System and method for improving and controlling flexibility of thermal power generating unit

Abstract

The invention belongs to the technical field of thermal power generating unit control, and discloses a thermal power generating unit flexibility improvement control system and method, and the method comprises the steps: collecting unit operation data of the same time sequence, carrying out the preprocessing of the unit operation data, and obtaining effective operation data; performing temperature correlation on the effective operation data based on the constructed boiler temperature prediction model to obtain the temperature at the future moment; lagging compensation is carried out on the effective operation data based on the future moment temperature and a preset target temperature, and a future adjustment strategy is obtained; performing environment disturbance learning on the future adjustment strategy to obtain an optimal adjustment strategy, and realizing flexible control on the thermal power generating unit based on the optimal adjustment strategy; and the stability and the flexibility of the thermal power generating unit are greatly enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of thermal power unit control, and more specifically, to a control system and method for improving the flexibility of a thermal power unit. Background Art

[0002] As an important part of the traditional power system, thermal power units mainly generate steam by burning coal to heat water to drive steam turbines to generate electricity. Although thermal power units have shown high stability and good efficiency in long-term operation, with the intensification of grid load fluctuations, especially in the context of the increasing proportion of renewable energy, the flexibility requirements of thermal power units have been significantly improved. In order to meet the grid's demand for load regulation, thermal power units must have strong rapid response capabilities and adaptability.

[0003] The combustion system of the boiler of a thermal power unit needs to adjust the ratio of air and fuel in real time according to load changes to ensure the stability and efficiency of combustion. However, due to the thermal inertia and fluid dynamics characteristics of the boiler combustion process, the adjustment of air flow and fuel flow often takes a certain amount of time to reach a new equilibrium state. This lag effect causes the unit to respond slowly, unable to quickly adjust the output power, and difficult to adapt to load fluctuations or emergencies, affecting the flexibility of the overall system; the relationship between air flow, fuel flow and combustion temperature in the boiler combustion process is highly nonlinear, and it is difficult for the control system to accurately adjust the ratio of air and fuel through a simple linear control algorithm. Even a small change in fuel or air flow will cause a large fluctuation in combustion temperature, especially in the case of large load changes. Traditional linear control methods often find it difficult to accurately respond to these changes in a short period of time, resulting in temperature overshoot or undershoot; the boiler combustion system is not only affected by the internal state of the boiler itself, but also easily disturbed by external factors. The traditional combustion control system has poor adaptability to these external disturbances, which affects the combustion temperature and combustion efficiency, resulting in an unstable combustion process, especially when the load increases or decreases suddenly, the boiler's adjustment ability is limited.

[0004] In view of this, the present invention proposes a control system and method for improving the flexibility of a thermal power unit to solve the above-mentioned problem. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: a control system for improving the flexibility of a thermal power unit, comprising:

[0006] Data acquisition and processing module: collects the unit operation data of the same time series, pre-processes the unit operation data, and obtains effective operation data;

[0007] Future prediction module: Based on the established boiler temperature prediction model, the effective operation data is correlated with the temperature to obtain the temperature at the future moment;

[0008] Input control module: performs hysteresis compensation on effective operation data based on the future temperature and the preset target temperature to obtain the future adjustment strategy;

[0009] Environmental compensation module: Conduct environmental disturbance learning for future regulation strategies, obtain the best regulation strategy, and realize flexible control of thermal power units based on the best regulation strategy.

[0010] Furthermore, the unit operation data includes: air flow, fuel flow and boiler pressure.

[0011] Furthermore, the method of preprocessing the unit operation data includes:

[0012] Each category of data in the unit operation data is fuzzy transformed to obtain the fuzziness of each category of data; based on the fuzziness of each category of data, the K-Means algorithm is used to perform cluster analysis on the fuzziness to obtain fuzzy membership clusters, and the fuzzy membership clusters are evaluated for rules to obtain fuzzy rules for each category of data in the unit operation data; each group of unit operation data is fuzzily described based on the fuzzy rules to obtain fuzzy expressions for each group of unit operation data; a rule recognition library is preset, and rule recognition is performed on the fuzzy expression based on the rule recognition library to obtain data labels, which include abnormal and normal data; the unit operation data is classified based on the data labels, and the data labels with abnormal labels are classified into one category as abnormal data, and the data labels with normal labels are classified into one category as normal data; abnormal correction is performed on the abnormal data based on the normal data to obtain time series correction data, and the time series correction data and normal data are merged to obtain effective operation data.

[0013] Furthermore, the method of performing abnormal correction on abnormal data includes:

[0014] According to the time series, the normal data is serialized to obtain the normal time series data; the observation matrix, the control matrix and the initial gain are preset, the initial gain is initialized to be infinitesimal, the time series prediction function is constructed based on the observation matrix and the control matrix, the time series normal data is predicted based on the time series prediction function to obtain the time series prediction data, the gain is calculated on the time series prediction data to obtain the prediction gain; the gain threshold is preset, when the prediction gain is greater than or equal to the gain threshold, the time series normal data at the corresponding moment is replaced by the time series prediction data to obtain the time series update data; the observation disturbance and the control disturbance are preset, the mean of the prediction gain is calculated to obtain the mean gain, when the mean gain is greater than or equal to the initial gain When , the observation matrix plus the observation disturbance is used as the new observation matrix, and the control matrix plus the control disturbance is used as the new control matrix. When the gain mean is less than the initial gain, the observation matrix minus the observation disturbance is used as the new observation matrix, and the control matrix minus the control disturbance is used as the new control matrix. The time series update data is used as the new time series normal data, and the process is repeated until the mean of the predicted gain converges. The observation matrix at this time is output as the optimal observation matrix, and the control matrix at this time is output as the optimal control matrix. The optimal observation matrix and the optimal control matrix are substituted into the time series prediction function to obtain the optimal prediction function. The abnormal data is predicted based on the optimal prediction function to obtain the time series corrected data.

[0015] Furthermore, the boiler temperature prediction model is constructed in the following manner:

[0016] Collect historical operation data and historical combustion temperature. The historical operation data includes: historical air flow, historical fuel flow and historical boiler pressure. The data type of the historical operation data is consistent with that of the data in the effective operation data. The historical air flow, historical fuel flow and historical boiler pressure in the historical operation data are used as the input of the boiler temperature prediction model, and the predicted combustion temperature is used as the output of the boiler temperature prediction model. The historical combustion temperature is used as the prediction target of the boiler temperature prediction model for model training to construct a boiler temperature prediction model. The boiler temperature prediction model is a long short-term memory model. The target loss function of the boiler temperature prediction model is a mean square error function. Minimizing the value of the target loss function is used as the training goal to obtain a boiler temperature prediction model with the smallest value of the target loss function.

[0017] Furthermore, the method of controlling and adjusting the effective operation data includes:

[0018] The time step error calculation is performed on the preset target temperature and the temperature at the future moment to obtain the error temperature at each moment, and the target adjustment function is constructed based on the error temperature. M groups of sequence control increments are initialized, where M is an integer greater than zero. The sequence control increment is the increment set at each moment, and the increment set contains the increment of each category of data. The optimal sequence control increment is initialized to be empty, and the optimal incremental fitness is initialized to be infinite. Each group of sequence control increments is regarded as a chromosome, and the genes in the chromosome correspond to the increment set at each moment in the sequence control increment. For each chromosome, the incremental fitness of each chromosome is calculated according to the target adjustment function. The smaller the incremental fitness, the better the chromosome. The chromosomes are selected based on the tournament selection algorithm to obtain the parent chromosome. The chromosomes in the parent chromosome set are paired with each other using the permutation and combination algorithm to obtain the parent combination. A gene in the parent combination is randomly selected as the exchange starting point, and the genes after the exchange starting point in the parent combination are exchanged to obtain the daughter chromosome. The incremental fitness of each daughter chromosome is calculated based on the target adjustment function, and the daughter chromosome with a fitness smaller than the parent combination is selected as the preferred generation.

[0019] A mutation factor is preset, and the mutation factor is a minimum value greater than zero. A gene in the preferred generation is randomly selected as the mutation node. The value of A mutation factors is increased or decreased for the mutation node with random probability to obtain the mutant chromosome. The incremental fitness of each mutant chromosome is calculated based on the target adjustment function. The mutant chromosome with an incremental fitness less than that of the preferred generation is taken as the preferred variant. The preferred variant with the smallest incremental fitness is selected as the candidate variant. When the incremental fitness of the candidate variant is less than the optimal incremental fitness, the candidate variant is taken as the new optimal sequence control increment, and the incremental fitness of the candidate variant is taken as the new optimal incremental fitness. The preferred variant is taken as the new parent chromosome, and the process is repeated until the value of the optimal incremental fitness no longer changes. The optimal sequence control increment at this time is output as the future adjustment strategy.

[0020] Furthermore, the formula of the objective adjustment function is:

[0021] Wherein, J represents the target adjustment function, Tep represents the predicted time step, e(t+a) represents the error temperature at the t+a moment, Δv(t+a) represents the control increment at the t+a moment, and the control increment includes: air flow increment, fuel flow increment and boiler pressure increment; represents the temperature adjustment weight, and β represents the incremental adjustment weight.

[0022] Furthermore, the method of performing environmental disturbance learning on the future adjustment strategy includes:

[0023] Collect real-time combustion temperature and real-time environmental data, the real-time environmental data includes environmental temperature and environmental humidity, build a boiler state space based on real-time combustion temperature, real-time environmental data and future adjustment strategy, the boiler state space includes the current combustion temperature, environmental temperature, environmental humidity and the air flow, fuel flow and boiler pressure at the corresponding time in the future adjustment strategy; preset parameter adjustment actions, the parameter adjustment actions include: increase or decrease air flow, increase or decrease fuel flow and increase or decrease boiler pressure, arrange and combine the parameter adjustment actions to obtain an adjustment action table; preset parameter adjustment factors, the parameter adjustment factors include: air flow adjustment factor, fuel flow adjustment factor and boiler pressure adjustment factor; build an action reward function based on real-time combustion temperature, select parameter adjustment actions based on target temperature and real-time combustion temperature, when the target temperature and real-time combustion temperature are not equal, randomly select an action in the adjustment action table for parameter adjustment, calculate the reward value after parameter adjustment based on the action reward function, repeat and record the adjustment action and boiler state space with the largest reward value at each moment as the time-series optimal action, all time-series optimal actions constitute the best adjustment strategy.

[0024] Furthermore, the formula of the action reward function is:

[0025] Re=-b1·|Tep now -Tep dest |-b2·ΔEmc; where Re represents the reward value, Tep now Represents the real-time combustion temperature, Tep dest represents the target temperature, b1 represents the temperature weight, ΔEmc represents the fuel increment, and b2 represents the fuel weight.

[0026] The present invention provides a control method for improving the flexibility of a thermal power unit, comprising:

[0027] S1. Collect the unit operation data of the same time series, pre-process the unit operation data, and obtain effective operation data;

[0028] S2. Based on the constructed boiler temperature prediction model, the effective operation data is temperature-correlated to obtain the temperature at the future moment;

[0029] S3, based on the future temperature and the preset target temperature, the effective operation data is delayed compensated to obtain the future adjustment strategy;

[0030] S4. Conduct environmental disturbance learning on future regulation strategies to obtain the best regulation strategy, and achieve flexible control of thermal power units based on the best regulation strategy.

[0031] The technical effects and advantages of a control system and method for improving the flexibility of a thermal power unit of the present invention are as follows:

[0032] The present invention classifies and regularizes the unit operation data through fuzzy transformation and cluster analysis, which can effectively eliminate the influence of noise data and abnormal points on system control. The abnormal identification and timing correction based on the rule base ensure the integrity and accuracy of the operation data, and provide a reliable data basis for subsequent control. The boiler temperature prediction model can capture the time series characteristics of the boiler operation data, and improve the prediction accuracy of the combustion temperature. Through the time step error calculation and the design of the target adjustment function, a dynamic adjustment strategy is generated, and multiple iterative optimizations are combined with the genetic algorithm to screen the optimal sequence control increment from multiple candidate strategies, thereby realizing accurate and efficient dynamic adjustment. Through active learning of environmental disturbances and real-time updating of the optimal adjustment strategy, the control accuracy and adaptability of the thermal power unit under complex environmental conditions are improved, the fuel consumption is effectively reduced, the combustion efficiency is improved, and the target temperature is ensured to be achieved quickly and stably, and the combustion process is optimized. The entire control system combines artificial intelligence algorithms and expert experience to realize autonomous decision-making and automatic adjustment functions, reduces the need for manual intervention, and improves the intelligence level of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A schematic diagram of a control system for improving the flexibility of a thermal power unit according to the present invention;

[0034] Figure 2 It is a schematic diagram of a control method for improving flexibility of a thermal power unit according to the present invention. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] Example 1

[0037] See also Figure 1 As shown, a thermal power unit flexibility improvement control system described in this embodiment includes:

[0038] Data acquisition and processing module: collects the unit operation data of the same time series, pre-processes the unit operation data, and obtains effective operation data;

[0039] Future prediction module: Based on the established boiler temperature prediction model, the effective operation data is correlated with the temperature to obtain the temperature at the future moment;

[0040] Input control module: performs hysteresis compensation on effective operation data based on the future temperature and the preset target temperature to obtain the future adjustment strategy;

[0041] Environmental compensation module: It conducts environmental disturbance learning on future regulation strategies, obtains the best regulation strategy, and realizes flexible control of thermal power units based on the best regulation strategy;

[0042] Each module is connected by wired and / or wireless means to achieve data transmission between modules;

[0043] The unit operation data includes: air flow, fuel flow and boiler pressure; the air flow is obtained through a vortex flowmeter, and the air flow refers to the amount of air entering the boiler combustion chamber. It is one of the key factors to ensure complete combustion of the fuel. The air supply directly determines the oxygen supply for combustion; the fuel flow is obtained through a mass flowmeter, and the fuel flow refers to the amount of fuel input into the boiler combustion chamber. It and the air flow together determine the combustion intensity and heat output of the boiler; the boiler pressure is obtained through a pressure transmitter, and the boiler pressure refers to the pressure of the steam inside the boiler. The boiler pressure affects the thermal efficiency and steam generation capacity.

[0044] The methods for preprocessing the unit operation data include:

[0045] Each category of data in the unit operation data is fuzzy transformed, and the formula for fuzzy transformation is: Among them, Va ij represents the fuzziness of the jth data in the i-th category of the unit operation data, x ij represents the jth data in the i-th category of the unit operation data, cen i represents the mean of the i-th type of data in the unit operation data, σ iRepresents the standard deviation of the i-th category of data in the unit operation data; based on the fuzziness of each category of data, the K-Means algorithm is used to perform cluster analysis on the fuzziness to obtain fuzzy membership clusters, and the fuzzy membership clusters are evaluated by rules to obtain fuzzy rules for each category of data in the unit operation data, and each group of unit operation data is fuzzily described based on the fuzzy rules to obtain fuzzy expressions for each group of unit operation data; a rule recognition library is preset, and rule recognition is performed on the fuzzy expression based on the rule recognition library to obtain data labels, which include abnormal and normal data. The unit operation data is classified based on the data labels, and the data with abnormal labels are classified into one category as abnormal data, and the data with normal labels are classified into one category as normal data; the unit operation data is fuzzy processed at multiple levels through fuzzy transformation and fuzzy rules, and the complex and multi-dimensional operation status is simplified into a fuzzy expression that is easy to analyze and identify, which greatly improves the accuracy and flexibility of data description, and the abnormal data and normal data are effectively separated through the fuzzy membership evaluation and rule recognition library based on expert rules, making the data input to the subsequent modules more reliable;

[0046] Based on normal data, the abnormal data is corrected, and the normal data is serialized according to the time series to obtain the normal time series data; the observation matrix, control matrix and initial gain are preset, and the initial gain is initialized to be infinitesimal. The time series prediction function is constructed based on the observation matrix and the control matrix. The formula of the time series prediction function is: Gu t =A·Tr t-1 +B·u; among them, Gu t Represents the time series forecast data at time t, Tr t-1 represents the normal time series data at the t-1th moment, A represents the observation matrix, B represents the control matrix, which is used to control the interference degree of the interference factor, and u represents the interference factor, which is used to simulate the interference of other factors; based on the time series prediction function, the time series normal data is predicted, the time series prediction data is obtained, and the gain calculation is performed on the time series prediction data. The formula for gain calculation is:

[0047] K t =(Gu t -Tr t )·H L ·(A·(Gu t -Tr t )·H L +B) -1 ; Among them, K t Represents the prediction gain at time t. The larger the prediction gain, the more we trust the actual measured value; the larger and smaller the prediction gain, the more we trust the predicted value. Tr trepresents the normal time series data at the tth moment, H represents the identity matrix, and L represents the total data category; a gain threshold is preset. When the predicted gain is greater than or equal to the gain threshold, the normal time series data at the corresponding moment is replaced by the time series prediction data to obtain the time series update data; the observation disturbance and control disturbance are preset, the mean of the predicted gain is calculated, and the gain mean is obtained. When the gain mean is greater than or equal to the initial gain, the observation matrix plus the observation disturbance is used as a new observation matrix, and the control matrix plus the control disturbance is used as a new control matrix. When the gain mean is less than the initial gain, the observation matrix minus the observation disturbance is used as a new observation matrix, and the control matrix minus the control disturbance is used as a new control matrix. The disturbance is used as a new control matrix; the time series update data is used as the new time series normal data, and the process is repeated until the mean of the prediction gain converges, and the observation matrix at this time is output as the optimal observation matrix, and the control matrix at this time is output as the optimal control matrix; the optimal observation matrix and the optimal control matrix are substituted into the time series prediction function to obtain the best prediction function, and the abnormal data is predicted based on the best prediction function to obtain the time series correction data, and the time series correction data and normal data are combined to obtain the effective operation data; by constructing the time series prediction function, the high-precision time series prediction and correction of the abnormal data of the thermal power unit are realized, and the quality of the effective operation data is improved;

[0048] The evaluation of fuzzy rules is based on the evaluation of fuzzy membership clusters by experts. Each type of data in the unit operation data has a set of fuzzy rules. Taking air flow as an example, the fuzzy rules for air flow are low flow, medium flow and high flow. The recognition rules in the rule recognition library are based on expert regulations. For example, the category label of high air flow and high combustion temperature is abnormal, and the category label of low air flow and high combustion temperature is normal. Each fuzzy expression contains multiple category labels, and the category label with the largest proportion is used as the final data label. The identity matrix is ​​a matrix with the value of the left pair of elements being 1 and the values ​​of the remaining elements being 0.

[0049] The boiler temperature prediction model is constructed in the following ways:

[0050] Collect historical operation data and historical combustion temperature. The historical operation data includes: historical air flow, historical fuel flow and historical boiler pressure. The data type of the historical operation data is consistent with that of the data in the effective operation data. The historical air flow, historical fuel flow and historical boiler pressure in the historical operation data are used as the input of the boiler temperature prediction model, and the predicted combustion temperature is used as the output of the boiler temperature prediction model. The historical combustion temperature is used as the prediction target of the boiler temperature prediction model for model training to construct a boiler temperature prediction model. The boiler temperature prediction model is a long short-term memory model. The target loss function of the boiler temperature prediction model is the mean square error function. Minimizing the value of the target loss function is used as the training target to obtain a boiler temperature prediction model with the smallest target loss function value. Preset the prediction time step, and predict the effective operation data according to the prediction time step based on the constructed boiler temperature prediction model to obtain the temperature at the future moment.

[0051] The methods for controlling and adjusting effective operation data include:

[0052] The time step error calculation is performed on the preset target temperature and the temperature at the future moment to obtain the error temperature at each moment. The target adjustment function is constructed based on the error temperature. The formula of the target adjustment function is: Wherein, J represents the target adjustment function, Tep represents the predicted time step, e(t+a) represents the error temperature at the t+a moment, Δv(t+a) represents the control increment at the t+a moment, and the control increment includes: air flow increment, fuel flow increment and boiler pressure increment; represents the temperature adjustment weight, and β represents the incremental adjustment weight;

[0053] Initialize M groups of sequence control increments, where M is an integer greater than zero. The sequence control increment is a set of increments at each moment. The increment set contains the increments of each category of data. The optimal sequence control increment is initialized to be empty, and the optimal incremental fitness is initialized to be infinite. Each group of sequence control increments is regarded as a chromosome, and the genes in the chromosome correspond to the set of increments at each moment in the sequence control increments. For each chromosome, the incremental fitness of each chromosome is calculated according to the target adjustment function. The smaller the incremental fitness, the better the chromosome. The chromosomes are selected based on the tournament selection algorithm to obtain the parent chromosome. The chromosomes in the parent chromosome set are paired with each other using the permutation and combination algorithm to obtain the parent combination. A gene in the parent combination is randomly selected as the exchange starting point, and the genes after the exchange starting point in the parent combination are exchanged to obtain the daughter chromosome. The incremental fitness of each daughter chromosome is calculated based on the target adjustment function, and the daughter chromosome with a fitness smaller than the parent combination is selected as the preferred generation.

[0054] Preset mutation factor, which is a minimum value greater than zero, randomly select a gene in the preferred generation as the mutation node, increase or decrease the value of A mutation factors for the mutation node with random probability, obtain the mutant chromosome, calculate the incremental fitness of each mutant chromosome based on the target adjustment function, take the mutant chromosome with an incremental fitness less than the preferred generation as the preferred variant, select the preferred variant with the smallest incremental fitness as the candidate variant, when the incremental fitness of the candidate variant is less than the optimal incremental fitness, take the candidate variant as the new optimal sequence control increment, take the incremental fitness of the candidate variant as the new optimal incremental fitness, take the preferred variant as the new parent chromosome, repeat until the value of the optimal incremental fitness no longer changes, and output the optimal sequence control increment at this time, which is the future adjustment strategy;

[0055] Based on the time step error between the preset target temperature and the temperature at future moments, the lag effect in the response of the thermal power unit is effectively eliminated by constructing the target adjustment function and optimizing the control increment. Through the initialization and evolutionary optimization of multiple groups of sequence control increments, the genetic algorithm is used to find the optimal control increment sequence to ensure the feasibility and efficiency of future adjustment strategies, realize the coordinated adjustment of multivariable control, and ensure the dynamic balance of air flow, fuel flow and boiler pressure.

[0056] Ways to learn environmental perturbations for future regulation strategies include:

[0057] Collect real-time combustion temperature and real-time environmental data, the real-time environmental data includes ambient temperature and ambient humidity, build a boiler state space based on real-time combustion temperature, real-time environmental data and future adjustment strategy, the boiler state space includes the current combustion temperature, ambient temperature, ambient humidity and the air flow, fuel flow and boiler pressure at the corresponding time in the future adjustment strategy; preset parameter adjustment actions, the parameter adjustment actions include: increase or decrease air flow, increase or decrease fuel flow and increase or decrease boiler pressure, arrange and combine the parameter adjustment actions to obtain an adjustment action table; preset parameter adjustment factors, the parameter adjustment factors include: air flow adjustment factor, fuel flow adjustment factor and boiler pressure adjustment factor; build an action reward function based on real-time combustion temperature, the formula of the action reward function is: Re = -b1·|Tep now -Tep dest |-b2·ΔEmc; where Re represents the reward value, Tep now Represents the real-time combustion temperature, Tep destRepresents the target temperature, b1 represents the temperature weight, which is used to adjust the proportion of temperature in the reward value, ΔEmc represents the fuel increment, and b2 represents the fuel weight, which is used to adjust the proportion of fuel increment in the reward value; the action reward function focuses on the target temperature and fuel increment to reduce the use of resources, achieve the lowest resource use and meet the target temperature; select the parameter adjustment action based on the target temperature and the real-time combustion temperature. When the target temperature and the real-time combustion temperature are not equal, randomly select an action in the adjustment action table for parameter adjustment, and calculate the reward value after parameter adjustment based on the action reward function. Repeat and record the adjustment action with the largest reward value at each moment and the boiler state space as the optimal action in time sequence. All optimal actions in time sequence constitute the optimal adjustment strategy, and the thermal power unit is flexibly controlled based on the optimal adjustment strategy.

[0058] This embodiment classifies and regularizes the unit operation data through fuzzy transformation and cluster analysis, which can effectively eliminate the influence of noise data and abnormal points on system control. The abnormal identification and timing correction based on the rule base ensure the integrity and accuracy of the operation data, and provide a reliable data basis for subsequent control; through the boiler temperature prediction model, the time series characteristics of the boiler operation data can be captured, and the prediction accuracy of the combustion temperature is improved; through the time step error calculation and the design of the target adjustment function, a dynamic adjustment strategy is generated, and multiple iterative optimizations are combined with the genetic algorithm to screen the optimal sequence control increment from multiple candidate strategies, thereby realizing accurate and efficient dynamic adjustment; through active learning of environmental disturbances and real-time updating of the optimal adjustment strategy, the control accuracy and adaptability of the thermal power unit under complex environmental conditions are improved, fuel consumption is effectively reduced, combustion efficiency is improved, and the target temperature is ensured to be achieved quickly and stably, and the combustion process is optimized; the entire control system combines artificial intelligence algorithms and expert experience to realize autonomous decision-making and automatic adjustment functions, reduces the need for manual intervention, and improves the intelligence level of the system.

[0059] Example 2

[0060] See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, and a method for improving the flexibility of a thermal power unit is provided, including:

[0061] S1. Collect the unit operation data of the same time series, pre-process the unit operation data, and obtain effective operation data;

[0062] S2. Based on the constructed boiler temperature prediction model, the effective operation data is temperature-correlated to obtain the temperature at the future moment;

[0063] S3, based on the future temperature and the preset target temperature, the effective operation data is delayed compensated to obtain the future adjustment strategy;

[0064] S4. Conduct environmental disturbance learning on future regulation strategies to obtain the best regulation strategy, and achieve flexible control of thermal power units based on the best regulation strategy.

[0065] Example 3

[0066] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the above-mentioned method for improving the flexibility of a thermal power unit is implemented.

[0067] Since the electronic device introduced in this embodiment is an electronic device used to implement a method for improving the flexibility of a thermal power unit in the embodiment of this application, based on the method for improving the flexibility of a thermal power unit introduced in the embodiment of this application, a person skilled in the art can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application will not be described in detail here. As long as a person skilled in the art implements the electronic device used in the method for improving the flexibility of a thermal power unit in the embodiment of this application, it belongs to the scope of protection of this application.

[0068] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0069] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A control system for improving the flexibility of a thermal power unit, characterized in that: include: Data acquisition and processing module: collects the unit operation data of the same time series, pre-processes the unit operation data, and obtains effective operation data; Future prediction module: Based on the established boiler temperature prediction model, the effective operation data is correlated with the temperature to obtain the temperature at the future moment; Input control module: performs hysteresis compensation on effective operation data based on the future temperature and the preset target temperature to obtain the future adjustment strategy; Environmental compensation module: Conduct environmental disturbance learning for future regulation strategies, obtain the best regulation strategy, and realize flexible control of thermal power units based on the best regulation strategy.

2. A thermal power unit flexibility improvement control system according to claim 1, characterized in that: The unit operation data includes: air flow, fuel flow and boiler pressure.

3. A thermal power unit flexibility improvement control system according to claim 2, characterized in that: The method of preprocessing the unit operation data includes: Each category of data in the unit operation data is fuzzy transformed to obtain the fuzziness of each category of data; based on the fuzziness of each category of data, the K-Means algorithm is used to perform cluster analysis on the fuzziness to obtain fuzzy membership clusters, and the fuzzy membership clusters are evaluated for rules to obtain fuzzy rules for each category of data in the unit operation data; each group of unit operation data is fuzzily described based on the fuzzy rules to obtain fuzzy expressions for each group of unit operation data; a rule recognition library is preset, and rule recognition is performed on the fuzzy expression based on the rule recognition library to obtain data labels, which include abnormal and normal data; the unit operation data is classified based on the data labels, and the data labels with abnormal labels are classified into one category as abnormal data, and the data labels with normal labels are classified into one category as normal data; abnormal correction is performed on the abnormal data based on the normal data to obtain time series correction data, and the time series correction data and normal data are merged to obtain effective operation data.

4. A thermal power unit flexibility improvement control system according to claim 3, characterized in that: The method of performing abnormal correction on abnormal data includes: According to the time series, the normal data is serialized to obtain the normal time series data; the observation matrix, the control matrix and the initial gain are preset, the initial gain is initialized to be infinitesimal, the time series prediction function is constructed based on the observation matrix and the control matrix, the time series normal data is predicted based on the time series prediction function to obtain the time series prediction data, the gain is calculated on the time series prediction data to obtain the prediction gain; the gain threshold is preset, when the prediction gain is greater than or equal to the gain threshold, the time series normal data at the corresponding moment is replaced by the time series prediction data to obtain the time series update data; the observation disturbance and the control disturbance are preset, the mean of the prediction gain is calculated to obtain the mean gain, when the mean gain is greater than or equal to the initial gain When , the observation matrix plus the observation disturbance is used as the new observation matrix, and the control matrix plus the control disturbance is used as the new control matrix. When the gain mean is less than the initial gain, the observation matrix minus the observation disturbance is used as the new observation matrix, and the control matrix minus the control disturbance is used as the new control matrix. The time series update data is used as the new time series normal data, and the process is repeated until the mean of the predicted gain converges. The observation matrix at this time is output as the optimal observation matrix, and the control matrix at this time is output as the optimal control matrix. The optimal observation matrix and the optimal control matrix are substituted into the time series prediction function to obtain the optimal prediction function. The abnormal data is predicted based on the optimal prediction function to obtain the time series corrected data.

5. A thermal power unit flexibility improvement control system according to claim 4, characterized in that: The boiler temperature prediction model is constructed in the following manner: Collect historical operation data and historical combustion temperature. The historical operation data includes: historical air flow, historical fuel flow and historical boiler pressure. The data type of the historical operation data is consistent with that of the data in the effective operation data. The historical air flow, historical fuel flow and historical boiler pressure in the historical operation data are used as the input of the boiler temperature prediction model, and the predicted combustion temperature is used as the output of the boiler temperature prediction model. The historical combustion temperature is used as the prediction target of the boiler temperature prediction model for model training to construct a boiler temperature prediction model. The boiler temperature prediction model is a long short-term memory model. The target loss function of the boiler temperature prediction model is a mean square error function. Minimizing the value of the target loss function is used as the training goal to obtain a boiler temperature prediction model with the smallest value of the target loss function.

6. A thermal power unit flexibility improvement control system according to claim 5, characterized in that: The method of controlling and adjusting the effective operation data includes: The time step error calculation is performed on the preset target temperature and the temperature at the future moment to obtain the error temperature at each moment, and the target adjustment function is constructed based on the error temperature. M groups of sequence control increments are initialized, where M is an integer greater than zero. The sequence control increment is the increment set at each moment, and the increment set contains the increment of each category of data. The optimal sequence control increment is initialized to be empty, and the optimal incremental fitness is initialized to be infinite. Each group of sequence control increments is regarded as a chromosome, and the genes in the chromosome correspond to the increment set at each moment in the sequence control increment. For each chromosome, the incremental fitness of each chromosome is calculated according to the target adjustment function. The smaller the incremental fitness, the better the chromosome. The chromosomes are selected based on the tournament selection algorithm to obtain the parent chromosome. The chromosomes in the parent chromosome set are paired with each other using the permutation and combination algorithm to obtain the parent combination. A gene in the parent combination is randomly selected as the exchange starting point, and the genes after the exchange starting point in the parent combination are exchanged to obtain the daughter chromosome. The incremental fitness of each daughter chromosome is calculated based on the target adjustment function, and the daughter chromosome with a fitness smaller than the parent combination is selected as the preferred generation. A mutation factor is preset, and the mutation factor is a minimum value greater than zero. A gene in the preferred generation is randomly selected as the mutation node. The value of A mutation factors is increased or decreased for the mutation node with random probability to obtain the mutant chromosome. The incremental fitness of each mutant chromosome is calculated based on the target adjustment function. The mutant chromosome with an incremental fitness less than that of the preferred generation is taken as the preferred variant. The preferred variant with the smallest incremental fitness is selected as the candidate variant. When the incremental fitness of the candidate variant is less than the optimal incremental fitness, the candidate variant is taken as the new optimal sequence control increment, and the incremental fitness of the candidate variant is taken as the new optimal incremental fitness. The preferred variant is taken as the new parent chromosome, and the process is repeated until the value of the optimal incremental fitness no longer changes. The optimal sequence control increment at this time is output as the future adjustment strategy.

7. A thermal power unit flexibility improvement control system according to claim 6, characterized in that: The formula of the objective adjustment function is: Wherein, J represents the target adjustment function, Tep represents the predicted time step, e(t+a) represents the error temperature at the t+a moment, Δv(t+a) represents the control increment at the t+a moment, and the control increment includes: air flow increment, fuel flow increment and boiler pressure increment; represents the temperature adjustment weight, and β represents the incremental adjustment weight.

8. A thermal power unit flexibility improvement control system according to claim 7, characterized in that: The method of performing environmental disturbance learning on the future adjustment strategy includes: Collect real-time combustion temperature and real-time environmental data, the real-time environmental data includes environmental temperature and environmental humidity, build a boiler state space based on real-time combustion temperature, real-time environmental data and future adjustment strategy, the boiler state space includes the current combustion temperature, environmental temperature, environmental humidity and the air flow, fuel flow and boiler pressure at the corresponding time in the future adjustment strategy; preset parameter adjustment actions, the parameter adjustment actions include: increase or decrease air flow, increase or decrease fuel flow and increase or decrease boiler pressure, arrange and combine the parameter adjustment actions to obtain an adjustment action table; preset parameter adjustment factors, the parameter adjustment factors include: air flow adjustment factor, fuel flow adjustment factor and boiler pressure adjustment factor; build an action reward function based on real-time combustion temperature, select parameter adjustment actions based on target temperature and real-time combustion temperature, when the target temperature and real-time combustion temperature are not equal, randomly select an action in the adjustment action table for parameter adjustment, calculate the reward value after parameter adjustment based on the action reward function, repeat and record the adjustment action and boiler state space with the largest reward value at each moment as the time-series optimal action, all time-series optimal actions constitute the best adjustment strategy.

9. A control system for improving the flexibility of a thermal power unit according to claim 8, characterized in that: The formula of the action reward function is: Re = -b1·|Tep now -Tep dest |-b2·ΔEmc; where Re represents the reward value, Tep now Represents the real-time combustion temperature, Tep dest represents the target temperature, b1 represents the temperature weight, ΔEmc represents the fuel increment, and b2 represents the fuel weight.

10. A method for improving the flexibility of a thermal power unit, which is implemented based on a thermal power unit flexibility improvement control system according to any one of claims 1 to 9, characterized in that: include: S1. Collect the unit operation data of the same time series, pre-process the unit operation data, and obtain effective operation data; S2. Predict the effective operation data based on the constructed boiler temperature prediction model to obtain the temperature at the future moment; S3, based on the preset target temperature and the temperature at the future time, hysteresis compensation is performed on the effective operation data to obtain the future adjustment strategy; S4. Conduct environmental disturbance learning on future regulation strategies to obtain the best regulation strategy, and achieve flexible control of thermal power units based on the best regulation strategy.

Images