CN106022542A - Enterprise gateway load prediction method based on operation load characteristics - Google Patents

Abstract

The invention discloses an enterprise gateway load prediction method based on operation load characteristics. The method comprises steps that (1), the sample data required for load prediction is acquired; (2), the influence factor information required for load prediction is acquired; (3), load characteristic analysis on an operation load is carried out, and prediction operation loads are divided on the basis of types according to a result; (4), prediction model matching is carried out, and sample set selection is carried out; (5), whether the model and the sample both have prediction conditions is determined, if yes, gateway component load prediction is carried out, prediction error statistics is further carried out, and component prediction results are corrected continuously according to prediction errors; and (6), all the operation component prediction results are superposed, whether the set conditions are satisfied is determined, if yes, the final gateway prediction result is outputted. The method is advantaged in that the final enterprises gateway load prediction result is acquired, taking consideration of load characteristics of each enterprise operation and load change trend, enterprise load prediction precision can be effectively improved.

Description

A kind of enterprise based on process load characteristic critical point load forecasting method

Technical field

The present invention relates to a kind of enterprise based on process load characteristic critical point load forecasting method, belong to industrial undertaking and produce The control technical field of process.

Background technology

Power load relates to national economy and the industry-by-industry of people's lives and field, the management of power load not only relation Power grid security, stable operation, meanwhile, be related to electric power enterprise and user at the moment and long-term interest, therefore, strengthen electricity consumption Management is significant.For large-scale electricity consumption enterprise, it has the most different with electrical feature from large area.Enterprise uses Electricity is mainly produced by enterprise and is determined, is affected by generation operating mode and rhythm of production, shows the spy significantly impacting wave mode Levy so that existing Predicting Technique can not directly apply in the electro-load forecast of large enterprise.The load of bulk power grid is pre- Survey precision higher, but affected by production status instable the biggest in the load prediction of large-scale electricity consumption enterprise, it was predicted that precision Relatively poor, it is therefore desirable to grasp the part throttle characteristics in all kinds of electricity consumption of enterprise workshop, power consumption characteristics, by effective technology hands Section is by good for the Resolving probiems of the critical point load prediction of industrial undertaking.

For a long time, load prediction is an important process content of electric power demand side.Grid company by intelligent meter meter, The Real-time Load of each electric terminal can be monitored by terminal, thus provides requirement forecasting more accurately, the most all right Being controlled the electrical equipment automatically run, load prediction accurately can make to provide advanced bearing into the management and running of electrical network Lotus change information, user can also be according to the load prediction results at critical point simultaneously, it is achieved autonomous wrong peak load, reduces the confession of enterprise Electricity cost.

Therefore, business electrical quality improves in industrial undertaking to be realized, and reduces the target of business electrical cost, needs to have set up Kind enterprise's critical point load prediction system, and combine the process load characteristic of enterprise, use the Forecasting Methodology of coupling, improve enterprise The accuracy of industry load prediction, provides scientific utilization of electricity, the load management platform of using electricity wisely for enterprise, improves the electricity consumption effect of enterprise Rate.

Currently, the Forecasting Methodology that the many employings of method of enterprise's critical point load prediction are identical with bulk power grid, this Forecasting Methodology master Will be according to the historical data of critical point future position, when enterprise's production schedule is altered or modified, it was predicted that result and actual deviation are relatively big, Cannot provide for the economic load dispatching of enterprise accurately can parameter evidence, it is impossible to meet utility power grid scheduling and the method for operation is worked out wants Ask.

Summary of the invention

The deficiency existed for prior art, it is an object of the present invention to provide a kind of enterprise based on process load characteristic critical point Load forecasting method, it is possible to obtain final enterprise's critical point load prediction results, owing to having considered each operation of enterprise Part throttle characteristics and load variations trend, can be effectively improved enterprise's load prediction precision.

To achieve these goals, the present invention is to realize by the following technical solutions:

A kind of based on process load characteristic the enterprise critical point load forecasting method of the present invention, including following step:

(1) obtaining the sample data that load prediction needs, described sample data includes (blast furnace, converter, oxygen processed, air blast, burning Knot, continuous casting, electric furnace refining, cold rolling, hot rolling, dining room, Administrative Area) the second DBMS of process load, minute load data, hour Electric degree data and day total electric degree data；

(2) obtain the influence factor's information needed during load prediction, described influence factor's information include weather information, Scheduling rule, the production schedule and repair schedule；

(3) according to time period in Pinggu, peak of electricity price, day minutes sequence curve is divided into three groups, during for difference Section, according to feature and the externalities factor information of the sample data of each operation, carries out part throttle characteristics to process load and divides Analysis, and according to Load Characteristic Analysis result prediction process load is divided into following five classes: life load, production load, impact Load, steady load and fluctuating load；

(4) set up be applicable to different load type forecast model (include weight moving average forecasting, trend slip authority Prediction, single exponential smoothing prediction, double smoothing prediction, Three-exponential Smoothing prediction, second order self-adaptive prediction, multinomial Matching prediction, correlation coefficient prediction, dynamic time warping predict these forecast models)；Part throttle characteristics according to each operation refers to Mark data are predicted Model Matching (matching process is prior art, and here is omitted), and according to current process load (nearly two hours) historical variations trend curve carries out the selection of sample set, and (system of selection is prior art, the most superfluous State)；

(5) according to the input requirements of forecast model, it is judged that model and the sample of operation prediction measuring point the most all possess prediction Condition, if not possessing predicted condition, turning to step (4), otherwise carrying out critical point component load prediction, and actuarial prediction error, Constantly component is predicted the outcome according to forecast error and be modified (modification method is prior art, and here is omitted), until Meet precision of prediction requirement；

(6) all process steps component predicted the outcome be overlapped, and by after superposition predict the outcome and critical point predicts the outcome Compare, the scheduling rule predicted the outcome with work out is compared simultaneously, it may be judged whether meet the condition set, if not Meeting condition set in advance, then turn to step (3), re-start prediction, otherwise predict the outcome output by final critical point, knot Restraint whole prediction process.

In step (1), the sampling period of the second DBMS of described process load can set according to operation data acquiring frequency For 3-60 second；

Within described minute, load data is by being weighted averagely obtaining to described second DBMS, the computing formula of minute load For:

$P_{Minute}\left(i\right)=\frac{\underset{j=1}{\overset{N}{\Σ}}\mathrm{\ω}\left(j\right)P_{Second}\left(j\right)}{\underset{j=1}{\overset{N}{\Σ}}\mathrm{\ω}\left(j\right)}---\left(1\right)$

In formula, P_MinuteI () is minute load value of i-th minute；

N is the sampling number of second DBMS in a minute；

The weight of the corresponding jth second DBMS of ω (j)；

P_SecondJ () is one minute interior jth second DBMS；

Within described hour, electric degree data obtain by being integrated described second DBMS calculating, the computing formula of hour electric degree For:

$E_{Hour}\left(h\right)=\underset{m=1}{\overset{M}{\Σ}}{P}_{\sec ond}\left(m\right)dt---\left(2\right)$

In formula, E_HourH () is the electricity of the h hour；

P_secondM () is the h hour interior m-th second DBMS.

Dt is the time difference of m second level sampling instant and the m-1 time second DBMS sampling instant；

M is sampling number total in 1 hour；

Described day, total electric degree data were by cumulative for electric degree on the same day hour acquisition；

Above four class data are all deposited into historical data by the data engine of enterprise's power scheduling secondary unified platform, For analyzing and prediction.

Within above-mentioned minute, load data is by being weighted averagely obtaining to described second DBMS, and wherein, weights omega (j) is really Surely have employed seasonal effect in time series morphological distance as determining the foundation of weight size, concrete grammar is as follows:

A () obtains time series S of all second DBMSs collected previous minute_n-1, and the institute gathered for current minute There is time series S of second DBMS_n, S_n-1With S_nSeasonal effect in time series length identical；

B () calculates S_n-1、S_nAverage value P_{n_avg}、P_{n-1_avg}；

C () obtains final ω (j),

$\mathrm{\ω}\left(j\right)=\left|\frac{S_n\left(j\right)-S_{n-1}\left(j\right)}{P_{n\_avg}-P_{n-1\_avg}}\right|---\left(3\right)$

Wherein, S_nJ () is the jth second DBMS of n-th minute；

S_n-1J () is the jth second DBMS of (n-1)th minute；

In step (2), described weather information includes weather pattern, hour temperature, hour wind-force, hour sunshine, humidity, and Employing below equation calculating effective temperature:

$ET=37-\frac{37-T_a}{0.68-0.14RH+\frac{1}{1.76+1.4V^{0.75}}}-0.29T_a(1-RH)---\left(4\right)$

In formula, ET is effective temperature；

T_aFor mean daily temperature；

RH is per day relative humidity；

V is per day wind speed；

Described scheduling rule includes the critical point load limit of enterprise, requirement limit value and powers the period in the flat difference of peak valley Load retrains；

The described production schedule wants clear and definite time started, end time, production product category, typical case's electricity consumption curve and electricity consumption total Amount；

Described repair schedule clearly to overhaul time started, end time, the production process having influence on, the allusion quotation of maintenance process Type load curve and electricity consumption increment.

The present invention utilizes the multivariate data that Power Secondary intergrated workbench provides, in conjunction with the power consumption characteristics of industrial undertaking With management and running requirement, complete the load prediction of each master operation of enterprise, and by the index number to each process load characteristic According to analysis, improve the accuracy of enterprise critical point load prediction, considered part throttle characteristics and the load of each operation of enterprise Variation tendency, can be effectively improved enterprise's load prediction precision, lays the first stone for working out more reasonably power supply plan.

Accompanying drawing explanation

Fig. 1 is the flowchart of enterprise based on process load characteristic critical point load prediction.

Detailed description of the invention

For the technological means making the present invention realize, creation characteristic, reach purpose and be easy to understand with effect, below in conjunction with Detailed description of the invention, is expanded on further the present invention.

See Fig. 1, a kind of based on process load characteristic the enterprise critical point load forecasting method of the present invention, specifically include with Under several steps:

(1) data engine and the application engine of enterprise's power scheduling secondary intergrated workbench offer are provided, obtain negative The sample data that lotus prediction needs, the most electric including: the second DBMS of process load, minute load data, hour electric degree data, day Degrees of data.

Wherein, the sampling period of the second DBMS of process load is set as 3 seconds, and minute load data is by second progression Averagely obtaining according to being weighted, hour electric degree data obtain by being integrated second DBMS calculating, day total electric degree data will The same day hour, electric degree increment accumulation obtained, and above four class data are all drawn by the data of enterprise's power scheduling secondary unified platform Hold up and be deposited into historical data, for analyzing and prediction；

The computing formula of minute load is:

$P_{Minute}\left(i\right)=\frac{\underset{j=1}{\overset{N}{\Σ}}\mathrm{\ω}\left(j\right)P_{Second}\left(j\right)}{\underset{j=1}{\overset{N}{\Σ}}\mathrm{\ω}\left(j\right)}---\left(1\right)$

In formula, P_MinuteI () is minute load value of i-th minute；

N is the sampling number of second DBMS in a minute, and in this method, the value of N is 20；

The weight of the corresponding jth second DBMS of ω (j)；

P_SecondJ () is one minute interior jth second DBMS；

The computing formula of hour electric degree is:

$E_{Hour}\left(h\right)=\underset{m=1}{\overset{M}{\Σ}}{P}_{\sec ond}\left(m\right)dt---\left(2\right)$

In formula, E_HourH () is the electricity of the h hour；

P_secondM () is the h hour interior m-th second DBMS；

Dt is the time difference of m second level sampling instant and the m-1 time second DBMS sampling instant；

M is sampling number total in 1 hour；

Within described minute, load data is by being weighted averagely obtaining to second DBMS, and wherein the determination of weights omega (j) is adopted With seasonal effect in time series morphological distance as determining the foundation of weight size, concrete grammar is as follows:

A () obtains time series S of all second DBMSs collected previous minute_n-1, and the institute gathered for current minute There is time series S of second DBMS_n, owing to using identical sample frequency, S_n-1With S_nSeasonal effect in time series length identical；

B () calculates S_n-1、S_nAverage value P_{n_avg}、P_{n-1_avg}, in order to embody the diversity factor in time series, work as the time Data deviation average in sequence is the most remote, and the weight that should give is the biggest；

C () utilizes formula (3) to calculate, obtain final ω (j).

$\mathrm{\ω}\left(j\right)=\left|\frac{S_n\left(j\right)-S_{n-1}\left(j\right)}{P_{n\_avg}-P_{n-1\_avg}}\right|---\left(3\right)$

Wherein, S_nJ () is the jth second DBMS of n-th minute；

S_n-1J () is the jth second DBMS of (n-1)th minute；

(2) obtain influence factor's information of needing during prediction, including: weather information, scheduling rule, the production schedule, Repair schedule.

Wherein, the weather information contained by influence factor's information includes weather pattern, hour temperature, hour wind-force, little time According to, humidity, and below equation is used to calculate effective temperature:

$ET=37-\frac{37-T_a}{0.68-0.14RH+\frac{1}{1.76+1.4V^{0.75}}}-0.29T_a(1-RH)---\left(4\right)$

In formula, ET is effective temperature (DEG C)；

T_aFor mean daily temperature (DEG C)；

RH is per day relative humidity；

V is per day wind speed (m/s)；

Wherein scheduling rule includes the critical point load limit of enterprise, requirement limit value and powers the period in the flat difference of peak valley Load retrains；

The production schedule wants clear and definite time started, end time, production product category, typical case electricity consumption curve and electricity consumption total amount etc. Information；

Repair schedule clearly to overhaul the time started, the end time, the production process having influence on, maintenance process typical case bear The information such as lotus curve and electricity consumption increment.

(3) according to feature and the external influence factors information of the sample data of each operation, process load is carried out load Specificity analysis, and prediction process load is divided into following four classes according to analysis result: life load, to produce load, impact negative Lotus, steady load, fluctuating load.

Wherein, the Load Characteristic Analysis of process load is according to the time period in Pinggu, peak of electricity price, by day minutes sequence Curve segmentation becomes three groups, respectively S_peak、S_plat、S_vally；Carry out the calculating of index system for different periods, wherein calculate Index includes feature and the external influence factors information of the middle sample data according to each operation, process load carries out load special Property analyze, and according to analysis result prediction process load is divided into following four classes: life load, production load, impact load, Steadily load, fluctuating load.

(4) forecast model is mated according to the achievement data of part throttle characteristics, and according to the variation tendency of current process load Carry out the selection of sample set.

(5) when operation predicts that the model of measuring point and sample all possess predicted condition, critical point component load prediction is carried out, and Actuarial prediction error, is constantly modified result according to forecast error, until meeting precision of prediction requirement.

Wherein, in Forecasting Methodology storehouse, set up forecast model and the algorithm being applicable to different load type, according to each work The Load characteristics index data of sequence carry out Model Matching automatically, and come according to current 30 minutes interior load variations trend curves Carry out the selection of sample set, determine final Forecasting Methodology, critical point component load is predicted and revises.

(6) all process steps component is predicted the outcome it is overlapped, and predict the outcome with critical point and compare, simultaneously with formulation Scheduling rule compare, if being unsatisfactory for condition set in advance, then enter the 3rd step, re-start prediction, the most satisfied During condition, just by the final output that predicts the outcome, whole prediction process terminates.

Wherein, operation component predict the outcome superposition time, the method that the determination of each subitem weight uses deviation maximization, constantly It is modified predicting the outcome.

To sum up, a kind of enterprise based on process load characteristic of present invention critical point load forecasting method, utilize Power Secondary one The multivariate data that body data platform provides, in conjunction with power consumption characteristics and the management and running requirement of industrial undertaking, completes enterprise each The load prediction of master operation, and by the analysis of the achievement data to each process load characteristic, improve enterprise's critical point load pre- The accuracy surveyed, has considered part throttle characteristics and the load variations trend of each operation of enterprise, can be effectively improved enterprise and bear Lotus precision of prediction, lays the first stone for working out more reasonably power supply plan.

The ultimate principle of the present invention and principal character and advantages of the present invention have more than been shown and described.The technology of the industry Personnel, it should be appreciated that the present invention is not restricted to the described embodiments, simply illustrating this described in above-described embodiment and description The principle of invention, without departing from the spirit and scope of the present invention, the present invention also has various changes and modifications, and these become Change and improvement both falls within scope of the claimed invention.Claimed scope by appending claims and Equivalent defines.

Claims (4)

1. enterprise based on a process load characteristic critical point load forecasting method, it is characterised in that include following step:

(1) obtaining the sample data that load prediction needs, described sample data includes the second DBMS of process load, minute load Data, hour electric degree data and day total electric degree data；

(2) obtaining the influence factor's information needed during load prediction, described influence factor's information includes weather information, scheduling Rule, the production schedule and repair schedule；

(3) according to time period in Pinggu, peak of electricity price, day minutes sequence curve is divided into three groups, for different periods, The feature of the sample data according to each operation and externalities factor information, carry out Load Characteristic Analysis to process load, And according to Load Characteristic Analysis result prediction process load being divided into following five classes: life load, to produce load, impact negative Lotus, steady load and fluctuating load；

(4) foundation is applicable to the forecast model of different load type；Load characteristics index data according to each operation carry out pre- Survey Model Matching, and carry out the selection of sample set according to the historical variations trend curve of current process load；

(5) according to the input requirements of forecast model, it is judged that model and the sample of operation prediction measuring point the most all possess predicted condition, If not possessing predicted condition, turning to step (4), otherwise carrying out critical point component load prediction, and actuarial prediction error, according to Component is constantly predicted the outcome and is modified by forecast error, until meeting precision of prediction requirement；

(6) all process steps component is predicted the outcome it is overlapped, and by predicting the outcome to predict the outcome with critical point and carry out after superposition Relatively, the scheduling rule predicted the outcome with work out is compared, it may be judged whether meet the condition set, if be unsatisfactory for simultaneously Condition set in advance, then turn to step (3), re-start prediction, and otherwise predict the outcome output by final critical point, terminates whole Individual prediction process.

A kind of enterprise based on process load characteristic the most according to claim 1 critical point load forecasting method, its feature exists In, in step (1), the sampling period of the second DBMS of described process load can be set as 3-according to operation data acquiring frequency 60 seconds；

Within described minute, load data is by being weighted averagely obtaining to described second DBMS, and the computing formula of minute load is:

$P_{Minute}\left(i\right)=\frac{\underset{j=1}{\overset{N}{\Σ}}\mathrm{\ω}\left(j\right)P_{Second}\left(j\right)}{\underset{j=1}{\overset{N}{\Σ}}\mathrm{\ω}\left(j\right)}---\left(1\right)$

In formula, P_MinuteI () is minute load value of i-th minute；

N is the sampling number of second DBMS in a minute；

The weight of the corresponding jth second DBMS of ω (j)；

P_SecondJ () is one minute interior jth second DBMS；

Within described hour, electric degree data obtain by being integrated described second DBMS calculating, and the computing formula of hour electric degree is:

$E_{Hour}\left(h\right)=\underset{m=1}{\overset{M}{\Σ}}{P}_{\sec ond}\left(m\right)dt---\left(2\right)$

In formula, E_HourH () is the electricity of the h hour；

P_secondM () is the h hour interior m-th second DBMS；

Dt is the time difference of m second level sampling instant and the m-1 time second DBMS sampling instant；

M is sampling number total in 1 hour；

Described day, total electric degree data were by cumulative for electric degree on the same day hour acquisition；

Above four class data are all deposited into historical data by the data engine of enterprise's power scheduling secondary unified platform, for dividing Analysis and prediction use.

A kind of enterprise based on process load characteristic the most according to claim 2 critical point load forecasting method, its feature exists In, within described minute, load data is by being weighted averagely obtaining to described second DBMS, and wherein, the determination of weights omega (j) is adopted With seasonal effect in time series morphological distance as determining the foundation of weight size, concrete grammar is as follows:

A () obtains time series S of all second DBMSs collected previous minute_n-1, and all seconds gathered for current minute Time series S of DBMS_n, S_n-1With S_nSeasonal effect in time series length identical；

B () calculates S_n-1、S_nAverage value P_{n_avg}、P_{n-1_avg}；

C () obtains final ω (j),

$\mathrm{\ω}\left(j\right)=\left|\frac{S_n\left(j\right)-S_{n-1}\left(j\right)}{P_{n\_avg}-P_{n-1\_avg}}\right|---\left(3\right)$

Wherein, S_nJ () is the jth second DBMS of n-th minute；

S_n-1J () is the jth second DBMS of (n-1)th minute.

A kind of enterprise based on process load characteristic the most according to claim 1 critical point load forecasting method, its feature exists In, in step (2), described weather information includes weather pattern, hour temperature, hour wind-force, hour sunshine, humidity, and uses Below equation calculating effective temperature:

$ET=37-\frac{37-T_a}{0.68-0.14RH+\frac{1}{1.76+1.4V^{0.75}}}-0.29T_a(1-RH)---\left(4\right)$

In formula, ET is effective temperature；

T_aFor mean daily temperature；

RH is per day relative humidity；

V is per day wind speed；

Described scheduling rule includes the critical point load limit of enterprise, requirement limit value and the load in the peak valley flat difference power supply period Constraint；

The described production schedule wants clear and definite time started, end time, production product category, typical case's electricity consumption curve and electricity consumption total amount；

Described repair schedule clearly to overhaul the time started, the end time, the production process having influence on, maintenance process typical case bear Lotus curve and electricity consumption increment.