CN103345662A - Commerce efficiency comprehensive evaluation method based on small-wave network method - Google Patents

Abstract

The invention relates to a commerce efficiency comprehensive evaluation method based on a small-wave network method. The method comprises the following steps that 1) an efficiency item comprehensive evaluation index system is established; 2) each item index value of the comprehensive evaluation index system is computed; 3) the best weight of each index is determined with the small-wave network method; 4) comprehensive evaluation values of alternative schemes are computed, and the best scheme is output. Network parameters wij, rj, bj and aj are computed in an iterative mode with the small-wave network method, and accordingly the parameter value which is closest to the expert score of each training scheme is found, finally, the comprehensive evaluation results of the efficiency schemes are obtained, different characteristics of different schemes are integrated, advantages and disadvantages of the schemes are measured comprehensively, quantitative evaluation results are obtained, and therefore an item decision maker can directly compare the overall advantages and disadvantages of the schemes and select the efficiency schemes. The small-wave network method is used in commerce efficiency comprehensive evaluation, computing errors are small, and convergence speed is high.

Description

A kind of commercial efficiency integrated evaluating method based on the wavelet network method

Technical field

The present invention relates to efficiency synthetic evaluation of projects method field, particularly relate to a kind of commercial efficiency integrated evaluating method based on the wavelet network method.

Background technology

The purpose of the efficiency project being carried out comprehensive evaluation is to intend related economic, environment and the social benefit that the enforcement project can be brought expection contribution effect and the clear and definite project of investor, grid company, terminal user and the whole society in order to weigh.Evaluation index is the basis of carrying out above-mentioned scientific evaluation.

Present stage is relatively ripe in the efficiency project evaluation method of industry, building field, production equipment commonly used in the medium and small sized enterprises, as boiler and vapour system, electric systems such as water pump, blower fan and air compressor, illuminator, heating ventilation air-conditioning system has all been set up corresponding efficiency calculation and analysis on energy saving effect model.But these computation models have been ignored efficiency and have been improved environmental benefit and the social benefit of bringing only from economize on electricity level and the economic degree of equipment, and evaluation content lacks integrality.

Similarly, the efficiency evaluation index of existing commercial field often is confined to the power saving capability of each terminal consumer, as air-conditioning, refrigerator, washing machine, the efficiency evaluation of transformer, do not consider after the efficiency project implementation influence to economy, environment and the social benefit of investor, grid company, or do not set corresponding quantized value, make each efficiency scheme not have unified judgment criteria, a certain purpose efficiency characteristics that various independently energy efficiency indexes reflect can not provide project decision person and investor to select effective suggestion of capital project.

Summary of the invention

The present invention has proposed first and can carry out comprehensive energy efficiency indexes system comprehensive, scientifical, objective evaluation to different efficiency schemes, adopts the wavelet network method to network parameter wi _j, r _j, b _j, a _jCarry out iterative computation, thereby find expert with each training program immediate parameter value of giving a mark, finally obtain the comprehensive evaluation result of each efficiency scheme, the different characteristics of comprehensive different schemes, the quality of comprehensive and systematic each scheme of measurement, provide the evaluation result of quantification, make project decision person can flat-footedly compare the overall good and bad situation of each scheme, thereby carry out the efficiency Scheme Choice.The wavelet network method is applied to commercial efficiency comprehensive evaluation, and the error of calculation is little, fast convergence rate.

The present invention makes best engineering decision at current efficiency limiting factor in order to obtain science, objective, correct evaluation result with guidance, the purpose that has held on to the efficiency engineering project designs evaluation index, and the aspects such as economy, rentability and outside social benefit that choosing of each index all is centered around the efficiency project and produces are considered.In addition, the design of index all has practical application, helps collection and the quantification of data like this, thus the evaluation index of generalities convert to we understand easily and the index that obtains is handled.

(wavelet network WN) is the product that wavelet theory combines with neural network to wavelet network, and it is the fusion of wavelet decomposition and feedforward neural network.Owing to had both good time-frequency localization character and the Neural Network Self-learning function of wavelet transformation, wavelet network progressively becomes the new method that evaluation, prediction field are adopted.Multiattribute comprehensive evaluation for the complex object system, on the basis of unified pointer type, utilize the dimensionless number certificate of evaluation index, by the study of wavelet network, obtain expertise, set up by the Nonlinear Mapping relation of evaluation index property value to the output comprehensive evaluation value.When other similar problems are estimated, import the achievement data vector of object to be evaluated, can obtain its comprehensive evaluation value through network calculations, thereby reach automatic operation, fast the purpose of evaluation and decision support.

Purpose of the present invention is achieved through the following technical solutions:

A kind of commercial efficiency integrated evaluating method based on the wavelet network method may further comprise the steps:

1) sets up efficiency synthetic evaluation of projects index system;

2) calculate each single index value of System of Comprehensive Evaluation;

3) method of utilization wavelet network is determined the optimal weight of each index;

4) calculate the comprehensive evaluation value of each alternatives, the output optimal case.

Described step (1) is set up efficiency synthetic evaluation of projects index system and is comprised first class index and corresponding two-level index thereof, each two-level index is extended to three grades of indexs that embody index particular content at the corresponding levels again separately, has namely formed the tertiary level structure of domination from top to bottom.

Described first class index content comprises: operation of power networks, environmental effect, economic benefit and four aspects of the level of resources utilization.

Described two-level index is estimated two aspect contents by electric network reliability evaluation and load shaping capability and is formed; Environmental effect in the described first class index partly comprises pollutant emission and carbon emission two aspect contents; Economic benefit in described first class index part is by comprising that mainly financial evaluation, gain on investments evaluation and three aspect contents of analysis of Life Cycle Cost form; The level of resources utilization in the described first class index comprises that partly energy-saving effect and grid side resource use two aspect contents.

The reliability evaluation of the two-level index of the System of Comprehensive Evaluation of described foundation comprises system's average short of electricity electric weight ASCI rate of change and two indexs of the total electric weight of system ENS rate of change in shortage; Load shaping capability in the described two-level index partly comprises daily load rate, day peak valley rate, yearly load factor and a year maximum peak valley rate and changes four indices; Pollutant emission in the described two-level index partly comprises oxides of nitrogen CER, oxysulfide CER, dust CER and PM2.5 CER four indices; Carbon emission one deck in the described two-level index corresponds to CO ₂The CER index; Financial evaluation part in the described two-level index is carried out systematic analysis by net present value (NPV), internal rate of return and investment payback time three indexs and is calculated; Gain on investments in the described two-level index is partly chosen income expense ratio PCR and can be avoided electric cost AC2 item index to weigh the gain on investments situation of efficiency project; Overall life cycle cost in the described two-level index partly adopts longevity current cost rate of descent as the evaluation index of weighing efficiency project life-cycle cycle economy; Energy-saving effect in the described two-level index partly expands to terminal year amount of electricity saving and two indexs of electricity consumption income efficient difference.

Three grades of index weightses adopt the method for wavelet network to find the solution in the described assessment indicator system, adopt the linear weighted function overall approach that alternatives is carried out comprehensive scoring then, choose the comprehensive grading soprano and are optimal case, and concrete implementation step is:

1) collects commercial efficiency project foundation data, calculate three grades of desired values in the System of Comprehensive Evaluation;

2) according to an expert view, tentatively give a mark for optional efficiency scheme, the expert who obtains each scheme appraises the mark set through discussion

3) with each the value attribute value vector { x that calculates in the step 1) _k(i) } be converted into the data { r of index attribute unanimity _k(i) };

Three types index is done following processing:

In the formula: r _IjBe j desired value in i the scheme, r _Ij ^*Be r _IjTreated value; M is the programme number; r _jIt is the desired quantity of j index;

4) give wavelet network parameter w _Ij, r _j, b _j, a _jInitial value at random and give the max calculation times N;

5) with the unification desired value r of each evaluation of programme _k(i) be input in the wavelet network computing formula, try to achieve corresponding comprehensive evaluation value y _iAnd calculate corresponding energy error E;

The wavelet network computing formula is:

$y_k=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{r}_{j}h\left[\frac{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{w}_{\mathrm{ij}}{r}_k\left(i\right)-b_j}{a_j}\right]$

The error energy function of network is:

$E=\frac{1}{2}\underset{k=1}{\overset{P}{\mathrm{\Σ}}}{({\hat{y}}_k-y_k)}^2$

6) gradient vector of calculating wavelet network;

Order

${\mathrm{\λ}}_k\left(j\right)=\frac{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{w}_{\mathrm{ij}}{r}_k\left(i\right)-b_j}{a_j}$

Have:

$g\left(w_{\mathrm{ij}}\right)=\frac{\∂E}{\∂w_{\mathrm{ij}}}=\underset{k=1}{\overset{P}{\mathrm{\Σ}}}({\hat{y}}_k-y_k)\left(\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{r}_j\frac{\∂h}{\∂{\mathrm{\λ}}_k\left(j\right)}\frac{r_k\left(i\right)}{a_j}\right)$

$g\left(r_j\right)=\frac{\∂E}{\∂r_j}=\underset{k=1}{\overset{P}{\mathrm{\Σ}}}({\hat{y}}_k-y_k)h\left({\mathrm{\λ}}_k\left(j\right)\right)$

$g\left(a_j\right)=\frac{\∂E}{\∂a_j}=-\underset{k=1}{\overset{P}{\mathrm{\Σ}}}({\hat{y}}_k-y_k)\left(\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{r}_j\frac{\∂h}{\∂{\mathrm{\λ}}_k\left(j\right)}\frac{{\mathrm{\λ}}_k\left(i\right)}{a_j}\right)$

$g\left(b_j\right)=\frac{\∂E}{\∂b_j}=-\underset{k=1}{\overset{P}{\mathrm{\Σ}}}({\hat{y}}_k-y_k)\left(\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{r}_j\frac{\∂h}{\∂{\mathrm{\λ}}_k\left(j\right)}\frac{1}{a_j}\right)$

Wherein,

$\frac{\∂h}{\∂{\mathrm{\λ}}_k\left(j\right)}=-\cos \left(1.75{\mathrm{\λ}}_k\left(j\right)\right)\exp (-\frac{{\mathrm{\λ}}_k^2\left(j\right)}{2}){\mathrm{\λ}}_k\left(j\right)-1.75\sin \left(1.75{\mathrm{\λ}}_k\left(j\right)\right)\exp (-\frac{{\mathrm{\λ}}_k^2\left(j\right)}{2})$

7) adopt method of conjugate gradient to adjust network parameter, t is iterations: order

$S_t\left(w_{\mathrm{ij}}\right)=\left\{\begin{array}{cc}-g_t\left(w_{\mathrm{ij}}\right),& t=1\\ -g_t\left(w_{\mathrm{ij}}\right)+\frac{\left|\right|g_t\left(w_{\mathrm{ij}}\right)\left|\right|}{\left|\right|g_{t-1}\left(w_{\mathrm{ij}}\right)\left|\right|}{S}_{t-1}\left(w_{\mathrm{ij}}\right),& t>1\end{array}\right.$

In like manner, can calculate S _t(r _j), S _t(a _j), S _t(b _j);

Then the adjusting of network parameter is as follows:

w _(t)ij＝w _(t-1)ij+αS _t-1(w _ij)

r _(t)j＝r _(t-1)j+βS _t-1(r _j)

a _(t)j＝a _(t-1)j+γS _t-1(a _j)

b _(t)j＝b _(t-1)j+ηS _t-1(b _j)

8) return step 4), till the error energy functional value of network is not more than set-point ε or calculation times and surpasses the max calculation times N.

9) according to the index weights of determining in the step 8), the method for employing linear weighted function is calculated the comprehensive evaluation value of each alternatives, and choosing the highest scheme of mark is optimal case.

The invention has the advantages that:

1) the present invention follows setting principle and the basic skills of assessment indicator system, characteristics in conjunction with the actual efficiency engineering of China, from operation of power networks, environmental effect, set out in four aspects of economic benefit and the level of resources utilization, by taking all factors into consideration, resident and commercial efficiency project evaluation index system that one cover quantitatively combines with qualitative analysis are at first proposed, comprehensively, objective, react the various aspects of each efficiency scheme exactly, fully having taken into account the efficiency project is power consumer, Utilities Electric Co., differences such as ESCO (ESCO) and the whole society participate in the interests situation that main body is brought, and can weigh the Expected Results whether the expection output can reach (comprising directly and indirect benefit) investment decision person comprehensively.

2) the present invention has overcome the subjectivity of traditional evaluation method, method by wavelet network, found each the bottom index weights that adapts with expertise, make evaluation result both meet expertise, by normalization the content of different index reactions is quantized to concrete numerical value again, for different alternativess, project investment person can directly draw optimal case from final comprehensive grading situation.

3) the present invention impels the power consumer that participates in the efficiency plan to pass through to adopt scientific management methods and advanced technical equipment save power and minimizing electricity needs, obtains maximum electricity charge saving in cycle life-cycle thereby be desirably in; The efficiency project implementation can change original part throttle characteristics effectively, realizes that " peak clipping " to reduce load level peak period, improves the network load rate, optimizes operation of power networks, improves stability, reliability and the economy of Operation of Electric Systems.The terminal power consumer can drop into the energy still less and satisfy himself to the demand of electric power effectiveness, and the reduction of energy use amount will promote high-quality and the security of national energy supply.

Description of drawings

Fig. 1 is wavelet network method step synoptic diagram;

Fig. 2 is concrete implementation step;

Fig. 3 is System of Comprehensive Evaluation.

Embodiment

The present invention is described in detail below in conjunction with accompanying drawing, and Fig. 1 is wavelet network method step synoptic diagram, among Fig. 1 with each property value vector { x _k(i) } (m altogether) carries out the processing of index unification, obtains the data { r of index attribute unanimity _k(i) }, with the unification desired value r of each evaluation of programme _k(i) be input in the wavelet network computing formula, try to achieve corresponding comprehensive evaluation value y _iAnd calculate corresponding energy error E.

The present invention is based on the commercial efficiency project evaluation method of wavelet network method as Fig. 2, wherein the integrated evaluating method of efficiency project should comprise at least: set up the energy efficiency indexes appraisement system, calculate each single index value of System of Comprehensive Evaluation, use the method for wavelet network to determine several partial contents such as the optimal weight of each index, the comprehensive evaluation value that calculates each alternatives, output optimal case:

One, the foundation of System of Comprehensive Evaluation.

Characteristics in conjunction with the actual efficiency engineering of China, the present invention is from four aspects such as operation of power networks, environmental effect, economic benefit and the level of resources utilizations, by taking all factors into consideration, resident and commercial efficiency project evaluation index system that one cover quantitatively combines with qualitative analysis are at first proposed, as shown in the table.Comprise 4 first class index as this system of Fig. 3,9 two-level index and 20 three grades of indexs.Wherein, then on this basis, again each class index is done brief analysis and explanation.

Commercial efficiency project evaluation index system

1. operation of power networks

The present invention mainly carries out evaluation analysis to the efficiency engineering to the influence of operation of power networks from reliability and 2 aspects of load shaping capability.

At the influence of efficiency project to operation of power networks reliability level, main average short of electricity electric weight (ASCI) rate of change of analytic system and two indexs of the total electric weight of system (ENS) in shortage rate of change among the present invention.The former is characterized in the official hour, the intensity of variation of the average damaged electric weight of each user in the system; The latter then refers to system's total electric weight undersupply quantitative change ratio at the appointed time.

(1) the average short of electricity electric quantity change of system rate

The computing formula of the average short of electricity electric weight of system (ASCI) is as follows:

$P_{\mathrm{ASCI}}=\frac{P_{\mathrm{ENS}}}{\mathrm{\Σ}{N}_i}=\frac{\mathrm{\Σ}{t}_{\mathrm{ui}}{P}_i}{\mathrm{\Σ}{N}_i}---(1-1)$

Wherein, N _iBe the total number of users (user) of load point i, P _ENSBe the total electric weight of system (WMh/) in shortage;

Based on formula 1-1, can the average short of electricity electric quantity change of computing system rate index as follows:

$I_{\mathrm{ASCI}}=\frac{P_{\mathrm{ASCI}}^2-{\mathrm{<figure-callout\; id="1"\; label="P\; ASCI"\; filenames="HDA00003506521300011.png"\; state="\{\{state\}\}">P</figure-callout>}}_{\mathrm{ASCI}}^{}}{\mathrm{\&Delta;t}}---(1-2)$

Wherein, Δ t is a stipulated time section (year),

Be respectively the efficiency project implementation front and back average short of electricity charge value of system (MWh/(user's year)).

(2) the total electric weight of system rate of change in shortage

The computing formula of the total electric weight of system (ENS) in shortage is as follows:

P _ENS＝Σt _uiP _i （1-3）

Wherein, P _iBe the average load (kW) of load point i, t _UiAnnual power off time (h/) for load point i.

Based on formula 1-formula 4, can the total electric weight of computing system rate of change index in shortage as follows:

$I_{\mathrm{ENS}}=\frac{P_{\mathrm{ENS}}^2-{\mathrm{<figure-callout\; id="1"\; label="P\; ENS"\; filenames="HDA00003506521300011.png"\; state="\{\{state\}\}">P</figure-callout>}}_{\mathrm{ENS}}^{}}{\mathrm{\&Delta;t}}---(1-4)$

Wherein,

Be respectively the not enough value (kWh/) of the total electric weight of efficiency project implementation front and back system.

The efficiency resource is mainly reflected on the change ability to rate of load condensate and peak-valley difference the impact effect of part throttle characteristics.(1) the daily load rate ratio of average power load and day peak load in a few days, it is calculated as follows:

γ＝P _d.av/P _d.max （1-5）

Wherein, P _D.avBe per day load, P _D.maxBe the day peak load;

(2) day peak valley rate ratio of peak-valley difference and day peak load in a few days, it is calculated as follows:

$\mathrm{\&sigma;}=\frac{P_{d.\max }-P_{d.\min }}{P_{d.\max }}---(1-6)$

Wherein, P _D.minBe the day minimum load;

(3) yearly load factor refers to the ratio of average of the whole year load and year peak load, and it is calculated as follows:

$\mathrm{\&delta;}=\frac{P_{y.\mathrm{av}}}{P_{y.\max }}---(1-7)$

Wherein, P _Y.avBe annual load, P _Y.maxBe the year peak load;

(4) year maximum peak valley rate refers to the mean value of annual day peak valley rate, and it is calculated as follows:

$\stackrel{\&OverBar;}{\mathrm{\&sigma;}}=\frac{\mathrm{\&Sigma;}{\mathrm{\&sigma;}}_i}{365}---(1-8)$

Wherein, σ _iIt is i days day peak valley rate.

2. environmental effect

Consider all kinds of disposals of pollutants to situations such as social influence degree and China's targets for energy-saving and emission-reduction, the present invention analyzes the environmental effect implementation evaluation that the efficiency engineering produces from pollutant emission and two aspects of carbon emission.

Its computing method are as follows:

(1) oxygen sulfur compound CER

Wherein:

A _SO2---cycle life-cycle SO ₂CER, t;

λ _SO2---the generated energy SO of unit ₂Discharging intensity, t/kWh;

W _Zc---cycle terminal life-cycle can be avoided electric weight, kWh;

L---the comprehensive electric quantity loss rate of electric power system, %.

(2) oxynitrides CER

Wherein:

A _FC---cycle life-cycle dust CER, t;

λ _FC---unit generated energy dust emission intensity, t/kWh;

(3) dust CER

A _FC＝λ _FCW _zc/(1-l) （1-11）

Wherein:

A _FC---cycle life-cycle dust CER, t;

λ _FC---unit generated energy dust emission intensity, t/kWh;

(4) PM2.5 CER

A _pm2.5＝λ _pm2.5W _zc/(1-l) （1-12）

Wherein:

Apm2.5---cycle life-cycle PM2.5 particle CER, t;

λ pm2.5---the generated energy PM2.5 of unit particulate emission intensity, t/kWh;

3. economic benefit

The present invention mainly comprises financial evaluation, gain on investments evaluation and three aspects of analysis of Life Cycle Cost to the economy evaluation of efficiency project.

(1) net present value (NPV)

In investment project was estimated, Index of Net Present Value NPV was one of most important index.Net present value (NPV) refers to the base earnings ratio i by power industry _k, with discount present worth sum at the beginning of time horizon of vestment of the free cash flow of each year in the efficiency project given period, its expression formula is:

$\mathrm{NPV}=\underset{t=0}{\overset{n}{\mathrm{\&Sigma;}}}{(\mathrm{CI}-\mathrm{CO})}_t{(1+i_K)}^{-t}---(1-13)$

Wherein:

CI _t---cash inflow;

CO _t---cash flow;

i _k---the base earnings ratio of power industry;

The lifetime of n---project.

(2) internal rate of return (IRR)

Internal rate of return irr is to instigate its expression formula of the null discount rate of net present value (NPV) of net cash flow to be:

$\underset{t=0}{\overset{n}{\text{\&Sigma;}}}{(\mathrm{CI}-\mathrm{CO})}_t{(1+\mathrm{IRR})}^{-t}=0---(1-14)$

Wherein:

CI---cash inflow;

CO---cash outflow;

(CI-CO) t---the net cash flow of t.

(3) static payback time (Pt)

Static payback time Pt is under the condition of not considering the time value of money, compensates for the project needed time of fully invested with the project net proceeds, and its expression formula is:

$\underset{t=0}{\overset{P_t}{\text{\&Sigma;}}}{(\mathrm{CI}-\mathrm{CO})}_t=0---(1-15)$

Gain on investments refers to enterprise and invests the economic interests that obtain.In assessment indicator system in this paper, to choose income expense ratio (PCR) and can avoid 2 indexs such as electric cost (AC) to weigh the gain on investments situation of efficiency project, its computing formula is as follows respectively:

$\mathrm{PCR}=\frac{d_{z,b}}{d_{z,s}}*100\%---(1-16)$

$\mathrm{AC}=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{W}_{z,i}{d}_i---(1-17)$

Wherein, d _{Z, b}With d _{Z, s}Be respectively terminal user unit's electricity consumption avoidable cost (unit/(kWh)); W _{Z, i}And d _iBe respectively the terminal user because of the efficiency transformation in the electric weight avoided (kWh) of period i and corresponding sale of electricity electricity price thereof (unit/(kWh)).N then is the total period number in cycle life-cycle.Generally speaking, resident and commercial user's efficiency project life-cycle period expense (LCC) model can be expressed as:

LCC＝C _I+C _O+C _M+C _F+C _D （1-18）

Wherein:

LCC---equipment overall life cycle cost;

CI---disposable input cost comprises the design cost of design phase, the equipment purchase cost of construction period, construction and installation cost;

CO---operating cost comprises equipment loss, operating staff training's cost and energy consumption cost;

CM---maintenance cost;

CF---failure cost comprises loss of outage, social influence loss etc.;

CD---obsolescence cost.

Based on following formula, the computing formula that further provides longevity current cost rate of descent (LCCR) is as follows:

$\mathrm{LCCR}=\frac{{\mathrm{LCC}}_0-\mathrm{LCC}}{{\mathrm{LCC}}_0}*100\%---(1-19)$

Wherein, LCC0 is common consumer overall life cycle cost expense; LCC is the overall life cycle cost expense of relevant device after the efficiency transformation.

The level of resources utilization:

Consider the efficiency project for the significance that improves terminal user and grid side resource utilization, this paper from 2 aspects such as energy-saving effect and grid side resource utilizations to this evaluation study.

Terminal year amount of electricity saving, refer to all kinds of efficiency equipment put into operation before and after all kinds of terminal users always use the difference of power consumption, its computing formula is as follows:

Wherein, Be electricity consumption rule coefficient of variation between different user, be used for adjusting electricity consumption simultaneity and consistance between different user; Total number of users that F relates to for the efficiency project; Q _{T, i, 0}Be former the use power consumption of terminal electricity consumption link i in period t; λ _iThe indicative function of expression electricity consumption link i, when carrying out the efficiency transformation in this link, this variable-value is 1, otherwise is 0; α _{EE, i}Be the put into operation consumption reduction rate of back electricity consumption link i of efficiency equipment; ζ _tFor each electricity consumption link among the terminal user at the simultaneity factor of period t.In addition, T is total period number in research cycle; K then is the total electricity consumption link number of terminal user.

The computing formula of electricity consumption income efficient difference is as follows:

$\mathrm{CE}=\left(\frac{Q_0-Q_e}{\mathrm{Ep}}\right)*100\%---(1-21)$

Q wherein ₀And Q _eBe respectively the put into operation total electricity consumption of terminal user in forward and backward research cycle of efficiency equipment.Ep is economic output total amount.For the commercial user, Ep can be taken as the sales volume in research cycle; For the resident, Ep can be taken as the corresponding GDP value of kinsfolk's number.

At the delayed-action of efficiency investment to the electrical network dilatation, proposition can be exempted from the power supply capacity index and be quantized to calculate, and its expression formula is as follows

Wherein, I _GridThe accumulative total electrical network capacity of expansion that reduces because of terminal power saving for power grid enterprises (comprising corresponding margin capacity);

Ratio for the electrical network margin capacity.

In System of Comprehensive Evaluation set forth above, fully having taken into account the efficiency project is that differences such as power consumer, Utilities Electric Co., ESCO (ESCO) and the whole society participate in the interests situation that main body is brought.In practical engineering application, need be according to the benefited condition of project implementation purpose and different subjects, therefrom select the index that is fit to flexibly, further implement computational analysis.

Two, calculate each three grades of desired value of System of Comprehensive Evaluation;

Collect commercial efficiency project foundation data, determine correlation computations parameter in three grades of index computing formula, calculate the bottom desired value in the System of Comprehensive Evaluation.

Three, the method for utilization wavelet network is determined the optimal weight of each index;

1) according to an expert view, tentatively give a mark for optional efficiency scheme, the expert who obtains each scheme appraises the mark set through discussion

2) with each the property value vector { x that calculates in the step (1) _k(i) } be converted into the data of index attribute unanimity $\left\{r_k\left(i\right)\right\}.$

3) give wavelet network parameter w _Ij, r _j, b _j, a _jInitial value at random and give the max calculation times N.

4) with the unification desired value r of each evaluation of programme _k(i) be input in the wavelet network computing formula, try to achieve corresponding comprehensive evaluation value y _iAnd calculate corresponding energy error E.

The wavelet network computing formula is:

$y_k=\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{r}_{j}h\left[\frac{\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ij}}{r}_k\left(i\right)-b_j}{a_j}\right]$

The error energy function of network is

$E=\frac{1}{2}\underset{k=1}{\overset{P}{\mathrm{\&Sigma;}}}{({\hat{y}}_k-y_k)}^2$

5) gradient vector of calculating wavelet network.

Order

${\mathrm{\&lambda;}}_k\left(j\right)=\frac{\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ij}}{r}_k\left(i\right)-b_j}{a_j}$

Have:

$g\left(w_{\mathrm{ij}}\right)=\frac{\&PartialD;E}{\&PartialD;w_{\mathrm{ij}}}=\underset{k=1}{\overset{P}{\mathrm{\&Sigma;}}}({\hat{y}}_k-y_k)\left(\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{r}_j\frac{\&PartialD;h}{\&PartialD;{\mathrm{\&lambda;}}_k\left(j\right)}\frac{r_k\left(i\right)}{a_j}\right)$

$g\left(r_j\right)=\frac{\&PartialD;E}{\&PartialD;r_j}=\underset{k=1}{\overset{P}{\mathrm{\&Sigma;}}}({\hat{y}}_k-y_k)h\left({\mathrm{\&lambda;}}_k\left(j\right)\right)$

$g\left(a_j\right)=\frac{\&PartialD;E}{\&PartialD;a_j}=-\underset{k=1}{\overset{P}{\mathrm{\&Sigma;}}}({\hat{y}}_k-y_k)\left(\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{r}_j\frac{\&PartialD;h}{\&PartialD;{\mathrm{\&lambda;}}_k\left(j\right)}\frac{{\mathrm{\&lambda;}}_k\left(i\right)}{a_j}\right)$

$g\left(b_j\right)=\frac{\&PartialD;E}{\&PartialD;b_j}=-\underset{k=1}{\overset{P}{\mathrm{\&Sigma;}}}({\hat{y}}_k-y_k)\left(\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{r}_j\frac{\&PartialD;h}{\&PartialD;{\mathrm{\&lambda;}}_k\left(j\right)}\frac{{\mathrm{\&lambda;}}_k\left(i\right)}{a_j}\right)$

Wherein,

$\frac{\&PartialD;h}{\&PartialD;{\mathrm{\&lambda;}}_k\left(j\right)}=-\cos \left(1.75{\mathrm{\&lambda;}}_k\left(j\right)\right)\exp (-\frac{{\mathrm{\&lambda;}}_k^2\left(j\right)}{2}){\mathrm{\&lambda;}}_k\left(j\right)-1.75\sin \left(1.75{\mathrm{\&lambda;}}_k\left(j\right)\right)\exp (-\frac{{\mathrm{\&lambda;}}_k^2\left(j\right)}{2})$

6) adopt method of conjugate gradient (Fletcher-Reeves formula) to adjust network parameter (t is iterations):

Order

$S_t\left(w_{\mathrm{ij}}\right)=\left\{\begin{array}{cc}-g_t\left(w_{\mathrm{ij}}\right),& t=1\\ -g_t\left(w_{\mathrm{ij}}\right)+\frac{\left|\right|g_t\left(w_{\mathrm{ij}}\right)\left|\right|}{\left|\right|g_{t-1}\left(w_{\mathrm{ij}}\right)\left|\right|}{S}_{t-1}\left(w_{\mathrm{ij}}\right),& t>1\end{array}\right.$

In like manner, can calculate S _t(r _j), S _t(a _j), S _t(b _j);

Then the adjusting of network parameter is as follows:

w _(t)ij＝w _(t-1)ij+αS _t-1(w _ij)

r _(t)j＝r _(t-1)j+βS _t-1(r _j)

a _(t)j＝a _(t-1)j+γS _t-1(a _j)

b _(t)j＝b _(t-1)j+ηS _t-1(b _j)

7) return step (4), till the error energy functional value of network is not more than set-point ε or calculation times and surpasses the max calculation times N.

Four, calculate the comprehensive evaluation value of each alternatives, the output optimal case

According to the index weights of determining before, the method for employing linear weighted function is calculated the comprehensive evaluation value of each scheme, and choosing the highest scheme of mark is optimal case.

Be illustrated in figure 3 as the wavelet network structural drawing that many index comprehensives are estimated, input layer, hidden layer and output layer have m, n and 1 cell node respectively among the figure.Here get

$y_k=\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{r}_{j}h\left[\frac{\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ij}}{r}_k\left(i\right){-b}_j}{a_j}\right]$

In the formula, x _k(i), r _k(i) raw data and the nondimensionalization data of the index i of expression input sample k respectively.Wi _j, r _jThe expression weight coefficient, b _j, a _jShift factor and the contraction-expansion factor of representing wavelet basis respectively.Basic small echo herein adopts the female small echo of the high bass wave of the external normal cosine modulation of using: Morlet.Its form is

$h\left(t\right)=\cos \left(1.75t\right)\exp (-\frac{t^2}{2})$

The corresponding index attribute value vector of evaluation sample k { r with the complex object system _k(i) } as the input of wavelet network, Dui Ying comprehensive evaluation value with it Desired output as network.The error energy function of define grid is:

$E=\frac{1}{2}\underset{k=1}{\overset{P}{\mathrm{\&Sigma;}}}{({\hat{y}}_k-y_k)}^2$

In the formula, y _kEstimate actual value for estimating sample k,

Be the network output valve.P is for estimating total sample number.By network parameter wi _j, r _j, b _j, a _jAdjustment, make the error energy function of network reach minimum.

It is as follows that the evaluation method based on the commercial efficiency project of wavelet network method that the present invention will propose is applied in the actual efficiency project evaluation example:

1. the present invention transform example as with certain sowntown efficiency, has proposed 8 cover efficiency modification schemes in the planning region, represents with P1～P8.

2. calculate each single index value of System of Comprehensive Evaluation

The content of this evaluation procedure selection environment effect, economic benefit, resource utilization three aspects is object as a comparison, relates to 8 of three grades of evaluation indexes altogether, represents with R1～R8.Give a mark according to the related data of collecting and expert, and respectively each single index of efficiency modification scheme is carried out the dimensionless processing according to preceding method, obtain following data, specifically as shown in table 1:

Each efficiency evaluate alternatives data of table 1

Index	R1	R2	R3	R4	R5	R6	R7	R8	Score
										P1	0.78	0.80	0.92	1.00	0.40	0.58	0.67	0.77	0.83
P2	0.88	0.67	1.00	0.56	0.69	0.29	0.50	0.91	0.78
										P3	0.92	0.89	0.94	0.88	1.00	0.86	0.98	0.78	0.93
P4	1.00	0.89	0.97	0.66	0.87	0.96	0.79	0.95	0.92
										P5	0.89	1.00	0.83	0.78	0.96	0.97	0.84	0.78	0.89
P6	0.83	0.78	0.79	0.68	0.93	0.89	0.94	0.88	0.83
										P7	0.93	0.98	0.86	0.89	0.93	0.79	0.83	1.00	0.95
P8	0.87	0.95	0.89	0.93	0.92	1.00	0.97	0.96	0.94

3. use the optimal weight of determining each index based on the commercial efficiency integrated evaluating method of wavelet network method

With preceding 5 groups of data as the efficiency transformation project of having selected in the table 1, train this wavelet network as training set, the back is object to be evaluated for 3 groups.Training result is as shown in table 2, and they are very approaching with the output of expectation.

Table 2 training result

The scheme numbering	P1	P2	P3	P4	P5
						Desired output	0.83	0.78	0.93	0.92	0.89
Training result	0.8277	0.7939	0.9312	0.9242	0.8921
						Relative error/%	0.277	0.782	0.129	0.457	0.236

4. calculate the comprehensive evaluation value of each alternatives, the output optimal case.

Result to unbred 3 groups of test set simulation evaluations is as shown in table 3, and wherein the efficiency transformation project is followed successively by p7 by excellent to bad order, p8, p6.

Table 3 test result and the expert comparison of giving a mark

The scheme numbering	P6	P7	P8
				Desired output	0.83	0.95	0.94
Training result	0.8277	0.9539	0.9312
				Relative error/%	0.277	0.411	0.936

Accordingly, obtaining optimum efficiency scheme is p7.

Whole process Matlab7.2 software programming Wavelet-network model, network training 1863 times, training time 2.37s consuming time, the result is comparatively desirable.

5. with the comparison of feedforward neural network method

Use wavelet network to carry out the superior character of commercial efficiency comprehensive evaluation in order to illustrate, error under same iterations compares with wavelet network and BP neural network, the average error of wavelet network is 0.00478, and least error is 0.00277, and maximum error is 0.00936; The BP networks average error is 0.028176, and least error is 0.026317, and maximum error is that 0.029628. is obvious, and the precision of wavelet network obviously is better than the BP neural network, also is more suitable for the numerous and diverse field of commercial this class assessment indicator system of the efficiency synthetic evaluation of projects

With same sample data two kinds of networks are carried out convergence research, getting the relative error index is 0.5 ‰, and wavelet network is through 732 iteration convergences, and the BP neural network is through 903 iteration convergences.The convergence of proof wavelet network also is better than the BP neural network.

Should be appreciated that the above detailed description of technical scheme of the present invention being carried out by preferred embodiment is illustrative and not restrictive.Those of ordinary skill in the art is reading on the basis of instructions of the present invention and can make amendment to the technical scheme that each embodiment puts down in writing, and perhaps part technical characterictic wherein is equal to replacement; And these modifications or replacement do not make the essence of appropriate technical solution break away from the spirit and scope of various embodiments of the present invention technical scheme.

Claims (6)

1. commercial efficiency integrated evaluating method based on the wavelet network method is characterized in that may further comprise the steps:

(1) sets up efficiency synthetic evaluation of projects index system;

(2) calculate each single index value of System of Comprehensive Evaluation;

(3) method of utilization wavelet network is determined the optimal weight of each index;

(4) calculate the comprehensive evaluation value of each alternatives, the output optimal case.

2. a kind of commercial efficiency integrated evaluating method based on the wavelet network method according to claim 1, it is characterized in that: described step (1) is set up efficiency synthetic evaluation of projects index system and is comprised first class index and corresponding two-level index thereof, each two-level index is extended to three grades of indexs that embody index particular content at the corresponding levels again separately, has namely formed the tertiary level structure of domination from top to bottom.

3. a kind of commercial efficiency integrated evaluating method based on the wavelet network method according to claim 2, it is characterized in that: described first class index content comprises: operation of power networks, environmental effect, economic benefit and four aspects of the level of resources utilization.

4. a kind of commercial efficiency integrated evaluating method based on the wavelet network method according to claim 2 is characterized in that: described two-level index is estimated two aspect contents by electric network reliability evaluation and load shaping capability and is formed; Environmental effect in the described first class index partly comprises pollutant emission and carbon emission two aspect contents; Economic benefit in described first class index part is by comprising that mainly financial evaluation, gain on investments evaluation and three aspect contents of analysis of Life Cycle Cost form; The level of resources utilization in the described first class index comprises that partly energy-saving effect and grid side resource use two aspect contents.

5. a kind of commercial efficiency project evaluation method based on the wavelet network method according to claim 2, it is characterized in that: the reliability evaluation of the two-level index of the System of Comprehensive Evaluation of described foundation comprises system's average short of electricity electric weight ASCI rate of change and two indexs of the total electric weight of system ENS rate of change in shortage; Load shaping capability in the described two-level index partly comprises daily load rate, day peak valley rate, yearly load factor and a year maximum peak valley rate and changes four indices; Pollutant emission in the described two-level index partly comprises oxides of nitrogen CER, oxysulfide CER, dust CER and PM2.5 CER four indices; Carbon emission one deck in the described two-level index corresponds to CO ₂The CER index; Financial evaluation part in the described two-level index is carried out systematic analysis by net present value (NPV), internal rate of return and investment payback time three indexs and is calculated; Gain on investments in the described two-level index is partly chosen income expense ratio PCR and can be avoided electric cost AC2 item index to weigh the gain on investments situation of efficiency project; Overall life cycle cost in the described two-level index partly adopts longevity current cost rate of descent as the evaluation index of weighing efficiency project life-cycle cycle economy; Energy-saving effect in the described two-level index partly expands to terminal year amount of electricity saving and two indexs of electricity consumption income efficient difference.

6. a kind of commercial efficiency project evaluation method based on the wavelet network method according to claim 2, it is characterized in that: three grades of index weightses adopt the method for wavelet network to find the solution efficiency in the described assessment indicator system, adopt the linear weighted function overall approach that alternatives is carried out comprehensive scoring then, choose the comprehensive grading soprano and be optimal case, concrete implementation step is:

1) collects commercial efficiency project foundation data, calculate three grades of desired values in the System of Comprehensive Evaluation;

2) according to an expert view, tentatively give a mark for optional efficiency scheme, the expert who obtains each scheme appraises the mark set through discussion

3) with each the value attribute value vector { x that calculates in the step 1) _k(i) } be converted into the data { r of index attribute unanimity _k(i) };

Three types index is done following processing:

In the formula: r _IjBe j desired value in i the scheme, r _Ij ^*Be r _IjTreated value; M is the programme number; r _jIt is the desired quantity of j index;

4) give wavelet network parameter wi _j, r _j, b _j, a _jInitial value at random and give the max calculation times N;

5) with the unification desired value { r of each evaluation of programme _k(i) } be input in the wavelet network computing formula, try to achieve corresponding comprehensive evaluation value y _iAnd calculate corresponding energy error E;

The wavelet network computing formula is:

$y_k=\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{r}_{j}h\left[\frac{\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ij}}{r}_k\left(i\right)-b_j}{a_j}\right]$

The error energy function of network is:

$E=\frac{1}{2}\underset{k=1}{\overset{P}{\mathrm{\&Sigma;}}}{({\hat{y}}_k-y_k)}^2$

6) gradient vector of calculating wavelet network;

Order

${\mathrm{\&lambda;}}_k\left(j\right)\frac{\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ij}}{r}_k\left(i\right)-b_j}{a_j}$

Have:

$g\left(w_{\mathrm{ij}}\right)=\frac{\&PartialD;E}{\&PartialD;w_{\mathrm{ij}}}=\underset{k=1}{\overset{P}{\mathrm{\&Sigma;}}}({\hat{y}}_k-y_k)\left(\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{r}_j\frac{\&PartialD;h}{\&PartialD;{\mathrm{\&lambda;}}_k\left(j\right)}\frac{r_k\left(i\right)}{a_j}\right)$

$g\left(r_j\right)=\frac{\&PartialD;E}{\&PartialD;r_j}=\underset{k=1}{\overset{P}{\mathrm{\&Sigma;}}}({\hat{y}}_k-y_k)h\left({\mathrm{\&lambda;}}_k\left(j\right)\right)$

$g\left(a_j\right)=\frac{\&PartialD;E}{\&PartialD;a_j}=-\underset{k=1}{\overset{P}{\mathrm{\&Sigma;}}}({\hat{y}}_k-y_k)\left(\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{r}_j\frac{\&PartialD;h}{\&PartialD;{\mathrm{\&lambda;}}_k\left(j\right)}\frac{{\mathrm{\&lambda;}}_k\left(i\right)}{a_j}\right)$

$g\left(b_j\right)=\frac{\&PartialD;E}{\&PartialD;b_j}=-\underset{k=1}{\overset{P}{\mathrm{\&Sigma;}}}({\hat{y}}_k-y_k)\left(\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{r}_j\frac{\&PartialD;h}{\&PartialD;{\mathrm{\&lambda;}}_k\left(j\right)}\frac{1}{a_j}\right)$

Wherein,

$\frac{\&PartialD;h}{\&PartialD;{\mathrm{\&lambda;}}_k\left(j\right)}=-\cos \left(1.75{\mathrm{\&lambda;}}_k\left(j\right)\right)\exp (-\frac{{\mathrm{\&lambda;}}_k^2\left(j\right)}{2}){\mathrm{\&lambda;}}_k\left(j\right)-1.75\sin \left(1.75{\mathrm{\&lambda;}}_k\left(j\right)\right)\exp (-\frac{{\mathrm{\&lambda;}}_k^2\left(j\right)}{2})$

7) adopt method of conjugate gradient to adjust network parameter, t is iterations: order

$S_t\left(w_{\mathrm{ij}}\right)=\left\{\begin{array}{cc}-g_t\left(w_{\mathrm{ij}}\right),& t=1\\ -g_t\left(w_{\mathrm{ij}}\right)+\frac{\left|\right|g_t\left(w_{\mathrm{ij}}\right)\left|\right|}{\left|\right|g_{t-1}\left(w_{\mathrm{ij}}\right)\left|\right|}{S}_{t-1}\left(w_{\mathrm{ij}}\right),& t>1\end{array}\right.$

In like manner, can calculate S _t(r _j), S _t(a _j), S _t(b _j);

Then the adjusting of network parameter is as follows:

w _(t)ij＝w _(t-1)ij+αS _t-1(w _ij)

r _(t)j＝r _(t-1)j+βS _t-1(r _j)

a _(t)j＝a _(t-1)j+γS _t-1(a _j)

b _(t)j＝b _(t-1)j+ηS _t-1(b _j)

8) return step 4), till the error energy functional value of network is not more than set-point ε or calculation times and surpasses the max calculation times N.

9) according to the index weights of determining in the step 8), the method for employing linear weighted function is calculated the comprehensive evaluation value of each scheme, and choosing the highest scheme of mark is optimal case.

Images