature measurement system under the MISO framework according to claim 2, characterized in that: The data cache module is used to store historical data and real-time data detected by the sensor module, as well as the boiler furnace temperature output by the data output module; At the same time, the data cache module can also transmit the data measured by the sensor to the data output module for error analysis and accuracy detection.

6. The furnace temperature measurement system under the MISO framework according to claim 1 is characterized in that: The preprocessing module is used to preprocess the acquired data. The preprocessing steps include: Firstly, the data is processed by threshold denoising method, and then the denoised data is normalized and dimensionless using Min-Max method.

7. The furnace temperature measurement system under the MISO framework according to claim 1 is characterized in that: After obtaining the pre-processed data, the HSP processing module constructs a cubic spline function, builds a Hammerstein modular model, performs PSO optimization, and performs denormalization processing to output the measurement results of the boiler furnace temperature. The specific steps include the following: S10: Obtain preprocessed data The main feed water flow, high-temperature superheater outlet steam temperature, reheat steam main pipe temperature, and B air preheater outlet primary air temperature processed by the pre-processing module are used as feature input data to construct the input matrix; S20: Constructing a cubic spline function The cubic spline function expression is as follows: Among them, (α0,α1,α2,α3,β1,β2...β j ...β n-1 ) is an arbitrary constant and is also the parameter to be identified in the nonlinear part of the Hammerstein model; Based on the four variable data of the pre-treated main feed water flow, the high-temperature superheater outlet steam temperature, the reheat steam main pipe temperature, and the B air preheater outlet primary air temperature, four cubic spline functions are constructed: The parameter superscripts (1), (2), (3), and (4) represent four variables respectively; the parameters to be identified in each spline function are independent of each other; S30: Constructing Hammerstein module model The difference equation of the Hammerstein model is expressed as: The nonlinear part is expressed as follows: The linear part expression is as follows: In the above formula, (u1(k), u2(k), u3(k), u4(k)) are the four input signals at time k, y(k) and e(k) are the output and noise at time k respectively, and x(k) is the output of the nonlinear part and the input of the linear part at time k. The above method uses a cubic spline function to approximate the memoryless nonlinear function f(·), where (γ1, γ2, γ3, γ4) are the weight coefficients corresponding to the four input variables, and the sum is 1. A(q -1 ) and B(q -1 ) are nth and mth order backshift operator polynomials respectively, d is the system delay, assuming z n +a1z n-1 +a2z n-2 +...+a n = 0 are all within the unit circle; S40: PSO Optimization The particle position X in the PSO algorithm i =[x i1 ,x i2 ,...,x id ,...,x iD ]for: The minimum fitness E of the Hammerstein-Spline subfunction is optimized by PSO A , that is, the error between the actual output of the system and the calculated output; the error is measured by the following criterion function: Among them, y(k) is the actual output of the system, is the output of the estimation model; the system identification problem is transformed into the value of minimizing formula (16), and then the model parameters are estimated; The PSO optimization steps are as follows: a) Initialize each particle in a population of size Q, including the position and velocity of the particle; the initial optimal position of the particle p i For itself, then, calculate the initial optimal position p of the group according to the fitness function g ; b) During each iteration, calculate the p of each particle in the group i and the p of the entire group g , then calculate the update speed of each particle and update the particle position; c) The conditions for the PSO algorithm to end are: reaching the maximum number of iterations or the Hammerstein-Spline objective function reaches a certain set value. If either of the two conditions is met, the algorithm ends. Otherwise, the cycle b) is repeated until the algorithm ends.

8. The furnace temperature measurement system under the MISO framework according to claim 1 is characterized in that: The data output module is electrically connected to the HSP processing module and can output the results of the HSP processing module and form a visual image.

9. The furnace temperature measurement system under the MISO framework according to claim 8, characterized in that: The data output module is a display, from which the boiler furnace temperature value can be directly read. The data output module can also apply to the data cache module to obtain the historical furnace temperature; The furnace temperature measured by the thermocouple obtained from the data cache module is checked for accuracy against the calculated value and a root mean square error curve is output. If the error is less than the specified accuracy, the calculation is considered reliable. The information of the data output module is also recorded in the data cache module so that it can be directly retrieved later.

Images