CN117452818A - Safe and economic coordination control method for depth peak shaving of 1000MW thermal power generating unit - Google Patents

Abstract

A safe economic coordination control method for depth peak shaving of a 1000MW thermal power generating unit obtains hearth pressure values at a plurality of preset time points in a preset time period and fire detection energy values at the preset time points; performing data preprocessing on the hearth pressure values at the plurality of preset time points and the fire detection energy values at the plurality of preset time points to obtain a hearth pressure time sequence input vector and a fire detection energy value time sequence input vector; performing association coding and feature extraction on the hearth pressure time sequence input vector and the fire detection energy value time sequence input vector to obtain a hearth pressure-fire detection energy value time sequence association feature diagram; and determining whether to throw the oil gun or not based on the hearth pressure-fire detection energy value time sequence correlation characteristic diagram. Therefore, the time sequence correlation change characteristics of the hearth pressure and the fire detection energy can be captured by using a deep learning algorithm, so that the intelligent control of the oil gun throwing is realized, and the running state of the unit is optimized.

Description

Safe and economic coordination control method for depth peak shaving of 1000MW thermal power generating unit

Technical Field

The application relates to the technical field of intelligent safe and economic control, and in particular relates to a safe and economic coordinated control method for deep peak shaving of a 1000MW thermal power generating unit.

Background

The 1000MW thermal power generating unit is a thermal power generating unit with the power generation capacity of 1000 megawatts, and generally consists of a steam turbine and a generator, wherein coal, gas or fuel oil is used as main fuel, high-temperature and high-pressure steam is generated through a boiler, and the steam turbine is driven to rotate to drive the generator to generate power.

The deep peak shaving refers to an operation mode of carrying out load lifting according to the fluctuation of the power grid load under the condition that the thermal power unit exceeds a basic peak shaving range. The purpose of deep peak shaving is to adapt to the large-scale grid connection of new energy, improve the regulation ability and the new energy consumption level of the power system, and simultaneously ensure the safety, stability and economy of the system.

In the deep peak shaving process, as the load is reduced, the input amount of coal dust is reduced, the lowest stable combustion load of the boiler is reduced, and an oil gun is required to be thrown at the moment to avoid fire extinguishment or explosion of the boiler. The oil gun can provide auxiliary fuel to ensure the continuous and stable combustion of the boiler. However, determining when an oil gun needs to be placed is a critical technical problem. The subjective judgment is performed manually in the prior art, but the method may have the problems of misjudgment or misjudgment and the like. Thus, an optimized coordinated control scheme is desired.

Disclosure of Invention

The present application has been made in order to solve the above technical problems. The embodiment of the application provides a safe and economic coordination control method for depth peak shaving of a 1000MW thermal power generating unit, which is used for acquiring hearth pressure values of a plurality of preset time points and fire detection energy values of the preset time points in a preset time period; performing data preprocessing on the hearth pressure values at the plurality of preset time points and the fire detection energy values at the plurality of preset time points to obtain a hearth pressure time sequence input vector and a fire detection energy value time sequence input vector; performing association coding and feature extraction on the hearth pressure time sequence input vector and the fire detection energy value time sequence input vector to obtain a hearth pressure-fire detection energy value time sequence association feature diagram; and determining whether to throw the oil gun or not based on the hearth pressure-fire detection energy value time sequence correlation characteristic diagram. Therefore, the time sequence correlation change characteristics of the hearth pressure and the fire detection energy can be captured by using a deep learning algorithm, so that the intelligent control of the oil gun throwing is realized, and the running state of the unit is optimized.

In a first aspect, a safe economic coordination control method for depth peak shaving of a 1000MW thermal power generating unit is provided, which comprises the following steps:

Acquiring hearth pressure values at a plurality of preset time points in a preset time period and fire detection energy values at the preset time points;

performing data preprocessing on the hearth pressure values at the plurality of preset time points and the fire detection energy values at the plurality of preset time points to obtain a hearth pressure time sequence input vector and a fire detection energy value time sequence input vector;

performing association coding and feature extraction on the hearth pressure time sequence input vector and the fire detection energy value time sequence input vector to obtain a hearth pressure-fire detection energy value time sequence association feature diagram; and

and determining whether to throw the oil gun or not based on the time sequence correlation characteristic diagram of the hearth pressure and fire detection energy value.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments or the description of the prior art will be briefly introduced below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a flowchart of a safe economic coordination control method for depth peaking of a 1000MW thermal power generating unit according to an embodiment of the present application.

Fig. 2 is a schematic architecture diagram of a safe economic coordination control method for depth peaking of a 1000MW thermal power generating unit according to an embodiment of the present application.

Fig. 3 is a block diagram of a safe economic coordination control system for depth peaking of a 1000MW thermal power unit according to an embodiment of the present application.

Fig. 4 is a schematic view of a scenario of a safe economic coordination control method for depth peaking of a 1000MW thermal power generating unit according to an embodiment of the present application.

Detailed Description

The following description of the technical solutions in the embodiments of the present application will be made with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.

Unless defined otherwise, all technical and scientific terms used in the examples of this application have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the present application is for the purpose of describing particular embodiments only and is not intended to limit the scope of the present application.

In the description of the embodiments of the present application, unless otherwise indicated and defined, the term "connected" should be construed broadly, and for example, may be an electrical connection, may be a communication between two elements, may be a direct connection, or may be an indirect connection via an intermediary, and it will be understood by those skilled in the art that the specific meaning of the term may be understood according to the specific circumstances.

It should be noted that, the term "first\second\third" in the embodiments of the present application is merely to distinguish similar objects, and does not represent a specific order for the objects, it is to be understood that "first\second\third" may interchange a specific order or sequence where allowed. It is to be understood that the "first\second\third" distinguishing objects may be interchanged where appropriate such that the embodiments of the present application described herein may be implemented in sequences other than those illustrated or described herein.

In recent years, the structure of the Chinese power supply is greatly changed, the new energy duty ratio is increased year by year, and the thermal power generating unit takes part in peak regulation and frequency modulation. Meanwhile, the peak-valley difference of the demand of the electricity market is larger and larger, and the thermal power unit is required to participate in the deep peak regulation task more and more. The thermal power generating unit needs to improve new energy consumption capability by improving flexibility and rapidity of the unit and bears the peak regulation and frequency modulation tasks of a power grid.

With the economic development of China entering a new normal state, the power production and consumption also presents a new normal state characteristic, and with the continuous and rapid increase of the power generation capacity of new energy and renewable energy, the power generation utilization hour of a thermal power generating unit continuously decreases, and the problem of surplus power supply capacity in local areas is further aggravated. The deep peak shaving phenomenon of the thermal power generating unit becomes normal in the next years. With the implementation of the Henan electric power peak shaving auxiliary service transaction rule, the deep peak shaving income is more obvious.

Under the deep peak regulation working condition, the unit load regulation range is wider, the requirement of the scheduling on the load regulation quality is stricter, the nonlinearity and time variability of the regulation object are more obvious, the influence of coal quality change, layered blending combustion and the like is further amplified, and the coordination control and the main automation under 50% Pe are not fully debugged, so that the regulation quality is required to be improved. Meanwhile, the main protection of low air quantity and low water supply flow rate MFT is close to an action value below 50% Pe, and the main protection needs to be specifically optimized under the deep peak regulation working condition. The operation of operators is increased under the deep peak-shaving working condition, the operation of one-key operation such as the operation of feeding pump withdrawal and pump sequential control, the operation of fan shutdown and fan doubling is needed to be increased, and the operation quantity of the operation under the deep peak-shaving working condition is reduced.

The expected target is that the load meets the regulation requirement when in dry running, the fluctuation range of main control parameters (main steam pressure, main steam temperature and the like) of the unit meets the regulation requirement, periodic fluctuation can not occur, the fluctuation range and the frequency of fuel quantity are reduced, and when combustion is deteriorated, the combustion can be rapidly judged and stabilized by feeding oil, so that the combustion safety under the deep peak regulation working condition is improved. And a one-key sequential control function group is designed by increasing frequent operation under the deep peak regulation working condition, so that the running operation quantity is greatly reduced.

Fig. 1 is a flowchart of a safe economic coordination control method for depth peaking of a 1000MW thermal power generating unit according to an embodiment of the present application. Fig. 2 is a schematic architecture diagram of a safe economic coordination control method for depth peaking of a 1000MW thermal power generating unit according to an embodiment of the present application. As shown in fig. 1 and 2, the safe economic coordination control method for depth peak shaving of a 1000MW thermal power generating unit includes: 110, acquiring hearth pressure values at a plurality of preset time points in a preset time period and fire detection energy values at the plurality of preset time points; 120, performing data preprocessing on the hearth pressure values at the plurality of preset time points and the fire detection energy values at the plurality of preset time points to obtain a hearth pressure time sequence input vector and a fire detection energy value time sequence input vector; 130, performing association coding and feature extraction on the hearth pressure time sequence input vector and the fire detection energy value time sequence input vector to obtain a hearth pressure-fire detection energy value time sequence association feature map; and 140, determining whether to throw the oil gun or not based on the hearth pressure-fire detection energy value time sequence correlation characteristic diagram.

Aiming at the technical problems, the technical concept of the method is to detect the hearth pressure value and the fire detection energy value, capture the time sequence correlation change characteristics of the hearth pressure and the fire detection energy by using a deep learning algorithm, so as to realize intelligent control of oil gun throwing and optimize the running state of a unit.

Based on the above, in the technical scheme of the application, firstly, furnace pressure values of a plurality of preset time points in a preset time period and fire detection energy values of the preset time points are obtained; and arranging the hearth pressure values at the plurality of preset time points and the fire detection energy values at the plurality of preset time points into a hearth pressure time sequence input vector and a fire detection energy value time sequence input vector according to the time dimension respectively.

In one specific embodiment of the present application, the data preprocessing is performed on the furnace pressure values at the plurality of predetermined time points and the fire detection energy values at the plurality of predetermined time points to obtain a furnace pressure time sequence input vector and a fire detection energy value time sequence input vector, including: and arranging the hearth pressure values at the preset time points and the fire detection energy values at the preset time points into the hearth pressure time sequence input vector and the fire detection energy value time sequence input vector according to the time dimension respectively.

And then, carrying out association coding and feature extraction on the hearth pressure time sequence input vector and the fire detection energy value time sequence input vector to obtain a hearth pressure-fire detection energy value time sequence association feature map. It should be understood that in the thermal power generation process, there is a complex, nonlinear relationship between the furnace pressure and the fire detection energy. Such an association relationship may be established by a correspondence relationship in the time dimension. Moreover, the association relation has important significance for judging when to put in the oil gun. Specifically, when the pressure fluctuation of the hearth is large, oil injection is required to be stable; in addition, when the fire detection energy is low, a corresponding oil gun is required to be thrown to realize stable combustion of the oil gun.

In one specific embodiment of the present application, performing association coding and feature extraction on the furnace pressure time sequence input vector and the fire detection energy value time sequence input vector to obtain a furnace pressure-fire detection energy value time sequence association feature map, including: performing association coding on the hearth pressure time sequence input vector and the fire detection energy value time sequence input vector to obtain a hearth pressure-fire detection energy value time sequence association matrix; and extracting correlation features of the hearth pressure-fire detection energy value time sequence correlation matrix by using a deep learning model to obtain a hearth pressure-fire detection energy value time sequence correlation feature map.

Further, the deep learning model is a time sequence correlation feature extractor based on a convolutional neural network model; the method for extracting the correlation characteristic of the hearth pressure-fire detection energy value time sequence correlation matrix by using a deep learning model to obtain the hearth pressure-fire detection energy value time sequence correlation characteristic map comprises the following steps: and passing the hearth pressure-fire detection energy value time sequence correlation matrix through the time sequence correlation feature extractor based on the convolutional neural network model to obtain the hearth pressure-fire detection energy value time sequence correlation feature map.

That is, in one specific example of the present application, the encoding process for performing association encoding and feature extraction on the furnace pressure time series input vector and the fire detection energy value time series input vector to obtain a furnace pressure-fire detection energy value time series association feature map includes: performing association coding on the hearth pressure time sequence input vector and the fire detection energy value time sequence input vector to obtain a hearth pressure-fire detection energy value time sequence association matrix; and then the hearth pressure-fire detection energy value time sequence correlation matrix passes through a time sequence correlation feature extractor based on a convolutional neural network model to obtain a hearth pressure-fire detection energy value time sequence correlation feature map.

Here, by arranging the furnace pressure values and the fire detection energy values at a plurality of predetermined time points in the time dimension as the furnace pressure time series input vector and the fire detection energy value time series input vector, respectively, they can be correlated in time. In this way, when the association coding is performed, the association relation between the hearth pressure time sequence input vector and the fire detection energy value time sequence input vector can be established by performing association operation on the corresponding time points of the hearth pressure time sequence input vector and the fire detection energy value time sequence input vector. Specifically, the correlation code may combine the furnace pressure value and the fire detection energy value at each point in time to form a time-series correlation matrix of furnace pressure-fire detection energy values. The correlation matrix can reflect the corresponding relation between the hearth pressure and the fire detection energy value at different time points.

Further, the hearth pressure-fire detection energy value time sequence correlation characteristic diagram is subjected to characteristic autocorrelation correlation strengthening module to obtain a strengthened hearth pressure-fire detection energy value time sequence correlation characteristic diagram. The characteristic autocorrelation correlation strengthening module utilizes similarity relations among characteristic values in the hearth pressure-fire detection energy value time sequence correlation characteristic diagram to gather complete information about hearth pressure-fire detection energy value time sequence correlation characteristic distribution. That is, after dynamically learning the element correlation between the furnace pressure-fire detection energy value timing correlation characteristic maps, the characteristic distribution of the furnace pressure-fire detection energy value timing correlation characteristic maps is optimized to highlight characteristic information on whether to throw an oil gun. And then, the corrected intensified hearth pressure-fire detection energy value time sequence correlation characteristic diagram is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether an oil gun is put in or not.

In a specific embodiment of the present application, determining whether to throw an oil gun based on the furnace pressure-fire detection energy value time sequence correlation feature map includes: performing information enhancement and distribution correction processing on the hearth pressure-fire detection energy value time sequence correlation characteristic map to obtain a corrected enhanced hearth pressure-fire detection energy value time sequence correlation characteristic map; and passing the corrected intensified furnace pressure-fire detection energy value time sequence correlation characteristic diagram through a classifier to obtain a classification result, wherein the classification result is used for indicating whether an oil gun is put in or not.

The method for carrying out information enhancement and distribution correction processing on the hearth pressure-fire detection energy value time sequence correlation characteristic diagram to obtain a corrected enhanced hearth pressure-fire detection energy value time sequence correlation characteristic diagram comprises the following steps: the hearth pressure-fire detection energy value time sequence correlation characteristic diagram is subjected to characteristic autocorrelation correlation strengthening module to obtain a strengthened hearth pressure-fire detection energy value time sequence correlation characteristic diagram; and carrying out characteristic distribution correction on the time sequence correlation characteristic map of the pressure-fire detection energy value of the enhanced hearth to obtain the time sequence correlation characteristic map of the pressure-fire detection energy value of the enhanced hearth after correction.

Further, in a specific embodiment of the present application, the step of passing the furnace pressure-fire detection energy value time sequence correlation characteristic map through a characteristic autocorrelation correlation strengthening module to obtain a strengthened furnace pressure-fire detection energy value time sequence correlation characteristic map includes: the characteristic diagram of the time sequence correlation of the hearth pressure and the fire detection energy value is obtained through a first convolution layer of the characteristic autocorrelation correlation strengthening module; the first characteristic diagram passes through a second convolution layer of the characteristic autocorrelation correlation strengthening module to obtain a second characteristic diagram; expanding each feature matrix of the second feature map along the channel dimension into feature vectors to obtain a sequence of first feature vectors; calculating cosine similarity between any two first feature vectors in the sequence of the first feature vectors to obtain a cosine similarity feature map; normalizing the cosine similarity feature map through a softmax function to obtain a normalized cosine similarity feature map; multiplying the normalized cosine similarity feature map and the cosine similarity feature map according to position points to obtain a similarity mapping optimization feature map; the similarity mapping optimization feature map passes through a first deconvolution layer of the feature autocorrelation correlation strengthening module to obtain a first deconvolution feature map; calculating element-by-element sums of the first deconvolution feature map and the first feature map to obtain a first fusion feature map; passing the first-fusion feature map through a second deconvolution layer of the feature autocorrelation correlation enhancement module to obtain a second deconvolution feature map; and calculating element-by-element sums of the second deconvolution feature map and the hearth pressure-fire detection energy value time sequence correlation feature map to obtain the enhanced hearth pressure-fire detection energy value time sequence correlation feature map.

After the hearth pressure time sequence input vector and the fire detection energy value time sequence input vector are subjected to association coding, the obtained hearth pressure-fire detection energy value time sequence association matrix expresses first-order association of hearth pressure values and fire detection energy values in a full time domain, so that after the hearth pressure-fire detection energy value time sequence association matrix passes through a time sequence association feature extractor based on a convolutional neural network model, each feature matrix of the hearth pressure-fire detection energy value time sequence association feature graph expresses high-order association features of hearth pressure values and fire detection energy values in a local time domain in the full time domain, and channel distribution of the convolutional neural network model is followed among the feature matrices. And after the hearth pressure-fire detection energy value time sequence correlation characteristic diagram passes through the characteristic autocorrelation correlation strengthening module, the channel distribution among the characteristic matrixes can be restrained based on the local time domain high-order time sequence correlation characteristic representation of each characteristic matrix, so that the whole strengthening hearth pressure-fire detection energy value time sequence correlation characteristic diagram follows the multi-order time sequence correlation characteristic distribution. However, considering that the time sequence correlation characteristic distribution differences of different orders can bring local characteristic distribution sparsification to the overall characteristic representation of the time sequence correlation characteristic diagram of the pressure-fire detection energy value of the enhanced hearth, namely, the sub-manifold is thinned out of the distribution relative to the overall high-dimensional characteristic manifold, the method can cause that when the time sequence correlation characteristic diagram of the pressure-fire detection energy value of the enhanced hearth is subjected to quasi-probability regression mapping through a classifier, the convergence from the time sequence correlation characteristic diagram of the pressure-fire detection energy value of the enhanced hearth to the predetermined quasi-probability class representation in a probability space is poor, and the accuracy of the classification result is affected.

Therefore, preferably, the position-by-position characteristic value optimization is performed on the enhanced hearth pressure-fire detection energy value time sequence correlation characteristic diagram, specifically: optimizing the position-by-position characteristic values of the enhanced hearth pressure-fire detection energy value time sequence correlation characteristic map by using the following optimization formula to obtain the corrected enhanced hearth pressure-fire detection energy value time sequence correlation characteristic map; wherein, the optimization formula is:

wherein f _i Is the characteristic value F 'of the characteristic diagram F of the time sequence correlation of the pressure and the fire detection energy value of the enhanced hearth' _i The characteristic value of the corrected enhanced hearth pressure-fire detection energy value time sequence correlation characteristic diagram is exp (·) and represents a natural exponential function value calculated by taking a numerical value as a power.

The sparse distribution in the high-dimensional feature space is processed through regularization based on the heavy probability so as to activate the natural distribution transfer from the geometric manifold of the enhanced hearth pressure-fire detection energy value time sequence correlation feature map F in the high-dimensional feature space to the probability space, so that the class convergence of the complex high-dimensional feature manifold with high spatial sparsity under the predetermined class probability is improved through a smooth regularization mode based on the heavy probability on the distributed sparse sub-manifold of the high-dimensional feature manifold of the enhanced hearth pressure-fire detection energy value time sequence correlation feature map F, and the accuracy of the classification result obtained by the classifier of the enhanced hearth pressure-fire detection energy value time sequence correlation feature map F is improved.

In one embodiment of the application, under the deep peak regulation working condition, the coordination and main control system regulation quality is improved through control optimization, the assessment of the load quality by scheduling is reduced, more AGC or load quality benefits are obtained, and meanwhile, the stability and safety of the unit are improved; the economy of the unit is improved through intelligent sliding pressure on-line optimization; through combustion state monitoring, when the combustion is worsened in early stage, oil is rapidly fed to stabilize combustion, so that combustion safety is improved; carding the protection of the main unit and the auxiliary unit aiming at the deep peak-shaving working condition, aiming at the phenomena that the total air quantity is low in MFT, the water supply flow is low in MFT, the misoperation of measurement signals is easy to occur, and the like, soft measurement calculation is required to be carried out on the total air quantity under the deep peak-shaving working condition, signals such as a fan state, a baffle opening degree, current, wind pressure and the like are required to correct the total air quantity, and signals such as a steam pump state, a rotating speed, a recirculation state, a pump inlet flow, a steam pump outlet pressure and the like are required to correct the water supply flow, so that the misoperation of the MFT caused by signal measurement reasons is prevented, and protection optimization is carried out in a targeted manner; one-key start-stop control is added aiming at the deep peak regulation working condition, so that the running operation quantity is reduced. The method specifically comprises the following steps:

(1) Energy balance-based coordinated control strategy optimization under the deep peak shaving working condition;

(2) Optimizing a combustion safety strategy under a deep peak-shaving working condition;

(3) Optimizing the control of the coal water ratio based on phase compensation;

(4) Intelligent slide pressure on-line control optimization;

(5) Intelligent coal quality quick correction;

(6) The protection and optimization of the main and auxiliary machines under the deep peak-shaving working condition;

(7) And one-key sequential control function group practicability such as pump returning and the like is applied;

(8) Dry-wet state conversion logic optimization;

(9) And a fan-reversing one-key sequential control function group.

Specifically, in the optimization of the coordination control strategy based on energy balance under the deep peak shaving working condition, the combustion approaches to a safety critical point, the requirement of scheduling on load indexes is met on the premise of ensuring the running safety of the unit, and a more stable and safer control strategy is needed to be adopted by the coordination control system. The hysteresis of fuel quantity and the change of coal quality are the biggest control problems in coordination control, and under the deep peak regulation working condition, the disturbance of a gas turbine valve regulation, water supply flow, air quantity and the like on main gas pressure is reduced, the passive regulation of fuel quantity is reduced, and the combustion safety is improved. Because the inertia time of the gas turbine regulating valve is smaller, the inertia time of the boiler combustion regulation is larger, the coordination control scheme based on energy balance reduces excessive disturbance of the gas turbine regulating valve to the main steam pressure, and more helps to adjust the main steam pressure deviation in a smaller load deviation range, when coal quality changes or other boilers are disturbed, the disturbance can be eliminated through the allowable load deviation based on the principle of energy balance, the main steam pressure is prevented from greatly fluctuating, further fluctuation of fuel quantity is reduced, and combustion safety is improved.

The main control of the steam turbine mainly controls the load of the unit, and the main steam pressure is controlled in a larger amount within the allowable range, so that the fluctuation range of the main steam pressure is smaller, and the main parameters of the unit are more stable to operate; the main control of the boiler controls the energy of the boiler, is responsible for controlling the load and the pressure of main steam of a unit, meets the AGC (automatic gain control) check requirement of the unit, realizes the load change process mainly through a feedforward loop, and respectively carries out load limiting and load limiting on the deviation adjustment quantity of the pressure of the main steam, so that the condition that the load adjustment precision and the adjustment speed do not meet the specified requirement when the load is changed due to the reverse direction of the deviation of the load adjustment quantity of a valve adjustment of a steam turbine; when the load deviation is large, the main control of the steam turbine only adjusts the load deviation, so that the assessment of the load precision by scheduling is avoided. The weighting coefficient of the main control adjusting pressure and the load of the steam turbine is gradually corrected according to different working conditions (load instructions) of the deep peak shaving, so that the requirements of load indexes can be met, and the aim of stabilizing the parameters of a unit can be fulfilled.

In the optimization of the combustion safety strategy under the deep peak regulation working condition, through the monitoring of the combustion state, the fuel is quickly fed to stabilize combustion in the early stage of combustion deterioration, so that the combustion safety is improved; the pressure fluctuation of the hearth is large (+ -250 Pa) and the oil is fed for stable combustion; detecting a corresponding fire detection energy value, and automatically throwing a corresponding oil gun when the fire detection energy value is low and flicker, so as to realize stable combustion of the oil gun; tripping the coal mill, and putting the coal mill into an oil gun in an interlocking way to burn stably; any one of the important auxiliary machines such as six fans trips, and the oil gun is stable in combustion; the coal is cut off, and the oil is fed for stable combustion. The oxygen amount suddenly increases.

In the coal water ratio control optimization based on phase compensation, as the fuel quantity is changed by changing the frequency conversion rotating speed of a coal feeder and then is sent into the coal mill for development, the coal powder reaching the specified fineness is blown into a hearth for combustion by primary air, and the temperature and the pressure of the water supply flow and the steam flow are influenced by radiation heat exchange and convection heat exchange, and the inertia time constant of the process is about 120 s. The change of the water supply flow is realized by changing the rotation speed of the steam pump through the change of the opening degree of the low-pressure regulating valve of the steam pump, and the time that the change of the outlet flow of the water supply pump influences the outlet temperature and the pressure of the water-cooled wall is far less than the fuel quantity. The water-supply flow command is changed along with the change of fuel quantity, and the fuel quantity is changed firstly after three inertia times of about 20-30s, namely, the water-supply flow is changed after three inertia times of 20-30s, and the fuel quantity and the energy of the water-supply flow synchronously reach the outlet of the water-cooled wall, so that the change of the superheat degree of the middle point is minimum, and the disturbance to a unit is minimum.

Because the dispatching has higher requirement on the load change speed, four regulating valves are changed into two regulating valves when the turbine through-flow is transformed, and the main steam pressure cannot reach the load command, the change amplitude of the turbine regulating valves is larger when the load is lifted, and particularly, the turbine regulating valves are always fully opened when the load is lifted, so that the regulating speed is lower when the load is lifted, and the dispatching and the checking are carried out.

The coal water ratio control based on phase compensation is to accelerate the change of the water supply flow, greatly reduce three inertia times, namely almost synchronously change the water supply flow when the load is lifted, and lead the phase of the change of the water supply flow to be advanced. However, the temperature of the middle point is greatly changed, and the stable operation of the unit is affected. Therefore, the water supply flow rate is changed while the fuel quantity is changed, and a plurality of fuel quantities are synchronously changed to counteract the change of the middle point temperature, namely, a plurality of fuel quantities are changed to exchange for the overheat variation caused by the rapid change of the water supply flow rate. The rapidity of the boiler can be improved through parameter adjustment, and the superheat degree (superheated steam temperature) of the middle point can be fluctuated within an allowable range.

In intelligent sliding pressure on-line control optimization, through-flow transformation changes the valve of the steam turbine from four to two, and the larger the opening of the valve of the steam turbine is obtained through a sliding pressure optimization test, the higher the economy of the unit is, but the valve of the steam turbine also meets the requirements of scheduling AGC and primary frequency modulation on load regulation quality, and not only meets the requirements of scheduling on load indexes, but also gives consideration to the economy of the unit. Online intelligent optimization of the sliding pressure is required.

And when the valve is smaller than 38%, the valve is overridden to reduce the set value bias of the main steam pressure, an operator can set the expected opening of the valve of the steam turbine to be 38-50%, and PID (proportion integration differentiation) operation is carried out on the expected opening of the valve and the actual valve position deviation to increase and decrease the set value bias of the main steam pressure, so that the periodic fluctuation of the set value bias is avoided through logic and parameter setting. When the load changes, the main steam pressure set value bias feedforward is increased according to the load change amplitude and the adjustment speed, so that the economy of the unit is more considered in a steady state, and the load adjustment quality is more considered in the process of changing the load.

In the intelligent rapid correction of coal quality, due to the control of combustion cost and combustion safety requirements, more coal sources are used, and various kinds of coal quality are mixed and burned according to fuel management, so that more coal quality change is caused, and the method becomes the most internal disturbance in unit operation. The long-time and large-scale change is reduced by the coal quality correction, but the rapid change of the coal quality is not well treated.

According to steady state calculation, dynamic application, quick correction and undisturbed tracking, the feedforward variable quantity such as the boiler main control and the water supply main control can be ensured to be closer to the real variable quantity as far as possible when the load changes each time. The operator can manually adjust the coal quality of each mill, so that the coal quality can be corrected more accurately and rapidly. And according to the judgment of the steady-state working condition, the change condition of the coal quality can be rapidly calculated by combining parameters such as the control output of the enthalpy value of the middle point, the actual coal quantity and the load. The coal quality correction coefficient is changed rapidly in the initial stage of load change, so that water, coal and wind keep unchanged, and the coal quality correction coefficient is changed rapidly.

In the protection optimization of the main and auxiliary machines under the deep peak shaving working condition, the water supply flow is close to the protection action value under the deep peak shaving working condition, and the water supply flow protection value is distinguished from the high load due to the reduction of the heat load. The low protection action value and the delay time of the water supply flow under the deep peak regulation working condition are evaluated, the suggestion of requesting the opinion of the boiler plant is provided, the water supply flow protection action value is correspondingly reduced according to the load instruction, and the delay action time is correspondingly increased. If no modification is required, the feedwater flow low alarm level should be adjusted to the highest level.

Under the deep peak regulation working condition, the total air quantity approaches to the protection action value, and when the low-load air quantity is low, the motor state of the blower and the position of the baffle plate are referred to for judgment, and meanwhile, the parameters such as wind pressure, current and the like are used for correction, so that the air quantity reliability is improved.

In the application of the practicability of the parallel and return pump and other one-key sequential control functional groups, the method comprises the following steps: water feeding and pumping, wherein the pumping time is as follows: the rated load is more than 40%, and one steam pump operates. And controlling front and back of the pump: the steam pump controls the water supply flow; the minimum flow recirculation of the steam pump is put into operation automatically.

The program control step sequence comprises the following steps: starting a small turbine remote control request; starting a small turbine throwing request; the starting steam pump is put into operation automatically, the differential pressure between the outlet of the regulating pump of the starting steam pump and the water supply main pipe is set to be 0.2MPa; opening an electric door at the outlet of the dynamic steam pump; the pump is started automatically; and the input flow leveling is automatic.

Further comprises: and (5) the water supply pump is back-up. Pump withdrawal time: and the rated load is below 45%, and the two steam pumps operate. And controlling front and back of the pump: the two steam pumps control the water supply flow; the minimum flow recirculation of the two steam pumps is put into operation automatically.

The program control step sequence comprises the following steps: cutting off the automatic (e.g. pump a) withdrawal pump, the small machine also being in a remote control position; gradually reducing the rotation speed command of the pump A, reducing the rotation speed command of the pump A at intervals of 2% for 30 seconds each time, and regulating the water supply flow rate by the pump B in an automatic mode. If the flow deviation is greater than 50t/h, suspending to reduce the A pump instruction; when the pressure of the pump A out of the main pipe is smaller than the pressure of the main pipe by 0.5MPa, the water supply flow is completely born by the pump B; and (5) putting the pump A for hot standby or continuing to reduce the instruction to 3000 turns, and finishing the pump withdrawal.

In dry-wet state transition logic optimization, wet-to-dry state admission conditions: the water supply is regulated automatically, the superheat degree of the outlet of the water-cooled wall is less than 5 ℃, and the load meets the requirements.

The program control step sequence comprises the following steps: the automatic start system of the direct current furnace is put into operation, and comprises a water feeding pump, a water feeding bypass valve and a water level regulating valve of a water storage tank; the feedwater flow is reduced to the current steam flow or the feedwater flow corresponding to the load (which should be greater than the boiler minimum feedwater flow command). The feedwater flow command is gradually reduced to the target value at 100t/h intervals of 30s (closed-loop flow adjustment time) each time. When deviation between the water supply flow and the instruction is large and delay is 10 seconds or fluctuation of main parameters (load, main steam pressure, main steam temperature and the like) of the unit is large, water reduction is suspended, after the water supply flow is adjusted in place or the fluctuation of the unit parameters is stable due to the water supply flow adjustment, water reduction is continued, the superheat degree of the middle point temperature is more than 5 ℃ or MFT occurs to finish water reduction, and if the middle point temperature meets the dry state characteristics, the step is skipped to the fifth step; the automatic load and the automatic fuel main control of the running coal mill are input, and if the manual cutting condition exists, the automatic input is forbidden; the amount of fuel is increased. The fuel quantity command is gradually increased at intervals of 60s (the time for the fuel to react to the middle point temperature after entering the combustion chamber of the furnace) of 5t/h each time until the superheat degree of the middle point temperature is more than 5 ℃. When the fluctuation of main parameters (power, main steam pressure, main steam temperature and the like) of the unit is large, the fuel quantity is stopped to be increased, the fuel quantity is continuously increased after the fluctuation of the parameter of the standby group is stable, the maximum fuel quantity is increased to 130 percent before conversion, and the phenomenon that the temperature of the steam at the outlet of the separator, the wall of a heated surface and the temperature of the main steam are over-temperature caused by the blind rapid increase of the fuel quantity is avoided; putting into the middle point enthalpy control or middle point temperature control automatically, and setting the enthalpy control/temperature control

In the parallel and reverse fan one-key sequential control function group, the method comprises the following steps: the blower is stopped, and the condition for allowing the blower to stop is as follows: the two blowers are operated with a load less than 45% Pe and the blower to be shut down (A, B) is selected.

The program control step sequence comprises the following steps: cutting off and selecting a blower moving blade to be stopped, and automatically cutting off the other blower moving blade; slowly setting a minimum value (0.2%/s) of a blower moving blade to be shut down, opening the other blower moving blade according to the balance weight, and performing total air volume closed-loop control; stopping the selected blower.

Further comprises: the blowers are operated in parallel, and the allowable condition of the blowers is started: one of the blowers operates, one blower is stopped, and the load is more than 35%Pe. Blower lube pumps a or B (default to a) are selected. Blower hydraulic oil pump a or B (default a) is selected.

The program control step sequence comprises the following steps: starting the selected blower lubricating oil pump and hydraulic oil pump, and putting into interlocking; closing an electric baffle at the outlet of the blower, and placing a blower movable blade at a minimum position; the blower is started. Interlocking to open an electric baffle door at the outlet of the blower; the other blower moving blade regulates the total air quantity in an automatic mode; memorizing the position of the other blower moving blade, and placing the blower moving blade which is just started to 30% of the memorized moving blade, wherein the speed is 0.5%/s; and (3) putting into a current leveling loop to enable the output of the two blowers to be equivalent.

Further, the technical difficulties and main innovations of the present application include, but are not limited to: energy balance-based coordinated control strategy optimization under the deep peak shaving working condition; optimizing a combustion safety strategy under a deep peak-shaving working condition; soft measurement and calculation of total air quantity under the deep peak regulation working condition; and (5) intelligent sliding pressure on-line control optimization.

In the embodiment of the application, the difficult problem of poor adjustment quality of the unit depth peak shaving automatic control system can be solved, the assessment of load quality by scheduling is reduced, and more AGC service compensation is acquired. The fluctuation range of main parameters (main steam pressure, main steam temperature and the like) of the unit is reduced, the operation parameters are improved, and the economy of the unit is further improved. And the main and auxiliary machines are protected, combed and optimized under the deep peak shaving working condition, so that risks of protection misoperation or refusal are effectively reduced, and the safety of the unit is improved. By detecting the combustion state, oil is fed in time to stabilize combustion, so that combustion safety is improved.

In summary, the safe and economic coordination control method for deep peak shaving of the 1000MW thermal power generating unit based on the embodiment of the application is clarified, the hearth pressure value and the fire detection energy value are detected, and the time sequence correlation change characteristics of the hearth pressure and the fire detection energy are captured by using a deep learning algorithm, so that intelligent control over oil gun throwing is realized, and the running state of the unit is optimized.

In one embodiment of the present application, fig. 3 is a block diagram of a safe economic coordination control system for depth peaking of a 1000MW thermal power unit according to an embodiment of the present application. As shown in fig. 3, a safe economic coordination control system 200 for depth peaking of a 1000MW thermal power generating unit according to an embodiment of the present application includes: a data acquisition module 210, configured to acquire furnace pressure values at a plurality of predetermined time points within a predetermined time period and fire detection energy values at the plurality of predetermined time points; the data preprocessing module 220 is configured to perform data preprocessing on the furnace pressure values at the plurality of predetermined time points and the fire detection energy values at the plurality of predetermined time points to obtain a furnace pressure time sequence input vector and a fire detection energy value time sequence input vector; the correlation encoding and feature extraction module 230 is configured to perform correlation encoding and feature extraction on the furnace pressure time sequence input vector and the fire detection energy value time sequence input vector to obtain a furnace pressure-fire detection energy value time sequence correlation feature map; and a gun delivery judging module 240, configured to determine whether to deliver a gun based on the furnace pressure-fire detection energy value time sequence correlation characteristic map.

Here, it will be understood by those skilled in the art that the specific functions and operations of the respective units and modules in the above-described safe economic coordination control system for depth peaking of 1000MW thermal power plant have been described in detail in the above description of the safe economic coordination control method for depth peaking of 1000MW thermal power plant with reference to fig. 1 to 2, and thus, repetitive descriptions thereof will be omitted.

As described above, the safe economic coordination control system 200 for depth peaking of a 1000MW thermal power generating unit according to the embodiment of the present application may be implemented in various terminal devices, for example, a server or the like for safe economic coordination control of depth peaking of a 1000MW thermal power generating unit. In one example, the safety economic coordinated control system 200 for depth peaking of a 1000MW thermal power plant according to embodiments of the present application may be integrated into a terminal device as one software module and/or hardware module. For example, the safe economic coordination control system 200 for depth peaking of a 1000MW thermal power generating unit may be a software module in an operating system of the terminal device, or may be an application developed for the terminal device; of course, the safe economic coordination control system 200 for the depth peaking of the 1000MW thermal power generating unit may also be one of numerous hardware modules of the terminal device.

Alternatively, in another example, the safe economic coordination control system 200 for deep peaking of the 1000MW thermal power plant and the terminal device may be separate devices, and the safe economic coordination control system 200 for deep peaking of the 1000MW thermal power plant may be connected to the terminal device through a wired and/or wireless network and transmit the interaction information according to a agreed data format.

Fig. 4 is a schematic view of a scenario of a safe economic coordination control method for depth peaking of a 1000MW thermal power generating unit according to an embodiment of the present application. As shown in fig. 4, in this application scenario, first, furnace pressure values (e.g., C1 as illustrated in fig. 4) at a plurality of predetermined time points within a predetermined period of time and fire detection energy values (e.g., C2 as illustrated in fig. 4) at the plurality of predetermined time points are acquired; the acquired furnace pressure value and fire detection energy value are then input into a server (e.g., S as illustrated in fig. 4) deployed with a safe economic coordination control algorithm for 1000MW thermal power plant depth peaking, where the server is capable of processing the furnace pressure value and the fire detection energy value based on the safe economic coordination control algorithm for 1000MW thermal power plant depth peaking to determine whether to deliver an oil gun.

It is also noted that in the apparatus, devices and methods of the present application, the components or steps may be disassembled and/or assembled. Such decomposition and/or recombination should be considered as equivalent to the present application.

The previous description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other aspects without departing from the scope of the application. Thus, the present application is not intended to be limited to the aspects shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

The foregoing description has been presented for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of the application to the form disclosed herein. Although a number of example aspects and embodiments have been discussed above, a person of ordinary skill in the art will recognize certain variations, modifications, alterations, additions, and subcombinations thereof.

Claims (7)

1. A safe economic coordination control method for depth peak shaving of a 1000MW thermal power generating unit is characterized by comprising the following steps:

acquiring hearth pressure values at a plurality of preset time points in a preset time period and fire detection energy values at the preset time points;

performing data preprocessing on the hearth pressure values at the plurality of preset time points and the fire detection energy values at the plurality of preset time points to obtain a hearth pressure time sequence input vector and a fire detection energy value time sequence input vector;

performing association coding and feature extraction on the hearth pressure time sequence input vector and the fire detection energy value time sequence input vector to obtain a hearth pressure-fire detection energy value time sequence association feature diagram; and

and determining whether to throw the oil gun or not based on the time sequence correlation characteristic diagram of the hearth pressure and fire detection energy value.

2. The safe economic coordination control method for depth peaking of a 1000MW thermal power generating unit according to claim 1, wherein performing data preprocessing on the furnace pressure values at the plurality of predetermined time points and the fire detection energy values at the plurality of predetermined time points to obtain a furnace pressure time sequence input vector and a fire detection energy value time sequence input vector comprises:

And arranging the hearth pressure values at the preset time points and the fire detection energy values at the preset time points into the hearth pressure time sequence input vector and the fire detection energy value time sequence input vector according to the time dimension respectively.

3. The safe economic coordination control method for depth peaking of a 1000MW thermal power generating unit according to claim 2, wherein performing association coding and feature extraction on the furnace pressure time sequence input vector and the fire detection energy value time sequence input vector to obtain a furnace pressure-fire detection energy value time sequence association feature map comprises:

performing association coding on the hearth pressure time sequence input vector and the fire detection energy value time sequence input vector to obtain a hearth pressure-fire detection energy value time sequence association matrix; and

and extracting correlation features of the hearth pressure-fire detection energy value time sequence correlation matrix by using a deep learning model to obtain a hearth pressure-fire detection energy value time sequence correlation feature map.

4. The safe economic coordination control method for deep peak shaving of a 1000MW thermal power generating unit according to claim 3, wherein the deep learning model is a time sequence correlation feature extractor based on a convolutional neural network model;

The method for extracting the correlation characteristic of the hearth pressure-fire detection energy value time sequence correlation matrix by using a deep learning model to obtain the hearth pressure-fire detection energy value time sequence correlation characteristic map comprises the following steps:

and passing the hearth pressure-fire detection energy value time sequence correlation matrix through the time sequence correlation feature extractor based on the convolutional neural network model to obtain the hearth pressure-fire detection energy value time sequence correlation feature map.

5. The safe economic coordination control method for depth peaking of a 1000MW thermal power generating unit according to claim 4, wherein determining whether to throw an oil gun based on the furnace pressure-fire detection energy value time sequence correlation feature map comprises:

performing information enhancement and distribution correction processing on the hearth pressure-fire detection energy value time sequence correlation characteristic map to obtain a corrected enhanced hearth pressure-fire detection energy value time sequence correlation characteristic map; and

and the corrected time sequence correlation characteristic diagram of the intensified hearth pressure-fire detection energy value is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether an oil gun is put in or not.

6. The safe economic coordination control method for depth peaking of a 1000MW thermal power generating unit according to claim 5, wherein performing information enhancement and distribution correction processing on the furnace pressure-fire detection energy value time sequence correlation characteristic map to obtain a corrected enhanced furnace pressure-fire detection energy value time sequence correlation characteristic map comprises:

The hearth pressure-fire detection energy value time sequence correlation characteristic diagram is subjected to characteristic autocorrelation correlation strengthening module to obtain a strengthened hearth pressure-fire detection energy value time sequence correlation characteristic diagram; and

and carrying out characteristic distribution correction on the time sequence correlation characteristic diagram of the pressure-fire detection energy value of the enhanced hearth to obtain the time sequence correlation characteristic diagram of the pressure-fire detection energy value of the enhanced hearth after correction.

7. The safe economic coordination control method for deep peak shaving of a 1000MW thermal power generating unit according to claim 6, wherein the step of passing the furnace pressure-fire detection energy value time sequence correlation characteristic map through a characteristic autocorrelation correlation strengthening module to obtain a strengthened furnace pressure-fire detection energy value time sequence correlation characteristic map comprises the following steps:

the characteristic diagram of the time sequence correlation of the hearth pressure and the fire detection energy value is obtained through a first convolution layer of the characteristic autocorrelation correlation strengthening module;

the first characteristic diagram passes through a second convolution layer of the characteristic autocorrelation correlation strengthening module to obtain a second characteristic diagram;

expanding each feature matrix of the second feature map along the channel dimension into feature vectors to obtain a sequence of first feature vectors;

calculating cosine similarity between any two first feature vectors in the sequence of the first feature vectors to obtain a cosine similarity feature map;

Normalizing the cosine similarity feature map through a softmax function to obtain a normalized cosine similarity feature map;

multiplying the normalized cosine similarity feature map and the cosine similarity feature map according to position points to obtain a similarity mapping optimization feature map;

the similarity mapping optimization feature map passes through a first deconvolution layer of the feature autocorrelation correlation strengthening module to obtain a first deconvolution feature map;

calculating element-by-element sums of the first deconvolution feature map and the first feature map to obtain a first fusion feature map;

passing the first-fusion feature map through a second deconvolution layer of the feature autocorrelation correlation enhancement module to obtain a second deconvolution feature map;

and calculating element-by-element sums of the second deconvolution feature map and the hearth pressure-fire detection energy value time sequence correlation feature map to obtain the enhanced hearth pressure-fire detection energy value time sequence correlation feature map.