Water-cooled wall soot blowing optimization method for preventing high-temperature corrosion based on GA-BP neural network

Abstract

The invention discloses a water-cooled wall soot blowing optimization method for preventing high-temperature corrosion based on a GA-BP neural network. The method comprises the following steps: 1, preprocessing operation data based on a weighted moving window; 2, selecting input and output parameters of the neural network; 3, setting GA-BP neural network parameters; 4, training a GA-BP neural network of the heat transfer capacity when the water cooling wall is theoretically cleaned and the actual heat transfer capacity of the water cooling wall; 5, superposition fitting is conducted on the heat loss change curve of the water cooling wall in the boiler operation process; 6, the soot blowing optimization model is solved based on the maximum unit time soot blowing net income, and the optimal soot blowing opportunity and the optimal soot blowing duration are obtained; and 7, dynamically adjusting the soot blowing opportunity and the soot blowing time according to the soot blowing strategy. The method can accurately monitor the heat loss change condition of the water cooling wall, and is suitable for practical engineering application.

Description

Technical Field

[0001] The invention relates to a boiler water-cooled wall soot blowing control method, and in particular to a water-cooled wall soot blowing control method based on neural network optimization. Background Art

[0002] When coal-fired boilers are in operation, slagging of water-cooled walls is inevitable. When slag adheres to the water-cooled walls, it will cause high-temperature corrosion of the water-cooled walls, affecting the safe operation of the boiler. Soot blowing is the most effective way to remove slagging of water-cooled walls, but before soot blowing, the slagging state of the water-cooled walls needs to be monitored in real time, and the original soot blowing strategy needs to be improved.

[0003] Soot blowers are usually composed of soot blowing pipes and operating valves. They are usually installed on the furnace walls of boiler furnaces, horizontal flues, and tail flues. According to the structural type, there are several types, such as gun type, rotary type, and telescopic type. According to the different media used, there are three types: air soot blowers, steam soot blowers, and hydraulic soot blowers. Air soot blowers use the impact force of high-pressure airflow of compressed air to blow away the accumulated ash on the heating surface; they are often used in industrial boilers. Steam soot blowers use the impact force of high-speed jets of superheated steam to remove the ash on the heating surface, and are mostly used in various power plant boilers. Hydraulic soot blowers, also known as "slag blowers", use high-pressure water as a medium to promote the removal of ash and slag bonded to the heating surface due to sudden cooling and brittle cracking; they are only suitable for individual special occasions. Ash accumulation on the heating surface will affect the heat transfer effect, increase the exhaust gas temperature, reduce the boiler efficiency, and in severe cases, cause the superheater to burst. Therefore, regular use of soot blowers to promptly remove the accumulated ash on various heating surfaces plays an important role in ensuring the safe and economical operation of the boiler.

[0004] The existing water-cooled wall slagging status monitoring model has a complex calculation process. When the boiler load changes rapidly, it will cause errors in the thermal parameter measurement point data, which will affect the calculation results of the model; using a fixed sootblowing plan may cause under-blowing or over-blowing of the water-cooled wall. Summary of the invention

[0005] Purpose of the invention: In view of the above-mentioned prior art, a water-cooled wall sootblowing optimization method based on GA-BP neural network to prevent high-temperature corrosion is proposed, which can realize fast and accurate monitoring of the slagging state of the water-cooled wall while dynamically adjusting the sootblowing timing and duration.

[0006] Technical solution: A water-cooled wall soot blowing optimization method based on GA-BP neural network to prevent high-temperature corrosion, including:

[0007] Step 1: Preprocess the running data based on the weighted moving window;

[0008] Step 2: Select the neural network input and output parameters;

[0009] Step 3: Set the GA-BP neural network parameters;

[0010] Step 4: training the GA-BP neural network of the theoretical heat transfer of the water-cooled wall when it is clean and the actual heat transfer of the water-cooled wall;

[0011] Step 5: Superimpose and fit the heat loss variation curve of the water-cooled wall during boiler operation;

[0012] Step 6: Solve the sootblowing optimization model based on the maximum net sootblowing benefit per unit time to obtain the optimal sootblowing timing and duration;

[0013] Step 7: Dynamically adjust the soot blowing timing and time according to the soot blowing strategy.

[0014] Furthermore, in step 1, historical data is collected from the DCS database based on a weighted moving window, the sampling interval is 1 minute, the size of the moving window is set to 4, and the weights of the observation values ​​at different time points from left to right are 0.1, 0.2, 0.3 and 0.4 respectively.

[0015] Furthermore, in step 2, the output parameters of the neural network are respectively the heat transfer q of the water-cooled wall when it is theoretically cleaned c and the actual heat transfer of the water-cooled wall q r ;q c The input parameters of the system are composed of moisture, ash, volatile matter, sulfur content and low calorific value; the combustion parameters are composed of burner coal feed, oxygen content, furnace primary inlet air temperature and furnace secondary inlet air temperature; the operation parameters are composed of screen superheater steam flow and high temperature reheater outlet flue gas temperature; q r The input parameters are in q c Based on the input parameters of , add the platen superheater inlet steam temperature, platen superheater inlet steam pressure, high-temperature superheater outlet steam temperature and high-temperature superheater outlet steam pressure.

[0016] Furthermore, in step 3, the encoding mode, crossover operator, genetic generation, population size, selection operator parameters and mutation operator parameters of the GA genetic algorithm are set, and the number of hidden layer neurons, maximum number of iterations, error threshold and learning rate of the BP neural network are set to complete the specific parameter settings of GA-BP.

[0017] Furthermore, in step 4, the heat loss of the water-cooled wall is the theoretical heat transfer of the water-cooled wall when it is clean minus the actual heat transfer of the water-cooled wall, and the slagging condition of the water-cooled wall is monitored in real time through the established GA-BP neural network.

[0018] Further, in step 5, according to the heat transfer q of the water-cooled wall during theoretical cleaning, c and the actual heat transfer of the water-cooled wall q rThe neural network obtains the heat loss change curve of the water-cooled wall under different working conditions, and the heat loss change curve q of the slagging process is obtained by superposition fitting. L-f (t) and the heat loss variation curve q during soot blowing L-s (t).

[0019] Furthermore, in step 6, a sootblowing optimization model is established with the maximum net sootblowing benefit per unit time as the sootblowing optimization index, and the sootblowing optimization model is solved to obtain the optimal sootblowing timing and sootblowing duration, wherein the mathematical expressions of the sootblowing optimization model and boundary conditions are as follows:

[0020]

[0021]

[0022] Among them, Q net is the net soot blowing income; t f and t s are slagging time and sootblowing time respectively; D is the sootblowing steam flow rate; Δh is the enthalpy of sootblowing steam consumption.

[0023] Furthermore, in step 7, the critical heat loss change value of the water-cooled wall during the slagging and soot blowing process is obtained, the heat loss value of the water-cooled wall is calculated in real time and accumulated, and when the cumulative heat loss value exceeds the critical heat loss value of the slagging process, the soot blowing operation is started at this time, and the corresponding time is the actual soot accumulation time; when the cumulative heat loss value exceeds the critical heat loss value of the soot blowing process, the soot blowing is stopped at this time, thereby realizing dynamic adjustment of the soot blowing timing and soot blowing duration.

[0024] Beneficial effects: The method of the present invention preprocesses the collected data of the boiler based on a weighted moving window, and introduces a GA-BP neural network to improve the water-cooled wall slagging monitoring model. The optimal sootblowing timing and sootblowing time are obtained by solving the sootblowing optimization model, and finally a sootblowing strategy that can dynamically adjust the sootblowing timing and sootblowing duration is formulated.

[0025] 1. Compared with the existing method, the present invention simplifies the calculation process and improves the calculation accuracy when the boiler is running under variable operating conditions.

[0026] 2. The present invention proposes a sootblowing strategy for dynamically adjusting the sootblowing timing and duration, which is more suitable for actual engineering applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 is a flow chart of the method of the present invention;

[0028] Figure 2 It is a schematic diagram of the sootblowing optimization model based on maximizing the net sootblowing benefit per unit time;

[0029] Figure 3 Schematic diagram of critical heat loss and cumulative heat loss;

[0030] Figure 4 Sootblowing strategy for dynamically adjusting the timing and duration of sootblowing; DETAILED DESCRIPTION

[0031] The present invention will be further explained below in conjunction with the accompanying drawings.

[0032] A water-cooled wall soot blowing optimization method for preventing high-temperature corrosion based on GA-BP neural network, the specific steps are as follows.

[0033] Step 1: Preprocess the running data based on the weighted moving window.

[0034] In step 1, a moving window is first defined to assign different weights to observations at different time points, where observations with larger weights have a greater influence on the calculated average, thereby eliminating singular points and outliers in the running data.

[0035] In this embodiment, historical data is collected from a DCS (Distributed Control System) database, the sampling interval is 1 minute, the size of the moving window is set to 4, and the weights of the observation values ​​at different time points from left to right are 0.1, 0.2, 0.3 and 0.4 respectively.

[0036] Step 2: Select the neural network input and output parameters.

[0037] The output parameters of the neural network are the heat transfer q of the water-cooled wall when it is theoretically clean. c and the actual heat transfer of the water-cooled wall q r When selecting input parameters, you should select the same as q c and q r The most relevant parameter, coal quality has a direct impact on the furnace effective heat and flue gas volume, which in turn affects the theoretical combustion temperature. In order to facilitate the online acquisition of coal quality parameters, industrial analysis of coal is selected as part of the input parameters. The furnace oxygen content and inlet air temperature will also affect the combustion of coal.

[0038] In this embodiment, q c The input parameters are composed of moisture, ash, volatile matter, sulfur content and low calorific value; the combustion parameters are composed of burner coal feed rate, oxygen content, furnace primary inlet air temperature and furnace secondary inlet air temperature; the operation parameters are composed of screen superheater steam flow and high temperature reheater outlet flue gas temperature.

[0039] q c and q r The difference between the neural network is that q rThe neural network input parameters need to reflect the slagging of the water-cooled wall. The parameter that directly affects the slagging of the water-cooled wall is the flue gas temperature at the furnace outlet. The flue gas temperature at the furnace outlet is calculated by combining the heat transfer process of the furnace heating surface, the platen superheater and the high-temperature superheater. Therefore, q r The input parameters should include the platen superheater inlet steam temperature, platen superheater inlet steam pressure, high temperature superheater outlet steam temperature and high temperature superheater outlet steam pressure.

[0040] Step 3: Set the GA-BP neural network parameters.

[0041] The encoding method of the GA algorithm is set to real number encoding, the crossover operator uses single-point crossover, the crossover point is a randomly generated number between (0,1), the genetic algorithm's genetic generation is 70, the population size is 50, the selection operator is linear sorting selection, and the mutation operator is non-uniform mutation. The number of hidden layer neurons of the BP neural network is set to 7, the maximum number of iterations is 1000, the error threshold is 1e-6, and the learning rate is 0.01.

[0042] Step 4: Train the GA-BP neural network of the theoretical heat transfer of the water-cooled wall during cleaning and the actual heat transfer of the water-cooled wall to obtain the optimal weights and thresholds.

[0043] The heat loss of the water-cooled wall is the theoretical heat transfer of the water-cooled wall when it is clean minus the actual heat transfer of the water-cooled wall. The GA-BP neural network established by the present invention can obtain good accuracy and monitor the slagging of the water-cooled wall in real time.

[0044] Step 5: Superimpose and fit the heat loss variation curve of the water-cooled wall during boiler operation.

[0045] Specifically, the heat loss variation curves of the water-cooled wall slagging process and the soot blowing process of the boiler under different operating conditions are obtained, and the universal variation curves of the water-cooled wall slagging process and the soot blowing process are obtained by superposition fitting.

[0046] In this embodiment, according to the theoretical heat transfer q of the water-cooled wall during cleaning c and the actual heat transfer of the water-cooled wall q r The neural network obtains the heat loss change curve of the water-cooled wall under different working conditions, and the heat loss change curve q of the slagging process is obtained by superposition fitting. L-f (t) and the heat loss variation curve q during soot blowing L-s (t).

[0047] Step 6: Based on the maximum net sootblowing benefit per unit time, solve the sootblowing optimization model to obtain the optimal sootblowing timing and duration.

[0048] The sootblowing optimization model is established with the maximum net sootblowing benefit per unit time as the sootblowing optimization index. The sootblowing optimization model is solved to obtain the optimal sootblowing timing and sootblowing duration. The mathematical expressions of the sootblowing optimization model and boundary conditions are as follows:

[0049]

[0050] Among them, Q net is the net soot blowing benefit, kW; t f and t s are slagging time and sootblowing time, s respectively; D is the sootblowing steam flow rate, kg / s; Δh is the enthalpy of sootblowing steam consumption, kJ / kg.

[0051] Step 7: Dynamically adjust the soot blowing timing and time according to the soot blowing strategy.

[0052] The critical heat loss change value during the slagging and soot blowing process of the water-cooled wall is obtained, and the heat loss value of the water-cooled wall is calculated and accumulated in real time. When the accumulated heat loss value exceeds the critical heat loss value of the slagging process, the soot blowing operation is started at this time, and the corresponding time is the actual soot accumulation time; when the accumulated heat loss value exceeds the critical heat loss value of the soot blowing process, the soot blowing is stopped at this time, thereby realizing dynamic adjustment of the soot blowing timing and soot blowing duration.

[0053] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A water-cooled wall soot blowing optimization method based on GA-BP neural network to prevent high-temperature corrosion, characterized in that: include: Step 1: Preprocess the running data based on the weighted moving window; Step 2: Select the neural network input and output parameters; Step 3: Set the GA-BP neural network parameters; Step 4: training the GA-BP neural network of the theoretical heat transfer of the water-cooled wall when it is clean and the actual heat transfer of the water-cooled wall; Step 5: Superimpose and fit the heat loss variation curve of the water-cooled wall during boiler operation; Step 6: Solve the sootblowing optimization model based on the maximum net sootblowing benefit per unit time to obtain the optimal sootblowing timing and duration; Step 7: Dynamically adjust the soot blowing timing and time according to the soot blowing strategy.

2. The water-cooled wall sootblowing optimization method for preventing high-temperature corrosion based on GA-BP neural network according to claim 1 is characterized in that: In step 1, historical data is collected from the DCS database based on a weighted moving window, the sampling interval is 1 minute, the size of the moving window is set to 4, and the weights of the observation values ​​at different time points from left to right are 0.1, 0.2, 0.3 and 0.4 respectively.

3. The water-cooled wall sootblowing optimization method for preventing high-temperature corrosion based on GA-BP neural network according to claim 1 is characterized in that: In step 2, the output parameters of the neural network are respectively the heat transfer q of the water-cooled wall when it is theoretically cleaned c and the actual heat transfer of the water-cooled wall q r ;q c The input parameters of the system are composed of moisture, ash, volatile matter, sulfur content and low calorific value; the combustion parameters are composed of burner coal feed, oxygen content, furnace primary inlet air temperature and furnace secondary inlet air temperature; the operation parameters are composed of screen superheater steam flow and high temperature reheater outlet flue gas temperature; q r The input parameters are in q c Based on the input parameters of , add the platen superheater inlet steam temperature, platen superheater inlet steam pressure, high-temperature superheater outlet steam temperature and high-temperature superheater outlet steam pressure.

4. The water-cooled wall soot blowing optimization method for preventing high temperature corrosion based on GA-BP neural network according to claim 1 is characterized in that: In step 3, the encoding mode, crossover operator, genetic generation, population size, selection operator parameters and mutation operator parameters of the GA genetic algorithm are set, and the number of hidden layer neurons, maximum number of iterations, error threshold and learning rate of the BP neural network are set to complete the specific parameter settings of GA-BP.

5. The water-cooled wall sootblowing optimization method for preventing high-temperature corrosion based on GA-BP neural network according to claim 1 is characterized in that: In step 4, the heat loss of the water-cooled wall is the theoretical heat transfer of the water-cooled wall when it is clean minus the actual heat transfer of the water-cooled wall, and the slagging of the water-cooled wall is monitored in real time through the established GA-BP neural network.

6. The water-cooled wall soot blowing optimization method for preventing high-temperature corrosion based on GA-BP neural network according to claim 1 is characterized in that: In step 5, according to the heat transfer q of the water-cooled wall during theoretical cleaning c and the actual heat transfer of the water-cooled wall q r The neural network obtains the heat loss change curve of the water-cooled wall under different working conditions, and the heat loss change curve q of the slagging process is obtained by superposition fitting. L-f (t) and the heat loss variation curve q during soot blowing L-s (t).

7. The water-cooled wall sootblowing optimization method for preventing high-temperature corrosion based on GA-BP neural network according to claim 1 is characterized in that: In step 6, a sootblowing optimization model is established with the maximum net sootblowing benefit per unit time as the sootblowing optimization index, and the sootblowing optimization model is solved to obtain the optimal sootblowing timing and sootblowing duration, wherein the mathematical expressions of the sootblowing optimization model and boundary conditions are as follows: Among them, Q net is the net soot blowing income; t f and t s are slagging time and sootblowing time respectively; D is the sootblowing steam flow rate; Δh is the enthalpy of sootblowing steam consumption.

8. The water-cooled wall sootblowing optimization method for preventing high-temperature corrosion based on GA-BP neural network according to claim 1 is characterized in that: In step 7, the critical heat loss change value during the slagging and soot blowing process of the water-cooled wall is obtained, the heat loss value of the water-cooled wall is calculated in real time and accumulated, and when the accumulated heat loss value exceeds the critical heat loss value of the slagging process, the soot blowing operation is started at this time, and the corresponding time is the actual soot accumulation time; When the accumulated heat loss value exceeds the critical heat loss value of the soot blowing process, the soot blowing is stopped, thereby realizing dynamic adjustment of the soot blowing timing and duration.

Images