Nothing Special   »   [go: up one dir, main page]

WO2022193569A1 - 基于稀疏大数据挖掘的火电机组汽轮机优化方法及系统 - Google Patents

基于稀疏大数据挖掘的火电机组汽轮机优化方法及系统 Download PDF

Info

Publication number
WO2022193569A1
WO2022193569A1 PCT/CN2021/116827 CN2021116827W WO2022193569A1 WO 2022193569 A1 WO2022193569 A1 WO 2022193569A1 CN 2021116827 W CN2021116827 W CN 2021116827W WO 2022193569 A1 WO2022193569 A1 WO 2022193569A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
frequent
steam turbine
list
item
Prior art date
Application number
PCT/CN2021/116827
Other languages
English (en)
French (fr)
Inventor
林金星
陆铭杰
胡志
Original Assignee
南京邮电大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 南京邮电大学 filed Critical 南京邮电大学
Priority to US18/548,447 priority Critical patent/US20240231295A9/en
Publication of WO2022193569A1 publication Critical patent/WO2022193569A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B15/00—Systems controlled by a computer
    • G05B15/02—Systems controlled by a computer electric
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24—Querying
    • G06F16/245—Query processing
    • G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
    • G06F16/2465—Query processing support for facilitating data mining operations in structured databases
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/08—Thermal analysis or thermal optimisation
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Definitions

  • the invention belongs to the technical field of steam turbine optimization, and in particular relates to a method and system for optimizing a steam turbine of a thermal power unit based on sparse big data mining.
  • the unit operation optimization method based on association rule mining is to obtain the main operating parameters and unit-related performance under the condition of good unit performance from the long-term operation data of the unit.
  • the strong association rules between indicators do not require the establishment of complex mechanism models, avoid the problems of being susceptible to working conditions and time-consuming and labor-intensive, and have high feasibility.
  • DCS systems and SIS systems have been widely deployed in power plants, and large-scale operation data have been accumulated, which has created favorable conditions for the optimization method of unit operation based on association rule mining.
  • the present invention provides a method and system for optimizing the steam turbine of a thermal power unit based on sparse big data mining, which can effectively overcome the time-consuming and memory occupation of the previous association rule mining algorithms when mining association rules of sparse massive data. It has high defects and can mine large-scale data with load balancing.
  • the technical scheme adopted in the present invention is: a method for optimizing a steam turbine of a thermal power unit, comprising: taking the reduction of the heat consumption of the steam turbine as an optimization goal, collecting historical operation data and preprocessing the data; The method selected from the Pearson correlation analysis, selects a set of controllable operating parameters of the unit that meet the set conditions with the correlation with the steam turbine heat consumption rate from the historical operating data as the optimization parameters; constructs a pattern growth class association rule for sparse data Mining algorithm; on the big data analysis and processing framework Apache Spark, a load balancing strategy based on matrix operations, a pattern growth-like association rule mining algorithm that realizes global computing balance in parallel; the fuzzy C-means clustering algorithm is used to discretize historical operating data, based on The parallelized pattern growth class association rule mining algorithm mines the discretized historical operation data to obtain the association rules, and de-discretes them to obtain the target value of the steam turbine optimization parameters under each boundary condition.
  • the data preprocessing refers to eliminating abnormal data and redundant data in the historical operation data and performing steady-state detection on the historical operation data.
  • the criterion for the steady state detection is: within a certain period of time, when the fluctuation value of the operating state parameter of the steam turbine is less than the set range, it can be considered that the unit is in a stable operating condition.
  • the construction of a sparse data-oriented pattern growth class association rule mining algorithm includes: S31, setting a minimum support threshold, traversing the sparse transaction data set, denoted as D, counting the frequency of each item, generating a frequent item list, denoting it.
  • F_List is F_List
  • S32 traverse F_List, label frequent items, and generate an item header table, denoted as H-Table, including item number, support count and link pointer
  • S33 filter out infrequent items in D, convert and store them as binary Matrix, denoted as PBM, where the element of "1" indicates that the element corresponding to a certain frequent item in the F-List is contained in a transaction, and the element of "0" indicates that the element corresponding to the element is not contained in a transaction A certain frequent item in the F-List
  • S34 scan the PBM, adjust the pointer in the H-Table, and link the position of the first "1" of each row in the PBM with the corresponding frequent item in the frequent item header table H-Table, Extract the row with the first "1" in the same position in the PBM, generate multiple sub-PBMs, and convert the task of mining all frequent itemsets into multiple subtasks of mining local frequent itemsets
  • S35 aggregate the local frequent itemsets,
  • the sub-task comprises the following steps: S341, scan the sub-PBM, sum up each column, update the support count of frequent items in the sub-item header table H-Table of the corresponding sub-PBM; S342, utilize the pointer to convert the sub-PBM The columns whose neutralization is greater than the minimum support threshold are linked with the corresponding frequent items in the sub-H-Table, and grow into a longer local frequent itemset; S343, recursively execute S341 and S342, until the sum of each column of the sub-PBM is less than the minimum support threshold.
  • Apache Spark based on the load balancing strategy of matrix operation, parallelize the pattern growth class association rule mining algorithm to realize global computing balance, including: S41, start Apache Spark, the main node reads Sparse transaction data set D, and horizontally cut D into P data blocks of equal size and continuous, and send them to P slave nodes respectively; S42, each slave node traverses its own data block once, and calculates the support count of all items , and sent to the master node; S43, the master node compares the support count of all items with the minimum support threshold, filters out frequent items, generates F_List and H-Table, and sends F_List and H-Table to P slave nodes; S44, each slave node traverses its own data block again according to F_List, converts and stores it into PBM according to step S33, and counts the number of the first row of "1" in the same position in the PBM, and the column number corresponds to the H-Table The item number is formed, the item number and the
  • the load balancing strategy based on matrix calculation includes: S451, the master node sorts the item number and the number of rows after adding the number of rows in the descending order of the number of rows; S452, the master node according to between F_List and H-Table There is a one-to-one correspondence between frequent items and item numbers, and the sorted item numbers and row numbers are converted into frequent items sorted in this order; S453, the main node combines the frequent items in the combination order from both ends, and is divided into P group; S454, the master node scans the frequent items in the P group in turn, and generates a grouping list G_List.
  • fuzzy C-means clustering algorithm to discretize the historical operation data includes: marking the discretized data interval in the form of letters and numbers, and replacing the numerical value with the label of the interval where each piece of data is located.
  • a steam turbine optimization system for a thermal power unit comprising: a first module for collecting historical operation data and data preprocessing with reducing the heat consumption rate of the steam turbine as an optimization goal; a second module for analyzing rough selection and peeling based on typical correlation The selected method of Mori correlation analysis, selects a set of controllable operating parameters of the unit whose correlation with the heat consumption rate of the steam turbine meets the set conditions from the historical operating data as the optimization parameters; the third module is used to construct a sparse data-oriented model Growth-type association rule mining algorithm; the fourth module is used in the big data analysis and processing framework Apache Spark, based on the load balancing strategy of matrix operation, parallelization to achieve global computing balance mode growth-type association rule mining algorithm; the fifth module, It is used to discretize historical operation data by using fuzzy C-means clustering algorithm, and mining the discretized historical operation data to obtain association rules based on the parallelized pattern growth class association rule mining algorithm, and de-discretize to obtain the optimization of steam turbine under various boundary conditions.
  • the present invention can effectively overcome the Apriori algorithm and the FP-Growth algorithm to mine sparse mining by using binary matrix and hyperlink technology to design a new pattern growth class association rule mining algorithm.
  • the invention also aims at the large scale of operation data under the full working condition and long cycle of the steam turbine of the thermal power unit, and is based on the analysis and processing of big data mainly based on memory calculation.
  • the framework Apache Spark realizes the parallelization of the designed association rule mining algorithm, avoids the shortcomings of MapReduce frequently reading and writing disks, and reduces a large amount of I/O overhead; and, in view of the characteristics of the binary matrix data storage structure that is easy to calculate the matrix, the present invention proposes an algorithm based on The load balancing strategy of matrix computing can more accurately allocate the tasks of cluster computing nodes, give full play to the performance advantages of the cluster, and efficiently mine the large-scale operation data of steam turbines to obtain the operation optimization target value when the heat consumption rate is low.
  • Fig. 1 is the main flow chart of a kind of thermal power unit steam turbine optimization method provided by the embodiment of the present invention
  • FIG. 3 is a flow chart of the parallelization of the association rule mining algorithm designed in the embodiment of the present invention on Apache Spark.
  • an optimization method for a steam turbine of a thermal power unit includes: taking reducing the heat consumption of the steam turbine as the optimization goal, collecting historical operation data and preprocessing the data; analyzing the correlation between rough selection and Pearson based on typical correlation The selected method is analyzed, and a set of controllable operating parameters of the unit whose correlation with the heat consumption rate of the steam turbine meets the set conditions are selected from the historical operating data as the optimization parameters.
  • Apache Spark On the data analysis and processing framework Apache Spark, a load balancing strategy based on matrix operations, a pattern growth-like association rule mining algorithm that realizes global computing balance in parallel; the fuzzy C-means clustering algorithm is used to discretize historical running data, and the pattern growth based on parallelization
  • the association rule-like mining algorithm mines the discretized historical operation data to obtain the association rules, and de-discretes them to obtain the target value of the steam turbine optimization parameters under each boundary condition.
  • Step 1 Take the low heat consumption rate of the steam turbine as the optimization goal, collect historical operation data and preprocess the data; data preprocessing refers to eliminating abnormal data and redundant data in the historical operation data and steady-state detection; The criterion is: within a certain period of time, when the fluctuation value of the operating state parameters of the steam turbine is less than the set range, it can be considered that the unit is in a stable operating condition;
  • the heat consumption rate is one of the indicators of thermal economy, which fully reflects the thermal economy of the unit. Therefore, the heat consumption rate is selected as the performance index, and the reduction of the heat consumption rate of the steam turbine is taken as the operation optimization goal; then,
  • the historical data of the steam turbine of a 1000MW thermal power unit from August 2018 to July 2019 is collected from the SIS and DCS system of a power plant.
  • the data range is a complete operation cycle between two shutdowns for maintenance, and the sampling frequency is 60s.
  • the The average value method can obtain a data that can fully and correctly reflect the actual state; finally, the parameters used for steady state detection are the unit load and the main steam pressure, and the basis for judging stability is that the difference between the maximum value and the minimum value of the data within 20 minutes is within within a certain stability threshold.
  • Step 2 According to the characteristics of high dimension and strong correlation of steam turbine operating data parameters, based on the methods of rough selection of typical correlation analysis and selection of Pearson correlation analysis, select a set of controllable operating parameters of the unit that are strongly related to the heat consumption rate of the steam turbine. as an optimization parameter;
  • a significance test is carried out on the typical correlation coefficient of the controllable variables of the steam turbine. If the correlation degree of a pair of controllable variables is not significant, it means that the pair of variables is not representative, and the pair of variables is discarded; For several pairs of controllable variables with higher typical correlation coefficients, the variables with larger absolute value of linear combination coefficients are selected as candidate operation optimization parameters; then, Pearson correlation analysis is used to further reduce the candidate operation optimization parameters, and the principle of reduction is to keep Pearson The correlation coefficient is greater than 0.8; finally, the optimal parameters of steam turbine operation are determined as: unit power, main steam flow, main steam pressure and temperature, feed water pressure, feed water temperature and condenser vacuum.
  • Step 3 Aiming at the characteristics of large transaction pattern differences and scattered item distribution in sparse data in frequent pattern mining, based on binary matrix and hyperlink technology, construct a new pattern growth class association rule mining algorithm for sparse data; including:
  • S34 Scan the PBM, adjust the pointer in the H-Table, link the position of the first "1" of each row in the PBM with the corresponding frequent item in the frequent item header table H-Table, and extract the first "1" in the PBM in the same
  • the row of the position generates multiple sub-PBMs, and converts the task of mining all frequent itemsets into multiple subtasks of mining local frequent itemsets; the subtasks include the following steps:
  • S341 scan the sub-PBM, sum up each column, and update the support count of frequent items in the sub-item header table H-Table of the corresponding sub-PBM;
  • S35 Aggregate local frequent itemsets, and output all frequent itemsets.
  • Step 4 Aiming at the problem that the serial association rule mining algorithm cannot mine large-scale data due to the limitation of the hardware resources of a single machine, on the big data analysis and processing framework Apache Spark, a load balancing strategy based on matrix operations is proposed, and a mode of parallelizing the global computing balance is proposed.
  • Growth class association rule mining algorithms including:
  • the master node reads the sparse transaction data set D, and horizontally cuts D into P data blocks of equal size and continuous, and sends them to P slave nodes respectively;
  • each slave node traverses its own data block once, calculates the support count of all items, and sends it to the master node;
  • the master node compares the support count of all items with the minimum support threshold, filters out frequent items, generates F_List and H-Table, and sends F_List and H-Table to P slave nodes;
  • each slave node traverses its own data block again according to F_List, converts and stores it into PBM according to step S33, and counts the number of the first row of "1" in the same position in the PBM, and the column number corresponds to the H-Table Item number, form a key-value pair (item number, row number), and send it to the master node;
  • the master node adds the number of rows with the same item number, groups according to the new load balancing strategy, generates a group list, denoted as G_List, and sends it to P slave nodes, including:
  • the master node sorts the key-value pairs (item number, number of rows) after adding the number of rows in the descending order of the number of rows;
  • the master node converts the sorted key-value pairs (item number, row number) into frequent items sorted in this order according to the one-to-one correspondence between frequent items and item numbers between F_List and H-Table;
  • the master node combines frequent items in order from both ends, and divides them into P groups;
  • the master node scans the frequent items in the P group in turn, and generates a grouping list G_List;
  • the slave nodes exchange data in the PBM between the slave nodes according to the G-List;
  • each slave node mines local frequent itemsets according to G-List and step S34;
  • the slave node sends the local frequent itemsets to the master node for aggregation, and obtains all frequent itemsets, that is, the frequent itemsets of the sparse transaction data set D (or the frequent patterns of discrete steady-state historical operation data).
  • Step 5 Use the fuzzy C-means clustering algorithm to discretize the historical operation data, and mine the frequent patterns of the discretized historical operation data based on the parallelized pattern growth class association rule mining algorithm, and de-discretize to obtain the steam turbine under each boundary condition.
  • the target value of optimization parameters Fuzzy C-Means (FCM) clustering algorithm can effectively classify objects with complex characteristics, and give relatively optimal classification results, which is more in line with objective reality. Therefore, the fuzzy C-means (FCM) clustering algorithm is used to discretize the steady-state historical data of each running optimization parameter, and the discretized data interval is marked in the form of letters + numbers, and the label of each data interval is used to replace its value. .
  • the significance of this strong association rule is that when the values of the adjustment operating parameters are within the range shown in Table 1, there is a probability of not less than 80% to make the heat consumption rate optimal. value.
  • the target value of these operation optimization parameters is the interval shown in the table, in order to make the target value more intuitive, the present invention selects the center value of the interval as the optimization target value, and the optimized target value is shown in Table 2:
  • the parameters can be adjusted according to the optimized target values in Table 2, so that the heat consumption rate can reach the optimal value, and then the unit can operate in the best condition.
  • the optimized value of the heat consumption rate is 6753.55kJ/kW ⁇ h, compared with the actual operating average of 6780.35kJ/kW ⁇ h, The heat consumption rate is reduced by 26.8kJ/kW ⁇ h. From the calculation formula of standard coal consumption rate:
  • the operation optimization method can save 1.01g of coal per 1kW ⁇ h of electricity, save economic expenses, and reduce the emission of air pollutants, which can effectively achieve the purpose of energy conservation and emission reduction.
  • the association rule mining algorithm is the analysis and processing of discrete data, but the operation data of steam turbines of thermal power units are continuous, and the discretization of continuous data will inevitably lead to large model differences and scattered item distribution; in addition, steam turbine is a multi-variable and high-dimensional System, the sparse characteristics of high-dimensional data are more obvious after discretization.
  • This embodiment uses binary matrix and hyperlink technology to design a new pattern growth class association rule mining algorithm, which can effectively overcome the Apriori algorithm and the FP-Growth algorithm to mine sparse mining.
  • the invention When the data is in frequent mode, it takes a long time and occupies too much memory; at the same time, the invention also aims at the large scale of operation data under the full working condition and long cycle of the steam turbine of the thermal power unit, and is based on the analysis and processing of big data mainly based on memory calculation.
  • the framework Apache Spark realizes the parallelization of the designed association rule mining algorithm, avoids the shortcomings of MapReduce frequently reading and writing disks, and reduces a large amount of I/O overhead; and, in view of the characteristics of the binary matrix data storage structure that is easy to calculate the matrix, the present invention proposes a new method.
  • the load balancing strategy can accurately allocate the tasks of cluster computing nodes, give full play to the performance advantages of the cluster, and efficiently mine the large-scale operation data of steam turbines to obtain the operation optimization target value when the heat consumption rate is low.
  • Embodiment 2 is a diagrammatic representation of Embodiment 1:
  • the present embodiment provides a steam turbine optimization system for a thermal power unit, including: a first module for collecting historical operating data and predicating the data with the reduction of the heat consumption of the steam turbine as the optimization goal. Processing; the second module is used to select a group of units whose correlation with the heat consumption rate of the steam turbine meets the set conditions from the historical operation data based on the method of rough selection of the canonical correlation analysis and the selection of the Pearson correlation analysis.
  • the parameters are used as optimization parameters; the third module is used to construct a pattern growth class association rule mining algorithm for sparse data; the fourth module is used for the load balancing strategy based on matrix operations on the big data analysis and processing framework Apache Spark, parallelization The pattern growth class association rule mining algorithm that realizes the global calculation balance; the fifth module is used to discretize the historical operation data by using the fuzzy C-means clustering algorithm, and the parallelized pattern growth class association rule mining algorithm is used to mine the discretized historical operation data.
  • the association rules are obtained and de-discretized to obtain the target value of the steam turbine optimization parameters under each boundary condition.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Probability & Statistics with Applications (AREA)
  • Mathematical Physics (AREA)
  • Fuzzy Systems (AREA)
  • Automation & Control Theory (AREA)
  • Computer Hardware Design (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • Control Of Turbines (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种基于稀疏大数据挖掘的火电机组汽轮机优化方法及系统，包括：采集机组历史运行数据；选取一组与汽轮机热耗率较强相关的可控运行参数作为优化参数；基于二进制矩阵和超链接技术，构建新型面向稀疏数据的模式增长类关联规则挖掘算法；在Apache Spark平台上，提出基于矩阵运算的负载均衡策略，并行化实现全局计算平衡的模式增长类关联规则挖掘算法；通过聚类离散化历史运行数据，并行挖掘离散化历史运行数据得到关联规则，并反离散化，得出各个边界条件下汽轮机优化参数的目标值。可有效解决火电机组汽轮机连续运行数据离散化导致的稀疏特性，并可高效挖掘汽轮机大规模运行数据以获取热耗率较低时的运行优化目标值。

Description

基于稀疏大数据挖掘的火电机组汽轮机优化方法及系统 技术领域
本发明属于汽轮机优化技术领域，具体涉及一种基于稀疏大数据挖掘的火电机组汽轮机优化方法及系统。
背景技术
近年来，在我国节能减排工作不断深入的大环境下，节能降耗成为各火电厂迫切需要完成的任务。汽轮机作为火电机组的重要组成部分，其运行效率直接决定火力发电厂电力资源的供应。因此，研究火电机组汽轮机运行优化至关重要，不仅提高发电效率，也能降低发电成本。关联规则挖掘是数据挖掘领域中应用广泛且实用意义较高的方法，能够发掘数据之间隐含着的关联性和相关性。相较于设计值法、最优运行试验法和变工况热力计算法，基于关联规则挖掘的机组运行优化方法是从机组长期运行数据中获取机组性能较好状态下主要运行参数与机组相关性能指标间的强关联规则，不需要建立复杂的机理模型，避免了易受工况局限以及耗时耗力的问题，具有较高的可行性。此外，随着自动化及信息化程度的日益提高，电厂广泛部署DCS系统和SIS系统，积累了大规模的运行数据，为基于关联规则挖掘的机组运行优化方法创造了有利条件。
现有技术中的算法大多是串行运行方式，处理的是小样本汽轮机数据。火电机组运行参数众多、系统工况多变且数据采集周期短，小样本运行数据难以体现汽轮机的实际特性。现有技术中采用Apriori算法或FP-Growth算法挖掘稀疏数据频繁模式时，因算法自身瓶颈导致耗时长、占用内存高的缺陷；串行关联规则挖掘算法受单机硬件资源限制，难以分析处理大规模数据。
发明内容
为解决现有技术中的不足，本发明提供一种基于稀疏大数据挖掘的火电机组汽轮机优化方法及系统，可有效克服挖掘稀疏海量数据关联规则时，以往关联规则挖掘算法耗时长、内存占用过高的缺陷，并能负载均衡地挖掘大规模数据。
为达到上述目的，本发明所采用的技术方案是：一种火电机组汽轮机优化 方法，包括：以降低汽轮机热耗率为优化目标，采集历史运行数据并数据预处理；基于典型相关性分析粗选与皮尔森相关性分析精选的方法，从历史运行数据中选取一组与汽轮机热耗率相关性符合设定条件的机组可控运行参数作为优化参数；构建面向稀疏数据的模式增长类关联规则挖掘算法；在大数据分析处理框架Apache Spark上，基于矩阵运算的负载均衡策略，并行化实现全局计算平衡的模式增长类关联规则挖掘算法；采用模糊C均值聚类算法离散化历史运行数据，基于并行化的模式增长类关联规则挖掘算法，挖掘离散化历史运行数据得到关联规则，并反离散化，得出各个边界条件下汽轮机优化参数的目标值。
进一步地，所述数据预处理，是指剔除历史运行数据中的异常数据和冗余数据并对历史运行数据进行稳态检测。
进一步地，所述稳态检测的判别标准是：在一定时间段内，当汽轮机的运行状态参数波动值小于设定范围时，可以认为机组处于稳定运行工况。
进一步地，所述构建面向稀疏数据的模式增长类关联规则挖掘算法，包括：S31、设定最小支持度阈值，遍历稀疏事务数据集，记为D，统计各项频数，生成频繁项列表，记为F_List；S32、遍历F_List，对频繁项标号，生成项头表，记为H-Table，包括项号、支持度计数和链接指针；S33、筛去D中的非频繁项，转化存储为二进制矩阵，记为PBM，其中为“1”的元素表示在某个事务中含有该元素对应F-List中的某个频繁项，为“0”的元素表示在某个事务中不含有该元素对应F-List中的某个频繁项；S34、扫描PBM，调整H-Table中的指针，将PBM中每行首个“1”所在位置和频繁项项头表H-Table中对应频繁项链接，提取PBM中首个“1”在相同位置的行，生成多个子PBM，将挖掘全部频繁项集的任务转化为多个挖掘局部频繁项集的子任务；S35、聚合局部频繁项集，输出全部频繁项集。
进一步地，所述子任务包括以下步骤：S341、扫描子PBM，对每列求和，更新对应子PBM的子项头表H-Table中频繁项的支持度计数；S342、利用指针将子PBM中和大于最小支持度阈值的列与子H-Table中对应频繁项链接起来，增长为更长的局部频繁项集；S343、递归执行S341和S342，直到子PBM每列的和小于最小支持度阈值。
进一步地，所述在大数据分析处理框架Apache Spark上，基于矩阵运算的负载均衡策略，并行化实现全局计算平衡的模式增长类关联规则挖掘算法，包括：S41、启动Apache Spark，主节点读取稀疏事务数据集D，并将D水平切割成大小相等且连续的P个数据块，分别发送到P个从节点；S42、每个从节点遍历一次各自的数据块，计算所有项的支持度计数，并发送至主节点；S43、主节点比较所有项的支持度计数与最小支持度阈值，筛选出频繁项，生成F_List和H-Table，并将F_List和H-Table发送到P个从节点；S44、每个从节点根据F_List，再次遍历各自的数据块，按步骤S33转化存储为PBM，并统计PBM中首个在相同位置的“1”的行的个数，列号对应H-Table中的项号，形成项号、行数，发送到主节点；S45、主节点将相同项号的行数相加，根据基于矩阵计算的负载均衡策略进行分组，生成分组列表，记为G_List，发送到P个从节点；S46、从节点根据G-List在从节点之间交换PBM中的数据；S47、数据交换完成后，各从节点根据G-List和步骤S34，挖掘局部频繁项集；S48、从节点将局部频繁项集发送至主节点进行汇总，得到全部频繁项集，即稀疏事务数据集D的频繁项集。
进一步地，所述基于矩阵计算的负载均衡策略，包括：S451、主节点将行数相加后的项号、行数按行数递减顺序进行排序；S452、主节点根据F_List和H-Table间频繁项与项号一一对应的关系，将排序后的项号、行数转化为按此顺序排序的频繁项；S453、主节点按从两端开始的组合顺序依次组合频繁项，分为P组；S454、主节点依次扫描P组中的频繁项，生成分组列表G_List。
进一步地，所述采用模糊C均值聚类算法离散化历史运行数据，包括：使用字母加数字的形式标记离散化后的数据区间，并用每条数据所在区间的标号替换其数值。
一种火电机组汽轮机优化系统，包括：第一模块，用于以降低汽轮机热耗率为优化目标，采集历史运行数据并数据预处理；第二模块，用于基于典型相关性分析粗选与皮尔森相关性分析精选的方法，从历史运行数据中选取一组与汽轮机热耗率相关性符合设定条件的机组可控运行参数作为优化参数；第三模块，用于构建面向稀疏数据的模式增长类关联规则挖掘算法；第四模块，用于 在大数据分析处理框架Apache Spark上，基于矩阵运算的负载均衡策略，并行化实现全局计算平衡的模式增长类关联规则挖掘算法；第五模块，用于采用模糊C均值聚类算法离散化历史运行数据，基于并行化的模式增长类关联规则挖掘算法，挖掘离散化历史运行数据得到关联规则，并反离散化，得出各个边界条件下汽轮机优化参数的目标值。
与现有技术相比，本发明所达到的有益效果：本发明通过运用二进制矩阵和超链接技术设计一种新的模式增长类关联规则挖掘算法，可有效克服Apriori算法和FP-Growth算法挖掘稀疏数据频繁模式时，耗时长、内存占用过高的缺陷；同时，本发明还针对火电机组汽轮机全工况、长周期下，运行数据规模大的特点，基于以内存计算为主的大数据分析处理框架Apache Spark实现所设计的关联规则挖掘算法并行化，避免MapReduce频繁读写磁盘的缺点，减少大量I/O开销；并且，本发明针对二进制矩阵数据存储结构易于矩阵计算的特点，提出一种基于矩阵计算的负载均衡策略，较精确地分配集群计算节点的任务，充分发挥集群的性能优势，可高效挖掘汽轮机大规模运行数据以获取热耗率较低时的运行优化目标值。
附图说明
图1是本发明实施例提供的一种火电机组汽轮机优化方法的主要流程示意图；
图2是本发明实施例中设计的关联规则挖掘算法的流程图；
图3是本发明实施例中设计的关联规则挖掘算法在Apache Spark上的并行化流程图。
具体实施方式
下面结合附图对本发明作进一步描述。以下实施例仅用于更加清楚地说明本发明的技术方案，而不能以此来限制本发明的保护范围。
实施例一：
如图1～图3所示，一种火电机组汽轮机优化方法，包括：以降低汽轮机热耗率为优化目标，采集历史运行数据并数据预处理；基于典型相关性分析粗选与皮尔森相关性分析精选的方法，从历史运行数据中选取一组与汽轮机热耗率 相关性符合设定条件的机组可控运行参数作为优化参数；构建面向稀疏数据的模式增长类关联规则挖掘算法；在大数据分析处理框架Apache Spark上，基于矩阵运算的负载均衡策略，并行化实现全局计算平衡的模式增长类关联规则挖掘算法；采用模糊C均值聚类算法离散化历史运行数据，基于并行化的模式增长类关联规则挖掘算法，挖掘离散化历史运行数据得到关联规则，并反离散化，得出各个边界条件下汽轮机优化参数的目标值。
步骤一：以汽轮机热耗率较低为优化目标，采集历史运行数据并数据预处理；数据预处理，是指剔除历史运行数据中的异常数据和冗余数据以及稳态检测；稳态检测的判别标准是：在一定时间段内，当汽轮机的运行状态参数波动值小于设定范围时，可以认为机组处于稳定运行工况；
首先，在汽轮机组中，热耗率作为热经济性指标之一，全面地反映了机组热经济性，因此选择热耗率作为性能指标，并以降低汽轮机热耗率作为运行优化目标；接着，本实施例从某电厂SIS和DCS系统中采集1000MW火电机组汽轮机从2018年8月到到2019年7月的历史数据，数据范围为两次停机检修之间的一个完整运行周期，采样频率为60s，除去极少缺失值，共采集525600条历史运行数据；然后，针对数据明显异常地死点、坏点，直接剔除；针对一个参数对应多个测点所测得的多个正常数据，采用求平均值的方法得到一个能够全面、正确反映实际状态的数据；最后，用于稳态检测的参数时机组负荷和主蒸汽压力，判稳依据是20分钟内数据的最大值和最小值之差在某个稳定阈值内。
步骤二：针对汽轮机运行数据参数高维和强相关的特点，基于典型相关性分析粗选与皮尔森相关性分析精选的方法，选取一组与汽轮机热耗率较强相关的机组可控运行参数作为优化参数；
首先，对汽轮机可控变量的典型相关系数进行显著性检验，若某对可控变量的相关程度不显著，说明这对变量不具有代表性，舍弃这对变量；接着，找出相关程度显著且典型相关系数较高的几对可控变量，选取线性组合系数绝对值较大的变量作为候选运行优化参数；然后，运用皮尔森相关性分析对候选运行优化参数进一步缩减，缩减的原则是保留皮尔森相关系数大于0.8的；最后，确定汽轮机运行优化参数为：机组功率，主蒸汽流量、主蒸汽压力及温度、给 水泵出水压力、给水温度和凝汽器真空。
步骤三：针对稀疏数据在频繁模式挖掘时具有事务模式差异度大和项分布零散的特点，基于二进制矩阵和超链接技术，构建新型面向稀疏数据的模式增长类关联规则挖掘算法；包括：
S31、设定最小支持度阈值，遍历稀疏事务数据集，记为D，统计各项频数，生成频繁项列表，记为F_List；
S32、遍历F_List，对频繁项标号，生成项头表，记为H-Table，包括项号、支持度计数和链接指针；
S33、筛去D中的非频繁项，转化存储为二进制矩阵，记为PBM，其中为“1”的元素表示在某个事务中含有该元素对应F-List中的某个频繁项，为“0”的元素表示在某个事务中不含有该元素对应F-List中的某个频繁项；
S34、扫描PBM，调整H-Table中的指针，将PBM中每行首个“1”所在位置和频繁项项头表H-Table中对应频繁项链接，提取PBM中首个“1”在相同位置的行，生成多个子PBM，将挖掘全部频繁项集的任务转化为多个挖掘局部频繁项集的子任务；所述子任务包括以下步骤：
S341、扫描子PBM，对每列求和，更新对应子PBM的子项头表H-Table中频繁项的支持度计数；
S342、利用指针将子PBM中和大于最小支持度阈值的列与子H-Table中对应频繁项链接起来，增长为更长的局部频繁项集；
S343、递归执行S341和S342，直到子PBM每列的和小于最小支持度阈值；
S35、聚合局部频繁项集，输出全部频繁项集。
步骤四：针对串行关联规则挖掘算法受单机硬件资源限制无法挖掘大规模数据的问题，在大数据分析处理框架Apache Spark上，提出基于矩阵运算的负载均衡策略，并行化实现全局计算平衡的模式增长类关联规则挖掘算法；包括：
S41、启动Apache Spark，主节点读取稀疏事务数据集D，并将D水平切割成大小相等且连续的P个数据块，分别发送到P个从节点；
S42、每个从节点遍历一次各自的数据块，计算所有项的支持度计数，并发送至主节点；
S43、主节点比较所有项的支持度计数与最小支持度阈值，筛选出频繁项，生成F_List和H-Table，并将F_List和H-Table发送到P个从节点；
S44、每个从节点根据F_List，再次遍历各自的数据块，按步骤S33转化存储为PBM，并统计PBM中首个在相同位置的“1”的行的个数，列号对应H-Table中的项号，形成键值对(项号，行数)，发送到主节点；
S45、主节点将相同项号的行数相加，根据新的负载均衡策略进行分组，生成分组列表，记为G_List，发送到P个从节点，包括：
S451、主节点将行数相加后的键值对(项号，行数)按行数递减顺序进行排序；
S452、主节点根据F_List和H-Table间频繁项与项号一一对应的关系，将排序后的键值对(项号，行数)转化为按此顺序排序的频繁项；
S453、主节点按从两端开始的组合顺序依次组合频繁项，分为P组；
S454、主节点依次扫描P组中的频繁项，生成分组列表G_List；
S46、从节点根据G-List在从节点之间交换PBM中的数据；
S47、数据交换完成后，各从节点根据G-List和步骤S34，挖掘局部频繁项集；
S48、从节点将局部频繁项集发送至主节点进行汇总，得到全部频繁项集，即稀疏事务数据集D的频繁项集(或离散稳态历史运行数据的频繁模式)。
步骤五：采用模糊C均值聚类算法离散化历史运行数据，基于并行化的模式增长类关联规则挖掘算法，挖掘离散化历史运行数据的频繁模式，并反离散化，得出各个边界条件下汽轮机优化参数的目标值；模糊C均值(FCM)聚类算法可有效地分类特性比较复杂的对象，给出相对最优的分类结果，更符合客观实际。因此，运用模糊C均值(FCM)聚类算法分别离散化各个运行优化参数的稳态历史数据，使用字母+数字的形式标记离散化后的数据区间，并用每条数据所在区间的标号替换其数值。
经计算，得到边界条件一定时，各典型负荷下的强关联规则如表1所示：
表1强关联规则
运行参数及性能指标 500MW 600MW 700MW 800MW 900MW 1000MW
机组功率/MW [494.8,500.5] [602.5,623.1] [703.1,715.6] [798.2,802.1] [892.4,899.5] [980.0,933.5]
主蒸汽流量/(t/h) [1070.4,1271.5] [1225.6,1514.8] [1780.5,1805.8] [1715.6,2052.3] [2115.2,2331.5] [2432.8,2552.5]
主蒸汽压力/MPa [13.54,13.84] [17.35,18.57] [19.56,19.80] [21.49,21.76] [23.76,24.05] [24.41,25.34]
主蒸汽温度/℃ [593.6,596.4] [593.8,596.0] [595.4,597.9] [594.6,596.9] [594.6,595.8] [593.8,595.7]
给水泵出水压力/MPa [17.12,17.45] [20.32,22.34] [22.98,23.24] [25.48,26.12] [28.68,29.08] [29.62,30.68]
给水温度/℃ [248.7,255.8] [265.4,272.9] [275.5,276.7] [282.9,283.7] [289.6,291.1] [296.5,297.4]
凝汽器真空(kPa) [2.67,3.58] [3.48,3.75] [3.87,3.96] [3.86,4.19] [2.89,3.67] [5.3,6.4]
热耗率(kJ/kWh) [7135.1,7465.8] [6902.3,7266.8] [6820.7,7158.2] [6802.5,7116.1] [6785.5,7057.8] [6690.3,6816.8]
以该汽轮机组的挖掘结果为例，此强关联规则的意义是当调节运行参数的值处于表1中所示的区间范围时，就有不低于80％的概率使热耗率达到最佳值。那么，这些运行优化参数的目标值即为表中所示的区间，为使目标值更为直观，本发明选择区间的中心值作为优化目标值，得优化目标值如表2所示：
表2优化目标值
运行参数及性能指标 500MW 600MW 700MW 800MW 900MW 1000MW
机组功率/MW 497.65 612.8 709.35 800.15 895.95 956.75
主蒸汽流量/(t/h) 1170.95 1370.2 1793.15 1883.95 2223.35 2492.65
主蒸汽压力/MPa 13.69 17.96 19.68 21.63 23.91 24.88
主蒸汽温度/℃ 595 594.9 596.65 595.75 595.2 594.75
给水泵出水压力/MPa 17.29 21.33 23.11 25.8 28.88 30.15
给水温度/℃ 252.25 269.15 276.1 283.3 290.35 296.95
凝汽器真空(kPa) 3.13 3.62 3.92 4.03 3.28 5.85
热耗率(kJ/kW·h) 7300.45 7084.55 6989.45 6959.3 6921.65 6753.55
针对该机组，在不同负荷下，可按照表2中优化目标值对各参数进行调整，使热耗率达到最优值，进而使机组在最佳状况下运行。以该机组1000MW负荷时的关联规则为例，将热耗率的优化值与实际运行平均值进行比较，优化值是6753.55kJ/kW·h，较之实际运行平均值6780.35kJ/kW·h，热耗率减少了26.8kJ/kW·h。由标准煤耗率的计算公式：
b b＝q×10 3/(29308η gdη g)    (1)
其中，b b为标准煤耗率，g/kW·h；η gd为管道效率，本发明取值0.98；η g为锅炉效率，本发明取值0.92。可算得优化值与实际运行平均值的标准煤耗率相差1.01g/kW·h，即当该机组以1000MW左右的负荷运行时，若采用本发明所述的一种基于稀疏大数据的火电机组汽轮机运行优化方法，可每发1kW·h电节省1.01g的煤，节省经济开支的同时，也减少空气污染物的排放，能有效达到节能减排的目的。
关联规则挖掘算法是对离散数据分析处理，但火电机组汽轮机运行数据都 是连续的，连续数据离散化后不可避免地会导致模式差异度大且项分布零散；另外，汽轮机是一个多变量高维系统，高维数据离散化后稀疏特性更为明显，本实施例通过运用二进制矩阵和超链接技术设计一种新的模式增长类关联规则挖掘算法，可有效克服Apriori算法和FP-Growth算法挖掘稀疏数据频繁模式时，耗时长、内存占用过高的缺陷；同时，本发明还针对火电机组汽轮机全工况、长周期下，运行数据规模大的特点，基于以内存计算为主的大数据分析处理框架Apache Spark实现所设计的关联规则挖掘算法并行化，避免MapReduce频繁读写磁盘的缺点，减少大量I/O开销；并且，本发明针对二进制矩阵数据存储结构易于矩阵计算的特点，提出一种新的负载均衡策略，较精确地分配集群计算节点的任务，充分发挥集群的性能优势，可高效挖掘汽轮机大规模运行数据以获取热耗率较低时的运行优化目标值。
实施例二：
基于实施例一所述的火电机组汽轮机优化方法，本实施例提供一种火电机组汽轮机优化系统，包括：第一模块，用于以降低汽轮机热耗率为优化目标，采集历史运行数据并数据预处理；第二模块，用于基于典型相关性分析粗选与皮尔森相关性分析精选的方法，从历史运行数据中选取一组与汽轮机热耗率相关性符合设定条件的机组可控运行参数作为优化参数；第三模块，用于构建面向稀疏数据的模式增长类关联规则挖掘算法；第四模块，用于在大数据分析处理框架Apache Spark上，基于矩阵运算的负载均衡策略，并行化实现全局计算平衡的模式增长类关联规则挖掘算法；第五模块，用于采用模糊C均值聚类算法离散化历史运行数据，基于并行化的模式增长类关联规则挖掘算法，挖掘离散化历史运行数据得到关联规则，并反离散化，得出各个边界条件下汽轮机优化参数的目标值。
以上所述仅是本发明的优选实施方式，应当指出，对于本技术领域的普通技术人员来说，在不脱离本发明技术原理的前提下，还可以做出若干改进和变形，这些改进和变形也应视为本发明的保护范围。

Claims (9)

  1. 一种火电机组汽轮机优化方法，其特征是，包括：
    以降低汽轮机热耗率为优化目标，采集历史运行数据并数据预处理；
    基于典型相关性分析粗选与皮尔森相关性分析精选的方法，从历史运行数据中选取一组与汽轮机热耗率相关性符合设定条件的机组可控运行参数作为优化参数；
    构建面向稀疏数据的模式增长类关联规则挖掘算法；
    在大数据分析处理框架Apache Spark上，基于矩阵运算的负载均衡策略，并行化实现全局计算平衡的模式增长类关联规则挖掘算法；
    采用模糊C均值聚类算法离散化历史运行数据，基于并行化的模式增长类关联规则挖掘算法，挖掘离散化历史运行数据得到关联规则，并反离散化，得出各个边界条件下汽轮机优化参数的目标值。
  2. 根据权利要求1所述的火电机组汽轮机优化方法，其特征是，所述数据预处理，是指剔除历史运行数据中的异常数据和冗余数据并对历史运行数据进行稳态检测。
  3. 根据权利要求1所述的火电机组汽轮机优化方法，其特征是，所述稳态检测的判别标准是：在一定时间段内，当汽轮机的运行状态参数波动值小于设定范围时，可以认为机组处于稳定运行工况。
  4. 根据权利要求1所述的火电机组汽轮机优化方法，其特征是，所述构建面向稀疏数据的模式增长类关联规则挖掘算法，包括：
    S31、设定最小支持度阈值，遍历稀疏事务数据集，记为D，统计各项频数，生成频繁项列表，记为F_List；
    S32、遍历F_List，对频繁项标号，生成项头表，记为H-Table，包括项号、支持度计数和链接指针；
    S33、筛去D中的非频繁项，转化存储为二进制矩阵，记为PBM，其中为“1”的元素表示在某个事务中含有该元素对应F-List中的某个频繁项，为“0”的元素表示在某个事务中不含有该元素对应F-List中的某个频繁项；
    S34、扫描PBM，调整H-Table中的指针，将PBM中每行首个“1”所在位 置和频繁项项头表H-Table中对应频繁项链接，提取PBM中首个“1”在相同位置的行，生成多个子PBM，将挖掘全部频繁项集的任务转化为多个挖掘局部频繁项集的子任务；
    S35、聚合局部频繁项集，输出全部频繁项集。
  5. 根据权利要求4所述的火电机组汽轮机优化方法，其特征是，所述子任务包括以下步骤：
    S341、扫描子PBM，对每列求和，更新对应子PBM的子项头表H-Table中频繁项的支持度计数；
    S342、利用指针将子PBM中和大于最小支持度阈值的列与子H-Table中对应频繁项链接起来，增长为更长的局部频繁项集；
    S343、递归执行S341和S342，直到子PBM每列的和小于最小支持度阈值。
  6. 根据权利要求1所述的火电机组汽轮机优化方法，其特征是，所述在大数据分析处理框架Apache Spark上，基于矩阵运算的负载均衡策略，并行化实现全局计算平衡的模式增长类关联规则挖掘算法，包括：
    S41、启动Apache Spark，主节点读取稀疏事务数据集D，并将D水平切割成大小相等且连续的P个数据块，分别发送到P个从节点；
    S42、每个从节点遍历一次各自的数据块，计算所有项的支持度计数，并发送至主节点；
    S43、主节点比较所有项的支持度计数与最小支持度阈值，筛选出频繁项，生成F_List和H-Table，并将F_List和H-Table发送到P个从节点；
    S44、每个从节点根据F_List，再次遍历各自的数据块，按步骤S33转化存储为PBM，并统计PBM中首个在相同位置的“1”的行的个数，列号对应H-Table中的项号，形成项号、行数，发送到主节点；
    S45、主节点将相同项号的行数相加，根据基于矩阵计算的负载均衡策略进行分组，生成分组列表，记为G_List，发送到P个从节点；
    S46、从节点根据G-List在从节点之间交换PBM中的数据；
    S47、数据交换完成后，各从节点根据G-List和步骤S34，挖掘局部频繁项集；
    S48、从节点将局部频繁项集发送至主节点进行汇总，得到全部频繁项集，即稀疏事务数据集D的频繁项集。
  7. 根据权利要求6所述的火电机组汽轮机优化方法，其特征是，所述基于矩阵计算的负载均衡策略，包括：
    S451、主节点将行数相加后的项号、行数按行数递减顺序进行排序；
    S452、主节点根据F_List和H-Table间频繁项与项号一一对应的关系，将排序后的项号、行数转化为按此顺序排序的频繁项；
    S453、主节点按从两端开始的组合顺序依次组合频繁项，分为P组；
    S454、主节点依次扫描P组中的频繁项，生成分组列表G_List。
  8. 根据权利要求1所述的火电机组汽轮机优化方法，其特征是，所述采用模糊C均值聚类算法离散化历史运行数据，包括：使用字母加数字的形式标记离散化后的数据区间，并用每条数据所在区间的标号替换其数值。
  9. 一种火电机组汽轮机优化系统，其特征是，包括：
    第一模块，用于以降低汽轮机热耗率为优化目标，采集历史运行数据并数据预处理；
    第二模块，用于基于典型相关性分析粗选与皮尔森相关性分析精选的方法，从历史运行数据中选取一组与汽轮机热耗率相关性符合设定条件的机组可控运行参数作为优化参数；
    第三模块，用于构建面向稀疏数据的模式增长类关联规则挖掘算法；
    第四模块，用于在大数据分析处理框架Apache Spark上，基于矩阵运算的负载均衡策略，并行化实现全局计算平衡的模式增长类关联规则挖掘算法；
    第五模块，用于采用模糊C均值聚类算法离散化历史运行数据，基于并行化的模式增长类关联规则挖掘算法，挖掘离散化历史运行数据得到关联规则，并反离散化，得出各个边界条件下汽轮机优化参数的目标值。
PCT/CN2021/116827 2021-03-15 2021-09-07 基于稀疏大数据挖掘的火电机组汽轮机优化方法及系统 WO2022193569A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US18/548,447 US20240231295A9 (en) 2021-03-15 2021-09-07 Optimization method and system for turbines of thermal power unit based on sparse big data mining

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110276819.0A CN112861379B (zh) 2021-03-15 2021-03-15 基于稀疏大数据挖掘的火电机组汽轮机优化方法及系统
CN202110276819.0 2021-03-15

Publications (1)

Publication Number Publication Date
WO2022193569A1 true WO2022193569A1 (zh) 2022-09-22

Family

ID=75994426

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/116827 WO2022193569A1 (zh) 2021-03-15 2021-09-07 基于稀疏大数据挖掘的火电机组汽轮机优化方法及系统

Country Status (3)

Country Link
US (1) US20240231295A9 (zh)
CN (1) CN112861379B (zh)
WO (1) WO2022193569A1 (zh)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115374384A (zh) * 2022-10-24 2022-11-22 北京长河数智科技有限责任公司 基于大数据分析的信息发布方法及装置
CN116304931A (zh) * 2023-05-12 2023-06-23 山东英伟电子技术有限公司 一种基于大数据的电力数据挖掘方法
CN116680090A (zh) * 2023-08-03 2023-09-01 睿至科技集团有限公司 一种基于大数据的边缘计算网络管理方法及平台
CN117032996A (zh) * 2023-10-09 2023-11-10 湖南中青能科技有限公司 一种电力元数据管理方法及系统
CN117724005A (zh) * 2024-02-05 2024-03-19 东莞市晟鼎精密仪器有限公司 一种中频宽幅等离子电源智能监控系统及方法

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112861379B (zh) * 2021-03-15 2022-08-12 南京邮电大学 基于稀疏大数据挖掘的火电机组汽轮机优化方法及系统
CN113670536B (zh) * 2021-07-06 2024-03-05 浙江浙能台州第二发电有限责任公司 火电厂用电用水监测和信息化管理方法
CN118365215B (zh) * 2024-06-14 2024-10-01 张家港大裕橡胶制品有限公司 基于数据挖掘的橡胶手套工艺参数优化方法及系统
CN118381408B (zh) * 2024-06-25 2024-08-27 自贡九天水利机械有限公司 基于同步驱动的启闭机智能变频控制方法及系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103401256A (zh) * 2013-06-09 2013-11-20 广东电网公司电力科学研究院 火力发电厂大型汽轮机功频电液控制系统参数的优化方法
CN103440528A (zh) * 2013-08-12 2013-12-11 国家电网公司 基于耗差分析的火电机组运行优化方法及装置
CN110442038A (zh) * 2019-07-25 2019-11-12 南京邮电大学 基于FP-Growth算法的火电机组运行优化目标值确定方法
WO2020240592A2 (en) * 2019-05-29 2020-12-03 Tata Consultancy Services Limited Method and system for optimum coal selection and power plant optimization
CN112861379A (zh) * 2021-03-15 2021-05-28 南京邮电大学 基于稀疏大数据挖掘的火电机组汽轮机优化方法及系统

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102636991A (zh) * 2012-04-18 2012-08-15 国电科学技术研究院 一种基于模糊集关联规则的火电机组运行参数优化方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103401256A (zh) * 2013-06-09 2013-11-20 广东电网公司电力科学研究院 火力发电厂大型汽轮机功频电液控制系统参数的优化方法
CN103440528A (zh) * 2013-08-12 2013-12-11 国家电网公司 基于耗差分析的火电机组运行优化方法及装置
WO2020240592A2 (en) * 2019-05-29 2020-12-03 Tata Consultancy Services Limited Method and system for optimum coal selection and power plant optimization
CN110442038A (zh) * 2019-07-25 2019-11-12 南京邮电大学 基于FP-Growth算法的火电机组运行优化目标值确定方法
CN112861379A (zh) * 2021-03-15 2021-05-28 南京邮电大学 基于稀疏大数据挖掘的火电机组汽轮机优化方法及系统

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115374384A (zh) * 2022-10-24 2022-11-22 北京长河数智科技有限责任公司 基于大数据分析的信息发布方法及装置
CN115374384B (zh) * 2022-10-24 2022-12-16 北京长河数智科技有限责任公司 基于大数据分析的信息发布方法及装置
CN116304931A (zh) * 2023-05-12 2023-06-23 山东英伟电子技术有限公司 一种基于大数据的电力数据挖掘方法
CN116304931B (zh) * 2023-05-12 2023-08-04 山东英伟电子技术有限公司 一种基于大数据的电力数据挖掘方法
CN116680090A (zh) * 2023-08-03 2023-09-01 睿至科技集团有限公司 一种基于大数据的边缘计算网络管理方法及平台
CN116680090B (zh) * 2023-08-03 2023-12-19 睿至科技集团有限公司 一种基于大数据的边缘计算网络管理方法及平台
CN117032996A (zh) * 2023-10-09 2023-11-10 湖南中青能科技有限公司 一种电力元数据管理方法及系统
CN117032996B (zh) * 2023-10-09 2023-12-22 湖南中青能科技有限公司 一种电力元数据管理方法及系统
CN117724005A (zh) * 2024-02-05 2024-03-19 东莞市晟鼎精密仪器有限公司 一种中频宽幅等离子电源智能监控系统及方法
CN117724005B (zh) * 2024-02-05 2024-05-31 东莞市晟鼎精密仪器有限公司 一种中频宽幅等离子电源智能监控系统及方法

Also Published As

Publication number Publication date
CN112861379B (zh) 2022-08-12
US20240134331A1 (en) 2024-04-25
US20240231295A9 (en) 2024-07-11
CN112861379A (zh) 2021-05-28

Similar Documents

Publication Publication Date Title
WO2022193569A1 (zh) 基于稀疏大数据挖掘的火电机组汽轮机优化方法及系统
CN110442038B (zh) 基于FP-Growth算法的火电机组运行优化目标值确定方法
CN110989360A (zh) 一种基于全量数据的火电机组稳态历史寻优方法
CN103742901B (zh) 一种循环流化床机组耗差分析优化目标值体系的确定方法
CN101893877A (zh) 一种发电厂基于能耗分析的优化运行方法及其系统
CN105701596A (zh) 一种基于大数据技术的配网抢修精益化方法以及管理系统
CN103440528A (zh) 基于耗差分析的火电机组运行优化方法及装置
CN102708180A (zh) 基于实时历史库的机组运行模式数据挖掘方法
CN109872012A (zh) 基于工况划分的火电厂运行多目标优化的确定方法
Xu et al. Data based online operational performance optimization with varying work conditions for steam-turbine system
CN101187803B (zh) 基于数据挖掘技术的氨合成装置生产优化方法
CN113515049A (zh) 一种燃气蒸汽联合循环发电机组运行调控系统及调控方法
CN104131847B (zh) 汽轮机组定滑压自动优化控制系统及其自动优化方法
Ma et al. Economic operation evaluation of active distribution network based on fuzzy Borda method
CN109469919A (zh) 一种基于权重聚类的电站空预器堵灰监测方法
CN113868836B (zh) 基于大数据的智慧热力系统在线专家分析平台
CN115952900A (zh) 基于ai分析的智慧能源运行优化方法
Li et al. Research and application of data mining in power plant process control and optimization
Li et al. Research and application of data mining technique in power plant
CN112200447A (zh) 一种基于大数据的电厂前馈控制优化方法
Song et al. Big data mining method of thermal power based on spark and optimization guidance
CN118409501A (zh) 一种基于工况寻优的燃煤机组智能决策运行控制方法及系统
Wen et al. Economic mining of thermal power plant based on improved Hadoop-based framework and Spark-based algorithms
Li et al. Real-time knowledge discovery for process objects
CN109241573B (zh) 一种汽轮机末级叶片选型方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21931144

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 18548447

Country of ref document: US

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21931144

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 21931144

Country of ref document: EP

Kind code of ref document: A1