Control method for deep peak regulation of coal-fired power plant

Abstract

The invention relates to the technical field of deep peak regulation, in particular to a control method for deep peak regulation of a coal-fired power plant. The method comprises the following steps: firstly, acquiring current load data, current weather forecast and current holiday information, inputting the current load data, the current weather forecast and the current holiday information into a trained load demand prediction model of a power grid to obtain a demand charge quantity, and determining combustion parameters of a boiler in a table look-up manner according to the demand charge quantity, therefore, intelligent cooperative control of deep peak regulation of the coal-fired power plant is realized. By solving the combustion parameters of the boiler, fuel (such as coal) consumed by combustion can be determined, so that emission of pollutants such as NOx and SO2 is reduced, and the combustion stability and efficiency of the boiler are improved. Therefore, the problems of efficiency reduction and emission increase in the deep peak regulation process of the coal-fired power plant can be solved.

Description

Technical Field

[0001] The present invention relates to the technical field of deep peak regulation, and in particular to a control method for deep peak regulation of a coal-fired power plant. Background Art

[0002] With the gradual opening of the power market and the large-scale grid connection of renewable energy, the power grid has put forward higher requirements on the peak-shaving capacity of coal-fired power plants, especially for deep peak-shaving during the load valley period. Traditional control methods often find it difficult to achieve efficient and low-emission operation while ensuring safe and stable operation. Therefore, it is particularly important to develop an intelligent collaborative control method that can intelligently adapt to changes in grid demand and achieve deep peak-shaving of coal-fired power plants. Summary of the invention

[0003] The present invention aims to solve the problems of reduced efficiency and increased emissions faced by coal-fired power plants during deep peak regulation, and provides a control method for deep peak regulation of coal-fired power plants.

[0004] An embodiment of the present invention provides a control method for deep peak regulation of a coal-fired power plant, comprising:

[0005] Get current load data, current weather forecast and current holiday information;

[0006] Inputting the current load data, the current weather forecast and the current holiday information into the trained load demand prediction model of the power grid to obtain the required charge amount;

[0007] Based on the required charge, the combustion parameters of the boiler are determined to achieve intelligent coordinated control of deep peak regulation of the coal-fired power plant.

[0008] The embodiment of the present invention provides a control method for deep peak regulation of a coal-fired power plant. First, the current load data, the current weather forecast and the current holiday information are obtained, and the current load data, the current weather forecast and the current holiday information are input into the load demand prediction model of the trained power grid to obtain the required charge. The combustion parameters of the boiler are determined by table lookup according to the required charge, so as to realize the intelligent coordinated control of deep peak regulation of the coal-fired power plant. By solving the combustion parameters of the boiler, the fuel consumed by the combustion (such as coal) can be determined to reduce the emission of pollutants such as NOx and SO2, and improve the stability and efficiency of the boiler combustion. Therefore, the present invention can solve the problems of reduced efficiency and increased emissions faced by coal-fired power plants during deep peak regulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0010] Figure 1 It is a flow chart of a control method for deep peak regulation of a coal-fired power plant provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0011] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0012] Please refer to Figure 1 , an embodiment of the present invention provides a control method for deep peak regulation of a coal-fired power plant, the method comprising:

[0013] Step 100: Obtain current load data, current weather forecast and current holiday information;

[0014] Step 102: inputting current load data, current weather forecast and current holiday information into the trained load demand prediction model of the power grid to obtain the required charge amount;

[0015] Step 104: Based on the required charge, determine the combustion parameters of the boiler to achieve intelligent coordinated control of deep peak regulation of the coal-fired power plant.

[0016] In an embodiment of the present invention, firstly, the current load data, the current weather forecast and the current holiday information are obtained, and the current load data, the current weather forecast and the current holiday information are input into the load demand prediction model of the trained power grid to obtain the required charge, such as the total charge required in the next day, and the combustion parameters of the boiler are determined by table lookup according to the required charge, so as to realize the intelligent coordinated control of deep peak regulation of coal-fired power plants. By accurately solving the combustion parameters of the boiler, the fuel consumed by the combustion (such as coal) can be determined to reduce the emission of pollutants such as NOx and SO2, and improve the stability and efficiency of boiler combustion. Therefore, the present invention can solve the problems of reduced efficiency and increased emissions faced by coal-fired power plants during deep peak regulation.

[0017] In one embodiment of the present invention, data can be obtained from the monitoring system of the power grid. These systems collect and upload relevant data of power load in real time through smart meters, power sensors and other devices installed in substations, transmission lines and various power consumption ends, such as active power, reactive power, current size and other information. It is common to obtain data through API interfaces provided by professional meteorological service agencies, such as China Meteorological Network and data service platforms related to the World Meteorological Organization, and data can also be obtained from some commercial meteorological data suppliers. In addition, some local meteorological monitoring stations will also upload real-time meteorological observation data to relevant network platforms for integration and utilization. They can be obtained from official release channels for national statutory holidays, such as the annual holiday schedule published on government department websites; some event processing software libraries or online services can also be used to determine the corresponding information by identifying whether the current date is a statutory holiday, weekend or adjusted holiday.

[0018] In one embodiment of the present invention, the combustion parameters include fuel parameters, air parameters, pressure parameters, temperature parameters and combustion product parameters.

[0019] In the embodiment of the present invention, the fuel parameters are specifically as follows: Fuel type: Different fuels have very different combustion characteristics. For example, the combustion process of solid fuels (such as coal and biomass) is relatively complicated and needs to go through stages such as drying, pyrolysis, and combustion; liquid fuels (such as diesel and heavy oil) are easy to atomize and evaporate, and burn relatively quickly; gaseous fuels (such as natural gas and coal gas) can be better mixed with air and have higher combustion efficiency. Fuel calorific value (calorific value): This is a key parameter to measure the amount of heat that can be released when a fuel is burned. The unit is usually kilojoules per kilogram (kJ / kg) or kilojoules per cubic meter (kJ / m 3 For example, the low calorific value of standard coal is about 29307 kJ / kg, and the low calorific value of natural gas is about 35588 kJ / m 3 . A higher calorific value means that a unit of fuel can release more heat, which affects the heat output of the furnace. Fuel composition: Carbon (C) content: In most fuels, carbon is the main heat-generating component. The higher the carbon content, the higher the calorific value of the fuel is usually, and the combustion process is relatively stable. For example, the carbon content of anthracite can be as high as 90% or more, and its flame is short when it burns, and the heat released is concentrated. Hydrogen (H) content: Hydrogen burns to produce water and releases a large amount of heat at the same time. The presence of hydrogen can increase the calorific value of the fuel and increase the water vapor content in the combustion products. For example, natural gas has a high hydrogen content, and its combustion products are mainly carbon dioxide and water vapor. Sulfur (S) content: The sulfur in the fuel will produce sulfur dioxide (SO 2 ) and other harmful gases, which not only cause environmental pollution, but may also corrode equipment such as furnaces. For example, high-sulfur coal will release more SO 2, desulfurization measures are needed to reduce harmful gas emissions. Oxygen (O) and nitrogen (N) content: Oxygen reduces the calorific value of the fuel because it has been partially oxidized with other elements in the fuel. Nitrogen generally does not participate in the combustion reaction during the combustion process, but it will form nitrogen oxides (NO x ), is also an air pollutant.

[0020] In the embodiment of the present invention, the air parameters are specifically: air volume (excess air coefficient): to ensure complete combustion of the fuel, the actual amount of air supplied is often greater than the theoretical amount of air required. The excess air coefficient is the ratio of the actual amount of air to the theoretical amount of air. A suitable excess air coefficient is crucial for complete combustion of the fuel and temperature control in the furnace. For example, in a coal-fired boiler, the excess air coefficient is generally between 1.1 and 1.3. If the excess air coefficient is too large, a large amount of heat will be taken away, reducing thermal efficiency; if it is too small, incomplete combustion of the fuel will result. Air temperature: The temperature of the air entering the furnace affects the combustion reaction rate. Preheating the air can improve the combustion efficiency because it provides additional energy for the combustion reaction. For example, in some high-temperature industrial furnaces, the air is preheated to several hundred degrees Celsius through a heat exchanger before entering the furnace, which can speed up the combustion reaction and reduce fuel consumption. Air humidity: The moisture content in the air affects the combustion process. High-humidity air absorbs more heat for water vaporization during the combustion process, and may dilute the combustion gas, affecting the combustion rate and temperature. However, in some special combustion processes, an appropriate amount of water vapor can participate in chemical reactions and improve the combustion effect.

[0021] In the embodiment of the present invention, the temperature parameters are specifically as follows: Furnace temperature: This is one of the key indicators reflecting the combustion conditions. The temperatures at different locations in the furnace may be different. For example, the temperature is highest at the center of the flame, while the temperature is relatively low near the furnace wall. The temperature in the furnace directly affects the combustion speed and completeness of the fuel. For example, in a metallurgical furnace, the furnace temperature needs to reach thousands of degrees Celsius to meet the requirements of metal smelting; while in an ordinary hot water boiler, the furnace temperature is generally between several hundred degrees and one thousand degrees. Flame temperature: The flame temperature depends on factors such as the type of fuel, air supply conditions, and combustion conditions. High-temperature flames can accelerate the combustion of fuel and heat transfer. For example, the temperature of an acetylene-oxygen flame can be as high as 3000°C or more, and is often used in high-temperature operations such as metal cutting.

[0022] In the embodiment of the present invention, the pressure parameter is specifically the pressure inside the furnace: it is crucial for the combustion process and safety that the pressure inside the furnace is maintained within an appropriate range. For a furnace operating at positive pressure, the pressure inside the furnace is higher than the external atmospheric pressure, which is beneficial to prevent the intrusion of cold air, but may cause problems such as flames emitting outward and hot gas leakage; for a furnace operating at negative pressure, the pressure inside the furnace is lower than the external atmospheric pressure, which is convenient for discharging the exhaust gas generated by combustion, but if the negative pressure is too large, a large amount of cold air will enter the furnace, reducing thermal efficiency. For example, in industrial boilers, it is generally required that the pressure inside the furnace be maintained at a slightly positive pressure or slightly negative pressure state. Ventilation pressure: includes the air supply pressure provided by the blower and the induced draft pressure provided by the induced draft fan. These pressures are used to ensure that the air can enter the furnace smoothly and the exhaust gas after combustion can be discharged in time. The size of the ventilation pressure is related to factors such as the furnace type, the amount of fuel, and the amount of air.

[0023] In the embodiment of the present invention, the combustion product parameters are specifically the furnace pressure: carbon dioxide (CO 2 ) content: It is one of the main products of complete combustion of fuel. 2 The content can indirectly determine whether the fuel is completely burned. For example, when natural gas is burned, if CO 2 The content is close to the theoretical value, indicating that the combustion is relatively complete; if CO 2 If the content is low and accompanied by an increase in other incomplete combustion products such as carbon monoxide (CO), it means that there may be problems in the combustion process. Carbon monoxide (CO) content: CO is a product of incomplete combustion of fuel. Its presence not only means that the heat is not fully utilized, but also causes harm to human health and the environment. During the combustion process, the CO content should be reduced as much as possible. This can be achieved by optimizing combustion conditions, such as adjusting the air supply and improving the mixing degree of fuel and air. Nitrogen oxides (NO x ) content: mainly includes nitric oxide (NO), nitrogen dioxide (NO 2 ), etc., which are formed under the conditions of high temperature combustion and the participation of nitrogen. x It is one of the main air pollutants and can cause environmental problems such as acid rain and photochemical smog. x To reduce emissions, low nitrogen combustion technologies can be used, such as staged combustion and flue gas recirculation. Particulate matter content: Various particulate matter, such as dust and carbon black, will be generated during the combustion process. These particulate matter will affect the air quality and may also be deposited in the furnace, affecting heat transfer and the normal operation of the equipment. For solid fuel combustion, special attention should be paid to the generation and control of particulate matter, such as reducing particulate matter emissions by installing dust removal equipment. Flue gas volume and flow rate: Flue gas volume depends on factors such as fuel type, combustion conditions and air supply. The flow rate of flue gas affects the residence time of combustion products in the furnace, which in turn affects the degree of complete combustion of the fuel and the efficiency of heat transfer. An appropriate flue gas flow rate can ensure that the combustion products are discharged in a timely manner while avoiding excessive heat loss.

[0024] In one embodiment of the present invention, the method further includes obtaining a trained load demand prediction model of a power grid according to the following steps:

[0025] Obtaining a historical data set; wherein the historical data set includes historical load data, historical weather forecasts, historical holiday information, and historical demand charge;

[0026] The initial load demand prediction model of the power grid is trained with historical load data, historical weather forecasts, and historical holiday information as inputs and historical demand charge as output to obtain a trained load demand prediction model of the power grid.

[0027] In the embodiment of the present invention, first, historical data sets are collected and integrated, and the data sets serve as the basis for subsequent model training, covering multi-dimensional key information. Among them, historical load data accurately records the power load conditions of each node in the power grid in different time periods in the past, such as the sequence of changes in specific values ​​such as active power and reactive power over time, which intuitively reflects the fluctuation trend of historical electricity demand; historical weather forecast data details the meteorological elements of the corresponding time period, including temperature, humidity, wind speed, wind direction, and weather conditions. Given that weather conditions have a significant impact on power load (for example, high temperature weather causes a surge in power consumption of refrigeration equipment, and cold weather drives up power consumption of heating appliances), it provides indispensable environmental related information for load forecasting; historical holiday information The categories of each date (working days, statutory holidays, and adjusted holidays, etc.) are clearly marked. Considering that the electricity consumption patterns of residents and enterprises during holidays are often very different from those on weekdays (such as changes in passenger flow and business hours in commercial areas during holidays, which lead to changes in electricity load, and residents use more electrical equipment for leisure and entertainment at home, etc.), this information dimension is of great significance for accurately predicting load demand; at the same time, the corresponding historical demand charge, as an output label, accurately quantifies the actual scale of electricity consumption in the past under specific historical load, weather, and holiday comprehensive scenarios, and constitutes a complete basis for training data benchmarking. Relying on the acquisition of a complete historical data set, the initial power grid load demand forecasting model is built based on a deep learning or machine learning algorithm framework (model structures such as long short-term memory networks (LSTMs), time series decomposition combined with regression algorithms, etc., which are suitable for processing problems with time series characteristics and multi-factor influences, can be selected). The sorted and normalized historical load data, historical weather forecast data, and historical holiday information are arranged in a predetermined format as model input feature vectors, and the corresponding historical demand charge is set as the target output of model training to start the model training iteration process. During the training phase, with the help of optimization algorithms (such as stochastic gradient descent, Adam optimizer, etc.), the internal parameters of the model (such as weights and biases in the neural network) are continuously adjusted according to the set loss function (for example, mean square error loss is used for regression prediction tasks to measure the degree of deviation between the predicted charge and the actual charge). This enables the model to learn and extract the mapping relationship between input information (load, weather, holidays) and output (demanded charge) from the complex rules contained in massive historical data. After multiple rounds of training and verification optimization, the model reaches the performance convergence standard on the verification data set (such as the loss value is lower than the set threshold, and evaluation indicators such as accuracy tend to be stable). At this point, a trained power grid load demand forecasting model with accurate prediction capabilities is successfully obtained, which lays a solid foundation for subsequent load and charge prediction in real-time operation scenarios and provides core support tools.

[0028] In one embodiment of the present invention, the historical sample data set includes a training set and a test set.

[0029] In one embodiment of the present invention, training the initial power grid load demand prediction model includes:

[0030] For each iteration training of updating the initial load demand prediction model of the power grid, the training set is input into the initial load demand prediction model of the power grid to obtain the prediction result of the current iteration training;

[0031] Obtain the loss of the current iteration training based on the prediction results and the historical required charge;

[0032] Back propagation is performed based on the loss to update the parameters of the initial power grid load demand prediction model to complete the current iterative training.

[0033] In an embodiment of the present invention, for each iterative training of the initial power grid load demand prediction model update, the training set is input into the initial power grid load demand prediction model, the prediction result of the current iterative training is obtained, the loss of the current iterative training is obtained based on the prediction result and the historical demand charge, back propagation is performed based on the loss, and the parameters of the initial power grid load demand prediction model are updated to complete the current iterative training.

[0034] In one embodiment of the present invention, after each iterative training, it is determined whether the initial load demand prediction model of the power grid has converged;

[0035] If yes, end the model training, and save the trained initial power grid load demand prediction model as the trained power grid load demand prediction model;

[0036] If not, continue to the next iteration of training.

[0037] In one embodiment of the present invention, it is determined whether the initial load demand forecasting model of the power grid has converged:

[0038] Determine whether the loss of the current iterative training is greater than the preset threshold;

[0039] If so, the initial power grid load demand forecasting model converges;

[0040] If not, the initial power grid load demand forecasting model has not converged.

[0041] In one embodiment of the present invention, the preset threshold is ninety-five percent.

[0042] In one embodiment of the present invention, the loss of the current iterative training is calculated by the following formula:

[0043]

[0044] Where S is the loss of the current iteration training, n is the number of samples in the training set, To predict the results, is the historical demand charge.

[0045] In one embodiment of the present invention, the trained load demand prediction model of the power grid is a convolutional neural network model.

[0046] A person of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiments; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk, etc., various media that can store program codes.

[0047] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A control method for deep peak regulation of a coal-fired power plant, characterized in that: include: Get current load data, current weather forecast and current holiday information; Inputting the current load data, the current weather forecast and the current holiday information into the trained load demand prediction model of the power grid to obtain the required charge amount; Based on the required charge, the combustion parameters of the boiler are determined to achieve intelligent coordinated control of deep peak regulation of the coal-fired power plant.

2. The method according to claim 1, characterized in that The combustion parameters include fuel parameters, air parameters, pressure parameters, temperature parameters and combustion product parameters.

3. The method according to claim 2, characterized in that The method further comprises obtaining the trained load demand prediction model of the power grid according to the following steps: Acquire a historical data set; wherein the historical data set includes historical load data, historical weather forecasts, historical holiday information, and historical demand charge; The historical load data, the historical weather forecast, and the historical holiday information are used as inputs, and the historical required charge amount is used as output to train the initial load demand prediction model of the power grid, so as to obtain the trained load demand prediction model of the power grid.

4. The method according to claim 3, characterized in that include: The historical sample data set includes a training set and a test set.

5. The method according to claim 4, characterized in that The training of the initial power grid load demand prediction model includes: For each iteration training of updating the initial load demand prediction model of the power grid, the training set is input into the initial load demand prediction model of the power grid to obtain the prediction result of the current iteration training; Obtaining a loss of current iterative training according to the prediction result and the historical required charge amount; Back propagation is performed according to the loss to update the parameters of the initial power grid load demand prediction model to complete the current iterative training.

6. The method according to claim 5, characterized in that Also includes: After each iteration of training, determining whether the initial load demand forecasting model of the power grid has converged; If yes, the model training is terminated, and the trained initial power grid load demand prediction model is saved as the trained power grid load demand prediction model; If not, continue to the next iteration of training.

7. The method according to claim 6, characterized in that The method further includes determining whether the initial load demand prediction model of the power grid has converged: Determine whether the loss of the current iterative training is greater than a preset threshold; If yes, the initial load demand forecasting model of the power grid converges; If not, the initial power grid load demand prediction model has not converged.

8. The method according to claim 7, characterized in that The preset threshold is ninety-five percent.

9. The method according to claim 8, characterized in that The loss of the current iteration training is calculated by the following formula: Where S is the loss of the current iteration training, n is the number of samples in the training set, is the prediction result, is the historical charge demand.

10. The method according to any one of claims 1 to 9, characterized in that The trained load demand prediction model of the power grid is a convolutional neural network model.

Images