CN103545846A - Microgrid economic operation method based on generalized load prediction - Google Patents

Abstract

The invention discloses a microgrid economic operation method based on generalized load prediction. A fan, a photovoltaic cell and a traditional load in a microgrid are seen as generalized load for conducting prediction, and therefore computation complexity and mechanism analysis complexity are greatly reduced. After prediction, the Markov principle is utilized to correct prediction errors, and therefore prediction accuracy of the generalized load is guaranteed. In the aspect of resolving an economic operation model of the microgrid, considering the variety of a target function and flexibility of constraint conditions, a conventional method is unsuitable for resolving the economic operation problems of the microgrid, and a particle swarm algorithm is suitable for resolving the economic operation problems of the microgrid due to the fact that the particle swarm optimization algorithm has the advantages of being simple in rule, easy to realize, few in adjustable parameters and high in convergence rate. In order to increase the iteration speed of the microgrid, the conventional particle swarm algorithm is improved, and therefore the obtained result is more rapid and more accurate.

Description

Microgrid economical operation method based on broad sense load prediction

Technical field

The invention belongs to microgrid economical operation technical field, be specifically related to a kind of microgrid economical operation method based on broad sense load prediction.

Background technology

Microgrid is a kind of miniature electric power networks being comprised of distributed power source and related load thereof, and possesses the ability that oneself controls, and it takes full advantage of the advantage of distributed power generation, has also coordinated the contradiction between distributed power source and large electrical network.Microgrid has dual role.For electrical network, the intelligent load that microgrid can change as a size, can dispatch load for local electric power system provides, and can within the several seconds, make response to meet system needs, provides strong support to large electrical network in good time; Can not affect at maintenance system client's load simultaneously; Can alleviate (prolongation) power distribution network updates.For user, microgrid, as a customizable power supply, can meet the diversified demand of user, for example, strengthen local power reliability, reduce feeder loss, support to work as ground voltage, by utilizing used heat to raise the efficiency, the correction of voltage sag is provided, or as interruptible power service not etc.The uncertainty of simultaneously exerting oneself due to diversity and contained uncontrollable micro-source of microgrid running status, this has brought very large difficulty to prediction and the energy-optimised management in uncontrollable micro-source.

Owing to containing uncontrollable micro-sources such as numerous blower fans, photovoltaic cell in microgrid, and uncontrollable micro-source exerts oneself and has the feature of intermittent and randomness, and the load of microgrid region also has the feature of the large poor controllability of general fluctuation in addition.Therefore quick and precisely predict that uncontrollable micro-source exerted oneself into the matter of utmost importance of microgrid Study on economical operation; And for the economical operation method of microgrid, existing a lot of scholars are studied it, but only resting on mostly, existing research analyzes economical operation under certain running status of microgrid, ignored the flexibility of microgrid running status, or the too theoretical property that makes a search, lacks directiveness to existing engineering practice.And along with the promulgation of the book > > of < < capital of a country association, each department are more and more paid attention to environment, formulate the feature of environmental protection that microgrid operational plan certainly will will be considered operation.

Summary of the invention

The invention provides a kind of microgrid economical operation method based on broad sense load prediction, the economical operation plan of varying environment Preference can be provided for microgrid under different running statuses.

To achieve these goals, the technical solution used in the present invention is:

The present invention optimizes for the load prediction of microgrid broad sense and operation, and described microgrid is a kind of miniature electric power networks being comprised of distributed power source and related load thereof; In described microgrid, contain uncontrollable micro-sources such as blower fan, photovoltaic cell and for the energy storage device of energy snubber; Described microgrid possesses isolated island and grid-connected two kinds of running statuses; During described formulation microgrid operational plan, need to consider its impact on environment; The described microgrid economical operation method based on broad sense load prediction, concrete steps are as follows:

Step 1, broad sense load prediction; Gather exert oneself information data and photovoltaic goes out force data and adopt BP neural network to predict of microgrid tradition load information data, blower fan, regarding blower fan, photovoltaic as consumed power is negative load; The known broad sense load being comprised of microgrid load, blower fan and photovoltaic is:

p ^LD＝p ^load-p ^w-p ^pv (1)

In formula: p ^lDfor broad sense load value, p ^loadfor microgrid tradition load value, p ^w, p ^pvbe respectively blower fan, the photovoltaic value of exerting oneself.

Step 2, the correction of broad sense load predicated error; Statistical history broad sense load actual value and broad sense load predicted value, can predict that relative residual error is:

In formula:

for the broad sense load predicted value obtaining by BP neural network prediction, p ^lDfor broad sense load actual value.

The random relative error sequence of the non-stationary with Markov chain feature is divided into n state, arbitrary state

S _i∈[s _i′，s _i″]。By adding up known, state S _ithe number of times occurring is m _i, by state S _ithrough k step, transfer to state S _jnumber of times be m _ij(k).Known, state transitions is p _ij(k)=m _ij(k)/m _i(i=1,2, Ln)

A step state transition probability matrix is:

$P=\left[\begin{array}{cccc}{p}_{11}& p_{12}& L& p_{1n}\\ p_{21}& p_{22}& L& p_{2n}\\ M& M& O& M\\ p_{n1}& p_{n2}& L& p_{\mathrm{nn}}\end{array}\right]---\left(3\right)$

Matrix P meets: 1. right

, n has 0≤p _ij(k)≤1

② $\&ForAll;i,j=\mathrm{1,2},L,$ N has $\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{p}_{\mathrm{ij}}\left(k\right)=1$

In formula: p _ijfor state S _ithrough a step, transfer to state S _jprobability.

By state-transition matrix, determined the transfering state [s in system future _i', s _i"], the correction predicted value of the h time is:

The target function of step 3, the economical operation of setting microgrid, the target function of described microgrid economical operation mainly comprises: operating cost function and Environmental costs function.

Described microgrid operating cost function comprises: the operating cost in controlled micro-source, with the mutual expense of electrical network and the energy storage device cost of exerting oneself, by operating cost function representation, be:

$F=\underset{j=1}{\overset{T}{\mathrm{\&Sigma;}}}[\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{f}_i\left(p_{\mathrm{ij}}\right)+\mathrm{\&rho;}({\mathrm{\&mu;}}_j^b{p}_j^b-{\mathrm{\&mu;}}_j^s{p}_j^s)+(1-\mathrm{\&rho;})\mathrm{\&mu;}|r_j\left|\right]---\left(5\right)$

In formula: N is the quantity in controlled micro-source in micro-grid system; T is p dispatching cycle _ijbe i controlled micro-source in period j power output: for the j period under grid-connected pattern is from power distribution network input power;

for the j period, to power distribution network, buy in electricity price;

for the j period under isolated island formula is to power distribution network power output;

for the j period, to power distribution network, sell electricity price; r _jfor exerting oneself of the energy storage device of j period; μ is the cost of exerting oneself of energy storage device.ρ is operational mode decision factor, ρ=0 micro-grid system islet operation, and in order to reduce the infringement of energy storage device, agreement energy storage device only moves under isolated island; ρ=1 micro-grid system is incorporated into the power networks.

F _i(p _ij) be the operating cost in each controlled micro-source, comprise energy consumption cost and maintenance cost.

f _i(p _ij)=h _i(p _ij)+w _i(p _ij) (6)

Wherein energy consumption cost is:

$h_i{p}_{\mathrm{ij}}=a_i{p}_{\mathrm{ij}}^2+b_i{p}_{\mathrm{ij}}+c_i---\left(7\right)$

In formula: a _i, b _i, c _ifor the fuel cost coefficient of unit i,

Maintenance cost is:

w _i(p _ij)=d _ip _ij(8)

Described microgrid is only considered the environmental impact in controlled micro-source.By Environmental costs function representation, be:

$E=\underset{j=1}{\overset{T}{\mathrm{\&Sigma;}}}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\underset{h=1}{\overset{k}{\mathrm{\&Sigma;}}}{e}_h\&CenterDot;K_{\mathrm{ih}}\&CenterDot;p_{\mathrm{ij}}---\left(9\right)$

In formula: e _hbe the blowdown price (unit/kg) of h kind pollutant, K _iit is the discharge capacity (kg/kw) of the controlled micro-source of i platform h kind pollutant.

Step 4, the constraints of microgrid economical operation is set.Comprise: power-balance constraint, controlled micro-source units limits, controlled micro-source climbing rate constraint.

1) power-balance constraint: in described microgrid, the sum of exerting oneself of controlled micro-source, electrical network input, energy storage device equals broad sense load value in the same time, arbitrary period j is had:

$\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{p}_{\mathrm{ij}}+\mathrm{\&delta;}(p_j^b-p_j^s)+(1-\mathrm{\&delta;})r_j=p_j^{\mathrm{LD}}---\left(10\right)$

In formula:

predicted value for j broad sense load constantly.

2) units limits in controlled micro-source

p _ij.min≤p _ij≤p _ij.max (11)

3) the climbing rate in controlled micro-source constraint

D _igΔt≤p _i(j+1)-p _ij≤U _igΔt (12)

D _i, U _tbe respectively maximum power climbing and the maximum power rate of descent of ramp-rate limits.△ t is adjacent dispatch interval time.

4) with the mutual power constraint of power distribution network

Input power constraint:

$0\&le;p_j^b\&le;p_{\max }^b---\left(13\right)$

Power output constraint:

$0\&le;p_j^s\&le;p_{\max }^s---\left(14\right)$

5) energy storage device operation constraint

r _min≤|r _j|≤r _max (15)

Step 5, according to above-mentioned, described microgrid economical operation is, an Environmental costs minimum Multiobjective Programming minimum with operating cost, introduces the environment preference factor Multiobjective Programming is converted into objective programming.

f=F+ηE (16)

In formula: η is the environment preference factor; In order to solve the above-mentioned planning problem that contains a plurality of constraintss, the present invention adopts improved particle cluster algorithm head it off.Standard particle group algorithm, its velocity location more new formula is as follows:

$v_i^{t+1}=w\&CenterDot;v_i^t+{\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HSA0000097357010000041.png,HSA0000097357010000042.png"\; state="\{\{state\}\}">c</figure-callout>}}_1\&CenterDot;{\mathrm{<figure-callout\; id="1"\; label="r"\; filenames="HSA0000097357010000041.png,HSA0000097357010000042.png"\; state="\{\{state\}\}">r</figure-callout>}}_1\&CenterDot;(p_i^t-x_i^t)+c_2\&CenterDot;r_2\&CenterDot;(p_g^t-x_i^t)---\left(17\right)$

$x_i^{t+1}=x_i^t+v_i^{t+1}---\left(18\right)$

In formula: v _i=[v _i1, v _i2l, v _in] be the speed of particle i, represent the distance between particle i present position and its next step position; x _i=[x _i1, x _i2l, x _in] be the current location of particle i, p _ifor current individual optimal value, p _gfor current colony optimal value, w is inertia weight, c ₁for individual cognition constant, c ₂for social recognition normal; Inertia of design weight of the present invention is pressed mode and is declined:

$w=(w_{\max }-w_{\min }){\left(\frac{t}{\mathrm{tnum}}\right)}^{\frac{1}{2}}---\left(19\right)$

In formula: w _maxfor initial inertia weight, w _minfor stopping inertia weight, tnum is termination of iterations number of times, and t is current iteration number of times; In order to increase the hunting zone of particle, eliminate equality constraint simultaneously, according to equation (10), obtain n platform unit output p _njabout the expression formula of its dependent variable, and by its substitution inequality (11) (12).In order to solve contained restricted problem in described microgrid economic operation problem, and do not reduce the search capability of particle.The present invention counts fitness function by constraints with the form of penalty function.Specific as follows:

By institute tangible as: a≤p (X)≤b inequality constraints, be converted into shape as g (X)≤0 inequality constraints, transform as follows:

$g\left(X\right)=|p\left(X\right)-\frac{a+b}{2}|-\frac{b-a}{2}---\left(20\right)$

, fitness function is:

Fit=f+M·∑(max(0，g _i(X) ²))=F+ηE+M·∑(max(0，g _i(X))) ² (21)

In formula: M is a positive number more much larger than f.

Particle population is divided into three parts, and portion solves search around in the same time at proxima luce (prox. luc), and portion solves search around at eve, and remaining portion carries out global search between confining region, that is:

$P(\mathrm{dnum},\mathrm{pnum})=\left\{\begin{array}{c}{p}^{\&prime;}+\frac{1}{10}\&CenterDot;(p_{\max }-p_{\min })\&CenterDot;(-1+2\&CenterDot;\mathrm{rand}(\mathrm{dnum},\frac{1}{3}\&CenterDot;\mathrm{pnum}))\\ p^{\&prime;\&prime;}+\frac{1}{5}\&CenterDot;(p_{\max }-p_{\min })\&CenterDot;(-1+2\&CenterDot;\mathrm{rand}(\mathrm{dnum},\frac{1}{3}\&CenterDot;\mathrm{pnum}))\\ p_{\min }+(p_{\max }-p_{\min })\&CenterDot;\mathrm{rand}(\mathrm{dnum},\frac{1}{3}\&CenterDot;\mathrm{pnum})\end{array}\right.---\left(22\right)$

In formula: p ' is proxima luce (prox. luc) controlled micro-source (electrical network, energy storage device) situation of exerting oneself in the same time, p, " be the controlled micro-source of previous moment (electrical network, the energy storage device) situation of exerting oneself, dnum is particle dimension, and pnum is population quantity.

The present invention has following beneficial effect: the present invention is directed to blower fan in microgrid, photovoltaic array power output is uncontrollable, and nearer with microgrid tradition load geographical separation, and with meteorotropic feature, so the present invention regarded as broad sense load, and adopts nonlinear method to predict.Simultaneously in order to make up the together large shortcoming of predicated error of three; The present invention introduces Markov Theory for revising broad sense load predicated error.On the basis of broad sense load prediction, the present invention considers economy and environment and two kinds of operational modes of grid-connected island, set up microgrid economical operation model, for fast accurate solving model, the present invention is according to the feature of microgrid, improved traditional particle cluster algorithm, made it be more suitable for solving microgrid economical operation model.

Accompanying drawing explanation

Fig. 1 is the microgrid economical operation method flow schematic diagram based on broad sense load prediction.

Fig. 2 is for solving microgrid economic operation algorithm schematic flow sheet.

Fig. 3 improves inertia weight schematic diagram in algorithm involved in the present invention.

Fig. 4 is environmental impact formulation operational plan the schematic diagram of exerting oneself off the net of considering in the specific embodiment of the invention that micro-grid system (environment preference factor η=1) produces completely.

Fig. 5 is environmental impact formulation operational plan the schematic diagram of exerting oneself off the net of not considering in the specific embodiment of the invention that micro-grid system (environment preference factor η=0) produces.

Fig. 6 considers in the specific embodiment of the invention that environmental impact that micro-grid system (environment preference factor η=1) produces formulates the schematic diagram of exerting oneself under operational plan islet operation completely.

Fig. 7 does not consider in the specific embodiment of the invention that environmental impact that micro-grid system (environment preference factor η=0) produces formulates the schematic diagram of exerting oneself under operational plan islet operation.

Embodiment

Microgrid described in the present embodiment is comprised of by certain topological structure controlled micro-source, uncontrollable micro-source, tradition load and energy storage device, and microgrid can with large grid network operation, also can be when electric network fault and major network disconnection islet operation.Along with the day by day attention of various places to environment, during microgrid operation, also should consider environmental impact.The present invention is based on the microgrid economical operation method flow diagram of broad sense load prediction, see Fig. 1; Specifically comprise the steps:

Step 1, broad sense load prediction; Gather exert oneself information data and photovoltaic of microgrid tradition load information data, blower fan and go out force data, regarding blower fan, photovoltaic as consumed power is negative load; The known broad sense load being comprised of microgrid load, blower fan and photovoltaic is:

p ^LD＝p ^lond-p ^w-p ^pv (1)

In formula: p ^lDfor broad sense load value, p ^loadfor microgrid tradition load value, p ^w, p ^pvbe respectively blower fan, the photovoltaic value of exerting oneself.Due to the broad sense load various factors such as temperature, wind speed, illumination and people's mechanics of being bullied, present nonlinear feature, therefore the prediction for broad sense load is relatively applicable to adopting analyzing non-linear method, as chaology, ANN, and be not suitable for using time series method.The present invention adopts BP neural net to predict broad sense load.

Step 2, the correction of broad sense load predicated error; Statistical history broad sense load actual value and broad sense load predicted value, can predict that relative residual error is:

In formula:

for the broad sense load predicted value obtaining by BP neural network prediction, p ^lDfor broad sense load actual value.

The random relative error sequence of the non-stationary with Markov chain feature is divided into n state, arbitrary state S _i∈ [s _i', s _i"].By adding up known, state S _ithe number of times occurring is m _i, by state s _ithrough k step, transfer to state S _jthe number of times of (wherein j=i+ (m-1) τ) is m _ij(k).Known, state transitions is p _ij(k)=m _ij(k)/m _i(i=1,2, Ln)

A step state transition probability matrix is:

$P=\left[\begin{array}{cccc}{p}_{11}& p_{12}& L& p_{1n}\\ p_{21}& p_{22}& L& p_{2n}\\ M& M& O& M\\ p_{n1}& p_{n2}& L& p_{\mathrm{nn}}\end{array}\right]---\left(3\right)$

Matrix P meets: 1. right

n has 0≤p _ij(k)≤1

② $\&ForAll;i,j=\mathrm{1,2},L,$ N has $\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{p}_{\mathrm{ij}}\left(k\right)=1$

In formula: p _ijfor state S _ithrough a step, transfer to state S _jprobability.

By state-transition matrix, determined the transfering state [s in system future _i', s _i"], the correction predicted value of the h time is:

The target function of step 3, the economical operation of setting microgrid, the target function of described microgrid economical operation mainly comprises: operating cost function and Environmental costs function.

Described microgrid operating cost function comprises: the operating cost in controlled micro-source, with the mutual expense of electrical network and the energy storage device cost of exerting oneself, by operating cost function representation, be:

$F=\underset{j=1}{\overset{T}{\mathrm{\&Sigma;}}}[\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{f}_i\left(p_{\mathrm{ij}}\right)+\mathrm{\&rho;}({\mathrm{\&mu;}}_j^b{p}_j^b-{\mathrm{\&mu;}}_j^s{p}_j^s)+(1-\mathrm{\&rho;})\mathrm{\&mu;}|r_j\left|\right]---\left(5\right)$

In formula: N is the quantity in controlled micro-source in micro-grid system; T is p dispatching cycle _ijbe i controlled micro-source in period j power output:

for the j period under grid-connected pattern is from power distribution network input power;

for the j period, to power distribution network, buy in electricity price;

for the j period under isolated island formula is to power distribution network power output;

for the j period, to power distribution network, sell electricity price; r _jfor exerting oneself of the energy storage device of j period; μ is the cost of exerting oneself of energy storage device.ρ is operational mode decision factor, ρ=0 micro-grid system islet operation, and in order to reduce the infringement of energy storage device, agreement energy storage device only moves under isolated island; ρ=1 micro-grid system is incorporated into the power networks.

F _i(p _ij) be the operating cost in each controlled micro-source, comprise energy consumption cost and maintenance cost.

f _i(p _ij)=h _i(p _ij)+w _i(p _ij) (6)

Wherein energy consumption cost is:

$h_i{p}_{\mathrm{ij}}=a_i{p}_{\mathrm{ij}}^2+b_i{p}_{\mathrm{ij}}+c_i---\left(7\right)$

In formula: a _i, b _i, c _ifor the fuel cost coefficient of unit i,

Maintenance cost is:

w _i(p _ij)=d _ip _ij (8)

The blower fan that described microgrid contains, photovoltaic array, energy storage device produce hardly pollutant in the middle of operation, and gas turbine, diesel generation chance in controlled micro-source produces oxynitrides, oxygen sulfur compound, hydrocarbon and dust etc. to environmentally hazardous substance.So described microgrid is only considered the environmental impact in controlled micro-source, and make Environmental costs minimum.By Environmental costs function representation, be:

$E=\underset{j=1}{\overset{T}{\mathrm{\&Sigma;}}}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\underset{h=1}{\overset{k}{\mathrm{\&Sigma;}}}{e}_h\&CenterDot;K_{\mathrm{ih}}\&CenterDot;p_{\mathrm{ij}}---\left(9\right)$

In formula: e _hbe the blowdown price (unit/kg) of h kind pollutant, K _iit is the discharge capacity (kg/kw) of the controlled micro-source of i platform h kind pollutant.

Step 4, the constraints of microgrid economical operation is set.Comprise: power-balance constraint, controlled micro-source units limits, controlled micro-source climbing rate constraint.

1) power-balance constraint: in described microgrid, the sum of exerting oneself of controlled micro-source, electrical network input, energy storage device equals broad sense load value in the same time, arbitrary period j is had:

$\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{p}_{\mathrm{ij}}+\mathrm{\&delta;}(p_j^b-p_j^s)+(1-\mathrm{\&delta;})r_j=p_j^{\mathrm{LD}}---\left(10\right)$

In formula:

predicted value for j broad sense load constantly.

2) units limits in controlled micro-source

p _ij.min≤p _ij≤p _ij.max (11)

3) the climbing rate in controlled micro-source constraint

D _igΔt≤p _i(j+1)-p _ij≤U _igΔt (12)

D _i, U _ibe respectively maximum power climbing and the maximum power rate of descent of ramp-rate limits.△ t is adjacent dispatch interval time.

4) with the mutual power constraint of power distribution network

Input power constraint:

$0\&le;p_j^b\&le;p_{\max }^b---\left(13\right)$

Power output constraint:

$0\&le;p_j^s\&le;p_{\max }^s---\left(14\right)$

5) energy storage device operation constraint

r _min≤|r _j|≤r _max (15)

Step 5, according to above-mentioned, described microgrid economical operation is, an Environmental costs minimum Multiobjective Programming minimum with operating cost, the present invention is according to the feature of institute's Solve problems, Environmental costs are identical with operating cost dimension, but various places are different to the attention degree of environment, therefore introducing the environment preference factor is converted into objective programming by Multiobjective Programming.

f＝F+ηE (16)

In formula: η is the environment preference factor; In order to solve the above-mentioned planning problem that contains a plurality of constraintss, the present invention adopts improved particle cluster algorithm head it off, and algorithm flow chart, is shown in Fig. 2.Standard particle group algorithm, its velocity location more new formula is as follows:

$v_i^{t+1}=w\&CenterDot;v_i^t+{\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HSA0000097357010000041.png,HSA0000097357010000042.png"\; state="\{\{state\}\}">c</figure-callout>}}_1\&CenterDot;{\mathrm{<figure-callout\; id="1"\; label="r"\; filenames="HSA0000097357010000041.png,HSA0000097357010000042.png"\; state="\{\{state\}\}">r</figure-callout>}}_1\&CenterDot;(p_i^t-x_i^t)+c_2\&CenterDot;r_2\&CenterDot;(p_g^t-x_i^t)---\left(17\right)$

$x_i^{t+1}=x_i^t+v_i^{t+1}---\left(18\right)$

In formula: v _i=[v _i1, v _i2l, v _in] be the speed of particle i, represent the distance between particle i present position and its next step position; x _i=[x _i1, x _i2l, x _in] be the current location of particle i, p _ifor current individual optimal value, p _gfor current colony optimal value, w is inertia weight, c ₁for individual cognition constant, c ₂for social recognition constant.Research discovery, particle search process is divided into two stages: free search phase and converged state.Freely search for inertia weight and decline soon, and converged state inertia weight declines slowly; Therefore Inertia of design weight of the present invention is pressed mode and is declined:

$w=(w_{\max }-w_{\min }){\left(\frac{t}{\mathrm{tnum}}\right)}^{\frac{1}{2}}---\left(19\right)$

In formula: w _maxfor initial inertia weight, w _minfor stopping inertia weight, tnum is termination of iterations number of times, and t is current iteration number of times, and described inertia weight decline mode and conventional linear decline comparison diagram, be shown in Fig. 3; In order to increase the hunting zone of particle, eliminate equality constraint simultaneously, according to equation (10), obtain n platform unit output p _njabout the expression formula of its dependent variable, and by its substitution inequality (11) (12).In order to solve contained restricted problem in described microgrid economic operation problem, and do not reduce the search capability of particle.The present invention counts fitness function by constraints with the form of penalty function.Specific as follows:

By institute tangible as: a≤p (X)≤b inequality constraints, be converted into shape as g (X)≤0 inequality constraints, transform as follows:

$g\left(X\right)=|p\left(X\right)-\frac{a+b}{2}|-\frac{b-a}{2}---\left(20\right)$

, fitness function is:

Fit=f+M·∑(max(0，g _i(X) ²))=F+ηE+M·∑(max(0，g _i(X))) ² (21)

In formula: M is a positive number more much larger than f.

On population arranges.In order to add the convergence of fast particle, the present invention also makes following improvement: according to the similitude of front and back day, and front and back continuity constantly, around particle finally should drop on last order in the same time or previous moment solves.Therefore in order to add the convergence rate of fast particle and the iterations that reduces particle, particle population is divided into three parts, portion solves search around in the same time at proxima luce (prox. luc), and portion solves search around at eve, and remaining portion carries out global search between confining region.Suppose that controlled micro-source in the same time of proxima luce (prox. luc) (or electrical network, energy storage device) situation of exerting oneself is p ', the controlled micro-source of previous moment (or electrical network, the energy storage device) situation of exerting oneself is p ", population quantity pnum;

$P(\mathrm{dnum},\mathrm{pnum})=\left\{\begin{array}{c}{p}^{\&prime;}+\frac{1}{10}\&CenterDot;(p_{\max }-p_{\min })\&CenterDot;(-1+2\&CenterDot;\mathrm{rand}(\mathrm{dnum},\frac{1}{3}\&CenterDot;\mathrm{pnum}))\\ p^{\&prime;\&prime;}+\frac{1}{5}\&CenterDot;(p_{\max }-p_{\min })\&CenterDot;(-1+2\&CenterDot;\mathrm{rand}(\mathrm{dnum},\frac{1}{3}\&CenterDot;\mathrm{pnum}))\\ p_{\min }+(p_{\max }-p_{\min })\&CenterDot;\mathrm{rand}(\mathrm{dnum},\frac{1}{3}\&CenterDot;\mathrm{pnum})\end{array}\right.---\left(22\right)$

The present invention regards the blower fan in microgrid, photovoltaic cell and tradition load as broad sense load and predicts, has greatly reduced numerous and diverse degree of computation complexity and Analysis on Mechanism.And adopt markov correction predicated error after prediction, guarantee the precision of prediction of broad sense load; Solving aspect microgrid economical operation model, consider the polytropy of target function and the flexibility of constraints, be unsuitable for conventional method, and particle cluster algorithm is easy to because it is simple in rule realize, adjustable parameter is few, the feature of fast convergence rate, is suitable for solving microgrid economic operation problem, and the present invention is in order to accelerate its iteration speed, conventional particle cluster algorithm is had made some improvements, make it more arrive result more rapid accurately.

Embodiment:

Now a micro-grid system that contains blower fan, photovoltaic array, diesel engine generator, miniature gas turbine and energy storage device is analyzed as example, the cycle of operation gets 1 day, is divided into 24 periods.In this microgrid, micro-grid system is carried out to dynamic dispatching.In micro-grid system, energy storage mode has lead-acid battery, superconducting energy storage, flywheel energy storage etc., but no matter be which kind of energy storage mode, according to the definition in the life-span of energy storage device, charging and discharging all can shorten the service time of energy storage device, now suppose energy storage device exert oneself cost be 0.16 yuan/(kwh).Controlled micro-source parameter as table 1, and to establish blowdown price be 0.068 yuan/kg.And suppose that micro-grid system and major network are signed an agreement and sell/buy electric maximum and can not surpass 40kw/h, sale of electricity electricity price be 0.65 yuan/(kwh), buy electricity price and be 1.0 yuan/(kwh).

The controlled micro-source of table 1 parameter

Under windows xp environment, adopt matlab2007 software as the solution musical instruments used in a Buddhist or Taoist mass of load prediction and economical operation.According to step 1, obtain the broad sense load data of 15 days a few days ago, totally 360 data.In these data of 15 days, within first 14 days, as input data, last day is as desired output data, training network.By the network training, only need the input broad sense load value of 2nd～15 days as input data, the broad sense load value that is prediction of output.After obtaining predicted value, adopt the method for step 2, utilize the exert oneself feature of stochastic problem of Markov Theory, predicated error is revised.

According to step 3, five, the target function that microgrid is set is:

$f=\underset{j=1}{\overset{24}{\mathrm{\&Sigma;}}}[\underset{i=1}{\overset{4}{\mathrm{\&Sigma;}}}{f}_i\left(p_{\mathrm{ij}}\right)+\mathrm{\&delta;}({\mathrm{\&mu;}}_j^b{p}_j^b-{\mathrm{\&mu;}}_j^s{p}_j^s)+(1-\mathrm{\&delta;})\mathrm{\&mu;}|r_j\left|\right]+-\mathrm{\&eta;}\underset{j=1}{\overset{24}{\mathrm{\&Sigma;}}}\underset{i=1}{\overset{4}{\mathrm{\&Sigma;}}}\underset{h=1}{\overset{k}{\mathrm{\&Sigma;}}}{e}_h\&CenterDot;K_{\mathrm{th}}\&CenterDot;p_{\mathrm{ij}}$

According to step 4, five, constraints (11)～(15) are set and by formula (20), are converted into penalty function, and solve p from equation (10) _4j(miniature gas turbine 2) about the expression formula of its dependent variable, and by the required penalty function of its substitution and target function.According to step 5, population quantity pnum=60 is set, iteration total degree tnum=200, individual cognition constant c ₁=2, the ripe c of social recognition ₂=2, initial inertia weight w _max=0.9, stop inertia weight w _min=0.1.And press the inertia weight that formula (19) generates iteration 200 times, according to the day before yesterday simultaneously segment data and when last segment data press formula (22) generation primary population.

Because micro-grid system has two kinds of operational modes of grid-connected island and various places different to the attention degree of environment, now scheme decision factor and the environment preference factor are got in the situation of different value 24 periods in micro-grid system one day are carried out to economic operation analysis.

1) under being incorporated into the power networks (operating scheme decision factor ρ=1), consider respectively micro-grid system (environment preference factor η=1) completely or do not consider the environmental impact formulation operational plan that micro-grid system (environment preference factor η=0) produces, knownly do not consider that microgrid environmental impact 24 hour operation cost is 4068.38 yuan, and consider that microgrid environmental impact 24 hour operation cost is 4480.61 yuan completely.And the schematic diagram of exerting oneself off the net is shown in respectively Fig. 4, Fig. 5.

1) under islet operation (operating scheme decision factor ρ=0), consider respectively micro-grid system (environment preference factor η=1) completely or do not consider the environmental impact formulation operational plan that micro-grid system (environment preference factor η=0) produces, knownly do not consider that microgrid environmental impact 24 hour operation cost is 4160.81 yuan, and consider that microgrid environmental impact 24 hour operation cost is 4607.45 yuan completely.The schematic diagram of exerting oneself under islet operation is shown in respectively Fig. 6, Fig. 7.

From above-mentioned four kinds of operating schemes, known operating scheme is different and be also different to the different scheduling result of environment attention degree, comparison diagram 4 and Fig. 5, and Fig. 6 and Fig. 7 are known, with the period, environment are paid attention to highlyer, and miniature gas turbine is exerted oneself just larger; And all very large from scheming known two kinds of miniature gas turbines ratio (actual motion is exerted oneself and accounted for the ratio of maximum output) of exerting oneself---miniature gas turbine 1 is exerted oneself ratio more than 80%, miniature gas turbine 2 is quota always, known miniature gas turbine not only has good environmental benefit but also has good economic benefit, compared with other distributed power sources, has larger competitive advantage.But the unit of miniature gas turbine installation cost is very high, with the installation cost of the miniature gas turbine of capacity, be approximately 2 times of diesel engine generator, 3 times of centralized thermal power generation, in addition miniature gas turbine operation expense and depreciable cost also very high, so the installed capacity of miniature gas turbine and installation scope are not very large, but along with more and more higher and novel CHP (cogeneration of heat and power) technology of the attention degree to environment can promote the application of miniature gas turbine to a certain extent.

Comparison diagram 4 and Fig. 6, Fig. 5 and Fig. 7 are known, no matter whether consider the impact of Environmental costs, the cost price being incorporated into the power networks is all smaller, and from Fig. 4, Fig. 5 is known while being incorporated into the power networks, and underload situation micro-grid system is to major network sale of electricity, overload or marginal generation cost are during higher than market electricity price, and micro-grid system is bought electricity to major network; And from Fig. 6, when Fig. 7 knows islet operation, micro-grid system, short of electricity can only be powered by energy storage device, and then to energy storage device, charges during underload, no matter charging or power supply, to energy storage device, it is all a kind of infringement, so strengthen the operating cost of micro-grid system, but consider the reliability of major network and microgrid, energy storage device still can not lack to a certain extent.

Claims (1)

1. the microgrid economical operation method of microgrid based on broad sense load prediction, described microgrid is a kind of miniature electric power networks being comprised of according to topological structure distributed power source and related load thereof; In described microgrid, contain blower fan, the uncontrollable micro-source of photovoltaic cell and for the energy storage device of energy snubber; Described microgrid possesses isolated island and grid-connected two kinds of running statuses; And while formulating microgrid operational plan, need to consider its impact on environment; It is characterized in that: the described microgrid economical operation method based on broad sense load prediction, concrete steps are as follows:

Step 1, broad sense load prediction; Gather exert oneself information data and photovoltaic of microgrid tradition load information data, blower fan and go out force data, regarding blower fan, photovoltaic as consumed power is negative load; The known broad sense load being comprised of microgrid load, blower fan and photovoltaic is:

p ^LD＝p ^load-p ^w-p ^pv (1)

In formula: p ^lDfor broad sense load value, p ^loadfor microgrid tradition load value, p ^wthe blower fan value of exerting oneself, p ^pvthe photovoltaic value of exerting oneself;

Step 2, the correction of broad sense load predicated error; Adopt BP neural net to predict broad sense load, statistical history broad sense load actual value and broad sense load predicted value, can predict that relative residual error is:

In formula:

for the broad sense load predicted value obtaining by BP neural network prediction, p ^lDfor broad sense load actual value;

The random relative error sequence of the non-stationary with Markov chain feature is divided into n state, and arbitrary state is S _i∈ [s _i', s _i"]; By adding up known, state S _ithe number of times occurring is m _i, by state S _ithrough k step, transfer to state S _jnumber of times be m _ij(k); Known, state transitions is P _ij(k)=m _ij(k)/mi (i=1,2, Ln)

A step state transition probability matrix is:

$P=\left[\begin{array}{cccc}{p}_{11}& p_{12}& L& p_{1n}\\ p_{21}& p_{22}& L& p_{2n}\\ M& M& O& M\\ p_{n1}& p_{n2}& L& p_{\mathrm{nn}}\end{array}\right]---\left(3\right)$

Matrix P meets: 1. right

n has 0≤p _ij(k)≤1

② $\&ForAll;i,j=\mathrm{1,2},L,$ N has $\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{p}_{\mathrm{ij}}\left(k\right)=1$

In formula: p _ijfor state S _ithrough a step, transfer to state S _jprobability;

By state-transition matrix, determined system future transfering state [s ' _i, s " _i], the correction predicted value of the h time is:

The target function of step 3, the economical operation of setting microgrid, the target function of described microgrid economical operation mainly comprises: operating cost function and Environmental costs function;

Described microgrid operating cost function comprises: the operating cost in controlled micro-source, with the mutual expense of electrical network and the energy storage device cost of exerting oneself, by operating cost function representation, be:

$F=\underset{j=1}{\overset{T}{\mathrm{\&Sigma;}}}[\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{f}_i\left(p_{\mathrm{ij}}\right)+\mathrm{\&rho;}({\mathrm{\&mu;}}_j^b{p}_j^b-{\mathrm{\&mu;}}_j^s{p}_j^s)+(1-\mathrm{\&rho;})\mathrm{\&mu;}|r_j\left|\right]---\left(5\right)$

In formula: N is the quantity in controlled micro-source in micro-grid system; T is p dispatching cycle _ijbe i controlled micro-source in period j power output:

for the j period under grid-connected pattern is from power distribution network input power;

for the j period, to power distribution network, buy in electricity price;

for the j period under isolated island formula is to power distribution network power output; for the j period, to power distribution network, sell electricity price; r _jfor exerting oneself of the energy storage device of j period; μ is the cost of exerting oneself of energy storage device; ρ is operational mode decision factor, ρ=0 micro-grid system islet operation, and in order to reduce the infringement of energy storage device, agreement energy storage device only moves under isolated island; ρ=1 micro-grid system is incorporated into the power networks;

F _i(p _ij) be the operating cost in each controlled micro-source, comprise energy consumption cost and maintenance cost;

f _i(p _ij)=h _i(p _ij)+w _i(p _ij) (6)

Wherein energy consumption cost is:

$h_i{p}_{\mathrm{ij}}=a_i{p}_{\mathrm{ij}}^2+b_i{p}_{\mathrm{ij}}+c_i---\left(7\right)$

In formula: a _ib _i, c _ifor the fuel cost coefficient of unit i,

Maintenance cost is:

w _t(p _ij)=d _ip _ij (8)

By Environmental costs function representation, be:

$E=\underset{j=1}{\overset{T}{\mathrm{\&Sigma;}}}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\underset{h=1}{\overset{k}{\mathrm{\&Sigma;}}}{e}_h\&CenterDot;K_{\mathrm{ih}}\&CenterDot;p_{\mathrm{ij}}---\left(9\right)$

In formula: e _hbe the blowdown price (unit/kg) of h kind pollutant, K _iit is the discharge capacity (kg/kw) of the controlled micro-source of i platform h kind pollutant;

Step 4, the constraints of microgrid economical operation is set, comprises: power-balance constraint, controlled micro-source units limits, controlled micro-source climbing rate constraint;

1) power-balance constraint: in described microgrid, the sum of exerting oneself of controlled micro-source, electrical network input, energy storage device equals broad sense load value in the same time, arbitrary period j is had:

$\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{p}_{\mathrm{ij}}+\mathrm{\&delta;}(p_j^b-p_j^s)+(1-\mathrm{\&delta;})r_j=p_j^{\mathrm{LD}}---\left(10\right)$

In formula: for the predicted value of j broad sense load constantly,

2) units limits in controlled micro-source

p _ij.min≤p _ij≤p _ij.max (11)

3) the climbing rate in controlled micro-source constraint

D _igΔt≤p _i(j+1)-p _ij≤U _igΔt (12)

D _i, U _ibe respectively maximum power climbing and the maximum power rate of descent of ramp-rate limits.Δ t is adjacent dispatch interval time;

4) with the mutual power constraint of power distribution network

Input power constraint:

$0\&le;p_j^b\&le;p_{\max }^b---\left(13\right)$

Power output constraint:

$0\&le;p_j^s\&le;p_{\max }^s---\left(14\right)$

5) energy storage device operation constraint

r _min≤|r _j|≤r _max (15)

Step 5, the introducing environment preference factor are converted into objective programming by Multiobjective Programming;

f=F+ηE (16)

In formula: η is the environment preference factor; Adopt standard particle group algorithm, its velocity location more new formula is as follows:

$v_i^{t+1}=w\&CenterDot;v_i^t+c_1\&CenterDot;r_1\&CenterDot;(p_i^t-x_i^t)+c_2\&CenterDot;r_2\&CenterDot;(p_g^t-x_i^t)---\left(17\right)$

$x_i^{t+1}=x_i^t+v_i^{t+1}---\left(18\right)$

In formula: v _i=[v _i1, v _i2l, v _in] be the speed of particle i, represent the distance between particle i present position and its next step position; x _i=[x _i1, x _i2l, x _in] be the current location of particle i, p _ifor current individual optimal value, p _gfor current colony optimal value, w is inertia weight, c ₁for individual cognition constant, c ₂for social recognition constant; Inertia of design weight declines in the following manner:

$w=(w_{\max }-w_{\min }){\left(\frac{t}{\mathrm{tnum}}\right)}^{\frac{1}{2}}---\left(19\right)$

In formula: w _maxfor initial inertia weight, w _minfor stopping inertia weight, tnum is termination of iterations number of times, and t is current iteration number of times; Constraints is counted to fitness function with the form of penalty function:

By institute tangible as: a≤p (X)≤b inequality constraints, be converted into shape as g (X)≤0 inequality constraints, transform as follows:

$g\left(X\right)=|p\left(X\right)-\frac{a+b}{2}|-\frac{b-a}{2}---\left(20\right)$

, fitness function is:

Fit=f+M·∑(max(0，g _i(X) ²))=F+ηE+M·∑(max(0，g _i(X))) ² (21)

In formula: M is a positive number more much larger than f;

Particle population is divided into three parts, and portion solves search around in the same time at proxima luce (prox. luc), and portion solves search around at eve, and remaining portion carries out global search between confining region, that is:

$P(\mathrm{dnum},\mathrm{pnum})=\left\{\begin{array}{c}{p}^{\&prime;}+\frac{1}{10}\&CenterDot;(p_{\max }-p_{\min })\&CenterDot;(-1+2\&CenterDot;\mathrm{rand}(\mathrm{dnum},\frac{1}{3}\&CenterDot;\mathrm{pnum}))\\ p^{\&prime;\&prime;}+\frac{1}{5}\&CenterDot;(p_{\max }-p_{\min })\&CenterDot;(-1+2\&CenterDot;\mathrm{rand}(\mathrm{dnum},\frac{1}{3}\&CenterDot;\mathrm{pnum}))\\ p_{\min }+(p_{\max }-p_{\min })\&CenterDot;\mathrm{rand}(\mathrm{dnum},\frac{1}{3}\&CenterDot;\mathrm{pnum})\end{array}\right.---\left(22\right)$

In formula: p ' is the situation of exerting oneself of controlled micro-source in the same time of proxima luce (prox. luc), electrical network, energy storage device, and " be the situation of exerting oneself of the controlled micro-source of previous moment, electrical network or energy storage device, dnum is particle dimension to p, and pnum is population quantity.

Images