CN103062076B - A kind of urban water plant pump station separate unit characteristic curve of pump rating method - Google Patents

Abstract

The invention provides a kind of urban water plant pump station separate unit characteristic curve of pump rating method, urban water plant pump station comprises multiple stage pump in parallel, described method comprises: (1) Monitoring Data collection: by being arranged on the monitoring instrument that pumping plant is imported and exported, obtain pumping station operation historical state data; (2) Monitoring Data process: utilize the pumping station operation historical state data recorded by step (1), obtains pumping plant total discharge, pumping plant total head, separate unit water pump operation state and pumping plant power consumption, generating process data; (3) Analysis on monitoring data: to the generation data in step (2), adopts sensitivity analytical method, according to given pumping plant performance curve equation, calculates separate unit pump's performance curre parameter in pumping plant; (4) parameter application: the separate unit pump's performance curre parameter obtained in step (3) is carried out simulation application, obtains separate unit pump's performance curre, the assessment of separate unit pump energy consumption, separate unit water pump improvement project and Optimal Operation of Pumping Stations.

Description

A kind of urban water plant pump station separate unit characteristic curve of pump rating method

Technical field

The present invention relates to the parameter estimation field of water supply network equipment, particularly relate to a kind of rating method of urban water plant pump station separate unit characteristic curve of pump.

Background technique

Pump is conveying liquid or the machinery making fluid pressurization.It sends the mechanical energy of prime mover or other external energies to liquid, and liquid energy is increased.Pump is mainly used to the liquid such as conveying water, oil, acid & alkali liquid, emulsified liquid, suspended emulsion and liquid metal, also can conveying liquid, gas mixture and the liquid containing suspended solids.Pump can be divided into positive displacement pump, driving type pump and other types pump three class by working principle usually.Except classifying by working principle, also can by additive method taxonomy and nomenclature.As, can motor-drive pump and turbine pump etc. be divided into by method for driving; Single-stage pump and multistage pump can be divided into by structure; Boiler feed pump and metering pump etc. can be divided into by purposes; Water pump, oil pump and slurry pump etc. can be divided into by the character of conveying liquid.

The basic performance parameter of water pump comprises flow, lift, power, efficiency, rotating speed etc., and these parameter flag water pump performance.Relation between each performance parameter of water pump and Changing Pattern, can express with one group of performance curve.Usually representing that the curve of relation between Specifeca tion speeification calls performance curve or the characteristic curve of centrifugal pump, in fact, centrifugal pump performance is the external presentation form of liquid characteristics of motion in pump, is tried to achieve by actual measurement.Characteristic curve comprises: flow-lift curve (Q-H), flow-efficiency curve (Q-η), flow-power curve (Q-N), flow-net positive suction head curve (Q-(NPSH) r), performance curve effect is the arbitrary flow point of pump, one group of lift corresponding thereto can be found out on curve, power, efficiency and net positive suction head value, this group parameter is called working state, be called for short operating mode or operating point, the operating mode of centrifugal pump best efficiency point is called optimal working point, and optimal working point is generally operating point for design.

According to separate unit pump's performance curre, operator can determine water pump operating range and in order to work out pump performance table, they hold water to select the foundation of water pump.Select flow and the lift of water pump, must be consistent or close with the flow required for water supply and lift, the maximal efficiency of competence exertion water pump.Departing from efficacious workaround as during water pump operation, is uneconomic, need regulate pump performance or use the water pump of more suitable types instead.

But in water supply pumping plant field, most water pumping station adopts the parallel running of multiple stage centrifugal pump, and when actual motion, actual water supply flow is often less than the design discharge of pumping plant, often needs partial devices, part water pump selection to run.Different pump characteristics is different, and water pump curve is provided by device fabrication producer.Along with the impact on equipment performance such as growth, the difference of installation quality, each pumping station operation operating mode of water pump service life, can be there is certain difference when comparatively dispatching from the factory in water lift of pump-rating curve, need re-starting identification to pumping plant with water pump curve.

Traditional method of checking pump characteristic is an important and loaded down with trivial details job, for obtaining the actual lift flow curve of every platform water pump, every platform water pump need be isolated, carry out independent experiment measuring, for obtaining the flow-heat-capacity curve of single pump accurately, when carrying out field measurement, need strictly to follow Survey control condition: (1) pumping plant runs by daily operating conditions, could implement after every platform water pump steady-state operation 3 ~ 5min to measure; (2) change water pump state, comprise the operating mode of all general pumps, variable frequency pump (different frequency), measuring flow-lift, to obtain complete water pump curve; (3) for avoiding the impact of current sudden change place hydraulic pressure drastic change and fluctuation, pressure measurement pipeline needs 2 times to the length of pipe diameter at distance unit fixed on water pump suction and outlet(discharge) flange Pan Chu; Should there be the straight section length being equivalent to 5 times ~ 10 times of pipe diameters flow conduit upstream, and downstream straight section length is 4 times.

The restriction that application traditional means of experiment carries out curve check is more, there is following shortcoming:

1) restriction of current conditions, experiment measuring had both needed to carry out independent experiment to water pump, must ensure that again test point and water pump have the straight length of certain length, but due to pump house compact in design, apparatus installation is subject to the restriction of condition, was not suitable for adopting conventional method;

2) water supply security is difficult to ensure, to obtain the actual lift flow curve of every platform water pump, every platform water pump need be isolated, but in order to ensure the safety water supply of pipe network, water pumping station, especially when multiple stage parallel connection of pumps works, is difficult to independent for water pump operation to carry out experiment measuring;

3) precision is low, pump house compact in design, and apparatus installation is subject to the restriction of condition, and it is many that estimating precision influence factor implemented by separate unit water pump, is difficult to large-scale promotion.

Current water pumping station is imported and exported and has been installed a large amount of monitoring instrument, is equipped with professional staff's record of patrol, have accumulated a large amount of pumping station operation creation data, but the anti-new method pushing away pump characteristic parameter of these data is but very rare about effectively utilizing.

Summary of the invention

For above-mentioned technical problem, the invention provides a kind of urban water plant pump station separate unit characteristic curve of pump rating method, wherein said urban water plant pump station comprises multiple stage pump, and described multiple stage pump is in parallel, and described method comprises the steps:

(1) Monitoring Data collection: by being arranged on the monitoring instrument that pumping plant is imported and exported, obtain pumping station operation historical state data;

(2) Monitoring Data process: utilize the pumping station operation historical state data recorded by step (1), obtains pumping plant total discharge, pumping plant total head, separate unit water pump operation state and pumping plant power consumption, generating process data;

(3) Analysis on monitoring data: to the generation data in step (2), adopts sensitivity analytical method, according to given pumping plant performance curve equation, calculates separate unit pump's performance curre parameter in pumping plant;

(4) parameter application: the separate unit pump's performance curre parameter obtained in step (3) is carried out simulation application, obtains separate unit pump's performance curre, the assessment of separate unit pump energy consumption, separate unit water pump improvement project and Optimal Operation of Pumping Stations;

Wherein, in described step (2), total head, the total discharge set of equation of pump when described multiple stage parallel connection of pumps are expressed as follows:

Lift: H=H _0ic _i ²-A _iq _i ²(i=1...n) 1.

Flow: $Q={\mathrm{\Σ}}_{i=1}^n{Q}_i$ ②

Wherein, H represents lift;

N represents the number of pump in parallel;

H _0irepresent the net lift of i-th water pump;

C _irepresent the rotating ratio of i-th water pump, if general pump, then c=1;

A _irepresent the frictional resistance of i-th water pump;

Q _irepresent the flow of i-th water pump;

And in described step (3), described sensitivity analytical method comprises:

A () separate unit parameters of pump is simulated: according to the record data of each water pump, and rule of thumb set the parameter space Ω possibility span of separate unit characteristic curve of pump, described parameter space Ω is expressed from the next:

Ω＝{(H ₀₁，…，H _0i，…，H _0n，…，A ₁，…，A _i，…，A _n)}︱H _0ia≦H _0i≦H _0ib，A _ia≦A _i≦A _ib，i＝1…n}③

Wherein:

H _0irepresent separate unit water lift of pump parameter, its value is H ₀₁~ H _0nin any one;

A _irepresent the frictional resistance parameter of separate unit water pump, its value is A ₁~ A _nin any one;

H _0ia, H _0ibfor separate unit water lift of pump parameter H _oithe upper and lower of valued space, A _ia, A _ibfor the frictional resistance parameter A of separate unit water pump _ithe upper and lower of valued space;

B () separate unit parameters of pump is assessed: the parameter space according to pump characteristic produces arbitrary group of random numbers at random, and substitutes into the formulae discovery solving lift-flow, adopts one group of (H _p, Q _p) coordinate replaces in pump curvilinear equation H ₀, A value, its conversion relation as shown in the formula:

$H_0=\frac{4}{3}{H}_p$ ④

$A=\frac{H_p}{3{Q_p}^2}$ ⑤

4. and 5. formula is substituted into formula 1. in calculation arrangement draw following equation:

${Q^{\′}}_{j_{k}k}=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{Q}_{\mathrm{pi}}\·\sqrt{S_{\mathrm{ik}}\·({4c}_{\mathrm{ik}}^2-\frac{{3H}_{j_{k}k}}{H_{\mathrm{pi}}})}$ ⑥

Random H is organized to each _pi, Q _pi, according to calculate , and with physical record value compare, choose calculated value with record value squared difference and assessment this group parameter quality, calculated by following formula:

$\mathrm{\ϵ}=\underset{k=1}{\overset{q}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{m_k}{\mathrm{\Σ}}}{({Q^{\′}}_{j_{k}k}-Q_{j_{k}k})}^2$ ⑦

Size according to ε sorts, and chooses the solving result of parameters combination as final water pump parameter of curve that ε is 1 ~ 2% correspondence;

Wherein, H _prepresent the rated head of water pump;

Q _prepresent the rated flow of water pump;

H ₀represent the net lift of water pump;

A represents the frictional resistance of water pump;

H _pirepresent the rated head of i-th water pump;

Q _pirepresent the rated flow of i-th water pump;

represent that pump combination sequence number is that under k, record number is j _kpump house work lift;

S _ikrepresent that pump combination sequence number is the status parameter of i-th water pump under k;

represent that pump combination sequence number is that under k, record number is j _kthe pump house flow calculated by random parameter;

represent that pump combination sequence number is that under k, record number is j _kpump house record flow;

C _ikrepresent that pump combination sequence number is the rotating ratio of lower i-th water pump of k;

ε represents with between error;

M _kvalue is pump combination sequence number is record number under k;

I represents that pump is numbered, and i=1...n, k represent pump combination sequence number, k=1...q, j _krepresent that pump combination sequence number is the record number under k, j _k=1 ... m _k, when pump is opened, S _ikbe 1, when pump cuts out, S _ikbe 0.

In addition, rating method of the present invention also comprises evaluates traffic error and assessment lift error step.

The rating method of a kind of urban water plant pump station separate unit characteristic curve of pump of the present invention is based on water factory's history total discharge and pressure online monitoring data, and anti-release separate unit water pump curve, is characterized in:

(1) depend on the accumulation of historical data, save a large amount of on-the-spot test work, workable; Ensure the Security of pipe network water supply, avoid complexity and the check of the experiment measuring of more difficult realization;

(2) by using sensitivity analytical method, rationally reducing the span of understanding, improve computational efficiency;

(3) precision is reliable, utilizes the analysis of parameter Posterior distrbutionp, has ensured the precision calculated, and can obtain comparatively reasonably parameter of curve.

Accompanying drawing explanation

Fig. 1 is the operating diagram of method of the present invention.

Fig. 2 is the basic step schematic diagram of the parameter calibration sensitivity analytical method used in method of the present invention.

Fig. 3 a, b are the parameter H according to pump _p, Q _pposterior distrbutionp situation map one example, Fig. 3 a represents pump H _pthe Posterior distrbutionp of parameter, Fig. 3 b represents pump Q _pthe Posterior distrbutionp of parameter.

Fig. 4 is that application method of the present invention carries out parameter calibration error analysis figure to pump combination (0, Isosorbide-5-Nitrae 0,0,0,0).

Fig. 5 is the basic step schematic diagram of evaluates traffic error of the present invention and assessment lift error.

Fig. 6 represents the tracing analysis figure of the XW-10 water pump of fallback.

Fig. 7 represents the efficiency schematic diagram of the XW-10 water pump of fallback.

Fig. 8 to be Fig. 7 RSA method with optimization method carry out parameter identification compare schematic diagram.

Embodiment

Below in conjunction with accompanying drawing, the rating method of a kind of urban water plant pump station separate unit characteristic curve of pump of the present invention is described in detail.

First, urban water plant pump station separate unit characteristic curve of pump rating method of the present invention, wherein said urban water plant pump station comprises multiple stage pump, and described multiple stage pump is in parallel, and described method comprises the steps:

(1) Monitoring Data collection: by being arranged on the monitoring instrument that pumping plant is imported and exported, obtain pumping station operation historical state data;

(2) Monitoring Data process: utilize the pumping station operation historical state data recorded by step (1), obtains pumping plant total discharge, pumping plant total head, separate unit water pump operation state and pumping plant power consumption, generating process data;

(3) Analysis on monitoring data: to the generation data in step (2), adopts sensitivity analytical method, according to given pumping plant performance curve equation, calculates separate unit pump's performance curre parameter in pumping plant;

(4) parameter application: the separate unit pump's performance curre parameter obtained in step (3) is carried out simulation application, obtains separate unit pump's performance curre, the assessment of separate unit pump energy consumption, separate unit water pump improvement project and Optimal Operation of Pumping Stations.

Fig. 1 is the operating diagram of method of the present invention.As shown in Figure 1, in Monitoring Data gathers, data capture is carried out by SCADA database, PUMP operation note and pumping plant ammeter reading.

In detection data processing step, by processing the data of above-mentioned collection, obtain the processing data such as pumping plant total discharge, pumping plant main-process stream, separate unit water pump operation state and pumping plant power consumption.

After this, utilize Analysis on monitoring data-RSA method, carry out the simulation of separate unit parameters of pump, the assessment of separate unit parameters of pump and separate unit parameters of pump posteriority.

Finally, in parameter applying step, the separate unit pump's performance curre parameter obtained in above-mentioned steps is carried out simulation application, obtain separate unit pump's performance curre, the assessment of separate unit pump energy consumption, separate unit water pump improvement project and Optimal Operation of Pumping Stations.

In described step (2), total head, the total discharge set of equation of pump when described multiple stage parallel connection of pumps are expressed as follows:

Lift: $H=H_{\mathrm{Qi}}\·c_i^2-A_i\·Q_i^2(i=1...n)$

Flow: $Q={\mathrm{\Σ}}_{i=1}^n{Q}_i$

Wherein, H represents head difference;

N represents the number of pump in parallel;

H _αrepresent the net lift of i-th water pump;

C _irepresent the rotating ratio of i-th water pump, if general pump, then c=1;

A _irepresent the frictional resistance of i-th water pump;

Q _irepresent the flow of i-th water pump.

Unknown number number due to this set of equation is n+2, and equation number is n+1, namely the degrees of freedom of set of equation is 1, in reality, do not know the parameter of every platform pump, the flow of single pump also cannot be measured, and the head of pump house and flow then have a large amount of records, calculates that the parameter of every platform pump is possible according to these measuring recordings.

For this reason, adopt above-mentioned RSA method (sensitivity analytical method), namely, wherein RSA method refers to and adopts sensitivity analysis (RegionalizedSensitivityAnalysis) method, carries out parameter identification by the method, and RSA method is that Parameter analysis is carried out in the binary division of Behavior-based control and non-behavior, i.e. given one group of parameter, if the simulation behavior of system meets the condition realizing setting, so this group parameter is exactly acceptable, otherwise is unacceptable.Adopting said method can reduce the interval of pump parameter, conveniently therefrom chooses one group of suitable parameter as pump.

Below, RSA method for solving and step are illustrated in greater detail.

RSA method is that the binary division of Behavior-based control and non-behavior carries out parameter identification, i.e. given one group of parameter, if the simulation behavior of system meets the condition realizing setting, so this group parameter is exactly acceptable, otherwise is unacceptable.Sensitivity analytical method (RSA) take into account the high correlation between parameter, have recognised the complexity of parameter space, carries out sampling simulation to parameter simultaneously.

Comparing when RSA method and optimization method carry out parameter identification sees Fig. 8.Fig. 8 (a), (b) represent Optimal Parameters identifying method, and in (b), mark point represents optimized parameter; C () and (d) represents RSA parameter identification method, in (d), period represents behavioral parameters, and cross represents non-behavioral parameters.As can be seen from Figure 8, RSA method has been abandoned optimization method and has been adopted single parameter to go to approach the parameter identification theory of actual observed value, and identification is a parameter set, or a region, and it is the set formed by a large amount of behavioral parameters.

The basic step that sensitivity analytical method realizes is as follows:

Determine the initial distribution F (θ) of parameter.If the understanding more indefinite to the distribution character of parameter, method the most normal is according to front testing knowledge and the possible span of document determination parameter, then supposes that parameter is be uniformly distributed on this interval, and is independently between parameter.If namely [ai, bi] is the span of parameter θ i, then θ i meets being uniformly distributed on interval [ai, bi].

According to the contrast of the analogue value and Observed value, determine the criteria for classifying of behavioral parameters (B) and non-behavioral parameters (NB).If the relative error of the such as analogue value and Observed value is less than certain level (such as 10%), then think that parameter is behavior, otherwise be non-behavior; Also the parameter can got corresponding to a certain proportion of optimal objective function divides.The definition of RSA method to behavioral parameters and non-behavioral parameters is very flexibly, and this greatly increases the Applicable scope of RSA method.

Sample from the initial distribution of parameter, obtain one group of parameter, bring model into and simulate, if the result of simulation is acceptable, so thinking that this group parameter is behavioral parameters, is acceptable; Otherwise be non-behavioral parameters, cannot accept.

Repeat step (2) and (3), until pre-determined sample number.

Obtain the sample distribution of behavioral parameters and non-behavioral parameters respectively, compare the difference between two sample distribution.Difference is larger, illustrates that parameter is more responsive; Otherwise, then more insensitive.

RSA Algorithm judges the sensitivity level of parameter by testing distribution and the gap of the Posterior distrbutionp that can accept parameter before parameter.If gap is larger, then representation parameter is sensitiveer, otherwise more insensitive.Determination methods divides quantitative and qualitative analysis two kinds.Method is cumulative distribution function by accepting parameter values and probability density function qualitatively, estimates itself and the front gap size testing distribution.Quantitative method is then calculated by the method for hypothesis testing, tests the gap of distribution and Posterior distrbutionp, and judge that whether it is sensitive with certain standard with numerical result before providing.

RSA method depends on random sampling technique.Stochastical sampling method quality depends primarily on two features, and one is the efficiency of sampling, and two is the precision of sampling, and both are conflicting often.Along with widely using of stochastic analysis, Sampling techniques also constantly make progress.Introduce relatively more conventional Sampling method below.

(grid sampling)

So-called grid sampling (GS, GridSampling), refer to and the parameter of every one dimension is divided into some deciles, sampled point just drops on grid node.GS method is only applicable to simple low-dimensional situation, such as a peacekeeping two dimension, and due to number and the parameter dimension exponentially relation of sampled point, when parameter dimension is very high, sampling nodes sharply rises.Such as to the situation of 10 dimensions, even if every one dimension only has 10 nodes, also have 10 altogether ¹⁰individual node, far beyond the computing capability of common computer.

Although GS sampling is not a kind of stochastical sampling truly, it discloses the sampling difficulty occurred when the dimension of parameter raises, " sampling disaster " namely alleged by some scholar.That is, when the dimension of parameter is very high, the information of the parameter space that we can collect is the information of the very little part in one, sample space all the time.But however, people still constantly finding Sampling method in higher dimensional space, effectively to improve the efficiency of stochastical sampling.

(Monte Carlo sampling)

Excite Monte Carlo sampling (MCS, MonteCarloSampling) thought the earliest be 1777 by GeorgesLouisLeclercContedeBuffon propose " Pu pin tests (Buffon ' sneedles).

The essence of MonteCarlo method obtains sample with being random independent from overall, and it is applicable to process any mathematical statistics problem.MonteCarlo method has two obvious features, and one is its simple structure, and two is that its sampling error is proportional to D/N, D constant here, and N is sampling number (Sabelfeld, 1991).

The sampling efficiency of MonteCarlo method is a major obstacle of its application, and research shows, if precision improves 10 times, then needs to extend computing time 100 times, so MonteCarlo can not produce very high-precision solution.As MonteCarlo method need not be special Accelerated Convergence Method, just can only obtain validity (Bu Silianke, 1964 that maximum error is 0.01 ~ 0.001; Sabelfeld, 1991).However, MonteCarlo method also has advantage (Sabelfeld, 1991), the particularly feature of its simple and flexible that other method a lot of does not have, and MonteCarlo method is still played an important role in statistical computation.

(Latin Hypercube Sampling)

Latin Hypercube Sampling (LHS, LatinHypercubeSampling) sampling is a kind of stochastical sampling method (McKay, etal., 1979) occurred for 1979, and obtains application widely in stochastic analysis afterwards.LHS sampling is a kind of stratified sampling technology of pair distribution function more more accurate than MonteCarlo method, and classical LHS sampling key step is (Chen Jining, 1993; Saltellietal., 2000):

(1) be first n intervals of equal probability by the probability distribution interval division of parameter vector.If be independently between parameter, and parameter is uniformly distributed on the interval of specifying, and is so in fact exactly the interval each parameter being divided into the sizes such as n;

(2) each parameter random selecting interval from parameter space, once after choosing between some parameter regions, is used between this interval candidate regions that just can not be re-used as sampling further;

(3) repeat (2) n time, until all intervals are all sampled once, and create n vector.These vectors comprise and are distributed in whole value spatially, but incoherent;

(4) (2) and (3) are repeated, until meet the requirements of sampling number.

(transfer Hammersley sampling)

Transfer Hammersley sampling (SHS, ShiftedHammersleySampling) obtains in recent years to be studied widely.The expansion of nonnegative integer k is obtained below the low different sequences that the sixties proposes in early days can adopt by HammersleyandHalton

k＝a ₀+a ₁p+a ₂p ²+…+a _rp ^r##################################

Here, p is a prime number (primenumber), a _iit is the integer quotient of the non-negative being less than p.These coefficients are used for establishment one number between zero and one:

${\mathrm{\Φ}}_p\left(k\right)=\frac{a_0}{p}+\frac{a_1}{p^2}+\frac{a_2}{p^3}+...+\frac{a_r}{p^{r+1}}$ #################################

Any one different Prime sequences p ₁..., p _m-1, cause the sequence of a N number of Hammersley point, be defined as:

$z_k=(\frac{k}{N},{\mathrm{\Φ}}_{p_1}\left(k\right),{\mathrm{\Φ}}_{p_2}\left(k\right),...,{\mathrm{\Φ}}_{p_{m-1}}\left(k\right))$ #################################

By p ₁, p ₂..., p _mthe sequence of the N number of Halton point caused is:

$z_k=({\mathrm{\Φ}}_{p_1}\left(k\right),{\mathrm{\Φ}}_{p_2}\left(k\right),...,{\mathrm{\Φ}}_{p_m}\left(k\right))$ #################################

Relatively Hammersley sequence, the benefit of Halton sequence is can expand larger sample easily to, and does not need to revise old sample point.

Below, the step of RSA Algorithm in parameters of pump solves is introduced further.

(random numbers)

RSA method is that the binary division of Behavior-based control and non-behavior carries out parameter identification, i.e. given one group of parameter, if the simulation behavior of system meets the condition realizing setting, so this group parameter is exactly acceptable, otherwise is unacceptable.Thus, the span that the parameter space of separate unit pump characteristic is possible can be expressed as

Ω

＝{(H ₀₁，…，H ₀₁，…，H _0n，A ₁，…，A ₁，…，A _n)H _01a≤H ₀₁≤H _01b，A _1a≤A ₁≤A _1b，i＝1…n}

Wherein: H _oa, H _obfor separate unit water lift of pump parameter H _oithe upper and lower of valued space, A _ia, A _ibfor the frictional resistance parameter A of separate unit water pump _ithe upper and lower of valued space.

First carry out Sobol Sequence pseudo stochastical sampling, generate the random number sequence U meeting certain distribution on [0,1] interval _i, then according to the interval of the parameter of model, U _iconvert the sequence of random variables X that model needs to _i, finally the stochastic parameter sequence obtained is applied in model, by the method random one group of parameter produced in parameter space, substitutes in following formula (3) and obtain calculated value

(assessment)

According to the contrast of the analogue value and Observed value, wherein relative error can be chosen and be less than the parameter value of certain level.According to any one group of random numbers that the parameter space of pump characteristic produces, substitute into the formulae discovery solving lift-flow .In Practical Calculation, for ease of estimating the interval of initial space, adopt one group of (H _p, Q _p) coordinate replaces in pump curvilinear equation H ₀, A value, its conversion relation as shown in the formula:

$H_0=\frac{4}{3}{H}_p---\left(1\right)$

$A=\frac{H_p}{{3Q}_p^2}---\left(2\right)$

Equation becomes:

${Q^{\′}}_{j_{k}k}=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{Q}_{\mathrm{pi}}\·\sqrt{S_{\mathrm{ik}}\·({4c}_{\mathrm{ik}}^2-\frac{{3H}_{j_{k}k}}{H_{\mathrm{pi}}})}---\left(3\right)$

Random H is organized to each _p1, Q _p1, according to calculate , and with physical record value compare, can calculating be chosen with record value squared difference assess the quality of this group parameter, following formula (4) can be used to assess the quality of this group parameter.

$\mathrm{\ϵ}=\underset{k=1}{\overset{q}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{m_k}{\mathrm{\Σ}}}{({Q^{\′}}_{j_{k}k}-Q_{j_{k}k})}^2---\left(4\right)$

Next, the size according to ε sorts, and chooses the solving result of parameters combination as final water pump parameter of curve that ε is 1 ~ 2% correspondence.

(Posterior distrbutionp)

Analyze its Posterior distrbutionp, with the space distribution of estimation model parameter.The sensitivity level of parameter is judged by testing distribution and the gap of the Posterior distrbutionp that can accept parameter before parameter.If gap is larger, then representation parameter is sensitiveer, otherwise more insensitive.Sensitive parameter adopts can accept parameter value scope, and insensitive parameter adopts the interval originally determined, completes model and parameters identification.As Fig. 3 a, 3b, with reference to H ₀, Q _pposterior distrbutionp situation, chooses rational parameters combination as pump parameter, finally determines the curvilinear equation of single pump.

In addition, rating method of the present invention also comprises subsequently by the evaluates traffic error that is introduced and assessment lift error step.

The rating method of a kind of urban water plant pump station separate unit characteristic curve of pump of the present invention is based on water factory's history total discharge and pressure online monitoring data, and anti-release separate unit water pump curve, is characterized in:

(1) depend on the accumulation of historical data, save a large amount of on-the-spot test work, workable; Ensure the Security of pipe network water supply, avoid complexity and the check of the experiment measuring of more difficult realization;

(2) by using sensitivity analytical method, rationally reducing the span of understanding, improve computational efficiency;

(3) precision is reliable, utilizes the analysis of parameter Posterior distrbutionp, has ensured the precision calculated, and can obtain comparatively reasonably parameter of curve.

Below, the specific embodiment of the present invention is described.

(embodiment 1)

Certain city intelligence water utilities project application method of the present invention, accurately obtains the characteristic curve of 11 water pumps (comprising 8 general pumps and 3 variable frequency pumps), and have identified inefficient water pump by the method and carried out changing transformation.

This city's water utilities system comprises Liang Ge water factory and has 11 pumps and often use, wherein, there is 1 variable frequency pump (XW6) in a water factory, 5 general pumps (XW2, XW4, XW8, XW9, XW10), the operation of these 6 pumps always has 11 kinds of compound modes, namely the situation that the situation that 1 to 5 general pump runs situation, 1 to 5 general pump and 1 variable frequency pump run and only 1 variable frequency pump run, 11 kinds altogether.

By the assembled state numeral of pump, general pump closedown duration is 0, unlatching duration is 1; It is 0 that variable frequency pump closes duration, is the numeral of 1 when opening, and " 40 " represent that pump is turned up full blast level, correspondence is adjusted a wage scale electric current 20mA, and " 39,38,37 " are corresponding " 19.5mA, 19mA, 18.5mA " respectively, such as, combine " 0; 1,37,0; 0; 0 " represent that XW1 general pump is closed, XW2 general pump is opened, and XW3 variable frequency pump electric current of adjusting a wage scale is 18.5mA, XW4 general pump is closed, and XW5 general pump is closed;

Table 1 below represents one example, within continuous three months, chooses historical data in stable nature hours for there being the water factory of above-mentioned 6 pumps:

Table 1

In the check of water pump, adopted 100000 points at random, by deployment area Sensitivity Analysis Method, 1000 points choosing ε value minimum carry out parameter H ₀, Q _pthe analysis of Posterior distrbutionp, the parameter H of acquisition ₀, Q _pposterior distrbutionp situation.

Wherein, sensitivity analytic method adopts method well known in the art, and concrete steps as shown in Figure 2, first determine that parameter value is interval, next determine that simulation can accept parameter and standard, then set frequency in sampling, next modeling is carried out to stochastic parameter sampling.

Now judge whether parameter can accept, if so, then record can accept parameter, if not, then judges whether to reach frequency in sampling.

Now if the judgment is No, then get back to and the step that stochastical sampling carries out modeling is carried out to parameter, if the judgment is Yes, then sensitivity analysis is carried out to the accepted parameter sets of parameter, obtain sensitive parameter and can interval be accepted, other parameters use former interval, thus terminate institute in steps.

After being analyzed by said method, with reference to its posteriority distribution situation as shown in Figure 4, choose rational parameters combination as pump parameter, this water factory every platform pump indicatrix that concrete final application method of the present invention obtains sees the following form 2.

Table 2 water factory pump characteristic calibration result

Pump is numbered	Frequency modulation	Pump characteristic
			XW2	Power frequency	H＝28.61-1.891×10 -6Q 2
XW4	Power frequency	H＝51.67-2.643×10 -6Q 2
			XW6 (frequency conversion)	20mA	H＝60.94-7.416×10 -7Q 2
XW8	Power frequency	H＝43.41-1.164×10 -6Q 2
			XW9	Power frequency	H＝45.33-1.185×10 -6Q 2
XW10	Power frequency	H＝44.00-3.429×10 -6Q 2

By the calibration result of every platform pump, wherein a kind of pump combination (0, Isosorbide-5-Nitrae 0,0,0,0) choosing above-mentioned table 2 total carries out parameter calibration error analysis, and the combination of this pump corresponds to pump, and analysis result as shown in Figure 4.

Fig. 4 is that application method of the present invention carries out parameter calibration error analysis figure to pump combination (0, Isosorbide-5-Nitrae 0,0,0,0)."+" in Fig. 4 represents actual monitoring value, and straight line represents the analogue value.With reference to Fig. 4, the known middle analogue value and actual monitoring value very close, thus illustrate that the rating method of urban water plant pump station separate unit characteristic curve of pump of the present invention has higher precision.

(embodiment 2)

Below the evaluates traffic error in rating method of the present invention and assessment lift error step are introduced.

Fig. 5 is the basic step schematic diagram of evaluates traffic error of the present invention and assessment lift error.

As seen from Figure 5, the basic step of evaluates traffic error of the present invention and assessment lift error is identical, and difference is it is the appointment lift utilizing working state recording in evaluates traffic error, carrys out calculated flow rate error with evaluate parameter; And be the target flow utilizing working state recording in assessment lift error, calculate lift error with evaluate parameter.Thus, referring to Fig. 5, only the evaluates traffic error step in the present invention is introduced.

As shown in Figure 5, described evaluates traffic error step:

(1) determine parameter and parameter search scope, and generate random parameter group: setup parameter and parameter search scope can be come according to the historgraphic data recording of water factory and experience, on this basis, generate random parameter group;

(2) one group of parameter is applied in pump group objects: in the random parameter group generated, random selecting one group of parameter is applied in pump group objects;

(3) select one group of working state recording to specify lift H according to working state recording: can according to the working state recording of record, Stochastic choice one group of working state recording specifies lift H, and herein, working state recording refers to the switch record, running state record etc. of pump in pumping plant;

(4) calculated flow rate Q, and the error of evaluates traffic Q: the error being carried out evaluates traffic Q by lift-flow process formula;

(5) judge whether to have assessed all operating modes; If so, then step (6) is entered; If not, then turn back to step (3);

(6) assess the mean error of whole operating mode, and judge whether that all population of parameter have been assessed, if so, then enter into step (7); If not, then turn back to step (2);

(7) screen the minimum population of parameter of error, and assess the requirement whether minimal error meets setting, if so, then enter into step (8); If not, then enter into step (9);

(8) satisfactory population of parameter is determined;

(9) to reappraise parameter search space or openly record the accuracy of data.

The difference of two kinds of error evaluation modes is mainly computational efficiency, when water pump adopts quadratic polynomial curve, lift can by flow explicit expression, concerning the total discharge of parallel pumps and lift, explicit expression can only represent flow and Q (H) with lift, namely specify H can explicit algorithm Q, and specify Q can only separate nonlinear equation to ask H, there is obvious efficiency difference between both.Example is as shown in following table 3 and 4.Table 3 is running environment and the Selecting parameter situation of two kinds of error evaluations, and table 4 is the interpretation of result of two kinds of error evaluations.

Table 3: running environment and Selecting parameter

Table 4: assessment result analysis

Thus, known by above-mentioned sample calculation analysis, method of the present invention is except algorithmically improving except computational efficiency, also comprise following advantage: the quadratic curve of single pump flow lift is two parameter by three parameter predigestings, reduce the complex nature of the problem, decrease amount of calculation, can calculation accuracy be ensured simultaneously.Parameter estimation adopts " rating value ", but not flow lift whose conic coefficient, more intuitively, be more convenient for estimating, greatly strengthen the practicability of the method.

(application examples 1)

Analyze the pump characteristic shown by Fig. 4 further, analysis result is as shown in Fig. 6 and Fig. 7.

Fig. 6 represents the tracing analysis figure of the XW-10 water pump of fallback; Fig. 7 represents the efficiency schematic diagram of the XW-10 water pump of fallback.

Known by the result shown by Fig. 6, Fig. 7, in this water factory, the long-time delivery pressure of XW-10 water pump is too low, is in fallback interval, and in conjunction with pipe network, recent and operating conditions at a specified future date is replaced by variable frequency pump in water factory.

The comprehensive unit power consumption of kiloton water of this water factory two pump house improved is by 88.74kwh/1000m ³reduce to 77.31kwh/1000m ³, kiloton water power consumption reduces 12.88%, reduces the successful of power consumption.

(application examples 2)

Obtain above-mentioned water factory (code XW) and another this water factory of city (code XS) every platform pump indicatrix basis on, carry out the water pump Optimized Operation of Liao Gai city water supply system, scheduling is divided into four times and completes, because running into the adjustment of a lot of actual used water amount in implementation process at the scene, water-use model changes, prediction water pump in pump station compound mode exists the realities such as restriction, reduce the on-off times of general pump, be connected by four-stage scheme and optimize and revise.Wherein, 7:00-18:00 period water pump regular job and the contrast of Optimized Operation table are as shown in following table 5 and table 6.

Wherein, XS water factory has XS1, XS2, XS3, XS5, XS6 5 pumps altogether, and wherein only XS1 pump is variable frequency pump.

Table 5 water supply system water pump regular job table

Table 6 water supply system water pump Optimized Operation table

Compared with Optimized Operation scheme, main difference is that in unoptimizable scheme, variable frequency pump XW6 is operated in highest frequency and without conversion, only compensates at the XW10 that water supply peak period unlatching efficiency is lower; And XW6 frequency of okperation changes according to actual conditions in optimized project, open identical with XW10 power but that efficiency is higher XW4 in the water use peak phase.In energy-saving analysis, also can see that energy-saving effect is mainly reflected in XW water factory.

According to the watering law of two water pumping stations, evaluation analysis is carried out to pumping plant power consumption in optimized project 24 hours and benefit, and survey has been done to the electricity consumption situation of pumping plant March.Two parts of power consumption statistics comparing results show, by researching and analysing the problem of 24 hours water pump Optimized Operations in this city's water engineering system, the Feedwater system optimization scheduling model of foundation, tentatively reduces the power consumption of pumping plant.

As shown in table 7 below, wherein kiloton water power consumption decline 0.24%, the XW water factory of XS water factory declines 9.63%, and kiloton water consumption adds up to reduction by 3.93%, obviously improves the economic benefit of water system.

The energy saving in pumping station analysis in 24 hours of table 7 water supply system

As can be seen here, described method can obtain the separate unit pump's performance curre of degree of precision, every platform water pump parameter of curve can be obtained fast, efficiently, be carry out water pumping station assessment, water factory save energy and reduce the cost and optimizing operation scheduling basic foundation, for production and practice meaning especially great.

In practice, for pumping plant managerial demand, in the import and export of pumping plant, a large amount of monitoring instruments is installed, has been equipped with professional staff's record of patrol, have accumulated a large amount of pumping station operation creation data.In addition, at water pump, pump variable frequency device, pumping plant ammeter reading, control valve are also provided with related sensor record water pump operation on off state, pump motor operation frequency, control valve aperture changes, these facility record pumping station operation states, and forms pumping plant day-to-day operation history data store in a database.

The historical data of the outlet pressure after the parallel connection of pumps of the measurement that method of the present invention utilizes water factory to record and total discharge, single pump switch record, adopts sensitivity data analysing method, calculates the parameter of curve of single pump, have good precision.The reasonable a series of instrument of arrangement and monitoring equipment on the basis of existing water supply pumping plant unit, according to Operational Limitss such as pumping plant total discharge, pumping plant total pressure, single pump switch records, instead push away separate unit water pump curve parameter, a large amount of manpowers and instrument expense can be saved, and avoid the interference to water factory's production.Particularly at the pumping plant without test condition, the historical data fully under a large amount of different pump switch combination condition of application, can obtain every platform water pump parameter of curve fast, efficiently by the method.

Finally it should be noted that above embodiment only in order to technological scheme of the present invention to be described and unrestricted.Although with reference to above-described embodiment to invention has been detailed description, those of ordinary skill in the art is to be understood that, still can modify to technological scheme of the present invention and be equal to replacement, and not departing from the spirit and scope of the technical program, it all should be encompassed in the middle of right of the present invention.

Claims (3)

1. a urban water plant pump station separate unit characteristic curve of pump rating method, wherein said urban water plant pump station comprises multiple stage pump, and described multiple stage pump is in parallel, and described method comprises the steps:

(1) Monitoring Data collection: by being arranged on the monitoring instrument that pumping plant is imported and exported, obtain pumping station operation historical state data;

(2) Monitoring Data process: utilize the pumping station operation historical state data recorded by step (1), obtains pumping plant total discharge, pumping plant total head, separate unit water pump operation state and pumping plant power consumption, generating process data;

(3) Analysis on monitoring data: to the generation data in step (2), adopts sensitivity analytical method, according to given pumping plant performance curve equation, calculates separate unit pump's performance curre parameter in pumping plant;

(4) parameter application: the separate unit pump's performance curre parameter obtained in step (3) is carried out simulation application, obtains separate unit pump's performance curre, the assessment of separate unit pump energy consumption, separate unit water pump improvement project and Optimal Operation of Pumping Stations;

Wherein, in described step (2), total head, the total discharge set of equation of pump when described multiple stage parallel connection of pumps are expressed as follows:

Lift: H=H _0ic _i ²-A _iq _i ²(i=1...n) 1.

Flow: $Q={\mathrm{\Σ}}_{i=1}^n{Q}_i$ ②

Wherein, H represents lift;

N represents the number of pump in parallel;

H _0irepresent the net lift of i-th water pump;

C _irepresent the rotating ratio of i-th water pump, if general pump, then c _i=1;

A _irepresent the frictional resistance of i-th water pump;

Q _irepresent the flow of i-th water pump;

And in described step (3), described sensitivity analytical method comprises:

A () separate unit parameters of pump is simulated: according to the record data of each water pump, and rule of thumb set the parameter space Ω possibility span of separate unit characteristic curve of pump, described parameter space Ω is expressed from the next:

Ω＝{(H ₀₁，…，H _0i，…，H _0n，A ₁，…，A _i，…，A _n)︱H _0ia≦H _0i≦H _0ib，A _ia≦A _i≦A _ib，i＝1…n}③

Wherein:

H _0irepresent separate unit water lift of pump parameter, its value is H ₀₁~ H _0nin any one;

A _irepresent the frictional resistance parameter of separate unit water pump, its value is A ₁~ A _nin any one;

H _0ia, H _0ibfor separate unit water lift of pump parameter H _oithe upper and lower of valued space, A _ia, A _ibfor the frictional resistance parameter A of separate unit water pump _ithe upper and lower of valued space;

B () separate unit parameters of pump is assessed: the parameter space according to pump characteristic produces arbitrary group of random numbers at random, and substitutes into the formulae discovery solving lift-flow, adopts one group of (H _p, Q _p) coordinate replaces in pump curvilinear equation H ₀, A value, its conversion relation as shown in the formula:

$H_0=\frac{4}{3}{H}_p$ ④

$A=\frac{H_p}{3{Q_p}^2}$ ⑤

4. and 5. formula is substituted into formula 1. in calculation arrangement draw following equation:

${Q^{\′}}_{j_{k}k}=\underset{i=1}{\overset{n}{\Σ}}{Q}_{pi}\·\sqrt{S_{ik}\·(4c_{ik}^2-\frac{3H_{j_{k}k}}{H_{pi}})}$ ⑥

Random H is organized to each _pi, Q _pi, according to calculate and with physical record value compare, choose calculated value with record value squared difference and assessment this group parameter quality, calculated by following formula:

$\mathrm{\ϵ}=\underset{k=1}{\overset{q}{\Σ}}\underset{j_k=1}{\overset{m_k}{\Σ}}{({Q^{\′}}_{j_{k}k}-Q_{j_{k}k})}^2$ ⑦

Size according to ε sorts, and chooses the solving result of parameters combination as final water pump parameter of curve that ε is 1 ~ 2% correspondence;

Wherein, H _prepresent the rated head of water pump;

Q _prepresent the rated flow of water pump;

H ₀represent the net lift of water pump;

A represents the frictional resistance of water pump;

H _pirepresent the rated head of i-th water pump;

Q _pirepresent the rated flow of i-th water pump;

represent that pump combination sequence number is that under k, record number is j _kpump house work lift;

S _ikrepresent that pump combination sequence number is the status parameter of i-th water pump under k;

represent that pump combination sequence number is that under k, record number is j _kthe pump house flow calculated by random parameter;

represent that pump combination sequence number is that under k, record number is j _kpump house record flow;

C _ikrepresent that pump combination sequence number is the rotating ratio of lower i-th water pump of k;

ε represents with between error;

M _kvalue is pump combination sequence number is record number under k;

I represents that pump is numbered, and i=1...n, k represent pump combination sequence number, k=1...q, j _krepresent that pump combination sequence number is the record number under k, j _k=1 ... m _k, when pump is opened, S _ikbe 1, when pump cuts out, S _ikbe 0.

2. urban water plant pump station separate unit characteristic curve of pump rating method according to claim 1, is characterized in that, also comprise following evaluates traffic error step:

(I) determine parameter and parameter search scope, and generate random parameter group;

(II) one group of parameter is applied in pump group objects;

(III) one group of working state recording is selected to specify lift H according to working state recording;

(IV) calculated flow rate Q, and the error of evaluates traffic Q;

(V) judge whether to have assessed all operating modes; If so, then step (VI) is entered; If not, then turn back to step (III);

(VI) assess the mean error of whole operating mode, and judge whether that all population of parameter have been assessed, if so, then enter into step (VII); If not, then turn back to step (II);

(VII) screen the minimum population of parameter of error, and assess the requirement whether minimal error meets setting, if so, then enter into step (VIII); If not, then enter into step (IX);

(VIII) satisfactory population of parameter is determined;

(IX) to reappraise parameter search space or openly record the accuracy of data.

3. urban water plant pump station separate unit characteristic curve of pump rating method according to claim 1, is characterized in that, also comprises following assessment lift error step:

I () determines parameter and parameter search scope, and generate random parameter group;

(ii) one group of parameter is applied in pump group objects;

(iii) one group of working state recording is selected to carry out target flow Q according to working state recording;

(iv) calculate lift H, and assess the error of lift H;

V () judges whether to have assessed all operating modes; If so, then step (vi) is entered; If not, then turn back to step (iii);

(vi) assess the mean error of whole operating mode, and judge whether that all population of parameter have been assessed, if so, then enter into step (vii); If not, then turn back to step (ii);

(vii) screen the minimum population of parameter of error, and assess the requirement whether minimal error meets setting, if so, then enter into step (viii); If not, then enter into step (ix);

(viii) satisfactory population of parameter is determined;

(ix) to reappraise parameter search space or openly record the accuracy of data.