CN106780141A - A kind of power transmission lines overhauling plan optimization method and system based on manifold learning - Google Patents

Abstract

The present invention relates to a kind of power transmission lines overhauling plan optimization method based on manifold learning and system, the method includes：Each state information of transmission line of electricity is extracted, state evaluation and risk assessment are carried out to transmission line of electricity；By transmission line status evaluation result and risk evaluation model, with reference to local O＆M policy library, O＆M strategy is formulated；Determine optimization aim, build transmission line of electricity O＆M Model for Multi-Objective Optimization；Dimensionality reduction solution is carried out to the decision variable in Mathematical Modeling using multiple target manifold learning, the transmission line of electricity O＆M strategy for being optimized；Research and development transmission line of electricity operational system, realizes the intelligent and high efficiency of O＆M strategy generating.Using the present invention, its economy, reliability, security can be improved with the excellent O＆M strategy for formulating simultaneously transmission line of electricity.

Description

A kind of power transmission lines overhauling plan optimization method and system based on manifold learning

Technical field

The present invention relates to a kind of optimization method and system, more particularly to a kind of power transmission lines overhauling meter based on manifold learning Draw optimization method and system

Background technology

Paces with China's power grid construction are accelerated, and traditional trouble hunting has engendered O＆M with regular O＆M mode The defects such as relative inadequacy of resources, extensive, the blindness maintenance of operation management.Generate in this context and commented with risk based on state evaluation O＆M strategy is divided into daily O＆M, special O＆M and based on quantity of state problem by the transmission line of electricity O＆M method estimated, the method O＆M strategy, compared to conventional method, the O＆M mode is with more scientific, reasonability and specific aim.Establishment power transmission line During the O＆M plan of road, there is the problems such as layout is extensive, and performance indications are not excellent enough.Therefore, need badly and set up suitable transmission line of electricity O＆M planning optimization model simultaneously studies practical method for solving original O＆M plan is optimized, it is ensured that O＆M plan is expired Sufficient current power enterprise operation is required with management.

Transmission line of electricity O＆M planning optimization algorithm generally comprises genetic algorithm, 2 layers of law of planning of chance constraint, Benders points Solution, multi-objective particle swarm algorithm, multiple target mimicry physics algorithm etc..Wherein Benders decomposition methods, 2 layers of rule of chance constraint The method of drawing is single goal model, it is necessary to the subjective importance for determining security and reliability factor in power system could be counted Enter object function.Genetic algorithm, multi-objective particle swarm fado processes constraint using penalty function method, does not propose suitable penalty factor , there is certain subjectivity, and the easy Premature Convergence of algorithm in local optimum, so as to occur that O＆M plan cannot be found in choosing method The situation of optimal solution.Multiple target mimicry physics is due to limited in reduction decision variable dimension, it is impossible to realize the quick receipts of algorithm Hold back.For current multiple target, Complex Constraints, the transmission line of electricity O＆M problem of higher-dimension decision variable need to find a kind of method and realize The dimensionality reduction and fast and accurate solution of Mathematical Modeling.

The content of the invention

For problem present in above-mentioned technology, the present invention proposes a kind of transmission line of electricity O＆M based on manifold learning arithmetic Optimization method and system.It can be with the efficient dimensionality reduction of transmission line of electricity O＆M Optimized model, Fast Convergent, and effectively treatment is constrained. The scientific and rational arrangement of transmission line of electricity O＆M plan is realized, the reasonable Arrangement of human and material resources resource is realized, O＆M plan is realized Adjust automatically slightly, so as to maximize the economy of maintenance work, security, reliability.

In order to achieve the above object, the invention discloses a kind of power transmission lines overhauling planning optimization side based on manifold learning Method, it is characterised in that

Step 1, extracts each state information of transmission line of electricity, and state evaluation and risk assessment are carried out to transmission line of electricity, sets up Transmission line status evaluation model, is specifically obtained by fuzzy evidence reasoning algorithm, wherein carrying out quantity of state according to Fault Tree Choose, analytic hierarchy process (AHP) obtains quantity of state weight, the quantity of state cracking severity that cloud model method is obtained；Transmission line of electricity risk evaluation model Obtained by FCM classification, the fault rate and power transmission line for specifically being occurred according to transmission line of electricity pollution flashover, icing, thunderbolt Road fixed assets value information, each circuit pollution flashover, icing, thunderbolt are calculated to the wind of transmission line of electricity using probability and severity product Danger value.Then, according to transmission line of electricity pollution flashover, the value-at-risk that icing is damaged, lightning damage failure occurs, using FCM point Circuit correspondence different faults are divided into severe risk, moderate risk, low risk by class method.

Step 2, by transmission line status evaluation result and risk evaluation model, with reference to local O＆M policy library, formulates fortune Dimension strategy, specifically by transmission line status evaluation result and risk evaluation model, divides transmission line of electricity risk class；According to work as Benchmark O＆M strategy corresponding to ground O＆M policy library risk grade, formulates daily patrolling and ties up Policy Table, including specialty patrols dimension (spy Patrolling) strategy, power failure maintenance strategy, risk change dynamic patrol dimension strategy, meteorological mutation dynamic and patrol dimension strategy, protect power supply strategy and meet The kurtosis summer dynamically patrols dimension strategy；

Step 3, determines optimization aim, transmission line of electricity O＆M Model for Multi-Objective Optimization is built, specifically according to current electric grid Safe operation requirement, the object function of the O＆M plan of transmission line of electricity is set to economy, reliability, safety indexes；Constraint bar Coordination between constraint of the part including O＆M resource, O＆M project, trend constraint,

Define decision variable：x_ktRepresent the O＆M state x of k circuit t periods_kt=1 represents O＆M, x_kt=0 expression is not transported Dimension；It directly affects maintenance work amoun, determines the formulation of O＆M plan；

The object function of Definition Model

In formula：C_ktThe operation and maintenance expenses for representing the circuit k t periods are used, different according to circuit and O＆M period difference value；δ_k It is value of the transmission line of electricity under life cycle management；L_ktIt is k circuits in the run time of t periods, C_KIt is the design of k circuits Service life；T_kIt is the cost of k circuits, including installation cost, design cost, material cost, cost of transportation；

Quantify the safety indexes of transmission line of electricity maintenance work, maintenance work safety evaluatio takes causality loss value and counted Calculate, including human element accident penalty values, environmental factor causality loss value and thing element accident penalty values；

In formula：M, E, H represent that maintenance work is artificial, the factor of environment, thing causes penalty values respectively；Point Biao Shi circuit k t periods O＆M be artificial, the factor loss probability of environment, thing；Circuit k t are represented respectively Three's loss probability of period non-O＆M；

Delivery EENS is lacked using expectation and evaluates transmission line of electricity O＆M reliability；

In formula, S_tIt is N-dimensional vector, refers to transmission line of electricity failure state set in t-th O＆M period；L_xIt is system failure shape The cutting load amount of state x；P_kIt is the stoppage in transit probability of element k；T_tIt is maintenance window duration；

The constraints of model includes：

Constraints one, the constraint of O＆M time

Constraints two, while the constraint of O＆M

Constraints three, the constraint of mutual exclusion O＆M

0≤x_it+x_jt≤1 (7)

Constraints four, the constraint of order O＆M

t_j=t_i+T_i (8)

Constraints five, the O＆M constraint that can not be changed

Constraints six, human resources constraints

Constraints seven, material resources constraint

Constraints eight, special weather constraint

Constraints nine, the maintenance work amoun distribution most proper restraint of day part

In formula：e_k、l_kThe earliest start periods and start periods the latest of circuit k O＆Ms are represented respectively；t_i, t_jIt is circuit i, j The O＆M time started；T_iThe time required to representing circuit i O＆Ms；h_ktThe human resources of input are needed for t period O＆M circuits k； H_tThe maximum of manpower can be put into for t period circuit O＆Ms；r_ktThe number of the material resources of input is needed for t period O＆M circuits k Amount；R_tThe maximum of O＆M resource can be put into for t period circuit O＆Ms；t_eIt is the special time period；m_kIt is circuit k O＆Ms needs Workload；M is the whole O＆M stage amount of work to be completed；α_tIt is proportionality coefficient, the work that the expression t periods can bear Amount accounts for the ratio of amount of work；

Constraints ten, transmission path constraint

Constraints 11, power-balance constraint

Some O＆M projects may be such that the transimission power of circuit increases even beyond limit value, it is therefore desirable to enter the tide of row line Stream verification；

A (PG+PC-PD)=T (16)

Constraints 12, risk level related constraint,

A certain period, the exceeded circuit of risk level, no more than limit value, must be luck more than the circuit of risk limit value Dimension eliminates risk；

Wherein, x_ktIt is circuit k O＆M states；R is the circuit sum in transmission path；b_rExpression can simultaneously implement maintenance Max line way；NG is the transmission line of electricity network generator set of node of O＆M；NC is load bus collection；PG, PC, PD are respectively Generator injection active power, cutting load amount, load vector；δ is the transmission line of electricity network exceeded limit value of risk value of O＆M, root Different according to circuit importance degree difference value, K is circuit total number；

Step 4, dimensionality reduction solution is carried out using multiple target manifold learning to the decision variable in Mathematical Modeling, obtains excellent The transmission line of electricity O＆M strategy of change, specifically using the Local Liner Prediction (LLE) in manifold learning, calculates each sample K Neighbor Points of this point；The k sample point closest relative to required sample point is defined as the neighbour of required sample point Point；K is a previously given value；Data in the present invention in the higher dimensional space of transmission line of electricity decision variable are nonlinear Distributions , employ Dijkstra distances；Dijkstra distances are a kind of geodesic distances, and it can keep the curved surface between sample point special Property；

Calculate the partial reconstruction weight matrix of sample point；Here an error function is defined, it is as follows

Wherein x_ij(j=1,2 k) is x_iK Neighbor Points, w_ijIt is x_iWith x_ijWeighted value, and to meet bar Part：Here W matrixes are asked for, it is necessary to construct a local covariance matrix Qⁱ；

By above formula withIt is combined, and uses method of Lagrange multipliers, you can obtains suboptimization and rebuild weights Matrix：

In actual operation, QⁱIt is probably a singular matrix, now must regularization Qⁱ, it is as follows：

Qⁱ=Qⁱ+rI (20)

Wherein r is regularization parameter, and I is a unit matrix of k × k；

All of sample point is mapped in lower dimensional space；Mapping condition meets as follows：

Wherein, ε (Y) is loss function value, y_iIt is x_iOutput vector, y_ij(j=1,2 k) is y_iK neighbour Point, and to meet two conditions, i.e.,：

Wherein I is m m matrix, hereCan store in N × N sparse matrixes W, work as x_jIt is x_i Neighbor Points when,

If W_i,j=0；Then loss function is rewritable is：

Wherein M is a symmetrical matrix of N × N, and its expression formula is:

M=(I-W)^T(I-W) (25)

Loss function value is reached minimum, then take the characteristic vector corresponding to the minimum m nonzero eigenvalue that Y is M； In processing procedure, the characteristic value of M is arranged from small to large, first characteristic value is nearly close to zero, then cast out first spy Value indicative；The characteristic vector corresponding to the characteristic value between 2~m-1 is generally taken as the transmission line of electricity after output result, i.e. dimensionality reduction O＆M decision variable；The constraints of decision variable is processed in conjunction with auto-adaptive function method, finally obtains economy, peace Quan Xing, reliability are maximized simultaneously, respectively optimal multiple alternatives.

A kind of power transmission lines overhauling planning optimization system based on manifold learning, it is characterised in that including：

Data acquisition unit, the quantity of state needed for collecting transmission line status evaluation, local transmission line of electricity O＆M policy library, And parameter needed for optimized mathematical model；

Data processing unit, state evaluation and risk assessment are carried out according to the quantity of state that data acquisition unit is collected；

Policy development module, the result according to obtained by data processing unit, correspondence locality transmission line of electricity O＆M policy library, system Fixed preliminary O＆M strategy；

Policy optimization module, is input into optimized mathematical model, to policy development mould by the relevant parameter in data acquisition unit The preliminary O＆M strategy that block is formulated is optimized；

Multiple optimisation strategies that preliminary O＆M strategy and optimization module are obtained are listed it economical by tactical comment module respectively Property, reliability, safety indexes.By manually being selected as needed.

Transmission line of electricity O＆M optimization method and system based on manifold learning of the invention have the advantages that：For most New O＆M transmission line of electricity maintenance work plan is scheduled that and carries out economy and section that scientific and reasonable layout improves maintenance work The property learned；Secondly, the system that this optimization method is used realizes the generation of transmission line of electricity O＆M and the automation of optimization process and carries out, And it is installed and operation is simple, facilitates operation maintenance personnel to grasp；Last this method of discrimination has considered transmission line status Evaluate and risk evaluation result, realize the high-quality and high efficiency of transmission line of electricity maintenance work.

Brief description of the drawings

Fig. 1 is the flow chart of the transmission line of electricity O＆M optimization method based on manifold learning of the invention.

Fig. 2 is the flow chart of the transmission line of electricity operational system based on manifold learning of the invention.

Specific embodiment

The embodiment of the present invention mainly provides a kind of transmission line of electricity O＆M method optimization method based on manifold learning and is System, to make the purpose of the present invention, technical scheme and effect clearer, clear and definite, referring to the drawings and gives an actual example to the present invention It is described further.

As shown in figure 1, the transmission line of electricity O＆M optimization method based on manifold learning of the embodiment, it includes following step Suddenly：

S101:Each state information of transmission line of electricity is extracted, state evaluation and risk assessment are carried out to transmission line of electricity.

S102:By transmission line status evaluation result and risk evaluation model, with reference to local O＆M policy library, O＆M is formulated Strategy.

S103:Determine optimization aim, build transmission line of electricity O＆M Model for Multi-Objective Optimization.

S104:Dimensionality reduction solution is carried out to the decision variable in Mathematical Modeling using multiple target manifold learning, obtains excellent The transmission line of electricity O＆M strategy of change.

As shown in Fig. 2 the embodiment also includes the transmission line of electricity O＆M optimization system based on manifold learning, including：Data Collecting unit, data processing unit, policy development module, policy optimization module, tactical comment module.

Data acquisition unit, the quantity of state needed for collecting transmission line status evaluation, local transmission line of electricity O＆M policy library, And parameter needed for optimized mathematical model.

Data processing unit, state evaluation and risk assessment are carried out according to the quantity of state that data acquisition unit is collected.

Policy development module, the result according to obtained by data processing unit, correspondence locality transmission line of electricity O＆M policy library, system Fixed preliminary O＆M strategy.

Policy optimization module, is input into optimized mathematical model, to policy development mould by the relevant parameter in data acquisition unit The preliminary O＆M strategy that block is formulated is optimized.

Multiple optimisation strategies that preliminary O＆M strategy and optimization module are obtained are listed it economical by tactical comment module respectively Property, reliability, safety indexes.By manually being selected as needed.

It is specific to implement as follows in detail：

By taking 12 transmission lines of electricity in somewhere as an example, the basic parameter such as each line status grade and risk class is shown in Table 1.

The somewhere transmission line of electricity basic parameter table of table 1

Step S101:Each state information of transmission line of electricity is extracted, state evaluation and risk assessment are carried out to transmission line of electricity.Tool Body is as follows：

Transmission line status evaluation and state information needed for risk assessment are extracted, obtains defeated by fuzzy evidence reasoning algorithm Electric line state evaluation model, wherein carrying out quantity of state selection according to Fault Tree, analytic hierarchy process (AHP) obtains quantity of state weight, cloud The quantity of state cracking severity that modelling is obtained.Equipment failure rate is obtained by state evaluation result, is damaged with reference to personal safety, environment Lose, power grid security, transmission line of electricity risk evaluation model is obtained by clustering weight vectors.

Step S102：By transmission line status evaluation result and risk evaluation model, with reference to local O＆M policy library, formulate O＆M strategy.It is specific as follows：

By transmission line status evaluation result and risk evaluation model, transmission line of electricity risk class is divided.Transported according to locality Benchmark O＆M strategy corresponding to dimension policy library risk grade, formulates daily patrolling and ties up Policy Table, including specialty patrols dimension (spy patrols) Strategy, power failure maintenance strategy, risk change dynamic are patrolled dimension strategy, meteorological mutation dynamic and patrol dimension strategy, protect power supply strategy and meet peak Dynamic of aestivating patrols dimension strategy.

Result to transmission line status evaluation carries out fault diagnosis, is drawn by entry and is based on quantity of state problem accordingly Transmission line of electricity O＆M strategy, such as table 2.

The transmission line of electricity O＆M original scheme table of table 2

Summary benchmark O＆M strategy and the strategy based on quantity of state problem, generate original according to O＆M maintenance scheduling chart O＆M strategy list., only by rough arrangement, without optimization, workload allocations are uneven, nothing for the original O＆M strategy list Method realizes the reasonable arrangement of maintenance work, and unrealized economy, reliability, security are optimal.

Step S103:Determine optimization aim, build transmission line of electricity O＆M Model for Multi-Objective Optimization.It is specific as follows：

Transmission line of electricity O＆M plan is a multiple target, multiple constraint, high-dimensional optimization problem.According to current electric grid safety Service requirement, the object function of the O＆M plan of transmission line of electricity is set to economy, reliability, safety indexes.Constraints bag Include the coordination between the constraint of O＆M resource, O＆M project, trend constraint etc..

Optimization aim：Economy, reliability, three indexs of security are first made to be lifted as far as possible, then according to power network reality Need, choose the optimal scheme of a certain index of deflection.

Decision variable：x_ktRepresent the O＆M state x of k circuit t periods_kt=1 represents O＆M, x_kt=0 represents non-O＆M. It directly affects maintenance work amoun, determines the formulation of O＆M plan.

1) object function of model

In formula：C_ktThe operation and maintenance expenses for representing the circuit k t periods are used, different according to circuit and O＆M period difference value.δ_k It is value of the transmission line of electricity under life cycle management.L_ktIt is k circuits in the run time of t periods, C_KIt is the design of k circuits Service life.T_kIt is the cost of k circuits, including installation cost, design cost, material cost, cost of transportation.

The present invention quantifies the safety indexes of transmission line of electricity maintenance work, and maintenance work safety evaluatio takes causality loss value Calculated, including human element accident penalty values, environmental factor causality loss value and thing element accident penalty values.

In formula：M, E, H represent that maintenance work is artificial, the factor of environment, thing causes penalty values respectively；Respectively Expression circuit k t periods O＆M is artificial, the factor loss probability of environment, thing； The circuit k t periods are represented respectively Three's loss probability of non-O＆M.

The present invention lacks delivery EENS and evaluates transmission line of electricity O＆M reliability using expectation.

In formula, S_tIt is N-dimensional vector, refers to transmission line of electricity failure state set in t-th O＆M period；L_xIt is system failure shape The cutting load amount of state x；P_kIt is the stoppage in transit probability of element k；T_tIt is maintenance window duration.

2) constraints of model

The constraint of O＆M time

The constraint of O＆M simultaneously

The constraint of mutual exclusion O＆M

0≤x_it+x_jt≤1 (7)

The constraint of order O＆M

t_j=t_i+T_i (8)

The O＆M constraint that can not be changed

Human resources constraints

Material resources is constrained

Special weather is constrained

The maintenance work amoun distribution most proper restraint of day part

In formula：e_k、l_kThe earliest start periods and start periods the latest of circuit k O＆Ms are represented respectively；t_i, t_jIt is circuit i, j The O＆M time started；T_iThe time required to representing circuit i O＆Ms；h_ktThe human resources of input are needed for t period O＆M circuits k； H_tThe maximum of manpower can be put into for t period circuit O＆Ms；r_ktThe number of the material resources of input is needed for t period O＆M circuits k Amount；R_tThe maximum of O＆M resource can be put into for t period circuit O＆Ms；t_eIt is the special time period；m_kIt is circuit k O＆Ms needs Workload；M is the whole O＆M stage amount of work to be completed；α_tIt is proportionality coefficient, the work that the expression t periods can bear Amount accounts for the ratio of amount of work.

Transmission path is constrained

Power-balance constraint some O＆M projects may be such that the transimission power of circuit increases even beyond limit value, it is therefore desirable to Enter the trend verification of row line.

A (PG+PC-PD)=T (16)

Risk level related constraint a certain period, the exceeded circuit of risk level no more than limit value, more than risk limit value Circuit must immediately O＆M eliminate risk.

Wherein, x_ktIt is circuit k O＆M states；R is the circuit sum in transmission path；b_rExpression can simultaneously implement maintenance Max line way；NG is the transmission line of electricity network generator set of node of O＆M；NC is load bus collection；PG, PC, PD are respectively Generator injection active power, cutting load amount, load vector.δ is the transmission line of electricity network exceeded limit value of risk value of O＆M, root Different according to circuit importance degree difference value, K is circuit total number.

Step S104:Dimensionality reduction solution is carried out to the decision variable in Mathematical Modeling using multiple target manifold learning, is obtained To the transmission line of electricity O＆M strategy of optimization.

Local Liner Prediction (LLE) in manifold learning, calculates k Neighbor Points of each sample point.Relative The Neighbor Points of required sample point are defined as in k closest sample point of required sample point.K is a previously given value.This Data in invention in the higher dimensional space of transmission line of electricity decision variable are nonlinear Distributions, employ Dijkstra distances. Dijkstra distances are a kind of geodesic distances, and it can keep the curved surface characteristic between sample point.

Calculate the partial reconstruction weight matrix of sample point.Here an error function is defined, it is as follows

Wherein x_ij(j=1,2 k) is x_iK Neighbor Points, w_ijIt is x_iWith x_ijWeighted value, and to meet bar Part：Here W matrixes are asked for, it is necessary to construct a local covariance matrix Qⁱ。

By above formula withIt is combined, and uses method of Lagrange multipliers, you can obtains suboptimization and rebuild weights Matrix：

In actual operation, QⁱIt is probably a singular matrix, now must regularization Qⁱ, it is as follows：

Qⁱ=Qⁱ+rI (20)

Wherein r is regularization parameter, and I is a unit matrix of k × k.

All of sample point is mapped in lower dimensional space.Mapping condition meets as follows：

Wherein, ε (Y) is loss function value, y_iIt is x_iOutput vector, y_ij(j=1,2 k) is y_iK

It is individual

Neighbor Points, and to meet two conditions, i.e.,：

Wherein I is m m matrix, hereCan store in N × N sparse matrixes W, work as x_jIt is x_i Neighbor Points when,

If W_i,j=0.Then loss function is rewritable is：

Wherein M is a symmetrical matrix of N × N, and its expression formula is:

M=(I-W)^T(I-W)(25)

Loss function value is reached minimum, then take the characteristic vector corresponding to the minimum m nonzero eigenvalue that Y is M. In processing procedure, the characteristic value of M is arranged from small to large, first characteristic value is nearly close to zero, then cast out first spy Value indicative.The characteristic vector corresponding to the characteristic value between 2~m-1 is generally taken as the transmission line of electricity after output result, i.e. dimensionality reduction O＆M decision variable.The constraints of decision variable is processed in conjunction with auto-adaptive function method, finally obtains economy, peace Quan Xing, reliability are maximized simultaneously, respectively optimal multiple alternatives.Table 3 is in prioritization scheme.

Transmission line of electricity O＆M planning chart after the optimization of table 3

The performance indications contrast table planned after the original scheme of table 4 and optimization

Tab.6 Optimiz ation result of daily maintenance and operation

From table 4, the economy of the plan after optimization, security, reliability index are better than original scheme, therefore this is specially The method of profit is practical.

Specific embodiment described herein is only to the spiritual explanation for example of the present invention.Technology neck belonging to of the invention The technical staff in domain can be made various modifications or supplement to described specific embodiment or be replaced using similar mode Generation, but without departing from spirit of the invention or surmount scope defined in appended claims.

Claims (2)

1. a kind of power transmission lines overhauling plan optimization method based on manifold learning, it is characterised in that

Step 1, extracts each state information of transmission line of electricity, and state evaluation and risk assessment are carried out to transmission line of electricity, sets up transmission of electricity Line status evaluation model, is specifically obtained by fuzzy evidence reasoning algorithm, wherein quantity of state selection is carried out according to Fault Tree, Analytic hierarchy process (AHP) obtains quantity of state weight, the quantity of state cracking severity that cloud model method is obtained；Transmission line of electricity risk evaluation model is by mould Paste C- average classification is obtained, and the fault rate and transmission line of electricity for specifically being occurred according to transmission line of electricity pollution flashover, icing, thunderbolt are solid Determine assets value information, each circuit pollution flashover, icing, thunderbolt are calculated to the risk of transmission line of electricity using probability and severity product Value；Then, according to transmission line of electricity pollution flashover, the value-at-risk that icing is damaged, lightning damage failure occurs, classified using FCM Circuit correspondence different faults are divided into severe risk, moderate risk, low risk by method；

Step 2, by transmission line status evaluation result and risk evaluation model, with reference to local O＆M policy library, formulates O＆M plan Slightly, specifically by transmission line status evaluation result and risk evaluation model, transmission line of electricity risk class is divided；Transported according to locality Benchmark O＆M strategy corresponding to dimension policy library risk grade, formulates daily patrolling and ties up Policy Table, including specialty patrols dimension (spy patrols) Strategy, power failure maintenance strategy, risk change dynamic are patrolled dimension strategy, meteorological mutation dynamic and patrol dimension strategy, protect power supply strategy and meet peak Dynamic of aestivating patrols dimension strategy；

Step 3, determines optimization aim, transmission line of electricity O＆M Model for Multi-Objective Optimization is built, specifically according to current electric grid safety Service requirement, the object function of the O＆M plan of transmission line of electricity is set to economy, reliability, safety indexes；Constraints bag Include the coordination between the constraint of O＆M resource, O＆M project, trend constraint,

Define decision variable：x_ktRepresent the O＆M state x of k circuit t periods_kt=1 represents O＆M, x_kt=0 represents non-O＆M； It directly affects maintenance work amoun, determines the formulation of O＆M plan；

The object function of Definition Model

$f_1=\underset{t=1}{\overset{T}{\Σ}}\underset{k=1}{\overset{N}{\Σ}}(C_{kt}+W_{kt}){\mathrm{\δ}}_k\·x_{kt}---\left(1\right)$

${\mathrm{\δ}}_k=\frac{L_{kt}}{C_K}{T}_k---\left(2\right)$

In formula：C_ktThe operation and maintenance expenses for representing the circuit k t periods are used, different according to circuit and O＆M period difference value；δ_kFor this Value of the transmission line of electricity under life cycle management；L_ktIt is k circuits in the run time of t periods, C_KIt is the design and operation of k circuits Life-span；T_kIt is the cost of k circuits, including installation cost, design cost, material cost, cost of transportation；

Quantify the safety indexes of transmission line of electricity maintenance work, maintenance work safety evaluatio takes causality loss value and calculated, Element accident penalty values including human element accident penalty values, environmental factor causality loss value and thing；

$\begin{array}{c}{f}_2\left(x_{kt}\right)=\underset{t=1}{\overset{T}{\mathrm{\Σ}}}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}\{\left(M\·{\mathrm{\α}}_{kt}^{+}+E\·{\mathrm{\β}}_{kt}^{+}+H\·{\mathrm{\γ}}_{kt}^{+}\right)x_{kt}+\\ \left(M\·{\mathrm{\α}}_{kt}^{-}+E\·{\mathrm{\β}}_{kt}^{-}+H\·{\mathrm{\γ}}_{kt}^{-}\right)\left(1-x_{kt}\right)\}\end{array}---\left(3\right)$

In formula：M, E, H represent that maintenance work is artificial, the factor of environment, thing causes penalty values respectively；Difference table Timberline road k t periods O＆M is artificial, the factor loss probability of environment, thing；The circuit k t periods are represented respectively Three's loss probability of non-O＆M；

Delivery EENS is lacked using expectation and evaluates transmission line of electricity O＆M reliability；

$f_3\left(x_{kt}\right)=\underset{t=1}{\overset{T}{\Σ}}\left(\underset{x\∈S_t}{\Σ}{L}_x\underset{k=1}{\overset{N}{\Π}}{P}_k^{x_{kt}}{\left(1-P_k\right)}^{1-x_{kt}}\right)T_t---\left(4\right)$

In formula, S_tIt is N-dimensional vector, refers to transmission line of electricity failure state set in t-th O＆M period；L_xIt is system fault condition x's Cutting load amount；P_kIt is the stoppage in transit probability of element k；T_tIt is maintenance window duration；

The constraints of model includes：

Constraints one, the constraint of O＆M time

$\left\{\begin{array}{cc}{x}_{kt}=1& e_k\&le;t\&le;l_k\\ x_{kt}=0& t<e_{k}ort>l_k\end{array}\right.---\left(5\right)$

Constraints two, while the constraint of O＆M

$\begin{array}{c}{x}_{k_{1}t}+x_{k_{2}t}+\mathrm{...}+x_{k_{n}t}=n\\ or\\ x_{k_{1}t}+x_{k_{2}t}+\mathrm{...}+x_{k_{n}t}=0\end{array}---\left(6\right)$

Constraints three, the constraint of mutual exclusion O＆M

0≤x_it+x_jt≤1 (7)

Constraints four, the constraint of order O＆M

t_j=t_i+T_i (8)

Constraints five, the O＆M constraint that can not be changed

$x_{{\mathrm{kt}}_1}=0or{x}_{{\mathrm{kt}}_2}=1---\left(9\right)$

Constraints six, human resources constraints

$\underset{t=1}{\overset{T}{\&Sigma;}}{h}_{kt}{x}_{kt}\&le;H_t---\left(10\right)$

Constraints seven, material resources constraint

$\underset{t=1}{\overset{T}{\&Sigma;}}{r}_{kt}{x}_{kt}\&le;R_t---\left(11\right)$

Constraints eight, special weather constraint

$x_{{\mathrm{kt}}_e}=1---\left(12\right)$

Constraints nine, the maintenance work amoun distribution most proper restraint of day part

$\underset{k=1}{\overset{N}{\&Sigma;}}{m}_k{x}_{kt}\&le;{\mathrm{\&alpha;}}_{t}M---\left(13\right)$

In formula：e_k、l_kThe earliest start periods and start periods the latest of circuit k O＆Ms are represented respectively；t_i, t_jIt is circuit i, the fortune of j The dimension time started；T_iThe time required to representing circuit i O＆Ms；h_ktThe human resources of input are needed for t period O＆M circuits k；H_tIt is t Period circuit O＆M can put into the maximum of manpower；r_ktThe quantity of the material resources of input is needed for t period O＆M circuits k；R_t The maximum of O＆M resource can be put into for t period circuit O＆Ms；t_eIt is the special time period；m_kFor the work that circuit k O＆Ms need Amount；M is the whole O＆M stage amount of work to be completed；α_tIt is proportionality coefficient, the workload that representing the t periods can bear is accounted for The ratio of amount of work；

Constraints ten, transmission path constraint

$\underset{k\&Element;r}{\&Sigma;}{x}_{kt}\&le;b_r---\left(14\right)$

Constraints 11, power-balance constraint

Some O＆M projects can be such that the transimission power of circuit increases even beyond limit value, it is therefore desirable to enter the trend school of row line Test；

$\underset{i\&Element;NG}{\&Sigma;}{\mathrm{PG}}_i+\underset{i\&Element;NC}{\&Sigma;}{C}_i=\underset{i\&Element;NC}{\&Sigma;}{\mathrm{PD}}_i---\left(15\right)$

A (PG+PC-PD)=T (16)

Constraints 12, risk level related constraint

A certain period, the exceeded circuit of risk level O＆M must disappear immediately no more than limit value, the circuit more than risk limit value Except risk；

${\mathrm{Count}}_{(R_k\&le;R_{st})}\&le;\mathrm{\&delta;}\&CenterDot;K---\left(17\right)$

Wherein, x_ktIt is circuit k O＆M states；R is the circuit sum in transmission path；b_rExpression can simultaneously implement maintenance most Big circuit number；NG is the transmission line of electricity network generator set of node of O＆M；NC is load bus collection；PG, PC, PD are respectively generating Machine injection active power, cutting load amount, load vector；δ is the transmission line of electricity network exceeded limit value of risk value of O＆M, according to line Road importance degree difference value is different, and K is circuit total number；

Step 4, dimensionality reduction solution is carried out using multiple target manifold learning to the decision variable in Mathematical Modeling, is optimized Transmission line of electricity O＆M strategy, specifically using the Local Liner Prediction (LLE) in manifold learning, calculates each sample point K Neighbor Points；The k sample point closest relative to required sample point is defined as the Neighbor Points of required sample point；K is One previously given value；Data in the present invention in the higher dimensional space of transmission line of electricity decision variable are nonlinear Distributions, are used Dijkstra distances；Dijkstra distances are a kind of geodesic distances, and it can keep the curved surface characteristic between sample point；

Calculate the partial reconstruction weight matrix of sample point；Here an error function is defined, it is as follows

$min\mathrm{\&epsiv;}\left(W\right)=\underset{i=1}{\overset{N}{\&Sigma;}}|x_i-\underset{j=1}{\overset{k}{\&Sigma;}}{w}_j^i{x}_{ij}{|}^2---\left(18\right)$

Wherein x_ij(j=1,2 k) is x_iK Neighbor Points, w_ijIt is x_iWith x_ijWeighted value, and to meet condition：Here W matrixes are asked for, it is necessary to construct a local covariance matrix Qⁱ；

By above formula withIt is combined, and uses method of Lagrange multipliers, you can obtains suboptimization and rebuild weights square Battle array：

$w_j^i=\frac{\underset{m-1}{\overset{k}{\&Sigma;}}{\left(Q^i\right)}_{jm}^{-1}}{\underset{p=1}{\overset{k}{\&Sigma;}}\underset{q=1}{\overset{k}{\&Sigma;}}{\left(Q^i\right)}_{pq}^{-1}}---\left(19\right)$

In actual operation, QⁱIt is probably a singular matrix, now must regularization Qⁱ, it is as follows：

Qⁱ=Qⁱ+rI (20)

Wherein r is regularization parameter, and I is a unit matrix of k × k；

All of sample point is mapped in lower dimensional space；Mapping condition meets as follows：

$min\mathrm{\&epsiv;}\left(Y\right)=\underset{i=1}{\overset{N}{\&Sigma;}}|y_i-\underset{j=1}{\overset{k}{\mathrm{\&Sigma;}}}{w}_j^i{y}_{ij}{|}^2---\left(21\right)$

Wherein, ε (Y) is loss function value, y_iIt is x_iOutput vector, y_ij(j=1,2 k) is y_iK Neighbor Points, And to meet two conditions, i.e.,：

$\underset{i-1}{\overset{N}{\&Sigma;}}{y}_i=0---\left(22\right)$

$\frac{1}{N}\underset{i-1}{\overset{N}{\&Sigma;}}{y}_i{y}_i^T=I---\left(23\right)$

Wherein I is m m matrix, hereCan store in N × N sparse matrixes W, work as x_jIt is x_iNeighbour During point,

If W_i,j=0；Then loss function is rewritable is：

$min\mathrm{\&epsiv;}\left(Y\right)=\underset{i=1}{\overset{N}{\&Sigma;}}\underset{j=1}{\overset{N}{\&Sigma;}}{M}_{i,j}{y}_i^T{y}_j---\left(24\right)$

Wherein M is a symmetrical matrix of N × N, and its expression formula is:

M=(I-W)^T(I-W) (25)

Loss function value is reached minimum, then take the characteristic vector corresponding to the minimum m nonzero eigenvalue that Y is M；In treatment During, the characteristic value of M is arranged from small to large, first characteristic value is nearly close to zero, then cast out first characteristic value； The characteristic vector corresponding to the characteristic value between 2~m-1 is generally taken to be determined as the transmission line of electricity O＆M after output result, i.e. dimensionality reduction Plan variable；The constraints of decision variable is processed in conjunction with auto-adaptive function method, finally obtain economy, security, Reliability is maximized simultaneously, respectively optimal multiple alternatives.

2. a kind of power transmission lines overhauling planning optimization system based on manifold learning, it is characterised in that including：

Data acquisition unit, the quantity of state needed for collecting transmission line status evaluation, local transmission line of electricity O＆M policy library, and Parameter needed for optimized mathematical model；

Data processing unit, state evaluation and risk assessment are carried out according to the quantity of state that data acquisition unit is collected；

Policy development module, the result according to obtained by data processing unit, correspondence locality transmission line of electricity O＆M policy library is formulated just Step O＆M strategy；

Policy optimization module, is input into optimized mathematical model, to policy development module system by the relevant parameter in data acquisition unit Fixed preliminary O＆M strategy is optimized；

Tactical comment module, multiple optimisation strategies that preliminary O＆M strategy and optimization module are obtained are listed respectively its economy, Reliability, safety indexes；By manually being selected as needed.