A bionic decision-driven intelligent power plant full life cycle optimization method and system

Abstract

The application provides a kind of bionic decision driven intelligent power plant full life cycle optimization method and system, it is related to intelligent power plant life cycle management technical field, including by immune algorithm fusion boiler wall temperature characteristic vector and flue gas-stress coupling characteristic vector, generate full life cycle characteristic vector set;Filter high sensitive area and generate damage hot spot coordinate set;In combination with damage hot spot coordinate set and SO2 Gradient data, crack propagation differential equation is solved by Paris formula of corrosion correction, and output residual strength field;With the failure time and cost of residual strength field as constraint, the fitness function is optimized by using bird swarm algorithm, and the pareto optimal maintenance sequence is generated to obtain the maintenance implementation result;According to the actual failure deviation in maintenance implementation result, KL divergence is calculated, and full life cycle closed loop optimization is formed.The application has the beneficial effect of significantly improving the intelligent level and full life cycle economy of power plant equipment health management.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent power plant life cycle management, in particular to a bionic decision-driven intelligent power plant full life cycle optimization method and system. BACKGROUND

[0002] At present, intelligent power plant construction has become the core direction of global power industry transformation and upgrading. Traditional power plant equipment health management mainly relies on periodic maintenance and manual experience judgment, which has significant limitations. The existing technology usually uses a SCADA system (such as a DCS alarm module) or a simple statistical model (such as a mean- range control chart) based on threshold alarm for equipment state monitoring. These methods can only identify explicit failures and cannot predict early damage evolution. At the same time, the traditional maintenance strategy adopts fixed cycle maintenance, which leads to excessive maintenance or frequent missed detection. To solve these problems, some power plants have introduced fault diagnosis systems based on vibration analysis (such as ISO 10816 standard) or offline finite element analysis. However, these technologies have defects such as single data dimension, lack of real-time performance, and failure to consider corrosion-fatigue coupling effects, making it difficult to meet the needs of full life cycle management of high-temperature and high-pressure pipelines (such as main steam pipelines).

[0003] There are still three major bottlenecks in the existing technology: multi-source heterogeneous data (such as wall temperature time series, flue gas composition, stress and strain) lack effective dynamic coupling mechanism; most systems rely on static model parameters and cannot correct predictions based on real-time feedback; and existing bionic algorithms (such as ant colony and bird swarm) focus on local optimization and do not cover the full life cycle. SUMMARY

[0004] The purpose of the present application is to provide a bionic decision-driven intelligent power plant full life cycle optimization method and system to improve the above problems. In order to achieve the above purpose, the technical solutions adopted by the present application are as follows:

[0005] In the first aspect, the present application provides a bionic decision-driven intelligent power plant full life cycle optimization method, comprising:

[0006] Step one, fuse the boiler wall temperature feature vector and the flue gas-stress coupling feature vector through the immune algorithm to generate a full life cycle feature vector set as input data for damage hot spot identification;

[0007] Step two, calculate the maximum Lyapunov index field based on the full life cycle feature vector set and construct a damage topology graph to screen high sensitivity areas to generate a damage hot spot coordinate set, and then drive the creep evolution modeling;

[0008] Step three, combining the damage hotspot coordinate set and the SO2 gradient data, solving the crack propagation differential equation by the Paris formula corrected by corrosion, and outputting the residual strength field, wherein the residual strength field includes position, crack length and failure time;

[0009] Step four, using the failure time and cost of the residual strength field as constraints, using the bird swarm algorithm to optimize the fitness function, generating a Pareto optimal maintenance sequence, and taking the Pareto optimal maintenance sequence as an input item of the entropy weight optimization engine to obtain a maintenance implementation result;

[0010] Step five, calculating the KL divergence according to the actual failure deviation in the maintenance implementation result, correcting the crack propagation model parameters by gradient descent, and feeding back to step three to form a whole life cycle closed loop optimization.

[0011] Preferably, in the step one, the boiler combustion chamber metal wall temperature time series data is reduced to 20-dimensional feature vectors by space-time tensor decomposition, and the energy entropy of the steam turbine vibration frequency spectrum is extracted to generate a device-level feature vector;

[0012] The flue gas composition spectrum and the pipeline stress and strain field are collected, and a 15-dimensional environment-mechanical feature vector is generated by thermal corrosion coupling analysis;

[0013] The 20-dimensional feature vector and the 15-dimensional environment-mechanical feature vector are fused by using the immune algorithm, and the data similarity is calculated by using the affinity matrix and iteratively convergent, and a whole life cycle feature vector set is output as an input of damage hotspot identification.

[0014] Preferably, in the step two, the step two includes:

[0015] Based on the wall temperature feature subset in the whole life cycle feature vector set, the Wolf algorithm is used to calculate the maximum Lyapunov index field, to quantify the wall temperature chaotic characteristics and the corrosion index, and to generate a damage sensitivity scalar field;

[0016] The damage sensitivity is associated with the spatial coordinates, and a damage topology graph including node coordinates, sensitivity values and connection relationships is constructed;

[0017] The nodes in the damage topology graph greater than 0.8 are screened out and defined as high sensitivity areas, and the damage hotspot coordinate set in the high sensitivity areas is taken as an input of creep evolution modeling.

[0018] Preferably, in the step three, the step three includes:

[0019] Combining the damage hotspot coordinate set and the SO2 concentration gradient, the crack propagation rate of each hotspot position is calculated by the Paris formula corrected by corrosion;

[0020] The fourth-order Runge-Kutta method is used to time-integrate the crack propagation rate to solve the crack length evolution path, and a function relationship of the crack depth changing with time is obtained.

[0021] Based on the function relationship, the residual strength is calculated.

[0022] It is judged whether the residual strength is less than or equal to 0. If the residual strength is less than or equal to 0, it is determined to be invalid, and the residual strength field data including the three-dimensional position of the crack, the final crack length and the failure time are output. If the residual strength is greater than 0, the pipeline is normal at the current time, and output is performed.

[0023] In a second aspect, the present application also provides a bionic decision-driven intelligent power plant full life cycle optimization system, comprising:

[0024] A first generation module is configured to generate a full life cycle feature vector set by fusing a boiler wall temperature feature vector and a flue gas-stress coupled feature vector through an immune algorithm, as input data for damage hot spot identification.

[0025] A screening module is configured to calculate a maximum Lyapunov index field based on the full life cycle feature vector set and construct a damage topology graph, screen a high sensitive area to generate a damage hot spot coordinate set, and further drive creep evolution modeling.

[0026] A solving module is configured to combine the damage hot spot coordinate set and SO2 gradient data, solve a crack propagation differential equation through a Paris formula corrected by corrosion, and output a residual strength field, wherein the residual strength field includes position, crack length and failure time.

[0027] A second generation module is configured to use the failure time and cost of the residual strength field as constraints, optimize an adaptability function using a bird swarm algorithm, generate a Pareto optimal maintenance sequence, use the Pareto optimal maintenance sequence as an input item of an entropy weight optimization engine, and obtain a maintenance implementation result.

[0028] A calculation module is configured to calculate KL divergence according to actual failure deviation in the maintenance implementation result, correct crack propagation model parameters through gradient descent, and feed back to step three to form a full life cycle closed loop optimization.

[0029] In a third aspect, the present application also provides a bionic decision-driven intelligent power plant full life cycle optimization device, comprising:

[0030] A memory is configured to store a computer program.

[0031] A processor is configured to execute the computer program to realize the steps of the bionic decision-driven intelligent power plant full life cycle optimization method.

[0032] In a fourth aspect, the present application also provides a readable storage medium, wherein the readable storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the above-mentioned method for whole life cycle optimization of a smart power plant based on bionic decision driving.

[0033] The present application has the following advantages:

[0034] The present application fuses spatio-temporal tensor decomposition, immune algorithm, corrosion correction Paris formula and bird swarm algorithm, and constructs a complete closed loop from multi-source data fusion, damage sensitivity field construction, crack propagation modeling and maintenance decision optimization. The present application first realizes the following: spatio-temporal coupling modeling of the chaotic characteristics of the boiler wall temperature and the SO2 corrosion rate; grid-level failure prediction based on the residual strength field; entropy weight closed loop correction of the maintenance decision and the model parameters. Compared with the prior art, the present application reduces the failure prediction error of the high-temperature pipeline to within 8%, reduces the maintenance cost by 18%, and has self-adaptive correction capability for coal quality fluctuations (±20%) and load mutations (±30%), thereby significantly improving the intelligent level and the whole life cycle economy of the power plant equipment health management.

[0035] Other features and advantages of the present application will be described in the following description, and some will become apparent from the description, or will be learned through implementation of the embodiments of the present application. The purposes and other advantages of the present application can be achieved and obtained through the structures specifically pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be regarded as limiting the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0037] Figure 1 The figure is a flowchart of the bionic decision-driven whole life cycle optimization method of a smart power plant described in the embodiments of the present application.

[0038] Figure 2 The figure is a structure diagram of the bionic decision-driven whole life cycle optimization system of a smart power plant described in the embodiments of the present application.

[0039] Figure 3 The figure is a structure diagram of the bionic decision-driven whole life cycle optimization device of a smart power plant described in the embodiments of the present application.

[0040] In the figure: 701, first generation module; 702, screening module; 703, solving module; 704, second generation module; 705, calculation module; 800, bionic decision-driven intelligent power plant full life cycle optimization equipment; 801, processor; 802, memory; 803, multimedia assembly; 804, I / O interface; 805, communication assembly. DETAILED DESCRIPTION

[0041] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art without creative labor based on the embodiments in the present application belong to the scope of protection of the present application.

[0042] It should be noted that: similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. Meanwhile, in the description of the present application, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0043] Embodiment 1

[0044] The present embodiment provides a bionic decision-driven intelligent power plant full life cycle optimization method.

[0045] Referring to Figure 1 , the present method includes steps S100, S200, S300, S400 and S500, as shown in the figure.

[0046] S100, fuse the boiler wall temperature feature vector and the flue gas-stress coupling feature vector through the immune algorithm to generate a full life cycle feature vector set as input data for damage hot spot identification.

[0047] It should be noted that step S100 includes S101, S102 and S103.

[0048] S101, reduce the boiler combustion chamber metal wall temperature time series data to 20-dimensional feature vectors through space-time tensor decomposition, and extract the energy entropy of the steam turbine vibration frequency spectrum to generate equipment-level feature vectors;

[0049] S102, collect flue gas composition spectrum and pipeline stress and strain field, generate 15-dimensional environment-mechanical characteristic vector through thermal corrosion coupling analysis;

[0050] S103, fuse 20-dimensional characteristic vector and 15-dimensional environment-mechanical characteristic vector by using immune algorithm, calculate data similarity through affinity matrix and iterate convergence, and output full life cycle characteristic vector set as input of damage hot spot identification.

[0051] It can be understood that in this step, the boiler combustion chamber temperature field matrix Tm×n(t) (m=50 rows, n=60 columns of grid, sampling period t=1s), steam turbine vibration frequency spectrum Fk(t) (k=1-10 order harmonic) are input, and Tm×n(t) is reduced to characteristic vector by space-time tensor decomposition (Tucker decomposition), and the energy entropy of Fk(t) is extracted by wavelet packet decomposition, and then the equipment level characteristic vector, that is, the wall temperature characteristic vector, is output.

[0052] It should be noted that the real-time wind speed, humidity and coal element analysis vector are collected, and an environment-fuel coupling matrix is constructed to obtain an environment-fuel characteristic vector; the equipment level characteristic vector and the environment-fuel coupling matrix are input into the improved immune algorithm, and the data similarity is calculated through the affinity matrix A ij = exp (-||D i -D j || 2 / σ 2 ), and the unified characteristic vector set is iteratively generated as the only input of damage hot spot identification.

[0053] This step fuses the boiler wall temperature space-time tensor decomposition characteristics, flue gas-stress coupling characteristics and immune algorithm through bionic decision-driven multi-source data fusion technology, constructs a 35-dimensional unified characteristic vector set, and compared with the traditional single-source monitoring method, the damage positioning accuracy is greatly improved, and the early damage identification time is advanced by 6-8 months.

[0054] S200, calculate the maximum Lyapunov index field based on the full life cycle characteristic vector set and construct a damage topology graph, select a high sensitivity area to generate a damage hotspot coordinate set, and then drive the creep evolution modeling.

[0055] It can be understood that in this step S200, S201, S202 and S203 are included.

[0056] S201, based on the wall temperature characteristic subset in the full life cycle characteristic vector set, calculate the maximum Lyapunov index field by using Wolf algorithm, quantify the wall temperature chaotic characteristics and corrosion index, and generate a damage sensitivity scalar field, wherein the formula for calculating the maximum Lyapunov index field by using Wolf algorithm is as follows:

[0057]

[0058] where T w (i) is the wall temperature vector at the i-th sampling time, is the reference orbit, t obs is the observation time length, and N is the number of sampling points;

[0059] S202, associate the damage sensitivity with the spatial coordinates, and construct a damage topology graph containing node coordinates, sensitivity values, and connection relationships;

[0060] S203, screen out nodes in the damage topology graph that are greater than 0.8, and define them as high-sensitivity areas, and take the damage hotspot coordinate set in the high-sensitivity area as the input for creep evolution modeling.

[0061] It should be noted that in the scenario of high-temperature steam pipelines in coal-fired power plants, the wall temperature field of the boiler combustion chamber (usually collected by a 50x60 thermocouple grid array with a sampling frequency of 1 Hz) presents significant nonlinear dynamic characteristics. Traditional analysis methods based on statistical mean or extreme value cannot capture the local mutation and chaotic evolution rules of the temperature field. When the maximum Lyapunov exponent is greater than 0, the system presents chaotic characteristics, and a small perturbation will lead to long-term unpredictability of the temperature field; the greater the maximum Lyapunov exponent, the higher the sensitivity of the system to initial conditions, and the greater the risk of thermal fatigue damage of the pipeline metal. Thus, the chaotic characteristics of the wall temperature are quantified, and the flue gas corrosion index is fused, and the calculation formula is as follows:

[0062]

[0063] where SO2 is the sulfur dioxide concentration, is the wall temperature gradient;

[0064] Define the sensitivity: S D = α·λ max + β·I corr , α is 0.6, β is 0.4, and the weights; then generate a damage sensitivity scalar field, and subsequently associate the scalar field with the pipeline three-dimensional coordinate system to construct a damage topology graph G D = {node coordinates, S D , connection relationships}; finally, determine the high-sensitivity area by sensitivity threshold screening (SD>0.8), and output the damage hotspot coordinate set (containing spatial coordinates, sensitivity values, and corrosion indexes) as the input data for creep evolution modeling in step three.

[0065] In this step, compared with the traditional single threshold method, the chaos-corrosion coupled model reduces the damage positioning error from ±0.5m to ±0.1m; the micro-damage of wall thickness thinning <0.1mm can be identified to provide early warning than artificial inspection; the hotspot coordinate set directly guides the detection path planning of the ultrasonic flaw detector, and reduces the invalid detection time.

[0066] In this embodiment, by modeling the damage sensitivity field coupled with chaos and corrosion, the Wolf algorithm is used to quantify the chaotic characteristics of wall temperature and fuse the SO2 corrosion index to generate a high-resolution damage topology map, so as to reduce the crack propagation rate prediction error.

[0067] S300, combine the damage hotspot coordinate set and the SO2 gradient data, solve the crack propagation differential equation by the Paris formula corrected by corrosion, and output the residual strength field, wherein the residual strength field includes position, crack length and failure time.

[0068] It can be understood that in this step S300, S301, S302, S303 and S304 are included.

[0069] S301, combine the damage hotspot coordinate set and the SO2 concentration gradient, and calculate the crack propagation rate at each hotspot position by the Paris formula corrected by corrosion, wherein the calculation formula of the Paris formula is as follows:

[0070]

[0071] In the formula, C and m are material constants, ΔK eff is the effective stress intensity factor range, γ is the corrosion acceleration coefficient, is the SO2 concentration gradient, Q f is the fatigue activation energy, R is the gas constant, and T is the pipe wall temperature.

[0072] S302, the crack propagation rate is time-integrated by using the fourth-order Runge-Kutta method, the crack length evolution path is solved, and the function relationship between the crack depth and time is obtained.

[0073] S303, based on the function relationship, the residual strength is calculated, and the calculation formula is as follows:

[0074]

[0075] In the formula, is the crack-induced bearing area reduction coefficient, σ u is the tensile strength of the material, σ hoop is the pipe hoop stress, R f (t) is the residual strength.

[0076] S304, judging whether the residual strength is less than or equal to 0, if the residual strength is less than or equal to 0, determining failure, and outputting residual strength field data including three-dimensional position of the crack, final crack length and failure time; if the residual strength is greater than 0, the pipeline is normal at the current time, and output is performed.

[0077] It should be noted that the function relationship of the crack depth changing with time is as follows:

[0078]

[0079] In the formula, a0 is the initial crack depth, and t is the running time. The crack starts from the initial depth a0 and expands with time t under the coupling effect of corrosion-fatigue. This function is the basis input for residual strength calculation, because the deeper the crack, the smaller the material carrying area; the stronger the stress concentration effect at the crack tip. Therefore, the residual strength directly reflects whether the current state of the material can withstand the working load. When the crack expansion Rf(t)≤0, the material carrying capacity is lower than the working stress, that is, failure occurs.

[0080] In summary, the crack depth function is the "time record" of the damage process, and the residual strength is the "ability evaluation" of the damage result. The two are closely related through the principle of material mechanics, and together constitute the core mathematical model of high-temperature pipeline life prediction. If the residual strength Rf(t) is greater than 0, it indicates that the pipeline is still in a safe operating state under the current working condition and has not reached the failure threshold. At this time, the system will continuously monitor the operating state of the pipeline according to the pre-set safety monitoring strategy, including real-time updating of the crack depth a(t) and the residual strength Rf(t), and dynamically evaluating the residual life of the pipeline in combination with the equipment operating conditions and environmental parameters, to ensure that sufficient time window is provided for subsequent maintenance decision-making before the crack expands to the critical size, and to ensure the safety and economy of power plant operation.

[0081] In this step, the Paris formula is used to model the whole cycle through corrosion correction, the crack length evolution path is integrated to solve the residual strength field, the residual life of the pipeline is predicted at the grid level, and the failure time prediction deviation is reduced.

[0082] S400, using the failure time and cost of the residual strength field as constraints, using the bird swarm algorithm to optimize the fitness function, generating a Pareto optimal maintenance sequence, taking the Pareto optimal maintenance sequence as an input item of the entropy weight optimization engine, and obtaining a maintenance implementation result.

[0083] It can be understood that the step S400 includes S401, S402 and S403.

[0084] S401, based on the residual strength field data, taking the failure time and maintenance cost as double constraints, constructing a multi-objective fitness function including a risk index function and a cost penalty term, and providing a quantitative evaluation standard for maintenance decision-making;

[0085] S402, adopt bird swarm algorithm to iteratively optimize the maintenance decision particle, fuse risk gradient and cost gradient information through particle position update formula, and screen out Pareto optimal maintenance sequence through non-dominated sorting;

[0086] S403, calculate information entropy of the Pareto solution set by using entropy weight method and select the optimal solution, and trigger parameter correction through actual failure time deviation analysis after maintenance implementation, forming a closed-loop optimization mechanism of maintenance decision and model parameters.

[0087] It should be noted that based on the residual strength field data output by step S300, the failure time t fail (unit: hour) and maintenance cost C repair (unit: ten thousand yuan) of each damage hot spot are extracted, and a double-objective fitness function is constructed:

[0088] (minimize failure risk)

[0089] (minimize maintenance cost)

[0090] Wherein, n is the number of damage hot spots, t now is the current time, and C budget is the upper limit of maintenance budget. The function amplifies the short-term failure risk through exponential decay term, and imposes quadratic penalty on the over-budget scheme, providing quantitative evaluation standard for subsequent optimization.

[0091] Taking the maintenance decision particle X i =[t start,i ,Loc i ,Method i ] (start time, position coordinates, maintenance method) as the optimization object, bird swarm algorithm is used for iterative search, and the particle position is updated. After each iteration, the non-dominated sorting is used to retain the Pareto optimal solution set, ensuring the optimal balance between risk and cost of the solution set.

[0092] Based on the above results, the information entropy of the Pareto solution set is calculated:

[0093]

[0094] Wherein, f j is the value of the jth objective function, ∑f j is the sum of all objective function values, and E is the information entropy of the Pareto solution set, which is used to measure the diversity of the solution set.

[0095] Select the solution with the smallest entropy value as the optimal maintenance sequence. After maintenance, the actual failure time is calculated by deviation, and the calculation formula is as follows:

[0096]

[0097] where t real is the actual failure time (unit: hour), t fail is the predicted failure time (unit: hour), and Δ represents the relative deviation between the actual failure time and the predicted failure time.

[0098] If Δ > 15%, the parameter correction (such as adjusting the risk coefficient β or the budget upper limit C budget ) of step S100 is triggered, forming a closed-loop optimization mechanism of the maintenance decision and the model parameters. This mechanism ensures the continuous effectiveness of the maintenance scheme in actual operation.

[0099] S500, according to the actual failure deviation in the maintenance implementation result, the KL divergence is calculated, the crack propagation model parameters are corrected by gradient descent, and feedback is formed to step three to form a whole life cycle closed-loop optimization.

[0100] Based on the actual failure time in the maintenance implementation result and the failure time predicted in step S300, the KL divergence formula is:

[0101]

[0102] where D KL is the prediction error, and n is the number of maintenance hotspots; the prediction deviation is quantified, where n is the number of maintenance hotspots, and the larger the DKL value is, the larger the prediction error is.

[0103] When D KL > 0.5, the Paris formula parameters are corrected by using the gradient descent method, and the formula is:

[0104]

[0105] where C new is the updated crack propagation rate coefficient, C old is the original crack propagation rate coefficient, m new is the updated material fatigue index, m old is the original material fatigue index, and η is the learning rate, is the partial derivative of the KL divergence to C, is the partial derivative of the KL divergence to m.

[0106] Subsequently, the crack propagation rate parameters are updated, and the corrected parameters C new and m new are fed back to the corrosion corrected Paris formula of step S300, the remaining strength field calculation module is updated synchronously, and the process of step S500 is repeated in the next maintenance cycle, forming a whole life cycle closed-loop optimization mechanism of “prediction-implementation-correction-re-prediction”.

[0107] It can be understood that in this step, the model parameters are dynamically updated by KL divergence through entropy weight closed loop parameter correction, forming a self-optimizing full life cycle closed loop, and the system has self-adaptive correction ability to coal quality fluctuation ± 20% and load mutation ± 30%, which significantly improves the intelligent level and full life cycle economy of power plant equipment health management.

[0108] Embodiment 2:

[0109] As shown in Figure 2 The embodiment provides a bionic decision-driven intelligent power plant full life cycle optimization system, which is shown in Figure 2 The system comprises:

[0110] The first generation module 701 is configured to generate a full life cycle feature vector set by fusing the boiler wall temperature feature vector and the flue gas-stress coupled feature vector through the immune algorithm, and the full life cycle feature vector set is used as input data for damage hot spot identification.

[0111] The screening module 702 is configured to calculate a maximum Lyapunov index field based on the full life cycle feature vector set, construct a damage topology graph, screen a high-sensitive area to generate a damage hot spot coordinate set, and drive creep evolution modeling.

[0112] The solving module 703 is configured to combine the damage hot spot coordinate set and SO2 gradient data, solve a crack propagation differential equation through a Paris formula with corrosion correction, and output a residual strength field, wherein the residual strength field comprises a position, a crack length and a failure time.

[0113] The second generation module 704 is configured to use the failure time and the cost of the residual strength field as constraints, optimize an adaptability function by using a bird swarm algorithm, generate a Pareto optimal maintenance sequence, use the Pareto optimal maintenance sequence as an input item of an entropy weight optimization engine, and obtain a maintenance implementation result.

[0114] The calculation module 705 is configured to calculate KL divergence according to an actual failure deviation in the maintenance implementation result, correct crack propagation model parameters through gradient descent, and feed back to step three to form a full life cycle closed loop optimization.

[0115] Specifically, the first generation module 701 comprises:

[0116] The extraction unit is configured to reduce the boiler combustion chamber metal wall temperature time series data to a 20-dimensional feature vector through space-time tensor decomposition, extract the energy entropy of the steam turbine vibration frequency spectrum, and generate a device-level feature vector.

[0117] The acquisition unit is configured to acquire a flue gas composition spectrum and a pipeline stress and strain field, and generate a 15-dimensional environment-mechanical feature vector through thermal corrosion coupling analysis.

[0118] Fusion unit: used to fuse 20-dimensional feature vector and 15-dimensional environment-mechanical feature vector by adopting immune algorithm, and to calculate data similarity and iterative convergence through affinity matrix, and to output full life cycle feature vector set as input of damage hotspot identification.

[0119] Specifically, the screening module 702 includes:

[0120] The first generation unit: for calculating the maximum Lyapunov exponent field by adopting Wolf algorithm based on the wall temperature feature subset in the full life cycle feature vector set, quantifying the wall temperature chaotic characteristics and the corrosion index, and generating the damage sensitivity scalar field, wherein the formula for calculating the maximum Lyapunov exponent field by adopting Wolf algorithm is as follows:

[0121]

[0122] In the formula, T w (i) is the wall temperature vector at the i-th sampling time, is the reference orbit, t obs is the observation time length, and N is the number of sampling points;

[0123] The construction unit: for associating the damage sensitivity with the spatial coordinates, and constructing the damage topology graph containing node coordinates, sensitivity values and connection relationships;

[0124] The screening unit: for screening out the nodes in the damage topology graph greater than 0.8, and defining as high sensitivity area, and taking the damage hotspot coordinate set in the high sensitivity area as the input of creep evolution modeling.

[0125] Specifically, the solving module 703 includes:

[0126] The first calculation unit: for calculating the crack propagation rate at each hotspot position by adopting the Paris formula corrected by corrosion, in combination with the damage hotspot coordinate set and the SO2 concentration gradient, wherein the calculation formula of the Paris formula is as follows:

[0127]

[0128] In the formula, C, m are material constants, ΔK eff is the effective stress intensity factor range, γ is the corrosion acceleration coefficient, is the SO2 concentration gradient, Q f is the fatigue activation energy, R is the gas constant, and T is the pipe wall temperature;

[0129] The solving unit: for time-integrating the crack propagation rate by adopting the fourth-order Runge-Kutta method, solving the crack length evolution path, and obtaining the functional relationship between the crack depth and time;

[0130] The second calculation unit is configured to calculate the residual strength based on a function relationship, and the calculation formula is as follows:

[0131]

[0132] wherein, is a reduction coefficient of the bearing area caused by the crack, and u is the tensile strength of the material, and hoop is the hoop stress of the pipeline, and f (t) is the residual strength.

[0133] The judging unit is configured to judge whether the residual strength is less than or equal to 0. If the residual strength is less than or equal to 0, it is determined that the pipeline fails, and residual strength field data including the three-dimensional position of the crack, the final crack length and the failure time are output. If the residual strength is greater than 0, the pipeline is normal at the current time, and the residual strength field data is output.

[0134] It should be noted that, as to the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment related to the method, and will not be described in detail here.

[0135] Embodiment 3

[0136] Corresponding to the above method embodiment, the embodiment also provides a bionic decision-driven smart power plant whole life cycle optimization device. The bionic decision-driven smart power plant whole life cycle optimization device described below can be correspondingly referred to the bionic decision-driven smart power plant whole life cycle optimization method described above.

[0137] Figure 3 Fig. 8 is a block diagram of a bionic decision-driven smart power plant whole life cycle optimization device 800 according to an example embodiment. As shown in Fig. 8, the bionic decision-driven smart power plant whole life cycle optimization device 800 includes a processor 801 and a memory 802. The bionic decision-driven smart power plant whole life cycle optimization device 800 also includes one or more of a multimedia component 803, an I / O interface 804, and a communication component 805. Figure 3

[0138] ​The processor 801 is configured to control overall operation of the bionic decision-driven smart power plant whole life cycle optimization device 800 to complete all or part of the steps in the bionic decision-driven smart power plant whole life cycle optimization method described above. The memory 802 is configured to store various types of data to support the operation of the bionic decision-driven smart power plant whole life cycle optimization device 800, which can include, for example, instructions for any application or method operating on the bionic decision-driven smart power plant whole life cycle optimization device 800, and application-related data, such as contact data, sent and received messages, pictures, audio, video, and the like. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk. The multimedia component 803 can include a screen and an audio component. The screen can be a touch screen, for example, and the audio component is configured to output and / or input audio signals. For example, the audio component can include a microphone configured to receive external audio signals. The received audio signals can be further stored in the memory 802 or transmitted through the communication component 805. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 804 provides an interface between the processor 801 and other interface modules, which can be a keyboard, mouse, or button, etc. These buttons can be virtual buttons or physical buttons. The communication component 805 is configured to enable wired or wireless communication between the bionic decision-driven smart power plant whole life cycle optimization device 800 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, near field communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 805 can include a Wi-Fi module, a Bluetooth module or an NFC module.

[0139] In an example embodiment, the bionic decision-driven smart power plant life cycle optimization device 800 can be implemented by one or more Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), controller, microcontroller, microprocessor or other electronic elements for executing the above-mentioned bionic decision-driven smart power plant life cycle optimization method.

[0140] In another example embodiment, a computer readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the above-mentioned bionic decision-driven smart power plant life cycle optimization method. For example, the computer readable storage medium can be the above-mentioned memory 802 including program instructions, which can be executed by the processor 801 of the bionic decision-driven smart power plant life cycle optimization device 800 to complete the above-mentioned bionic decision-driven smart power plant life cycle optimization method.

[0141] Embodiment 4:

[0142] Corresponding to the above method embodiments, the present embodiment also provides a readable storage medium, which can be referred to in conjunction with the above-mentioned bionic decision-driven smart power plant life cycle optimization method.

[0143] The computer program stored on the readable storage medium, when executed by a processor, implements the steps of the above-mentioned bionic decision-driven smart power plant life cycle optimization method of the method embodiments.

[0144] The readable storage medium can be specifically a U disk, a mobile hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and various readable storage media that can store program codes.

[0145] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0146] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for full life cycle optimization of a smart power plant driven by a biomimetic decision, characterized in that, The method comprises the following steps: Step 1: The boiler wall temperature feature vector and the flue gas-stress coupling feature vector are fused by an immune algorithm to generate a full life cycle feature vector set as input data for damage hot spot identification; Step 2: The maximum Lyapunov index field is calculated based on the full life cycle feature vector set, and a damage topology graph is constructed to screen a damage hot spot coordinate set and drive creep evolution modeling; Step 3: The corrosion-corrected Paris formula is used to solve the crack propagation differential equation by combining the damage hot spot coordinate set and the SO2 gradient data, and the residual strength field is output, wherein the residual strength field includes position, crack length and failure time; Step 4: The failure time and cost of the residual strength field are taken as constraints, and the bird swarm algorithm is used to optimize the fitness function to generate a Pareto optimal maintenance sequence, which is taken as an input item of an entropy weight optimization engine to obtain a maintenance implementation result; Step 5: The KL divergence is calculated according to the actual failure deviation in the maintenance implementation result, the crack propagation model parameters are corrected by gradient descent, and the full life cycle closed loop optimization is formed by feedback to step 3; In step 1, the following steps are included: The time series data of the boiler combustion chamber metal wall temperature is reduced to a 20-dimensional feature vector by space-time tensor decomposition, and the energy entropy of the steam turbine vibration spectrum is extracted to generate a device-level feature vector; The flue gas composition spectrum and the pipeline stress and strain field are collected, and a 15-dimensional environment-mechanical feature vector is generated through thermal corrosion coupling analysis; The 20-dimensional feature vector and the 15-dimensional environment-mechanical feature vector are fused by an immune algorithm, and the data similarity is calculated by an affinity matrix and iteratively converged to output a full life cycle feature vector set as input for damage hot spot identification; In step 3, the following steps are included: The crack propagation rate at each hot spot position is calculated by combining the damage hot spot coordinate set and the SO2 concentration gradient through the corrosion-corrected Paris formula, and the calculation formula of the Paris formula is as follows: where Cm is a material constant, is the effective stress intensity factor range, is the corrosion acceleration factor, is the SO2 concentration gradient, is the fatigue activation energy, R is the gas constant, and T is the pipe wall temperature; The crack propagation rate is time-integrated by using the fourth-order Runge-Kutta method to solve the crack length evolution path, and a function relationship between the crack depth and time is obtained; Based on the function relationship, the residual strength is calculated, and the calculation formula is as follows: wherein is the reduction factor for the load bearing area due to the crack, is the tensile strength of the material, is the hoop stress of the pipe, is the residual strength; It is judged whether the residual strength is less than or equal to 0, if the residual strength is less than or equal to 0, it is judged that the failure occurs, and the residual strength field data including the three-dimensional position of the crack, the final crack length and the failure time are output; if the residual strength is greater than 0, the pipeline is normal at the current time, and the output is performed.

2. The method of claim 1, wherein, In step 2, the following steps are included: Based on the wall temperature feature subset in the full life cycle feature vector set, the Wolf algorithm is used to calculate the maximum Lyapunov index field to quantify the wall temperature chaotic characteristics and the corrosion index, and a damage sensitivity scalar field is generated, wherein the formula for calculating the maximum Lyapunov index field by using the Wolf algorithm is as follows: In the formula, is the wall temperature vector at the i-th sampling time, is a reference orbit, is an observation time length, and N is the number of sampling points. The damage sensitivity is associated with the spatial coordinates to construct a damage topology graph including node coordinates, sensitivity values and connection relationships; The nodes with a value greater than 0.8 in the damage topology graph are screened out and defined as high sensitivity areas, and the damage hot spot coordinate set in the high sensitivity areas is taken as input for creep evolution modeling.

3. A bionic decision-driven intelligent power plant whole life cycle optimization system based on the bionic decision-driven intelligent power plant whole life cycle optimization method of claim 1, characterized in that, The method comprises the following steps: The first generation module is configured to generate a full life cycle feature vector set by fusing the boiler wall temperature feature vector and the flue gas-stress coupling feature vector through an immune algorithm, and the full life cycle feature vector set is used as input data for damage hot spot identification. The screening module is configured to calculate a maximum Lyapunov index field based on the full life cycle feature vector set, construct a damage topology graph, screen a high sensitive area to generate a damage hot spot coordinate set, and drive creep evolution modeling. The solving module is configured to solve a crack propagation differential equation by using a Paris formula corrected by corrosion in combination with the damage hot spot coordinate set and SO2 gradient data, and output a residual strength field including position, crack length and failure time. The second generation module is configured to generate a Pareto optimal maintenance sequence by optimizing an adaptability function using a bird swarm algorithm with the failure time and cost of the residual strength field as constraints, and use the Pareto optimal maintenance sequence as an input item of an entropy weight optimization engine to obtain a maintenance implementation result. The calculation module is configured to calculate a KL divergence according to an actual failure deviation in the maintenance implementation result, correct crack propagation model parameters through gradient descent, and feed back to step three to form a full life cycle closed loop optimization. The first generation module includes: The extraction unit is configured to reduce the boiler combustion chamber metal wall temperature time series data to a 20-dimensional feature vector through space-time tensor decomposition, and extract the energy entropy of the steam turbine vibration spectrum to generate a device-level feature vector. The acquisition unit is configured to acquire a flue gas composition spectrum and a pipeline stress and strain field, and generate a 15-dimensional environment-mechanical feature vector through thermal corrosion coupling analysis. The fusion unit is configured to fuse the 20-dimensional feature vector and the 15-dimensional environment-mechanical feature vector using an immune algorithm, calculate the data similarity through an affinity matrix, and iteratively converge to output a full life cycle feature vector set as input for damage hot spot identification. The solving module includes: The first calculation unit is configured to calculate the crack propagation rate of each hot spot position by using a Paris formula corrected by corrosion in combination with the damage hot spot coordinate set and the SO2 concentration gradient, and the calculation formula of the Paris formula is as follows: where Cm is a material constant, is the effective stress intensity factor range, is the corrosion acceleration factor, is the SO2 concentration gradient, is the fatigue activation energy, R is the gas constant, and T is the pipe wall temperature; The solving unit is configured to use a fourth-order Runge-Kutta method to time-integrate the crack propagation rate to solve the crack length evolution path, and obtain a function relationship between the crack depth and time. The second calculation unit is configured to calculate the residual strength based on the function relationship, and the calculation formula is as follows: wherein is the reduction factor for the load bearing area due to the crack, is the tensile strength of the material, is the hoop stress of the pipe, is the residual strength; The judgment unit is configured to determine whether the residual strength is less than or equal to 0, and if the residual strength is less than or equal to 0, it is determined that the pipeline has failed, and the residual strength field data including the three-dimensional position of the crack, the final crack length and the failure time are output; if the residual strength is greater than 0, the pipeline is normal at the current time, and the output is performed.

4. The biologically inspired decision driven, intelligent power plant, life cycle optimization system of claim 3, wherein, The screening module includes: The first generation unit is configured to calculate a maximum Lyapunov index field using a Wolf algorithm based on a wall temperature feature subset in the full life cycle feature vector set, quantify the wall temperature chaotic characteristics and corrosion index, and generate a damage sensitivity scalar field, and the formula for calculating the maximum Lyapunov index field using the Wolf algorithm is as follows: In the formula, is the wall temperature vector at the i-th sampling time, is a reference orbit, is an observation time length, and N is the number of sampling points. A construction unit is configured to associate the damage sensitivity with the spatial coordinates, and construct a damage topology graph including node coordinates, sensitivity values and connection relationships; A screening unit is configured to screen out nodes greater than 0.8 in the damage topology graph, and define the nodes as high sensitivity areas, and set a damage hotspot coordinate set in the high sensitivity areas as an input of creep evolution modeling.

5. A kind of bionic decision-driven wisdom power plant full life cycle optimization equipment, it is characterized in to, It comprises: a memory for storing a computer program; a processor for implementing the bionic decision-driven whole life cycle optimization method of the smart power plant according to any one of claims 1-2 when executing the computer program.

6. A readable storage medium characterized by: The readable storage medium has a computer program stored thereon, and the computer program is executed by the processor to implement the bionic decision-driven whole life cycle optimization method of the smart power plant according to any one of claims 1-2.

Images