Multi-objective energy consumption monitoring and optimization method and system for double-reheat unit

Abstract

The present application belongs to the field of monitoring and control of reheat units, in particular to a multi-objective energy consumption monitoring and optimization method and system for double-reheat units. The method first collects unit operation data through integrated sensors, and after preprocessing, determines the sub-decision variables of the main target factor using principal component analysis. Second, by constructing a multi-objective optimization model, combining angle penalty distance and genetic algorithm, the single and mixed main target optimal point set is obtained. Third, according to the obtained single and mixed main target optimal point set, the servo control model trained by reinforcement learning and PID control algorithm is used to obtain the output power fluctuation rate. If the optimal output power fluctuation rate is not met, the real-time minimum fluctuation rate is fed back to the servo control model and the multi-objective optimization model for secondary optimization until the optimal output condition is met. The present application realizes multi-objective monitoring and optimization of double-reheat unit energy consumption, and improves the unit operation efficiency and stability.

Description

Technical Field

[0001] The present invention belongs to the field of reheat unit monitoring and control, and in particular relates to a multi-objective energy consumption monitoring and optimization method and system for a secondary reheat unit. Background Art

[0002] In the current energy sector, double-reheat coal-fired power generation units are widely used due to their efficient ability to convert coal into electricity. These units utilize a double steam reheat process to improve thermal cycle efficiency, reduce coal consumption, and minimize greenhouse gas emissions. However, with increasing environmental standards and fluctuating energy costs, optimizing the operation of double-reheat units, achieving refined energy consumption management and multi-objective balancing, has become a pressing issue.

[0003] For example, the patent with publication number CN114488798A discloses a method for performance monitoring and operation optimization of a secondary reheat unit based on data coordination, wherein the method includes: studying the dominant factors and action mechanisms that affect the characteristics of the main components of the thermal system, obtaining the component characteristic curves within the full operating range, and establishing a high-precision mathematical model of the key components of the thermal system under all operating conditions; utilizing the measurement redundant information of the unit's thermal parameters and adopting a data coordination algorithm to reduce the measurement uncertainty of key data; providing accurate expected values ​​for component health status monitoring, and realizing various performance monitoring functions by comparing monitoring values ​​with expected values; on the basis of the high-precision mathematical model under all operating conditions, obtaining a system optimization model through overall integration, obtaining the optimization potential of the unit and parameter adjustment through scenario hypothesis calculation to realize real-time operation optimization of the unit.

[0004] The above existing technologies have the following problems: 1. How to establish an accurate dynamic model based on the numerous controllable and uncontrollable variables involved in the operation of the secondary reheat unit and implement the optimal control strategy under multi-objective functions is the primary difficulty; 2) How to maintain power output balance while considering optimal energy consumption is also a challenge to the existing technologies. To solve the above problems, the present invention proposes a multi-objective energy consumption monitoring and optimization method and system for secondary reheat units. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention proposes a multi-objective energy consumption monitoring and optimization method and system for secondary reheat units. The method first collects unit operation data through integrated sensors, and after preprocessing, uses principal component analysis to determine the sub-decision variables of the main objective factors. Secondly, by constructing a multi-objective optimization model, combined with angle penalty distance and genetic algorithm, single and mixed main objective optimal point sets are obtained. Thirdly, a servo control model trained with reinforcement learning and PID control algorithm is used to obtain output power fluctuation rate. If the optimal output power fluctuation rate is not met, the real-time minimum fluctuation rate is fed back to the servo control model and the multi-objective optimization model for secondary optimization until the optimal output condition is met. The present invention realizes multi-objective monitoring and optimization of the energy consumption of the secondary reheat unit, and improves the operating efficiency and stability of the unit.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A multi-objective energy consumption monitoring and optimization method for a double reheat unit, including:

[0008] Step S1: obtaining controllable and uncontrollable variables of the secondary reheat unit operation, and preprocessing the collected controllable and uncontrollable variables to obtain a variable data set in a unified format;

[0009] Step S2: Energy consumption, heat conversion efficiency, and environmental factors are set as primary target factors. Based on the set primary target factors and the obtained variable data set, the principal component analysis method is used to delineate sub-decision variables corresponding to the primary target factors. The delineated sub-decision variables are then used to construct three single sub-decision variable sequences and one mixed sub-decision variable sequence.

[0010] Step S3: construct a multi-objective optimization model using the angle penalty distance and genetic algorithm, input the obtained three single sub-decision variables into the multi-objective optimization model to obtain the single main objective optimal point set, and simultaneously input the mixed sub-decision variable set into the multi-objective optimization model to obtain the mixed main objective optimal point set;

[0011] Step S4: A servo control model is constructed and trained by combining reinforcement learning with a PID control algorithm. The obtained single main objective optimal point set and the mixed main objective optimal point set are respectively input into the servo control model to obtain the output power fluctuation rate 1 and the output power fluctuation rate 2 obtained by monitoring the corresponding secondary reheat unit.

[0012] Step S5, set the optimal output power fluctuation rate to 0, compare the obtained output power fluctuation rate 1 with the output power fluctuation rate 2, obtain the real-time minimum output power fluctuation rate, and compare the real-time minimum output power fluctuation rate with the optimal output power fluctuation rate. If they are not satisfied, the real-time minimum output power fluctuation rate is fed back to the servo control model, and at the same time, the main target optimal point set corresponding to the real-time minimum output power fluctuation rate is fed back to the multi-objective optimization model for secondary optimization, so that the output main target optimal point set corresponding to the real-time minimum output power fluctuation rate meets the optimal output power fluctuation rate.

[0013] Specifically, the specific steps of step S2 include:

[0014] S201. Set the principal component factor contribution rate threshold, input the energy consumption main target factor and the obtained variable data set into the principal component analysis algorithm, calculate the sum of the contribution rates corresponding to different sub-decision variables, and compare the calculated sum of the contribution rates with the contribution rate threshold, obtain the valid sub-decision variables whose sum of contribution rates is greater than the contribution rate threshold, and use the obtained sub-decision variables to construct the first sub-decision variable sequence ;in represents the nth decision variable in the first sub-decision variable sequence;

[0015] S202: Repeat step S201 to obtain the second sub-decision variable sequence corresponding to thermal efficiency The third sub-decision variable sequence corresponding to the environmental factors ;in Indicates the first m decision variables, Indicates the first s decision variables

[0016] S203, using the obtained first sub-decision variable sequence, the second sub-decision variable sequence and the third sub-decision variable sequence, and constructing a mixed sub-decision variable sequence through a random filtering algorithm , represents the first k decision variables, K represents the total number of decision variables in the mixed sub-decision variable sequence, .

[0017] Specifically, the specific steps of constructing the multi-objective optimization model in step S3 include:

[0018] S301, construct a multi-objective optimization function by setting the main objective factor and the first sub-decision variable sequence, the second sub-decision variable sequence and the third sub-decision variable sequence ,in represents the energy consumption objective function, represents the thermal efficiency objective function, represents the environmental factor objective function;

[0019] S302: Constructing an initial population of the genetic algorithm using the obtained first sub-decision variable sequence, second sub-decision variable sequence, and third sub-decision variable sequence ;

[0020] S303, according to the constructed initial population, calculate the corresponding fitness value of each decision variable through the fitness function and the constraint condition set, where the fitness function Specifically:

[0021] ,

[0022] The constraint set U is specifically:

[0023]

[0024] in, Indicates the maximum heat conversion efficiency corresponding to the boiler in the secondary machinery group, represents the environmental factor threshold; Indicates the difference between the optimal output power fluctuation rate and the real-time minimum output power fluctuation rate, Indicates the constraint condition set u The penalty function of the constraint condition, the main objective factor or decision variable violates the u When a constraint , otherwise 0; represents the penalty coefficient corresponding to the u-th constraint in the constraint set, is the weight coefficient of the corresponding factor and variable; Indicates the minimum energy consumption threshold per unit time. Indicates the maximum energy consumption threshold per unit time, Indicates the minimum thermal conversion efficiency threshold per unit time, Indicates the maximum thermal conversion efficiency threshold per unit time; Represents environmental factors Thresholds of environmental factors The difference, represents the constraint constant; Represents the angle penalty distance balance coefficient corresponding to a single sub-decision variable sequence, It represents the angle penalty distance penalty values ​​corresponding to the first sub-decision variable sequence, the second sub-decision variable sequence, and the third sub-decision variable sequence respectively.

[0025] Specifically, the specific steps of constructing the multi-objective optimization model in step S3 also include:

[0026] S304. Set an angle penalty distance threshold, introduce the NSGA algorithm, use the non-dominated sorting in the NSGA algorithm to sort the corresponding decision variable values ​​in the first sub-decision variable sequence, the second sub-decision variable sequence, and the third sub-decision variable sequence in the initial population, and use the angle penalty distance formula to output the angle penalty distance of the sorted decision variables;

[0027] S305: The first sub-decision variable sequence, the second sub-decision variable sequence and the third sub-decision variable sequence are n decision variables, m decision variables and k The angle penalty distance corresponding to the decision variable is compared with the angle penalty distance threshold. n decision variables, m decision variables and k The angle penalty distances corresponding to the decision variables are less than the angle penalty distance threshold, then the n decision variables, m decision variables and k decision variables are retained to obtain the quadratic optimal population in the genetic algorithm;

[0028] S306, setting the stopping condition and iteration cycle of the genetic algorithm, integrating the multi-objective optimization function constructed in S301-S305, the initial population, the fitness function, the constraint condition set, and the secondary optimal population obtained by the angle penalty distance into the original genetic algorithm to obtain a multi-objective optimization model;

[0029] S307: According to the set stop condition or iteration cycle, the variable data set obtained by real-time monitoring is input into the multi-objective optimization model for iteration, and the corresponding single main objective optimal point set is calculated.

[0030] Specifically, the specific steps of constructing the multi-objective optimization model in step S3 also include:

[0031] S308. When the input variables of the multi-objective optimization model are the main objective factors and the mixed sub-decision variable sequence, the corresponding multi-objective optimization function is , the initial population , the corresponding fitness function Specifically:

[0032]

[0033] in, represents the energy consumption objective function corresponding to the hybrid sub-decision variable sequence, represents the thermal efficiency objective function corresponding to the hybrid sub-decision variable sequence, Represents the environmental factor objective function corresponding to the sequence of mixed sub-decision variables; represents the collinearity penalty term between all variables in the mixed sub-decision variable sequence, represents the collinearity penalty coefficient, , Where represents the number of collinear decision variables in the mixed sub-decision variable sequence; represents the angle penalty distance balance coefficient corresponding to the mixed sub-decision variable sequence, Represents the angle penalty distance penalty value corresponding to the mixed sub-decision variable sequence, Represents the environmental factors in the sequence of mixed sub-decision variables Thresholds of environmental factors further, in this embodiment, the collinearity calculation between all variables and the number of collinear sub-decision variables are calculated by factor analysis algorithm.

[0034] S309: Replace the multi-objective optimization function, initial population and fitness function constructed in S308 with the corresponding variables and functions in S301-S303, and repeat the process of S301-S307 to calculate and obtain the optimal point set of the hybrid main objective.

[0035] This process constructs a multi-objective optimization model that integrates energy consumption, thermal efficiency, and environmental factors. By introducing the NSGA algorithm and angle penalty distance, the diversity and quality of the optimized solutions are effectively improved. In particular, when dealing with mixed sub-decision variable sequences, the design of collinearity penalty terms further avoids collinearity between variables, ensuring the accuracy and reliability of the optimization results. This method not only improves energy efficiency but also reduces environmental pollution, providing an effective decision support tool for industrial production and energy management.

[0036] Furthermore, the variable data set includes historical and real-time output power values ​​of the secondary reheat unit, primary steam temperature and secondary reheat steam temperature of the secondary reheat unit, main steam pressure and reheat steam pressure, energy consumption, heat conversion efficiency, environmental factors, and execution action values ​​corresponding to the historical optimal point set of the secondary reheat unit and the power value required by the power grid;

[0037] The specific steps of constructing and training the servo control model in step S4 include:

[0038] S401, using the variable data set to build the enhanced adjustment layer in the servo control model, the enhanced adjustment layer includes the current t Input status at all times , perform the action and current t Moment reward function ,

[0039] current t The input status at all times is:

[0040] , the execution action is: ,current t The moment reward function is: ;in, Indicates the current t The initial steam temperature of the secondary reheat unit at time Indicates the current t Secondary reheat steam temperature of the secondary reheat unit at time Indicates the current t Main steam pressure of secondary reheat unit at time Indicates the current t The reheat steam pressure of the secondary reheat unit at this moment, Indicates that when the condition is not met When , the feedback action of finding the optimal point set is triggered; Indicates the environmental factor corresponding to the mixed sub-decision variable sequence k Decision variables and environmental factor thresholds The difference.

[0041] Specifically, the output power fluctuation rate in step S5 is The specific calculation formula is:

[0042] ,

[0043] in, Indicates the current t The power value required by the grid at any moment, Indicates the current t The power value output by the secondary reheat unit at this moment.

[0044] A multi-objective energy consumption monitoring and optimization system for secondary reheat units, including: data processing module, data screening module, optimization model module, and servo control model module;

[0045] A data processing module is used to obtain the controllable and uncontrollable variables of the secondary reheat unit operation from the deployed integrated sensors and pre-process these variables to obtain a variable data set in a unified format;

[0046] The data screening module includes a variable screening unit and a sub-decision variable sequence construction unit;

[0047] The variable screening unit is used to obtain valid sub-decision variables based on the set main target factor and the obtained variable data set using the configured principal component analysis algorithm; the sub-decision variable sequence construction unit is used to construct the sub-decision variable sequence and mixed sub-decision variable sequence corresponding to the main target factor using the obtained valid sub-decision variables;

[0048] The optimization model module includes an optimization model building unit and an optimal point set unit;

[0049] The optimization model construction unit is used to construct a multi-objective optimization model through angle penalty distance and genetic algorithm; the optimal point set unit is used to calculate the optimal point set corresponding to the main objective factor through the constructed multi-objective optimization model based on the sub-decision variable sequence and mixed sub-decision variable sequence corresponding to the main objective factor.

[0050] Specifically, the monitoring and optimization system also includes: a servo control model module;

[0051] The servo control model module includes a servo control model construction training unit and a monitoring feedback unit;

[0052] The servo control model construction and training unit is used to construct a servo control model using reinforcement learning combined with the PID control algorithm, and to train the constructed servo control model using the acquired historical optimal point set; the monitoring and feedback unit is used to generate a control signal based on the optimal point set acquired in real time and the trained servo control model, to control the operation of the secondary reheat unit, and to monitor the output minimum output power fluctuation rate in real time, and to perform feedback optimization on the servo control model and the multi-objective optimization model through the set optimal output power fluctuation rate and the monitored minimum output power fluctuation rate.

[0053] A computer-readable storage medium stores computer instructions, which, when executed, execute a multi-objective energy consumption monitoring and optimization method for a secondary reheat unit.

[0054] An electronic device comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the steps of the multi-objective energy consumption monitoring and optimization method for a secondary reheat unit.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] This invention addresses the shortcomings of existing technologies. First, through an integrated sensor network, a large number of controllable and uncontrollable variables in unit operation are monitored and preprocessed in real time. This establishes an accurate dynamic model, providing a solid data foundation for multi-objective optimization. Second, through principal component analysis, key sub-decision variables are effectively extracted, reducing dimensionality while improving the model's interpretability and computational efficiency. Third, a multi-objective optimization model constructed using angle-penalized distance and a genetic algorithm finds the optimal point set that satisfies multi-objective constraints. This model can simultaneously address multiple core objectives, such as energy consumption, thermal efficiency, and environmental factors, thereby solving and implementing an optimal control strategy under multi-objective functions. This invention addresses the shortcomings of existing technologies by integrating reinforcement learning with PID control algorithms to construct a servo control model that dynamically adjusts the control strategy based on the multi-objective optimization results, achieving precise control of the secondary reheat unit and effectively reducing output power fluctuation. This approach not only improves the intelligent control level but also enhances the system's adaptability and stability, ensuring high quality and stability of power output while reducing energy consumption. This invention implements a closed-loop control strategy by setting the optimal output power fluctuation rate to zero and continuously comparing the real-time minimum output power fluctuation rate with the optimal target. This mechanism promptly identifies and provides feedback on control deficiencies, driving the servo control model and multi-objective optimization model to self-adjust and perform secondary optimization, gradually approaching or even achieving the optimal output power fluctuation rate. This dynamic feedback mechanism significantly improves the system's responsiveness to complex operating conditions and optimization efficiency, ensuring high efficiency and cost-effectiveness in long-term operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is a flow chart of a multi-objective energy consumption monitoring and optimization method for a secondary reheat unit according to embodiment 1 of the present invention;

[0058] Figure 2 This is a system module architecture diagram of a multi-objective energy consumption monitoring and optimization method for a secondary reheat unit according to Example 2 of the present invention. DETAILED DESCRIPTION

[0059] Example 1

[0060] See also Figure 1 The present invention provides an embodiment of a multi-objective energy consumption monitoring and optimization method for a secondary reheat unit, the specific steps of which include:

[0061] Step S1: Obtain controllable and uncontrollable variables of the secondary reheat unit operation through the deployed integrated sensors, and pre-process the collected controllable and uncontrollable variables to obtain a variable data set in a unified format;

[0062] Step S2: Energy consumption, heat conversion efficiency, and environmental factors are set as primary target factors. Based on the set primary target factors and the obtained variable data set, sub-decision variables corresponding to the primary target factors are delineated using the principal component analysis method. The delineated sub-decision variables are then used to construct three single sub-decision variable sequences and one mixed sub-decision variable sequence. Furthermore, in this embodiment, the controllable variables corresponding to energy consumption include fuel supply, steam flow, reheat steam pressure, and air-fuel ratio; the uncontrollable variables corresponding to energy consumption include fuel quality, ambient temperature, and equipment aging. In addition, in this embodiment, fuel supply is used as the primary target factor in the calculation of energy consumption.

[0063] Controllable variables corresponding to heat conversion efficiency include: the change in primary steam temperature and secondary reheat steam temperature within the secondary reheat unit per unit time, excess air coefficient, turbine inlet steam temperature, and number of regenerative stages. Uncontrollable variables corresponding to heat conversion efficiency include: combustion reaction rate and flue gas emission loss. In this embodiment, the change in secondary reheat steam temperature is used as the primary target factor in heat conversion efficiency calculation.

[0064] The controllable variables corresponding to the environmental factors include: the real-time concentration of carbon dioxide in the exhaust gas, the efficiency of the desulfurization and denitrification system, and the dust concentration in the exhaust gas. The uncontrollable variables corresponding to the environmental factors include: the atmospheric diffusion rate. In this embodiment, the real-time concentration of carbon dioxide in the exhaust gas is used as the main target factor in the calculation of the environmental factors.

[0065] Step S3: construct a multi-objective optimization model using the angle penalty distance and genetic algorithm, input the obtained three single sub-decision variables into the multi-objective optimization model to obtain the single main objective optimal point set, and simultaneously input the mixed sub-decision variable set into the multi-objective optimization model to obtain the mixed main objective optimal point set;

[0066] Step S4: A servo control model is constructed and trained by combining reinforcement learning with a PID control algorithm. The obtained single main objective optimal point set and the mixed main objective optimal point set are respectively input into the servo control model to obtain the output power fluctuation rate 1 and the output power fluctuation rate 2 obtained by monitoring the corresponding secondary reheat unit.

[0067] Step S5, set the optimal output power fluctuation rate to 0, compare the obtained output power fluctuation rate 1 with the output power fluctuation rate 2, obtain the real-time minimum output power fluctuation rate, and compare the real-time minimum output power fluctuation rate with the optimal output power fluctuation rate. If they are not satisfied, the real-time minimum output power fluctuation rate is fed back to the servo control model, and at the same time, the main target optimal point set corresponding to the real-time minimum output power fluctuation rate is fed back to the multi-objective optimization model for secondary optimization, so that the output main target optimal point set corresponding to the real-time minimum output power fluctuation rate meets the optimal output power fluctuation rate.

[0068] Furthermore, the specific steps of step S2 include:

[0069] S201. Set the principal component factor contribution rate threshold, input the energy consumption main target factor and the obtained variable data set into the principal component analysis algorithm, calculate the sum of the contribution rates corresponding to different sub-decision variables, and compare the calculated sum of the contribution rates with the contribution rate threshold, obtain the valid sub-decision variables whose sum of contribution rates is greater than the contribution rate threshold, and use the obtained sub-decision variables to construct the first sub-decision variable sequence ;in represents the nth decision variable in the first sub-decision variable sequence; in this embodiment, the principal component factor contribution rate threshold is 0.8;

[0070] S202: Repeat step S201 to obtain the second sub-decision variable sequence corresponding to thermal efficiency The third sub-decision variable sequence corresponding to the environmental factors ;in Indicates the first m decision variables, Indicates the first s decision variables

[0071] S203, using the obtained first sub-decision variable sequence, the second sub-decision variable sequence and the third sub-decision variable sequence, and constructing a mixed sub-decision variable sequence through a random filtering algorithm , represents the first k decision variables, K represents the total number of decision variables in the mixed sub-decision variable sequence, .

[0072] Principal component analysis effectively reduces dimensionality and screens key variables, constructing a decision sequence focused on energy consumption, thermal efficiency, and environmental impact. A comprehensive hybrid sequence is then generated through a random fusion strategy, significantly enhancing the decision accuracy and efficiency of the multi-objective optimization model and promoting the comprehensive optimal allocation of resources. This embodiment uses a random filtering algorithm to merge and eliminate duplicate variables in the hybrid sub-decision variable sequence. The algorithm includes cosine similarity, Euclidean distance, Jaccard similarity, clustering algorithms, and the distinct function in the database.

[0073] Furthermore, the specific steps of constructing the multi-objective optimization model in step S3 include:

[0074] S301, construct a multi-objective optimization function by setting the main objective factor and the first sub-decision variable sequence, the second sub-decision variable sequence and the third sub-decision variable sequence ,in represents the energy consumption objective function, represents the thermal efficiency objective function, represents the environmental factor objective function;

[0075] S302: Constructing an initial population of the genetic algorithm using the obtained first sub-decision variable sequence, second sub-decision variable sequence, and third sub-decision variable sequence ;

[0076] S303, according to the constructed initial population, calculate the corresponding fitness value of each decision variable through the fitness function and the constraint condition set, where the fitness function Specifically:

[0077] ,

[0078] The constraint set U is specifically:

[0079]

[0080] in, Indicates the maximum heat conversion efficiency corresponding to the boiler in the secondary machinery group, represents the environmental factor threshold; Indicates the difference between the optimal output power fluctuation rate and the real-time minimum output power fluctuation rate, Indicates the constraint condition set u The penalty function of the constraint condition, the main objective factor or decision variable violates the u When a constraint , otherwise 0; represents the penalty coefficient corresponding to the u-th constraint in the constraint set, is the weight coefficient of the corresponding factor and variable; Indicates the minimum energy consumption threshold per unit time. Indicates the maximum energy consumption threshold per unit time, Indicates the minimum thermal conversion efficiency threshold per unit time, Indicates the maximum thermal conversion efficiency threshold per unit time; Represents environmental factors Thresholds of environmental factors The difference, represents the constraint constant; Represents the angle penalty distance balance coefficient corresponding to a single sub-decision variable sequence, The angle penalty distance penalty values ​​corresponding to the first sub-decision variable sequence, the second sub-decision variable sequence, and the third sub-decision variable sequence are represented in sequence; in this embodiment, the maximum heat conversion efficiency is set by those skilled in the art based on the basic attribute data of the boiler in the secondary machinery group; in this embodiment, the environmental factor threshold is determined by the concentration of carbon dioxide in the exhaust gas. The setting of is determined by those skilled in the art according to the environmental emission standards; the heat conversion efficiency is calculated by the mass of fuel burned per unit time, the combustion coefficient, and the temperature change of the corresponding mass of water in the boiler of the secondary machinery group per unit time according to the laws of thermodynamics. The laws of thermodynamics are: ,in Q Indicates the heat value absorbed by the corresponding mass of water in the boiler of the secondary machinery group per unit time. M Indicates the quality of water in the boiler of the secondary machinery group at the current moment, represents the specific heat capacity of water, Indicates the temperature change of the corresponding mass water in the boiler of the secondary machinery group per unit time; represents a very small constraint constant, which is set by those skilled in the art based on experience and is set to 0.01 in this embodiment;

[0081] and The specific steps to obtain it include:

[0082] S3031. Collect the fuel supply corresponding to the historical energy consumption and other controllable and uncontrollable sub-decision variables, as well as the minimum energy consumption threshold and the maximum energy consumption threshold corresponding to the fuel supply;

[0083] S3032. Build a prediction model using a support vector machine, input the collected data into the prediction model for training, and obtain a trained prediction model;

[0084] S3032. Energy consumption data collected during the real-time operation of the secondary reheat turbine wheel group is input into the trained prediction model to obtain the minimum energy consumption threshold and the maximum energy consumption threshold corresponding to each moment;

[0085] S304. Set an angle penalty distance threshold, introduce the NSGA algorithm, use the non-dominated sorting in the NSGA algorithm to sort the corresponding decision variable values ​​in the first sub-decision variable sequence, the second sub-decision variable sequence, and the third sub-decision variable sequence in the initial population, and use the angle penalty distance formula to output the angle penalty distance of the sorted decision variables;

[0086] S305: The first sub-decision variable sequence, the second sub-decision variable sequence and the third sub-decision variable sequence are nThe angle penalty distances corresponding to the nth decision variable, the mth decision variable, and the kth decision variable are compared with the angle penalty distance threshold. If the angle penalty distances corresponding to the nth decision variable, the mth decision variable, and the kth decision variable are respectively less than the angle penalty distance threshold, the nth decision variable, the mth decision variable, and the kth decision variable are retained to obtain the quadratic optimal population in the genetic algorithm. Furthermore, the NSGA algorithm is a multi-objective optimization algorithm based on the Pareto dominance relationship. It uses non-dominated sorting to stratify the individuals in the population. The individuals in each layer are not dominated by each other. The non-dominated sorting is used for each individual in the initial population (that is, each group of the first sub-decision variable sequence, the second sub-decision variable sequence, and the third sub-decision variable sequence) to calculate its dominance relationship with other individuals and perform stratification. For example, assuming that the objective function is , there are 4 individuals in the initial population (i.e., a combination of 4 decision variable sequences), and the coordinates of the 4 individuals in the target space are:

[0087] Instance 1: = (10, 80, 20),

[0088] Instance 2: = (12, 78, 22),

[0089] Instance 3: = (11, 82, 19),

[0090] Instance 4: = (10, 79, 21),

[0091] Assume that the calculated angle penalty distance threshold is For the above four individuals, calculate the angle formed by them and other individuals in the target space, and calculate the angle penalty distance. Assume that the angle formed by them and other individuals in the target space is calculated by the cosine formula, and the angle penalty distance formula is used to calculate the corresponding angle penalty distances of the four individuals:

[0092] Individual 1: distance1 = 0.5, individual 2: distance2 = 0.7, individual 3: distance3 = 0.2, individual 4: distance4 = 0.3;

[0093] By comparing with the angle penalty distance threshold, if the angle penalty distance between individual 3 and individual 4 is less than the threshold, then individual 3 and individual 4 are retained to form the secondary optimal population; the angle penalty distance formula in this embodiment is the prior art and will not be repeated here;

[0094] S306, setting the stopping condition and iteration cycle of the genetic algorithm, integrating the multi-objective optimization function constructed in S301-S305, the initial population, the fitness function, the constraint condition set, and the secondary optimal population obtained by the angle penalty distance into the original genetic algorithm to obtain a multi-objective optimization model;

[0095] S307: According to the set stop condition or iteration cycle, the variable data set obtained by real-time monitoring is input into the multi-objective optimization model for iteration, and the corresponding single main objective optimal point set is calculated.

[0096] S308. When the input variables of the multi-objective optimization model are the main objective factors and the mixed sub-decision variable sequence, the corresponding multi-objective optimization function is , the initial population , the corresponding fitness function Specifically:

[0097]

[0098] in, represents the energy consumption objective function corresponding to the hybrid sub-decision variable sequence, represents the thermal efficiency objective function corresponding to the hybrid sub-decision variable sequence, Represents the environmental factor objective function corresponding to the sequence of mixed sub-decision variables; represents the collinearity penalty term between all variables in the mixed sub-decision variable sequence, represents the collinearity penalty coefficient, , Where represents the number of collinear decision variables in the mixed sub-decision variable sequence; represents the angle penalty distance balance coefficient corresponding to the mixed sub-decision variable sequence, Represents the angle penalty distance penalty value corresponding to the mixed sub-decision variable sequence, Represents the environmental factors in the sequence of mixed sub-decision variables Thresholds of environmental factors The difference between

[0099] S309: Replace the multi-objective optimization function, initial population and fitness function constructed in S308 with the corresponding variables and functions in S301-S303, and repeat the process of S301-S307 to calculate and obtain the optimal point set of the hybrid main objective.

[0100] Furthermore, the variable data set also includes historical and real-time output power values ​​of the secondary reheat unit, primary steam temperature and secondary reheat steam temperature of the secondary reheat unit, main steam pressure and reheat steam pressure, energy consumption, heat conversion efficiency, environmental factors, and execution action values ​​corresponding to the historical optimal point set of the secondary reheat unit and power demand value of the power grid;

[0101] The specific steps of constructing the servo control model in step S4 are:

[0102] S401, constructing the current value in the enhanced adjustment layer of the servo control model according to the acquired variable data set t Input status at all times , perform the action and current t Moment reward function , among which, the current t The input status at all times is:

[0103] , perform the action ,current t Moment reward function ;in, Indicates the current t The initial steam temperature of the secondary reheat unit at time Indicates the current t Secondary reheat steam temperature of the secondary reheat unit at time Indicates the current t Main steam pressure of secondary reheat unit at time Indicates the current t The reheat steam pressure of the secondary reheat unit at this moment, Indicates that when the condition is not met When , the feedback action of finding the optimal point set is triggered; Indicates the environmental factor corresponding to the mixed sub-decision variable sequence k Decision variables and environmental factor thresholds The difference between

[0104] S402: Construct a reinforcement adjustment sub-model in the reinforcement adjustment layer using a reinforcement learning algorithm, and input the constructed current time t input state, execution action, and current time t reward function into the reinforcement adjustment model to obtain a trained reinforcement adjustment sub-model. In this embodiment, reinforcement learning algorithms include: Q-Learning, SARSA algorithm, Deep Q-Network, Trust Region Policy Optimization, SAC, and Hierarchical Reinforcement Learning.

[0105] S403, constructing a control sub-model in the control layer of the servo control model by using the PID algorithm, and parameterizing the parameters in the control sub-model using the output power fluctuation rate. The parameters in the control sub-model further include: proportional coefficient Kp , integral coefficient Ki and differential coefficients Kd;

[0106] S404: Integrate the enhanced adjustment sub-model trained in S402 and the enhanced adjustment sub-model initialized in S403 into the servo control model, input the variable data set obtained from the real-time operation monitoring of the secondary reheat turbine group and the three single main objective optimal point sets into the servo control model, and output corresponding triple single control instructions for the secondary reheat turbine group. Simultaneously, input the variable data set obtained from the real-time operation monitoring of the secondary reheat turbine group and the hybrid main objective optimal point set into the servo control model, and output corresponding hybrid control instructions for the secondary reheat turbine group.

[0107] S405: Input the acquired triple single control instruction and hybrid control instruction into the control unit for controlling the secondary reheat turbine to generate electricity, monitor the power value 1 corresponding to the triple single control instruction and the power value 2 corresponding to the hybrid control instruction in real time, and simultaneously monitor the power value required by the power grid;

[0108] S406, using the obtained power value 1 and power value 2 and the real-time monitored power value of the grid demand, use the calculation formula in step S5 below to calculate the corresponding output power fluctuation rate 1 and output power fluctuation rate 2, output power fluctuation rate 6

[0109] ,

[0110] in, Indicates the current t The power value required by the grid at any moment, Indicates the current t The power value output by the secondary reheat unit at the moment;

[0111] S407: Compare the calculated output power fluctuation rate 1 and the calculated output power fluctuation rate 2 to obtain the real-time minimum output power fluctuation rate;

[0112] S408. Compare the real-time minimum output power fluctuation rate with the optimal output power fluctuation rate set in step S5. If they are equal, maintain the existing secondary reheat unit operating state. If they are not equal, feed the difference between the real-time minimum output power fluctuation rate and the optimal output power fluctuation rate back to the multi-objective optimization model to calculate the secondary main objective optimal point set. The calculated secondary main objective optimal point set is input into the servo control model to generate secondary control instructions to control the secondary reheat unit to operate and generate electricity.

[0113] S409. Repeat step S408 so that the secondary reheat unit always operates under the control instructions generated by the optimal point set, and keeps the output power in a stable state in real time.

[0114] Example 2

[0115] See also Figure 2 , another embodiment provided by the present invention: a multi-objective energy consumption monitoring and optimization system for a secondary reheat unit, comprising: a data processing module, a data screening module, an optimization model module, a servo control model module and an interaction module;

[0116] A data processing module is used to obtain the controllable and uncontrollable variables of the secondary reheat unit operation from the deployed integrated sensors and pre-process these variables to obtain a variable data set in a unified format;

[0117] The data screening module is used to set the main target factor and obtain the corresponding sub-decision variable sequence based on the set main target factor. The data screening module includes a variable screening unit and a sub-decision variable sequence construction unit. The variable screening unit is used to obtain valid sub-decision variables based on the set main target factor and the obtained variable data set using the configured principal component analysis algorithm.

[0118] A sub-decision variable sequence construction unit is used to construct a sub-decision variable sequence and a mixed sub-decision variable sequence corresponding to the main target factor through the obtained valid sub-decision variables;

[0119] The optimization model module is used to construct a multi-objective optimization model and obtain the optimal point set of the main objective based on the constructed multi-objective optimization model. The optimization model module includes an optimization model construction unit and an optimal point set unit. The optimization model construction unit is used to construct a multi-objective optimization model using angle penalty distance and genetic algorithm. The optimal point set unit is used to calculate the optimal point set corresponding to the main objective factor through the constructed multi-objective optimization model based on the sub-decision variable sequence and mixed sub-decision variable sequence corresponding to the obtained main objective factor.

[0120] The servo control model module is used to build a servo control model and input the obtained optimal point set into the servo control model to generate a control signal to control the operation of the secondary reheat unit, monitor the power output in real time, and provide optimization feedback. The servo control model module includes a servo control model construction training unit and a monitoring feedback unit.

[0121] The servo control model construction and training unit is used to construct a servo control model using reinforcement learning combined with a PID control algorithm, and to train the constructed servo control model using the acquired historical optimal point set. The monitoring and feedback unit is used to generate a control signal based on the acquired optimal point set and the trained servo control model to control the operation of the secondary reheat unit, and to monitor the output minimum output power fluctuation rate in real time. The servo control model and the multi-objective optimization model are optimized through feedback based on the set optimal output power fluctuation rate and the monitored minimum output power fluctuation rate.

[0122] The interactive module is used to set the interactive interface, through which the parameters of the secondary reheat unit are set and the real-time monitoring parameters are visualized.

[0123] Example 3

[0124] A computer-readable storage medium stores computer instructions, which, when executed, execute a multi-objective energy consumption monitoring and optimization method for a secondary reheat unit.

[0125] Example 4

[0126] An electronic device comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement a multi-objective energy consumption monitoring and optimization method for a secondary reheat unit.

[0127] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also change, modify, replace and modify the above-mentioned embodiments without departing from the scope of protection of the purpose of the present invention and the claims, and all of these are protected by the present invention.

Claims

1. A multi-objective energy consumption monitoring and optimization method for a secondary reheat unit, characterized in that: include: Step S1: obtaining controllable and uncontrollable variables of the secondary reheat unit operation, and preprocessing the collected controllable and uncontrollable variables to obtain a variable data set in a unified format; Step S2: Energy consumption, heat conversion efficiency, and environmental factors are set as primary target factors. Based on the set primary target factors and the obtained variable data set, the principal component analysis method is used to delineate sub-decision variables corresponding to the primary target factors. The delineated sub-decision variables are then used to construct three single sub-decision variable sequences and one mixed sub-decision variable sequence. Step S3: construct a multi-objective optimization model using the angle penalty distance and genetic algorithm, input the obtained three single sub-decision variables into the multi-objective optimization model to obtain the single main objective optimal point set, and simultaneously input the mixed sub-decision variable set into the multi-objective optimization model to obtain the mixed main objective optimal point set; Step S4: A servo control model is constructed and trained by combining reinforcement learning with a PID control algorithm. The obtained single main objective optimal point set and the mixed main objective optimal point set are respectively input into the servo control model to obtain the output power fluctuation rate 1 and the output power fluctuation rate 2 obtained by monitoring the corresponding secondary reheat unit. Step S5: Set the optimal output power fluctuation rate to 0, compare the obtained output power fluctuation rate 1 with the output power fluctuation rate 2, obtain the real-time minimum output power fluctuation rate, and compare the real-time minimum output power fluctuation rate with the optimal output power fluctuation rate. If they are not satisfied, feed the real-time minimum output power fluctuation rate back to the servo control model, and at the same time feed the main objective optimal point set corresponding to the real-time minimum output power fluctuation rate back to the multi-objective optimization model for secondary optimization, so that the output main objective optimal point set corresponding to the real-time minimum output power fluctuation rate satisfies the optimal output power fluctuation rate. The specific steps of step S2 include: S201. Set the principal component factor contribution rate threshold, input the energy consumption main target factor and the obtained variable data set into the principal component analysis algorithm, calculate the sum of the contribution rates corresponding to different sub-decision variables, and compare the calculated sum of the contribution rates with the contribution rate threshold, obtain the valid sub-decision variables whose sum of contribution rates is greater than the contribution rate threshold, and use the obtained sub-decision variables to construct the first sub-decision variable sequence ;in represents the nth decision variable in the first sub-decision variable sequence; S202: Repeat step S201 to obtain the second sub-decision variable sequence corresponding to thermal efficiency The third sub-decision variable sequence corresponding to the environmental factors ;in Indicates the first m decision variables, Indicates the first s decision variables; S203, using the obtained first sub-decision variable sequence, the second sub-decision variable sequence and the third sub-decision variable sequence, and constructing a mixed sub-decision variable sequence through a random filtering algorithm , represents the first k decision variables, K represents the total number of decision variables in the mixed sub-decision variable sequence, ; The specific steps of constructing the multi-objective optimization model in step S3 include: S301, construct a multi-objective optimization function by setting the main objective factor and the first sub-decision variable sequence, the second sub-decision variable sequence and the third sub-decision variable sequence ,in represents the energy consumption objective function, represents the thermal efficiency objective function, represents the environmental factor objective function; S302: Constructing an initial population of the genetic algorithm using the obtained first sub-decision variable sequence, second sub-decision variable sequence, and third sub-decision variable sequence ; S303, according to the constructed initial population, calculate the corresponding fitness value of each decision variable through the fitness function and the constraint condition set, where the fitness function Specifically: ， Constraint Set U Specifically: in, Indicates the maximum heat conversion efficiency corresponding to the boiler in the secondary machinery group, represents the environmental factor threshold; Indicates the difference between the optimal output power fluctuation rate and the real-time minimum output power fluctuation rate, Indicates the constraint condition set u The penalty function of the constraint condition, the main objective factor or decision variable violates the u When a constraint , otherwise 0; Indicates the constraint condition set u The penalty coefficient corresponding to the constraint condition is is the weight coefficient of the corresponding factor and variable; Indicates the minimum energy consumption threshold per unit time. Indicates the maximum energy consumption threshold per unit time, Indicates the minimum thermal conversion efficiency threshold per unit time, Indicates the maximum thermal conversion efficiency threshold per unit time; Represents environmental factors Thresholds of environmental factors The difference, represents the constraint constant; Represents the angle penalty distance balance coefficient corresponding to a single sub-decision variable sequence, It represents the angle penalty distance penalty values ​​corresponding to the first sub-decision variable sequence, the second sub-decision variable sequence, and the third sub-decision variable sequence respectively.

2. The multi-objective energy consumption monitoring and optimization method for a secondary reheat unit according to claim 1, characterized in that: The specific steps of constructing the multi-objective optimization model in step S3 also include: S304. Set an angle penalty distance threshold, introduce the NSGA algorithm, use the non-dominated sorting in the NSGA algorithm to sort the corresponding sub-decision variable values ​​in the first sub-decision variable sequence, the second sub-decision variable sequence, and the third sub-decision variable sequence in the initial population, and use the angle penalty distance formula to output the angle penalty distance of the sorted decision variables; S305: The first sub-decision variable sequence, the second sub-decision variable sequence and the third sub-decision variable sequence are n decision variables, m decision variables and k The angle penalty distance corresponding to the decision variable is compared with the angle penalty distance threshold. n decision variables, m decision variables and k The angle penalty distances corresponding to the decision variables are less than the angle penalty distance threshold, then the n decision variables, m decision variables and k decision variables are retained to obtain the quadratic optimal population in the genetic algorithm; S306, setting the stopping condition and iteration cycle of the genetic algorithm, integrating the multi-objective optimization function constructed in S301-S305, the initial population, the fitness function, the constraint condition set, and the secondary optimal population obtained by the angle penalty distance into the original genetic algorithm to obtain a multi-objective optimization model; S307: According to the set stop condition or iteration cycle, the variable data set obtained by real-time monitoring is input into the multi-objective optimization model for iteration, and the corresponding single main objective optimal point set is calculated.

3. The multi-objective energy consumption monitoring and optimization method for a secondary reheat unit according to claim 2, characterized in that: The specific steps of constructing the multi-objective optimization model in step S3 also include: S308. When the input variables of the multi-objective optimization model are the main objective factors and the mixed sub-decision variable sequence, the corresponding multi-objective optimization function is , the initial population , the corresponding fitness function Specifically: ； in, represents the energy consumption objective function corresponding to the hybrid sub-decision variable sequence, represents the thermal efficiency objective function corresponding to the hybrid sub-decision variable sequence, Represents the environmental factor objective function corresponding to the mixed sub-decision variable sequence; represents the collinearity penalty term between all variables in the mixed sub-decision variable sequence, represents the collinearity penalty coefficient, , Where represents the number of collinear decision variables in the mixed sub-decision variable sequence; represents the angle penalty distance balance coefficient corresponding to the mixed sub-decision variable sequence, Represents the angle penalty distance penalty value corresponding to the mixed sub-decision variable sequence, Represents the environmental factors in the sequence of mixed sub-decision variables Thresholds of environmental factors The difference between S309: Replace the multi-objective optimization function, initial population and fitness function constructed in S308 with the corresponding variables and functions in S301-S303, and repeat the process of S301-S307 to calculate and obtain the optimal point set of the hybrid main objective.

4. The multi-objective energy consumption monitoring and optimization method for a secondary reheat unit according to claim 3, characterized in that: The servo control model in step S4 includes an enhanced adjustment layer, and the enhanced adjustment layer includes the current t Input status at all times , perform the action and current t Moment reward function The variable data set includes historical and real-time output power values ​​of the secondary reheat unit, primary steam temperature and secondary reheat steam temperature of the secondary reheat unit, main steam pressure and reheat steam pressure, energy consumption, heat conversion efficiency, environmental factors, and execution action values ​​corresponding to the historical optimal point set of the secondary reheat unit and the power value required by the power grid; The current t The input status at all times is: , the execution action is: , the current t The moment reward function is: ;in, Indicates the current t The initial steam temperature of the secondary reheat unit at time Indicates the current t Secondary reheat steam temperature of the secondary reheat unit at time, Indicates the current t Main steam pressure of secondary reheat unit at time Indicates the current t The reheat steam pressure of the secondary reheat unit at this moment, Indicates that when the condition is not met When , the feedback action of finding the optimal point set is triggered; Indicates the environmental factor corresponding to the mixed sub-decision variable sequence k Decision variables and environmental factor thresholds The difference.

5. A multi-objective energy consumption monitoring and optimization system for a secondary reheat unit, which is implemented based on the multi-objective energy consumption monitoring and optimization method for a secondary reheat unit according to any one of claims 1 to 4, characterized in that: include: Data processing module, data screening module, optimization model module, servo control model module; The data processing module is used to obtain the controllable and uncontrollable variables of the secondary reheat unit operation from the deployed integrated sensors, and pre-process these variables to obtain a variable data set in a unified format; The data screening module includes a variable screening unit and a sub-decision variable sequence construction unit; The variable screening unit is used to obtain valid sub-decision variables based on the set main target factor and the obtained variable data set using the configured principal component analysis algorithm; the sub-decision variable sequence construction unit is used to construct the sub-decision variable sequence and the mixed sub-decision variable sequence corresponding to the main target factor using the obtained valid sub-decision variables; The optimization model module includes an optimization model construction unit and an optimal point set unit; The optimization model construction unit is used to construct a multi-objective optimization model through angle penalty distance and genetic algorithm; the optimal point set unit is used to calculate the optimal point set corresponding to the main objective factor through the constructed multi-objective optimization model based on the sub-decision variable sequence and mixed sub-decision variable sequence corresponding to the acquired main objective factor.

6. The multi-objective energy consumption monitoring and optimization system for a secondary reheat unit according to claim 5, characterized in that: The monitoring and optimization system further includes: a servo control model module; The servo control model module includes a servo control model construction training unit and a monitoring feedback unit; The servo control model construction training unit is used to use reinforcement learning combined with PID control algorithm to construct a servo control model, and use the acquired historical optimal point set to train the constructed servo control model; the monitoring feedback unit is used to generate a control signal based on the optimal point set acquired in real time and the trained servo control model, control the operation of the secondary reheat unit, and monitor the output minimum output power fluctuation rate in real time, and perform feedback optimization on the servo control model and the multi-objective optimization model through the set optimal output power fluctuation rate and the monitored minimum output power fluctuation rate.

7. A computer-readable storage medium, characterized in that Computer instructions are stored thereon, and when the computer instructions are executed, the multi-objective energy consumption monitoring and optimization method for a secondary reheat unit according to any one of claims 1 to 4 is executed.

8. An electronic device, characterized in that: It comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the multi-objective energy consumption monitoring and optimization method for a secondary reheat unit as described in any one of claims 1 to 4.

Images