Turboset operation pressure optimization method based on improved particle swarm optimization

Abstract

The invention belongs to the technical field of turboset key operation parameter optimization, and particularly relates to a turboset operation pressure optimization method based on an improved particle swarm optimization algorithm, which comprises the following steps: firstly, establishing a heat rate mechanism model, and determining 11 input parameters such as active power load and main steam pressure and heat rate output parameters; the convergence performance of the particle swarm algorithm is improved through an inertia weight decline strategy, and the search boundary is dynamically adjusted in combination with elite reverse learning to enhance the optimization ability, so that the weight and threshold of an extreme learning machine are optimized, and a high-precision heat rate prediction model is constructed; and finally, in the feasible interval of the main steam pressure, with the minimum heat consumption rate as the target, optimizing and determining the optimal pressure by using an improved algorithm. The method effectively solves the problems that a traditional algorithm is insufficient in optimization precision and low in prediction model reliability, can quickly find the optimal main steam pressure under complex working conditions, remarkably reduces the heat consumption rate of a unit, improves the operation economy, and is high in applicability.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of key operation parameter optimization of steam turbine units, and particularly relates to a steam turbine unit operation pressure optimization method based on an improved particle swarm algorithm. BACKGROUND

[0002] As the core equipment of energy systems such as thermal power generation and heat supply, the operation efficiency of steam turbine units directly affects the energy utilization level and production cost. Optimizing key operation parameters of steam turbine units to reduce heat rate and improve power generation efficiency has become an important research direction in the energy industry. Among them, the main steam pressure, as the core parameter affecting the cycle efficiency of the steam turbine, its optimal value changes dynamically with the unit load, environmental conditions and other working conditions, so it is of great significance to realize the accurate optimization of the main steam pressure for the economic operation of the unit.

[0003] The existing steam turbine unit operation pressure optimization methods mainly include the following categories: 1. Traditional test method: Adjust the main steam pressure on site and test the heat rate to determine the optimal value under different working conditions. This method relies on experience and is time-consuming and laborious, and is difficult to adapt to complex and variable working conditions, and the test process may affect the stable operation of the unit.

[0004] 2. Mechanism model method: Establish a unit heat balance model based on the laws of thermodynamics, and derive the relationship between pressure and heat rate through theoretical calculation. Although this model has clear physical meaning, it ignores the nonlinear and coupled factors in the actual operation of the unit, resulting in limited optimization accuracy.

[0005] 3. Intelligent algorithm optimization method: With the development of artificial intelligence technology, intelligent optimization algorithms such as particle swarm algorithm and genetic algorithm are applied to parameter optimization. Among them, the particle swarm algorithm is widely used because of its simple principle and fast convergence speed, but the conventional particle swarm algorithm has problems such as easy to fall into local optimum and insufficient convergence accuracy in the later period, especially in the high-dimensional and strongly constrained unit parameter optimization scene, it is difficult to balance the optimization efficiency and global optimality.

[0006] In addition, the accuracy of the heat rate prediction model is the premise of pressure optimization. Existing prediction models mostly use neural networks, support vector machines and other methods, but the setting of model parameters (such as connection weights and threshold values) relies on experience or random initialization, resulting in unstable prediction accuracy, which further affects the reliability of the optimization results. At the same time, the search space boundary in the traditional optimization algorithm is mostly fixed value, which cannot be dynamically adjusted with the evolution of the population, which easily leads to the reverse solution out of the effective search range, reducing the optimization ability of the algorithm.

[0007] Therefore, in view of the problems of insufficient optimization accuracy, poor algorithm adaptability, and low reliability of the prediction model in the prior art, there is an urgent need for a steam turbine unit operation pressure optimization method capable of dynamically adjusting the search space, improving the optimization accuracy and efficiency, and combining high-precision heat rate prediction to realize economic operation of the unit under full working conditions. SUMMARY

[0008] In view of the technical problems existing in the above-mentioned steam turbine unit operation pressure optimization method, the present application provides a steam turbine unit operation pressure optimization method based on an improved particle swarm algorithm.

[0009] To solve the above technical problems, the technical scheme adopted by the present application is as follows: A steam turbine unit operation pressure optimization method based on an improved particle swarm algorithm, comprising the following steps: S1, establishing a unit heat rate mechanism model: analyzing the logical relationship between the operating parameters and the heat rate, determining the input parameters and output parameters of the prediction model, wherein the input parameters include active power load, main steam pressure, main steam temperature, reheater steam inlet pressure, reheater steam inlet temperature, reheater steam outlet pressure, reheater steam outlet temperature, turbine back pressure, feed water flow, reheated steam flow, heat supply extraction steam coefficient, and the output parameter is the heat rate; S2, improving the particle swarm algorithm: proposing an elite reverse learning particle swarm algorithm based on decreasing inertia weight, and the decreasing inertia weight is realized by a decreasing formula: wherein is the inertia weight, and are the maximum and minimum values in the change range, is the maximum number of steps, and t is the current number of steps; the elite reverse learning is realized by generating reverse solutions of the local extreme value and the global extreme value; S3, establishing a heat rate prediction model: using the improved particle swarm algorithm in S2 to optimize the connection weight and threshold of the extreme learning machine, the extreme learning machine is a three-layer network structure, the number of input layer nodes is 11, the number of hidden layer nodes is 50, and the number of output layer nodes is 1, the input weight and the threshold boundary of the hidden layer are limited to [-1, 1]; S4, main steam pressure optimization: determining the feasible interval of the main steam pressure, and combining the improved particle swarm algorithm in S2 to optimize the main steam initial pressure under the given working condition, taking the minimum heat rate as the target to determine the optimal main steam pressure value.

[0010] ​In the S1, the absolute internal efficiency and heat rate of the steam turbine are obtained by energy conservation and mass conservation calculation of the extraction steam and condensate flowing through the heater, and then the heat rate mechanism model is established.

[0011] The inertia weight in the S2 The maximum value And the minimum value Is selected in the range of 0.8-1.2.

[0012] The method for realizing the elite reverse learning in the S2 is that the reverse solution of the own extreme value and the global extreme value is generated. Assume is a point in the N-dimensional space, and take it as a feasible solution , then the corresponding reverse point is defined by the following formula: Wherein: represents the feasible solution The component of the corresponding reverse point in the jth-dimensional space, , respectively represent the fixed upper limit and the fixed lower limit of the jth-dimensional search space, represents the component of the feasible solution in the jth-dimensional space. Let be a general particle in the population, then the corresponding own extreme value is the elite particle , and the elite reverse solution is defined by the following formula: Wherein: represents the generalization coefficient, , and respectively represent the upper limit and the lower limit which are always changing in the jth-dimensional search space. The fixed boundary in the search space is transformed into a boundary which is always changing, so that the generated reverse solution always exists in the search space which is continuously narrowing.

[0013] The population size of the improved particle swarm algorithm in the S2 is 50, the maximum iteration number is 100, and the local learning factor c1 and the global learning factor c2 are both 1.5.

[0014] The operation data of the power plant in the past 7 days and 168 hours are collected as training samples when the heat consumption rate prediction model is established in S3, and the operation data includes active power load, main steam pressure, main steam temperature, reheater steam inlet pressure, reheater steam inlet temperature, reheater steam outlet pressure, reheater steam outlet temperature, turbine back pressure, feed water flow, reheated steam flow, heat supply extraction steam coefficient and heat consumption rate.

[0015] The determination method of the feasible interval of the main steam pressure in S4 is that, for a given load The minimum pressure is calculated by the following formula: Wherein, is the main steam pressure under the rated load , and is the minimum main steam pressure under the load .

[0016] The population number of the improved particle swarm algorithm in S4 is 20, the iteration number is 50, the local learning factor c1 and the global learning factor c2 are both 1.5, the initial value of the weight is 0.9, and the weight range is [0.2, 0.9].

[0017] The mathematical model of the main steam pressure optimization in S4 is: Wherein: represents the steam turbine heat consumption rate, is the main steam pressure, is the unit load, is other influencing factors.

[0018] The prediction performance of the heat consumption rate is evaluated by the contribution coefficient, the average relative percentage error and the root mean square error.

[0019] Compared with the prior art, the present application has the beneficial effects that: The present application establishes a precise unit heat consumption rate mechanism model, combines the improved particle swarm algorithm based on inertia weight reduction and elite reverse learning, solves the problems of easy falling into local optimum and insufficient convergence precision of the traditional particle swarm algorithm, ensures that the reverse solution is always in the effective optimization range through dynamic adjustment of the search space boundary, improves the global search ability and optimization efficiency of the algorithm, optimizes the parameters of the extreme learning machine by using the improved particle swarm algorithm, significantly improves the accuracy of the heat consumption rate prediction model, provides a reliable foundation for subsequent main steam pressure optimization, and finally can quickly determine the optimal main steam pressure value under different working conditions, effectively reduces the heat consumption rate of the steam turbine unit, improves the operation economy, and has strong adaptability and can meet the complex and variable unit operation requirements. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only exemplary, and for those skilled in the art, other drawings can be derived from the provided drawings without creative labor.

[0021] The structures, proportions, sizes, etc. shown in the specification are only used to cooperate with the content disclosed in the specification, to be understood and read by those skilled in the art, and are not used to limit the conditions that the present application can be implemented, so they do not have technical significance. Any modification of structure, change of proportion relationship or adjustment of size, without affecting the effects and purposes that the present application can produce, should still fall within the scope of the technical content disclosed by the present application.

[0022] Figure 1 The principle heat system calculation diagram of the steam turbine of the present application; Figure 2 The flow chart for establishing the OBL-IPSO-ELM model of the present application; Figure 3 The feasible pressure interval diagram of the present application; Figure 4 The optimal initial pressure calculation flow chart of the present application. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical scheme and advantages of the embodiments of the present application more clear, the technical scheme in the embodiments of the present application will be clearly and completely described below. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. These descriptions are only for further illustrating the features and advantages of the present application, but not for limiting the claims of the present application; based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of the present application.

[0024] The specific embodiments of the present application will be further described in detail below in combination with the drawings and embodiments. The following embodiments are used to illustrate the present application, but not to limit the scope of the present application.

[0025] The present embodiment provides a steam turbine unit operation pressure optimization method based on an improved particle swarm algorithm, comprising the following steps: Step one, as Figure 1The shown is a principle thermodynamic system calculation diagram of steam turbine, the main steam first flows into the high pressure cylinder of the steam turbine, two stages of extraction are arranged in the high pressure cylinder, the main steam flows out from the high pressure cylinder and is heated again in the high temperature reheater, the reheated steam continues to flow into the medium pressure cylinder of the steam turbine to complete the energy conversion process, the third stage, the fourth stage and the fifth stage of extraction are arranged in the medium pressure cylinder, the remaining main steam enters the low pressure cylinder after the extraction is completed, the sixth stage and the seventh stage of extraction are arranged in the low pressure cylinder of the steam turbine, the steam after the energy conversion ends is discharged from the low pressure cylinder and enters the condenser, and the condensate enters the economizer after being heated by the extraction steam of each stage in sequence. The energy conservation and mass conservation of the extraction steam and the condensate flowing through the heater are carried out, and the absolute internal efficiency and the heat rate of the steam turbine are calculated in combination with the principle thermodynamic calculation diagram.

[0026] Step two, collect the active power load, main steam pressure, main steam temperature, reheater steam inlet pressure, reheater steam inlet temperature, reheater steam outlet pressure, reheater steam outlet temperature, turbine back pressure, feed water flow, reheated steam flow, heating extraction steam coefficient and heat rate of the power plant in the past 7 days for 168 hours.

[0027] Step three, optimize the intelligent algorithm and establish a prediction model.

[0028] 1. Particle swarm algorithm with decreasing inertia weight In order to find the appropriate inertia weight value , in order to enhance the performance of the algorithm, this paper puts forward the decreasing method, that is, at the beginning of the algorithm, we can let be a little larger, so that the algorithm can search more comprehensively in the whole range, and select a small value of in the later stage of the algorithm, so as to maintain effective global search while improving the convergence performance of the algorithm. The decreasing formula (1) is: (1) and means the two extreme values in the change range, usually selected within 0.8~1.2, and t represent the maximum forward step number of the particle and the current step number.

[0029] 2. Elite reverse learning particle swarm algorithm based on decreasing inertia weight The central idea of the reverse learning strategy is: when a feasible solution appears, its reverse solution is also calculated, evaluated, and then the best value in the side is selected as the next generation group. Assuming is a point in N-dimensional space, which is taken as a feasible solution , then its corresponding inverse point can be defined as follows: (2) In the velocity update formula of particles, we can analyze that the trajectory of particles is mainly determined by the velocity of the last step, the local extreme value and the global extreme value . When the velocity of a particle is given, the local extreme value and the global extreme value become the main factors in the search space, which directly determine the new flight position of the particle. The inverse learning strategy is used to generate the inverse solution of and . It has been proved in some documents that the optimization effect of generating the inverse solution of the local extreme value and the global extreme value is usually stronger than that of the particle itself. In order to distinguish them from ordinary particles in the population, they are called elite particles. Let be an ordinary particle in the population, then its corresponding local extreme value is the elite particle , and the elite inverse solution is which can be defined as follows: (3) In the formula, the represents the generalization coefficient, , and are the boundary values in the j-dimensional search space, which are always changing. They can be given by the following formula (4): (4) The fixed boundary in the search space is transformed into a changing boundary, so that the generated inverse solution can exist in the search space with a continuously shrinking range.

[0030] Step four, use the improved particle swarm algorithm to optimize the extreme learning machine in this model (OBL-IPSO-ELM) The elite inverse learning particle swarm algorithm based on the inertia weight reduction has a built-in population size of 50, a maximum iteration number of 100, and local and global learning factors of , The initial weights are 0.9, with a given limit of [0.9, 0.2]. The Extreme Learning Machine (ELM) is configured with a three-layer network structure: 11 nodes in the input layer, 50 nodes in the hidden layer, and 1 node in the output layer. The input weights and the thresholds for the hidden layers are bounded to [-1, 1]. The flowchart for establishing the OBL-IPSO-ELM model is shown below. Figure 2 As shown.

[0031] Step 5: Predictive Performance Verification The OBL-IPSO algorithm is used to optimize the input weights and thresholds of the Extreme Learning Machine (ELM), and then a model for predicting the heat rate of a steam turbine unit is constructed. To verify the impact of the algorithm improvement on the accuracy of heat rate prediction, the improvement effect is expressed in the evaluation data and its simulation curves on the training and test sets. This invention uses the following three indicators to reflect the performance impact of the algorithm improvement on the Extreme Learning Machine.

[0032] 1. Coefficient of determination: (5) 2. Average relative percentage error: (6) 3. Root mean square error (7) In the formula for calculating the index, y refers to the actual value of the heat rate. The average value of the actual heat rate. This refers to the predicted heat rate, and N refers to the number of sample data.

[0033] Step 6: Calculation of Optimal Initial Pressure for the Unit To fulfill the grid's peak-shaving directives, participating generating units need to adjust their load while maintaining grid connection. Load changes can be implemented by the boiler or turbine control system. The adjustable operating modes on the boiler and turbine sides differ; considering the advantages and disadvantages, a constant-slip-slip-constant-pressure composite operating mode is the most suitable.

[0034] 1. Description of the optimal initial pressure problem of the unit Constant-sliding-constant-pressure combined operation requires determining under which load conditions constant-pressure operation and under which load conditions sliding-pressure operation should be used, and clearly stating when the two operation modes should be switched. Many other factors influence the heat rate value, such as the performance of turbine components, reheat steam parameters, back pressure, and ambient temperature. The mathematical description of the heat rate can be expressed as equation (8): (8) In the expression, q represents the turbine heat rate. , , This refers to the main steam flow rate, pressure, and temperature. This refers to the unit load. Z refers to the reheat steam temperature, while Z represents other influencing factors.

[0035] During variable load operation, the temperature of the main steam generated by the boiler and the reheater outlet steam temperature remain at their rated values, while the main steam flow rate is adjusted according to the main steam pressure. Therefore, the mathematical expression for the unit's heat rate can be simply described as follows: (9) 2. Optimal initial pressure range for the unit is defined. In actual production, the main steam under any load will have a pressure range that meets both safety and economic performance requirements. Each pressure value also corresponds to a heat rate value. For a given load, we select the lowest heat rate value; the main steam pressure corresponding to this value is the optimal value we need to find. For a given load, the feasible pressure range can be determined by… Figure 3 The analysis yielded the following results.

[0036] Analysis of the minimum pressure POM using the principle of triangle similarity yields the following: (10) 3) Optimal Initial Pressure Mathematical Model of the Unit The unit's thermal system is a complex system characterized by nonlinearity and strong coupling. A mathematical model is established to quantitatively analyze the relationship between influencing variables and the target optimization quantity. Based on the above mathematical description and the initial pressure range limitation, the optimal operating initial pressure mathematical model for the unit under known specific load conditions can be calculated using the following formula: (11) Step 7: Determining the optimal pressure of the main steam. Historical data was collected from the SIS system, with 167 sets of data collected every hour over a 7-day period. This data included various parameters required for model building, and the collected operating conditions covered 50%-100% of the unit's rated load. Based on the turbine heat rate prediction model established using OBL-IPSO-ELM, the improved particle swarm optimization algorithm's excellent global and local search capabilities were then utilized to optimize the unit's initial operating pressure.

[0037] The detailed parameter settings for the improved particle swarm optimization algorithm are as follows: population size is 20, number of iterations is 50, and local and global learning factors are respectively... , The initial weight is 0.9, with a given constraint of [0.9, 0.2]. Iteration stops after 50 iterations. The optimal initial pressure calculation process is as follows: Figure 4 As shown.

[0038] The above description only illustrates the preferred embodiments of the present invention. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention, and all such changes should be included within the protection scope of the present invention.

Claims

1. A method for optimizing the operating pressure of a steam turbine unit based on an improved particle swarm optimization algorithm, characterized in that, Includes the following steps: S1. Establish a heat rate mechanism model for the unit: Analyze the logical relationship between operating parameters and heat rate, and determine the input and output parameters of the prediction model. The input parameters include active power load, main steam pressure, main steam temperature, reheater steam inlet pressure, reheater steam inlet temperature, reheater steam outlet pressure, reheater steam outlet temperature, turbine back pressure, feedwater flow rate, reheat steam flow rate, and heating extraction coefficient. The output parameter is the heat rate. S2. Improved Particle Swarm Optimization Algorithm: An elite back-learning particle swarm optimization algorithm based on decreasing inertia weights is proposed. The decreasing inertia weights are achieved through a decreasing formula: in For inertial weights, and They are respectively The maximum and minimum values ​​within the range of variation. Let t be the maximum number of steps a particle can take, and t be the number of steps it has already taken. The elite backward learning is achieved by generating backward solutions to its own extrema and the global extrema. S3. Establish a heat consumption rate prediction model: Optimize the connection weights and thresholds of the extreme learning machine using the improved particle swarm optimization algorithm described in S2. The extreme learning machine is a three-layer network structure with 11 input layer nodes, 50 hidden layer nodes, and 1 output layer node. The boundary between the input weights and the threshold of the hidden layer is limited to [-1, 1]. S4. Main Steam Pressure Optimization: Determine the feasible range of main steam pressure, and combine the improved particle swarm optimization algorithm described in S2 to optimize the initial pressure of main steam under given operating conditions. With the goal of minimizing heat consumption rate, determine the optimal main steam pressure value.

2. The method for optimizing the operating pressure of a steam turbine unit based on an improved particle swarm optimization algorithm according to claim 1, characterized in that: In step S1, by performing energy and mass conservation calculations on the extracted steam and condensate flowing through the heater, the absolute internal efficiency and heat rate of the steam turbine are obtained, and then the heat rate mechanism model is established.

3. The method for optimizing the operating pressure of a steam turbine unit based on an improved particle swarm optimization algorithm according to claim 1, characterized in that: The inertial weight in S2 maximum value and minimum value Choose within the range of 0.8 to 1.

2.

4. The method for optimizing the operating pressure of a steam turbine unit based on an improved particle swarm optimization algorithm according to claim 1, characterized in that, The method used in S2 to achieve elite back-learning by generating back-inverse solutions for its own extrema and the global extrema is as follows: Assumption It is a point in N-dimensional space, and we treat it as a feasible solution. Then its corresponding reverse point Defined by the following formula: in: Indicates a feasible solution The component of the corresponding reverse point in the j-th dimension. , Let these represent the fixed upper bound and fixed lower bound of the j-th dimension search space, respectively. Indicates a feasible solution The components in the j-th dimension; set up If a particle is an ordinary particle in the population, then its corresponding extreme value is an elite particle. The elite reverse solution is Defined by the following formula: in: Represents the generalization coefficient. , and These represent the upper and lower bounds of the j-th dimension search space, which are constantly changing. Transform the constant boundaries in the search space into constantly changing boundaries, so that the generated inverse solutions always exist in the increasingly narrow search space.

5. The method for optimizing the operating pressure of a steam turbine unit based on an improved particle swarm optimization algorithm according to claim 1, characterized in that, The improved particle swarm optimization algorithm in S2 has a population size of 50, a maximum number of iterations of 100, and both the local learning factor c1 and the global learning factor c2 are 1.

5.

6. The method for optimizing the operating pressure of a steam turbine unit based on an improved particle swarm optimization algorithm according to claim 1, characterized in that: When establishing the heat rate prediction model in S3, the operating data of the power plant over the past 7 days (168 hours) is collected as training samples. The operating data includes active power load, main steam pressure, main steam temperature, reheater steam inlet pressure, reheater steam inlet temperature, reheater steam outlet pressure, reheater steam outlet temperature, turbine back pressure, feedwater flow rate, reheat steam flow rate, heating extraction coefficient, and heat rate.

7. The method for optimizing the operating pressure of a steam turbine unit based on an improved particle swarm optimization algorithm according to claim 1, characterized in that, The method for determining the feasible range of the main steam pressure in S4 is as follows: for a given load Minimum pressure Calculated by the following formula: in, Rated load The main steam pressure below, For load The minimum main steam pressure.

8. The method for optimizing the operating pressure of a steam turbine unit based on an improved particle swarm optimization algorithm according to claim 1, characterized in that, The improved particle swarm optimization algorithm in S4 has a population size of 20, an iteration count of 50, a local learning factor c1 and a global learning factor c2 of 1.5, an initial weight value of 0.9, and a weight range of [0.2, 0.9].

9. The method for optimizing the operating pressure of a steam turbine unit based on an improved particle swarm optimization algorithm according to claim 1, characterized in that, The mathematical model for optimizing the main steam pressure in S4 is as follows: in: This indicates the steam turbine heat consumption rate. Main steam pressure, For unit load, Other influencing factors.

10. The method for optimizing the operating pressure of a steam turbine unit based on an improved particle swarm optimization algorithm according to claim 1, characterized in that, The predictive performance of the heat rate is evaluated using the coefficient of determination, mean relative percentage error, and root mean square error.

Images