CN113988390A - Unit maintenance optimization method based on support vector machine - Google Patents
Unit maintenance optimization method based on support vector machine Download PDFInfo
- Publication number
- CN113988390A CN113988390A CN202111210633.1A CN202111210633A CN113988390A CN 113988390 A CN113988390 A CN 113988390A CN 202111210633 A CN202111210633 A CN 202111210633A CN 113988390 A CN113988390 A CN 113988390A
- Authority
- CN
- China
- Prior art keywords
- unit
- overhaul
- branch
- constraint
- support vector
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
- 238000012423 maintenance Methods 0.000 title claims abstract description 64
- 238000005457 optimization Methods 0.000 title claims abstract description 52
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000012706 support-vector machine Methods 0.000 title claims abstract description 39
- 238000004364 calculation method Methods 0.000 claims abstract description 17
- 230000008569 process Effects 0.000 claims abstract description 14
- 230000006870 function Effects 0.000 claims description 58
- 238000012549 training Methods 0.000 claims description 19
- 238000010801 machine learning Methods 0.000 claims description 12
- 230000007717 exclusion Effects 0.000 claims description 10
- 239000000126 substance Substances 0.000 claims description 7
- 230000005540 biological transmission Effects 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 6
- 238000012546 transfer Methods 0.000 claims description 6
- 238000010586 diagram Methods 0.000 claims description 5
- 230000017105 transposition Effects 0.000 claims description 2
- 238000007689 inspection Methods 0.000 description 12
- 230000000694 effects Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000010248 power generation Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/20—Administration of product repair or maintenance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/04—Constraint-based CAD
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/10—Numerical modelling
-
- 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
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Evolutionary Computation (AREA)
- Tourism & Hospitality (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- General Business, Economics & Management (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Marketing (AREA)
- Artificial Intelligence (AREA)
- Entrepreneurship & Innovation (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Mathematical Physics (AREA)
- Evolutionary Biology (AREA)
- Game Theory and Decision Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Development Economics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a unit maintenance optimization method based on a support vector machine, which comprises the following steps of establishing a unit maintenance plan safe reliability index based on system standby; constructing a target function of a unit maintenance model based on the safety and reliability indexes; constructing a unit overhaul optimization model based on a target function and constraint conditions of the unit overhaul model; constructing a branch-and-bound calculation framework of an improved branch strategy based on a support vector machine; the method for optimizing the unit overhaul based on the support vector machine accelerates the solving process of the integer programming model with numerous integer variables and complex constraint conditions, and optimizes the execution of the unit overhaul scheme on the basis of ensuring the safety and the reliability of the unit overhaul plan, thereby ensuring the safe and stable operation of the power system.
Description
Technical Field
The invention relates to a unit overhaul optimization method, in particular to a unit overhaul optimization method based on a support vector machine.
Background
The orchestration of the unit maintenance schedule has a significant impact on the operation of the power system. For example, if the dispatch center schedules an unreasonable crew service plan, it may risk a power shortage, especially for those power systems where spare capacity is not sufficient. The arrangement of the maintenance plan of the power system unit is directly related to the reserve capacity of the power system, and can influence the composition of medium and short term power grid units, and further influence the arrangement of medium and short term power generation plans of the power system, such as medium and short term unit combination, economic dispatching and the like.
Most of traditional unit maintenance scenes are manually established, system operation constraint conditions such as system standby constraint and supply and demand constraint are not considered, concentrated unit maintenance in certain time periods can be caused, and the system standby capacity is insufficient; the method is characterized in that a unit maintenance optimization model is compiled by part of scheduling centers based on the safety and reliability requirements of the power system, a mature commercial solver is mostly called for solving, the problems of calculation efficiency of a unit maintenance plan optimization algorithm and nonlinearity of unit maintenance safety and reliability indexes are rarely considered, and the method is not suitable for the maintenance optimization problem of the large-scale power system. Therefore, a solution algorithm capable of dealing with the unit maintenance plan problems with numerous decision variables and complex constraint conditions needs to be designed.
Disclosure of Invention
In order to solve the problems, the invention provides a unit maintenance optimization method based on a support vector machine, which can efficiently and accurately solve a unit maintenance optimization model through a scientific method to realize the optimization of a unit maintenance scheme.
The technical scheme adopted by the invention for overcoming the technical problems is as follows: the invention provides a unit maintenance optimization method based on a support vector machine, which comprises the following steps: s1, establishing a unit maintenance plan safe reliability index based on the system standby; s2, constructing a target function of the unit maintenance model based on the safe reliability index; s3, constructing a unit overhaul optimization model based on the objective function and constraint conditions of the unit overhaul model; s4, constructing a branch and bound calculation framework of the improved branch strategy based on the support vector machine; and S5, calculating the constructed unit maintenance optimization model based on the branch-and-bound calculation framework to obtain a unit maintenance plan.
Furthermore, the safety and reliability index formula is as follows,
wherein T is the total time period, StFor system standby for the period t, the formula is as follows,
MTithe sum of the overhaul times of the unit i in the T period,
is the overhaul state of the overhaul unit i in the jth overhaul time period t,
is an integer variable from 0 to 1, if
Representing the unit i is in maintenance state if
The unit i is in a non-overhaul state; NG is the total number of the units; gi，maxThe upper limit of the output of the unit is set; dtThe system load for the time period t.
The system standby is to minimize the variance of the standby or standby rate of the power system in all time periods so as to achieve the effect of waiting for standby or waiting for standby rate. As can be seen from equation (1), the safety and reliability index is determined by the amount of system sparing, and the safety and reliability index ensures that the system sparing in adjacent time periods is as close as possible, so as to achieve the purpose of sparing in each time period. System standby
Further, step S2 specifically includes: s21, taking the maximum safe reliability index as the objective function of the unit maintenance model, wherein the objective function of the unit maintenance model is as follows,
the absolute value of the system standby difference between the t time period and the t-1 time period;
s22, linearizing the formula (3), the expression of the linearization is as follows,
wherein S istFor system standby during time period t, St-1For the system standby during the t-1 period.
Furthermore, the constraint conditions at least comprise a unit inspection time constraint, a unit inspection continuity constraint, a unit inspection plan non-overlapping constraint, a unit inspection mutual exclusion constraint, a unit inspection simultaneous constraint, a system standby constraint, a power balance constraint, a unit output upper and lower limit constraint, an inspection unit output upper and lower limit constraint, a line power flow constraint and a variable integer constraint.
Further, the model constructed based on the unit overhaul time constraint is as follows,
wherein the content of the first and second substances,
the time required for the jth overhaul of the unit i;
the model constructed based on the crew overhaul time constraints is as follows,
wherein the content of the first and second substances,
a model constructed based on the unit overhaul plan non-overlapping constraints is as follows,
the model constructed based on the unit overhaul mutual exclusion constraint is as follows,
wherein phi iseThe method comprises the steps of (1) setting a mutual exclusion overhaul unit set;
the model constructed based on the crew simultaneous constraints is as follows,
wherein，ΦsFor servicing sets of units simultaneously, Ns，tOverhauling the upper limit of the number of units for the set s in the time period t;
the model built based on the system standby constraints is as follows,
St≥DtR，t＝1，2，...，T (11)
wherein R is the system standby rate, StFor system standby during time t, DtSystem load for time period t;
a model constructed based on the electronic balance constraints is as follows,
wherein, Pi，tThe active power output of the unit i in the time period t is obtained;
the model constructed based on the upper and lower unit output limit constraints is as follows,
wherein, i 1m；t＝1，2，...，T，Gi，minLower limit of unit output, Gi，maxThe upper limit of the output of the unit is set;
the model built based on the line flow constraints is as follows,
wherein, l 1., NL; t1, 2, T, NN is the total number of load nodes in the system; NL is the total number of lines in the network topology; pl，maxUpper limit for transmission power of line l; pl，minThe lower limit of the transmission power for line l; dn，tA node load predicted value of a node n in a power grid in a time period t is obtained; gl-iA node-line power transfer distribution factor of the unit i to the line l; gl-nA node-line power transfer distribution factor for load n pairs of lines l;
the model constructed based on the variable integer constraint is as follows,
the unit maintenance model based on safety composed of formulas (3) to (14) is ProbR-MS。
Further, step S4 specifically includes: s41, simulating a branch process of a candidate branch variable set C of theta nodes based on a strong branch strategy to obtain a training data set based on candidate variable characteristics and candidate variable labels of the former theta nodes; s42, inputting the training data into a machine learning algorithm of a support vector machine, and fitting to obtain a branch strategy function f by taking a loss function of a training data set as a minimum target; and S43, branching the nodes behind the theta nodes based on the branch strategy function.
Further, step S42 specifically includes: based on the machine learning algorithm of the support vector machine, a following branch strategy function f is fitted by taking a loss function of training data as a minimum target,
f[φ(xj，Ni)]＝wTφ(xj，Ni) (16)
wherein, wTFor the transposition of the parameter vector w, the training data comprises at least a set of nodes of the search solution
A group of existing nodes
Candidate variable set of
Labels of candidate variables at each node i
Characteristic diagram phi:
Ω is the label yiThe domain of (a) is selected,
φ(xj，Ni) Refers to a branch variable xjAt node NiThere are p features;
text setting
For each one
The variable passes through a function f [ phi (x)j，Ni)]Calculated result, loss function
According to a pair variable at a node Ni
The penalty resulting from sorting is taken as the regularization parameter.
The candidate variable characteristics mainly refer to the current state and the historical performance of the variable in the solution of the current node linear programming problem, and comprise the coefficient of the variable in an objective function, the number of involved constraint conditions, the statistic of the coefficients involved in different constraints, including the number, the maximum value, the minimum value, the average value, the standard deviation and the like, the number of the variables involved in the constraint of the current linear programming problem and the like.
The variable in the loss function is a parameter vector w, and the parameter vector w is solved through the minimization of the loss function, so that the w required by the branch strategy function f is obtainedT。
Further, step S43 specifically includes: s431, after the branch-and-bound strong branch strategy carries out theta nodes, switching the variable branch strategy from the strong branch strategy to a branch function strategy f based on a support vector machine; s432, for each node
Each variable
Computing a feature vector phi (x)j，Ni) And a branch strategy calculation score is obtained as follows,
wherein S isj＝f[φ(xj，Ni)]。
Further, step S5 specifically includes: s51, obtaining a target function and constraint conditions based on the unit maintenance optimization model; s52, presetting the number theta of branch nodes of the strong branch strategy and the upper limit sigma of the error of the unit overhaul optimization model; s53, the former theta nodes adopt a strong branch strategy to branch, the score calculated by each node is used as a label of a candidate variable, and the candidate variable extraction characteristics are obtained based on the target function and the constraint condition; s54, inputting the labels of the candidate variables and the extraction characteristics of the candidate variables into a machine learning algorithm of a support vector machine as a training data set, and fitting to obtain a branch strategy function f; s55, branching all the variables of the nodes in the branching process based on the branch strategy function; and S56, if the obtained optimal solution is within the error upper limit sigma range, solving the unit overhaul optimization model to obtain a unit overhaul plan, otherwise, repeating the steps S53-S56 by taking the calculated optimal solution as an initial solution of the unit overhaul optimization model until the obtained optimal solution is within the error upper limit sigma range.
The invention has the beneficial effects that:
the branch-and-bound solving method is optimized through the method of the support vector machine, the solving efficiency of the integer programming algorithm is improved, and therefore the safety, reliability and economy of the unit maintenance plan decision mode are effectively improved.
Drawings
FIG. 1 is a flow chart of an embodiment of the present invention;
FIG. 2 is a block diagram of a safety and reliability-based unit maintenance schedule according to an embodiment of the present invention;
FIG. 3 is a network topology diagram of an IEEE118 node according to an embodiment of the present invention;
FIG. 4 is basic information of a maintenance unit for performing maintenance optimization calculation according to an embodiment of the present invention;
FIG. 5 is a graph illustrating an exemplary annual load of an IEEE118 node in accordance with an embodiment of the present invention;
fig. 6 is a calculation result of the continuous overhaul optimization algorithm in the embodiment of the present invention.
Detailed Description
In order to facilitate a better understanding of the invention for those skilled in the art, the invention will be described in further detail with reference to the accompanying drawings and specific examples, which are given by way of illustration only and do not limit the scope of the invention.
As shown in fig. 1, the flowchart of the unit overhaul optimization method based on the support vector machine according to this embodiment at least includes S1, and the safety and reliability index of the unit overhaul plan is established based on the system backup; s2, constructing a target function of the unit maintenance model based on the safe reliability index; s3, constructing a unit overhaul optimization model based on the objective function and constraint conditions of the unit overhaul model; s4, constructing a branch and bound calculation framework of the improved branch strategy based on the support vector machine; and S5, calculating the constructed unit maintenance optimization model based on the branch-and-bound calculation framework to obtain a unit maintenance plan.
Embodiments of the invention are based on the ubntun18.04 system, Cplex12.6 integer programming solver, and pytorch1.1.0 depth scienceThe method is completed in an experimental system for setting up a software environment of a learning environment, and an upper error limit sigma is set to 10-5The experimental verification is carried out on the unit maintenance optimization method based on the support vector machine. According to the unit maintenance plan model based on the safety and reliability, the scheduling center schedules and obtains a unit maintenance plan based on the safety and reliability, wherein fig. 2 shows a specific maintenance time period of each maintenance unit. And performing 14-thermoelectricity overhaul plan optimization calculation on the power grid topology structure of the 118 node shown in the figure 3. Fig. 4 is basic information of the service unit for performing service optimization calculation.
And S1, establishing a unit maintenance plan safe reliability index based on the system standby.
In the embodiment of the invention, the variance of the standby or standby rate of the power system in all time periods is minimized to achieve the effect of waiting for the standby or standby rate, and the safety and reliability indexes of the unit maintenance plan are established as follows:
wherein T is the total time period, StFor the system standby for the period t. The index ensures that the system standby in the adjacent time period is as close as possible, so as to achieve the purpose of standby in each time period and the like.
the system standby for time period t is:
wherein, MTiThe sum of the overhaul times of the unit i in the T time period is obtained;
is the overhaul state of the overhaul unit i in the jth overhaul time period t,
is an integer variable from 0 to 1, and is,
representing that the unit i is in a maintenance state, otherwise, the unit i is in a non-maintenance state; NG is the total number of the units; gi，maxThe upper limit of the output of the unit is set; dtThe system load for the time period t.
And S2, constructing a target function of the unit overhaul model based on the safe reliability index.
In one embodiment of the present invention, the maximum safe reliability index is taken as the objective function of the unit maintenance model, and the objective function of the unit maintenance model is determined by the following formula,
wherein the content of the first and second substances,
system standby difference S for time period t and time period t-1t-St-1Absolute value of (a).
Equation (21) is a non-linear form that can be linearized as:
and S3, constructing a unit overhaul optimization model based on the objective function and the constraint condition of the unit overhaul model.
The constraint conditions include but are not limited to a unit inspection time constraint, a unit inspection continuity constraint, a unit inspection plan non-overlapping constraint, a unit inspection mutual exclusion constraint, a unit inspection simultaneous constraint, a system standby constraint, a power balance constraint, a unit output upper and lower limit constraint and a variable integer constraint to establish a unit inspection optimization model. Each constraint is explained in detail below. The IEEE118 node arithmetic annual load curve shown in fig. 5 is a constraint.
(1) Time constraints for unit maintenance
Because each unit may need to be overhauled more than once in the period of T, the jth overhaul of overhauled unit i:
wherein the content of the first and second substances,
the time required for the jth overhaul of the unit i.
(2) Unit maintenance continuity constraint
Once the unit starts to be overhauled, the unit cannot be stopped before the overhaul is finished, namely the overhaul continuity constraint must be met:
wherein the content of the first and second substances,
this means that the variable is 0 outside the service interval.
(3) Non-overlapping constraint of unit maintenance plan
Obviously, for the same overhaul unit i, different times of overhaul plans cannot be overlapped:
(4) unit overhaul mutual exclusion constraint
Power plants typically require that their affiliated units cannot be serviced at the same time; in addition, the dispatching center can also set some specific sets, and the machine sets in the sets cannot be overhauled simultaneously. Therefore, these units should satisfy the overhaul exclusion constraint:
wherein phieThe method comprises the steps of (1) setting a mutual exclusion overhaul unit set;
(5) unit maintenance simultaneous restraint
Power generation companies will typically require that the number of service units in the same time period not exceed their specified upper limit; in addition, the power grid dispatching center also sets certain sets, and the number of the overhaul units in the same time period set cannot exceed the specified upper limit of the overhaul units. Therefore, these units should meet the service and maintenance constraints:
wherein phisFor servicing sets of units simultaneously, Ns，tAnd overhauling the upper limit of the number of the units for the set s in the period t.
(6) System backup constraints
St≥DtR，t＝1，2，...，T (30)
Wherein R is the system standby rate, StFor system standby during time t, DtThe system load for the time period t.
(7) Power balance constraint
Wherein, Pi，tAnd the active output of the unit i in the time period t.
(8) Upper and lower limit restraint of unit output
The unit combination optimization is not considered, that is, all units are assumed to be in a starting state in a non-overhaul time period, so that the upper and lower output limits of the overhaul unit are constrained as shown in the following formula:
wherein i is 1..，NGm；t＝1，2，...，T，Gi，minLower limit of unit output, Gi，maxThe upper limit of the output of the unit is set;
(9) constraint conditions such as line power flow constraint and the like
Wherein, l 1., NL; t1, 2, T, NN is the total number of load nodes in the system; NL is the total number of lines in the network topology; pl，maxUpper limit for transmission power of line l; pl，minThe lower limit of the transmission power for line l; dn，tA node load predicted value of a node n in a power grid in a time period t is obtained; gl-iA node-line power transfer distribution factor of the unit i to the line l; gl-nA node-to-line power transfer distribution factor for load n versus line l.
(10) Variable integer part constraint
the unit maintenance model based on safety composed of formulas (22) to (34) is ProbR-MS。
S4, constructing a branch and bound computation framework of the improved branch strategy based on the support vector machine specifically comprises the following steps:
and S41, simulating the branch process of the candidate branch variable set C of the theta nodes based on the strong branch strategy to obtain a training data set based on the candidate variable characteristics and the candidate variable labels of the previous theta nodes.
In the integer programming problem solving process, for the limited number of branch nodes theta, a strong branch strategy is adopted as a branch strategy.
Branch morphing in the usual process of delimiting branchesQuantity xjThe metric for branch strategy to improve quality is the amount of improvement to the dual boundary of the objective function. Considering the branch variable xjPost-relaxation linear programming objective function values
Post-relaxation linear programming solution
Candidate branch variable set C. After node N branches, two subproblems
And
the corresponding linear programming objective function values are respectively
And
then the corresponding change in the objective function value is respectively
And
the strong branch strategy scoring mechanism is calculated as follows:
wherein, SBjIs a branch variable xjThe score obtained by the strong branch strategy, score (a, b) refers to the product of a and b, and epsilon is a small constant and takes the value of 10-6. The strong branch strategy is used for finding the variable with the highest score for branching by simulating the branch process in the candidate branch variable set C and calculating the score according to the above formula.
The training data mainly consisted of:
a. set of nodes for search solution
b. Set of candidate variables
For the existing node
c. Label (R)
Is for the candidate variable at each node i, where Ω is the label yiThe domain of (2). d. A characteristic diagram phi:
wherein phi (x)j，Ni) Finger variable xjAt node NiThere are p features.
Wherein the label is assigned to
The value of each variable in (a). We use a simple binary tagging mechanism, as follows:
where α ∈ [0, 1] is the 1 taken in the strong branching strategy.
In the process of branching and delimiting, each node takes the score calculated by the strong branching strategy as a label of a candidate variable and analyzes the corresponding variable characteristic.
And S42, inputting the training data into a machine learning algorithm of the support vector machine, and fitting to obtain a branch strategy function f by taking a loss function of the training data set as a minimum target.
The method comprises a model learning stage, wherein training data in a data acquisition stage is input into a machine learning algorithm based on a support vector machine to train a loss function of a data set
And fitting a branch variable ordering function simulating a strong branch strategy by taking the minimum as a target. The training data set is based on candidate variable characteristics and candidate variable labels of nodes before the strong branch strategy stage.
Based on the training data, we fit a linear branch strategy function f,
f[φ(xj，Ni)]＝wTφ(xj，Ni) (37)
definition of
For each one
The variable passes through a function f [ phi (x)j，Ni)]The calculated result is:
the loss function is then the following equation:
wherein the loss function
Is based on at node Ni
The penalty of ordering the variables, λ > 0, is a regularization parameter that helps to avoid overfitting.
The method comprises the steps of converting a sequencing problem into a pairwise classification problem by using a machine learning algorithm based on a vector machine, and then learning and solving by using a traditional support vector machine classification model.
And S43, branching the variables in the branching process after the theta nodes based on the branch strategy function.
The step is a branch stage of the machine learning period, and a branch strategy function f [ phi (x) which is fitted by a support vector machine is adoptedj，Ni)]And (4) scoring the variables in the branching process, and selecting the variable direction with the highest score for branching each time.
After the strong branch strategy of branch and bound is carried out on theta nodes, a machine learning algorithm based on a support vector machine learns the weight vector omega, and a variable branch strategy is switched from the strong branch strategy to a branch function strategy f based on the support vector machine. For each node
For each variable
We compute the feature vector φ (x)j，Ni) Then, a score is calculated at the branch policy.
Wherein S isj＝f[φ(xj，Ni)]。
And S5, calculating the constructed unit overhaul optimization model based on the branch-and-bound calculation framework to obtain a unit overhaul plan, and specifically comprising the following steps.
S51, establishing a unit maintenance optimization model based on safety and reliability based on the step S3, and forming a target function and constraint conditions;
s52, setting the number theta of branch nodes of the strong branch strategy and the calculation error upper limit sigma of the optimization model;
s53, in the process of branch delimitation of the unit overhaul optimization model solution based on safety and reliability, the former theta nodes adopt a strong branch strategy to branch, each node takes the score calculated by the strong branch strategy as a label of a candidate variable, and the characteristics of the candidate variable, such as the current state, the historical performance and the like of the variable in the current node linear programming problem solution, such as the objective function coefficient, the number of involved constraint conditions and the like, are taken as candidate variable extraction characteristics.
S54, inputting the labels of the candidate variables and the extraction characteristics of the candidate variables into a machine learning algorithm of a support vector machine as a training data set, and fitting to obtain a branch strategy function f;
s55, branching all the variables of the nodes in the branching process based on the branch strategy function;
and S56, if the obtained optimal solution is within the error upper limit sigma range as shown in the figure 1, solving the unit overhaul optimization model to obtain a unit overhaul plan, otherwise, taking the calculated optimal solution as the initial solution of the unit overhaul optimization model, and repeating the steps S53-S56 until the obtained optimal solution is within the error upper limit sigma range.
Through the steps, the system safety and reliability index is 0.022, and the result of the unit maintenance plan obtained through calculation by the unit maintenance optimization method based on the vector machine is shown in fig. 6.
The foregoing merely illustrates the principles and preferred embodiments of the invention and many variations and modifications may be made by those skilled in the art in light of the foregoing description, which are within the scope of the invention.
Claims (9)
1. A unit maintenance optimization method based on a support vector machine is characterized by comprising the following steps:
s1, establishing a unit maintenance plan safe reliability index based on the system standby;
s2, constructing a target function of the unit maintenance model based on the safe reliability index;
s3, constructing a unit overhaul optimization model based on the objective function and constraint conditions of the unit overhaul model;
s4, constructing a branch and bound calculation framework of the improved branch strategy based on the support vector machine;
and S5, calculating the constructed unit maintenance optimization model based on the branch-and-bound calculation framework to obtain a unit maintenance plan.
2. The overhaul optimization method of the support vector machine based unit according to claim 1, wherein the safety and reliability index formula is as follows,
wherein T is the total time period, StFor system standby for the period t, the formula is as follows,
MTithe sum of the overhaul times of the unit i in the T period,
is the overhaul state of the overhaul unit i in the jth overhaul time period t,
is an integer variable from 0 to 1, if
Representing the unit i is in maintenance state if
The unit i is in a non-overhaul state; NG is the total number of the units; gi,maxThe upper limit of the output of the unit is set; dtThe system load for the time period t.
3. The overhaul optimization method for the unit based on the support vector machine according to claim 2, wherein the step S2 specifically comprises:
s21, taking the maximum safe reliability index as the objective function of the unit maintenance model, wherein the objective function of the unit maintenance model is as follows,
the absolute value of the system standby difference between the t time period and the t-1 time period;
s22, linearizing the formula (3), the expression of the linearization is as follows,
wherein S istFor system standby during time period t, St-1For the system standby during the t-1 period.
4. The support vector machine-based unit overhaul optimization method according to claim 1, wherein the constraint conditions at least include a unit overhaul time constraint, a unit overhaul continuity constraint, a unit overhaul plan non-overlapping constraint, a unit overhaul mutual exclusion constraint, a unit overhaul simultaneous constraint, a system standby constraint, a power balance constraint, a unit output upper and lower limit constraint, an overhaul unit output upper and lower limit constraint, a line power flow constraint and a variable integer constraint.
5. The support vector machine-based overhaul optimization method according to claim 4, wherein the model constructed based on the overhaul time constraint is as follows,
wherein the content of the first and second substances,
the time required for the jth overhaul of the unit i;
the model constructed based on the crew overhaul time constraints is as follows,
wherein the content of the first and second substances,
a model constructed based on the unit overhaul plan non-overlapping constraints is as follows,
the model constructed based on the unit overhaul mutual exclusion constraint is as follows,
wherein phieThe method comprises the steps of (1) setting a mutual exclusion overhaul unit set;
the model constructed based on the crew simultaneous constraints is as follows,
wherein phisFor servicing sets of units simultaneously, Ns,tOverhauling the upper limit of the number of units for the set s in the time period t;
the model built based on the system standby constraints is as follows,
St≥DtR,t＝1,2,…,T (11)
wherein R is the system standby rate, StFor system standby during time t, DtSystem load for time period t;
a model constructed based on the electronic balance constraints is as follows,
wherein, Pi,tThe active power output of the unit i in the time period t is obtained;
the model constructed based on the upper and lower unit output limit constraints is as follows,
where, i is 1, …, NGm；t＝1,2,…,T，Gi,minLower limit of unit output, Gi,maxThe upper limit of the output of the unit is set;
the model built based on the line flow constraints is as follows,
where l ═ 1, …, NL; t is 1,2, …, and T, NN is the total number of load nodes in the system; NL is the total number of lines in the network topology; pl,maxUpper limit for transmission power of line l; pl,minThe lower limit of the transmission power for line l; dn,tA node load predicted value of a node n in a power grid in a time period t is obtained; gl-iA node-line power transfer distribution factor of the unit i to the line l; gl-nA node-line power transfer distribution factor for load n pairs of lines l;
the model constructed based on the variable integer constraint is as follows,
the unit maintenance model based on safety composed of formulas (3) to (14) is ProbR-MS。
6. The overhaul optimization method for the unit based on the support vector machine according to claim 1, wherein the step S4 specifically comprises:
s41, simulating a branch process of a candidate branch variable set C of theta nodes based on a strong branch strategy to obtain a training data set based on candidate variable characteristics and candidate variable labels of the former theta nodes;
s42, inputting the training data into a machine learning algorithm of a support vector machine, and fitting to obtain a branch strategy function f by taking a loss function of a training data set as a minimum target;
and S43, branching the nodes behind the theta nodes based on the branch strategy function.
7. The overhaul optimization method for the unit based on the support vector machine according to claim 6, wherein the step S42 specifically comprises:
machine learning algorithm based on support vector machine to train loss function of data
For the minimum objective, a branch strategy function f is fitted,
f[φ(xj，Ni)]＝wTφ(xj,Ni) (16)
wherein, wTFor the transposition of the parameter vector w, the training data comprises at least a set of nodes of the search solution
A group of existing nodes
Candidate variable set of
Labels of candidate variables at each node i
Characteristic diagram
Ω is the label yiX ═ x1,…,xn}，φ(xj,Ni) Refers to a branch variable xjAt node NiThere are p features;
definition of
For each one
The variable passes through a function f [ phi (x)j,Ni)]Calculated result, loss function
According to a pair variable at a node Ni
The penalty resulting from sorting is taken as the regularization parameter.
8. The overhaul optimization method for the unit based on the support vector machine according to claim 6, wherein the step S43 specifically comprises:
s431, after the branch-and-bound strong branch strategy carries out theta nodes, switching the variable branch strategy from the strong branch strategy to a branch function strategy f based on a support vector machine;
s432, for each node
Each variable
Computing a feature vector phi (x)j,Ni) And a branch strategy calculation score is obtained as follows,
wherein S isj＝f[φ(xj,Ni)]。
9. The overhaul optimization method for the unit based on the support vector machine according to claim 6, wherein the step S5 specifically comprises:
s51, obtaining a target function and constraint conditions based on the unit maintenance optimization model;
s52, presetting the number theta of branch nodes of the strong branch strategy and the upper limit sigma of the error of the unit overhaul optimization model;
s53, the former theta nodes adopt a strong branch strategy to branch, the score calculated by each node is used as a label of a candidate variable, and the candidate variable extraction characteristics are obtained based on the target function and the constraint condition;
s54, inputting the labels of the candidate variables and the extraction characteristics of the candidate variables into a machine learning algorithm of a support vector machine as a training data set, and fitting to obtain a branch strategy function f;
s55, branching all the variables of the nodes in the branching process based on the branch strategy function;
and S56, if the obtained optimal solution is within the error upper limit sigma range, solving the unit overhaul optimization model to obtain a unit overhaul plan, otherwise, repeating the steps S53-S56 by taking the calculated optimal solution as an initial solution of the unit overhaul optimization model until the obtained optimal solution is within the error upper limit sigma range.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111210633.1A CN113988390A (en) | 2021-10-18 | 2021-10-18 | Unit maintenance optimization method based on support vector machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111210633.1A CN113988390A (en) | 2021-10-18 | 2021-10-18 | Unit maintenance optimization method based on support vector machine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113988390A true CN113988390A (en) | 2022-01-28 |
Family
ID=79739114
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111210633.1A Pending CN113988390A (en) | 2021-10-18 | 2021-10-18 | Unit maintenance optimization method based on support vector machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113988390A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116736780A (en) * | 2023-08-15 | 2023-09-12 | 贵州汇通华城股份有限公司 | Startup and shutdown control optimization method and system for regional energy station |
| CN117422183A (en) * | 2023-12-18 | 2024-01-19 | 国网四川省电力公司 | Unit overhaul optimization method, device, equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102738833A (en) * | 2012-06-20 | 2012-10-17 | 湖北省电力公司 | Multi-time-scale rolling coordination scheduling method for electric power system with wind power |
| CN103488869A (en) * | 2013-08-23 | 2014-01-01 | 上海交通大学 | Wind power generation short-term load forecast method of least squares support vector machine |
| CN109117570A (en) * | 2018-08-24 | 2019-01-01 | 国网安徽省电力有限公司岳西县供电公司 | A kind of power distribution network optimized maintenance method based on distributed photovoltaic |
| CN112633529A (en) * | 2020-12-24 | 2021-04-09 | 浙江大学 | Distribution line maintenance optimization method based on deep reinforcement learning |
| US20210117586A1 (en) * | 2019-05-24 | 2021-04-22 | Tsinghua University | Method, apparatus, and storage medium for planning power distribution network |
-
2021
- 2021-10-18 CN CN202111210633.1A patent/CN113988390A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102738833A (en) * | 2012-06-20 | 2012-10-17 | 湖北省电力公司 | Multi-time-scale rolling coordination scheduling method for electric power system with wind power |
| CN103488869A (en) * | 2013-08-23 | 2014-01-01 | 上海交通大学 | Wind power generation short-term load forecast method of least squares support vector machine |
| CN109117570A (en) * | 2018-08-24 | 2019-01-01 | 国网安徽省电力有限公司岳西县供电公司 | A kind of power distribution network optimized maintenance method based on distributed photovoltaic |
| US20210117586A1 (en) * | 2019-05-24 | 2021-04-22 | Tsinghua University | Method, apparatus, and storage medium for planning power distribution network |
| CN112633529A (en) * | 2020-12-24 | 2021-04-09 | 浙江大学 | Distribution line maintenance optimization method based on deep reinforcement learning |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116736780A (en) * | 2023-08-15 | 2023-09-12 | 贵州汇通华城股份有限公司 | Startup and shutdown control optimization method and system for regional energy station |
| CN116736780B (en) * | 2023-08-15 | 2023-11-24 | 贵州汇通华城股份有限公司 | Startup and shutdown control optimization method and system for regional energy station |
| CN117422183A (en) * | 2023-12-18 | 2024-01-19 | 国网四川省电力公司 | Unit overhaul optimization method, device, equipment and storage medium |
| CN117422183B (en) * | 2023-12-18 | 2024-03-19 | 国网四川省电力公司 | Unit overhaul optimization method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Glotić et al. | Short-term combined economic and emission hydrothermal optimization by surrogate differential evolution | |
| CN113988390A (en) | Unit maintenance optimization method based on support vector machine | |
| US7489989B2 (en) | System for dispatching and controlling of generation in large-scale electric power systems | |
| CN102521652B (en) | Evaluation and decision method for operation efficiency of power grid | |
| CN110163450A (en) | A kind of distribution network planning bi-level optimal model construction method limited based on operation | |
| Han et al. | Train timetabling in rail transit network under uncertain and dynamic demand using advanced and adaptive NSGA-II | |
| CN104077651B (en) | Maintenance scheduling for power systems optimization method | |
| Zaretalab et al. | A multi-objective model for optimizing the redundancy allocation, component supplier selection, and reliable activities for multi-state systems | |
| CN111062514A (en) | Power system planning method and system | |
| CN115374692B (en) | Double-layer optimization scheduling decision method for regional comprehensive energy system | |
| CN111709632A (en) | Power failure plan automatic arrangement method based on artificial intelligence and multi-target constraint | |
| CN114897379A (en) | Dynamic production plan scheduling system and method based on information feedback | |
| CN107145968A (en) | Photovoltaic apparatus life cycle cost Forecasting Methodology and system based on BP neural network | |
| Morquecho et al. | Security constrained AC dynamic transmission expansion planning considering reactive power requirements | |
| CN113452025B (en) | Model-data hybrid driven power grid expected fault assessment method and system | |
| CN112990584B (en) | Automatic production decision system and method based on deep reinforcement learning | |
| Pinto et al. | Reliability‐constrained robust expansion planning of active distribution networks | |
| Han et al. | Learning-based topology optimization of power networks | |
| CN111723334B (en) | Power generation plan correction decision method and system based on safety distance sensitivity | |
| Yuan et al. | Research on real-time prediction of completion time based on AE-CNN-LSTM | |
| CN112784439A (en) | Energy internet planning method and device based on discretization model | |
| Qiu et al. | Local integrated energy system operational optimization considering multi‐type uncertainties: A reinforcement learning approach based on improved TD3 algorithm | |
| CN111062513B (en) | Distributed community energy trading system and method based on self-adaptive consensus mechanism | |
| Yu | Modelling for emergency manufacturing resources schedule to unexpected events | |
| CN116579801B (en) | Electric power clearing method and device based on double-layer iteration and related equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |