CN103761580A - Energy consumption supervision method capable of achieving energy dynamic prediction for beer brewing enterprises - Google Patents

Abstract

The invention relates to an energy consumption system and an energy consumption supervision method capable of achieving energy dynamic prediction for beer brewing enterprises. The energy consumption system comprises an energy data input unit, a data processing unit, an energy prediction unit, an energy management unit and an energy consumption alarming and processing unit, wherein the data processing unit performs normalization processing on the data through a [0, 1] normalized method. When the energy consumption is abnormal, the energy consumption alarming and processing unit gives out an alarm through alarming sound or a light-emitting diode. Compared with the existing method depending experience-based judgment of managers, the energy consumption supervision method capable of achieving energy dynamic prediction for beer brewing enterprises has the advantages of being simple in usage and capable of adding objective evaluation standards.

Description

A kind of beer producers energy resource consumption measure of supervision of energy performance prediction

Technical field

The present invention relates to a kind of energy consumption system and measure of supervision, be specifically related to a kind of beer producers energy consumption system and measure of supervision of energy performance prediction.

Background technology

Beer enterprise is always energy consumption rich and influential family, according to statistics, between Dun Pihao grain domestic enterprise, differs 30 kilograms, and beer total loss rate differs 5%～6%, 20 tons of the domestic average out to of ton beer water consumption, and the world is 6 tons; The power consumption of ton beer, domestic is 130 degree, the world is 110 degree; Ton beer consumption standard coal equivalent, domestic is 170 kilograms, the world is 80 kilograms.In recent years, Domestic Beer industrial development is swift and violent, and beer production is every year with the speed increase of 20% left and right, how energy-saving and cost-reducing as early as possible when improving product, increases the benefit, and day by day becomes the focus that each beer producers and whole beer industry are paid close attention to.The main energy resource consumption of brewery is steam, water, electricity, carbon dioxide.2010, national beer production was 6,483 ten thousand tons, and according to 1500 yuan of calculating of average ton wine cost, the annual cost of producing beer in the whole nation is 87,200,000,000 yuan, uses the energy to account for 17% of cost calculate, produce every year beer and use approximately 14,800,000,000 yuan of the energy according to beer.Because national beer enterprise does not generally have effective quantitative analysis means, so energy dissipation situation is serious, there is very large energy-saving and cost-reducing space.

The energy-saving and cost-reducing problem of beer enterprise relates to many-sided reason, the energy supervision and management of applying efficient energy-efficient equipment, adopt advanced production technology, improving operator's technical ability and sense of responsibility, science is all effectively to realize the energy-saving and cost-reducing effective ways of beer enterprise.Wherein, utilize efficient energy-efficient equipment and advanced production technology, can realize energy-saving and emission-reduction, and successful, but need to drop into a large amount of fund costs; The technical ability and the sense of responsibility that improve operator also can realize the energy-saving and cost-reducing order ground of beer enterprise.Utilize the energy supervision and management of science, can, on the basis of existing beer equipment and production technology, realize the supervisory role that the energy is used, consumed, promote that operator sharpens one's skills and sense of responsibility subjective, avoid unnecessary energy resource consumption and waste.Visible, effectively implement the energy resource consumption supervision and management of science, do not drop under the prerequisite of a large amount of fund costs, can realize the maximization that the energy effectively utilizes.

The energy resource consumption measure of supervision of traditional beer enterprise is to utilize gerentocratic experience, realize the prediction of each workshop to energy service condition, thereby judge that whether energy-output ratio is normal, gerentocratic energy forecast experience accumulation, on affecting the subjective promptness of understanding in the aspect such as market sale situation variability of Beer Brewage, capital produces directly impact to energy supervision and management, this management method has subjectivity, and affect the many factors of beer enterprise energy resource consumption, between each factor, often interact and restrict, supvr will judge the effective ways of energy resource consumption supervision from numerous influence factors, conventionally there is the property of being difficult for.In existing energy forecast method, the energy forecast that utilizes neural network, time series models, statistical model etc. to carry out, has the advantages that required sample is many, do not embody performance prediction; The support vector machine of utilizing having occurred is carried out the method for energy forecast, has the advantages that required sample is few, but does not embody the feature of performance prediction.In order to reduce the dependence to supvr's subjectivity in science energy resource consumption measure of supervision, to increase the objective evaluation standard of science energy resource consumption supervision and can reflect in time the trend that the energy changes, by utilizing the energy performance prediction of beer producers, be embodied as when supvr carries out energy resource consumption supervision important objective reference frame is provided, have important practical significance.

Summary of the invention

Technical matters to be solved by this invention is to have overcome the subjective judgement of the too much dependence supvr who exists in existing beer enterprise energy resource consumption supervision on energy forecast, on affect the many factors of energy resource consumption, change and be difficult to react in time and the problem of energy forecast dynamic, and a kind of beer producers energy resource consumption measure of supervision based on energy performance prediction is provided.

The beer producers energy resource consumption measure of supervision of a kind of energy performance prediction of the present invention, mainly pass through support vector machine, support vector machine utilizes grid-search algorithm to find best penalty parameter c and gamma value, adopt respectively polynomial kernel function, linear kernel function, gaussian kernel function, sigmoid kernel function, the packing shop to beer enterprise, power houses, the required energy of brewing workshop are predicted respectively.Energy forecast precision when more each workshop adopts different IPs function, utilization realizes the kernel function that precision of prediction is the highest the future source of energy consumption in this workshop is predicted.By support vector machine, constantly find and utilize the highest kernel function of energy resource consumption precision of prediction, constantly increase the input value of forecast model, forecast model is dynamically changed, and improved precision of prediction, realized the performance prediction of the energy, by forecast consumption value and actual consumption value, realize the energy resource consumption supervisory role of beer producers, improved workshop personnel's energy-conservation subjective initiative.

The beer producers energy resource consumption measure of supervision of a kind of energy performance prediction of the present invention, comprises the following steps:

1), the structure of energy consumption system: energy consumption system comprises energy data input cell, data processing unit, energy forecast unit, energy management unit, energy consumption warning and processing unit, described data processing unit is by [0,1] normalized method, data unification is normalized, energy forecast is provided with support vector machine in unit, energy service condition is predicted, if when energy resource consumption is undesired, energy consumption is reported to the police and processing unit sends alarm by chimes of doom or light emitting diode;

2), statistics: the actual consumption Value Data of W (x) workshop E (y) energy by n month before the mode of complicate statistics is input in energy data input cell;

3), data processing unit deal with data: the input data of the mode by complicate statistics using the actual consumption value of the W in front (n-1) individual month (x) workshop E (y) energy as W (x) workshop E (y) energy forecast model, this unit is EXCEL table; Data processing unit, by [0,1] normalized method, is normalized data unification;

4), energy forecast unit is predicted energy service condition: pass through support vector machine, respectively by PK, LK, GK, SK is as kernel function, support vector machine utilizes grid-search algorithm to find best penalty parameter c and gamma value, support vector machine is by the study to input data, obtain W (x) workshop E (y) energy forecast model, utilize energy forecast model can obtain W (x) workshop E (y) energy resource consumption predicted value M (n) _ W (x) _ E (the y) _ PV in n month, calculate M (n) _ W (x) _ E (y) _ DV, relatively PK, LK, GK, the size of M (n) _ W (x) _ E (the y) _ DV value producing under SK kernel function condition, find out and make the minimum kernel function K1 of M (n) _ W (x) _ E (y) _ DV value, M (n) _ W (x) _ E (y) _ RV is increased in the input data of forecast model, support vector machine utilizes grid-search algorithm to find best penalty parameter c and gamma value, utilize kernel function K1 to predict W (x) workshop E (y) energy resource consumption in (n+1) month, obtain W (x) workshop E (y) the energy resource consumption predicted value in (n+1) month, be M (n+1) _ W (x) _ E (y) _ PV, establish Times (n)=0,

5), if M (n+1) _ W (x) _ E (y) _ DV≤5%*M (n+1) _ W (x) _ E (y) _ RV, beer producers energy resource consumption monitor system thinks that the energy-output ratio in (n+1) month is in normal range, the value of cumulative number Times (n+1) is constant, if M (n+1) _ W (x) _ E (y) _ DV>5%*M (n+1) _ W (x) _ E (y) _ RV, the value of cumulative number Times (n+1) increases by 1 time, M (n+1) _ W (x) _ E (y) _ RV is increased to the input value of W (x) workshop E (y) energy forecast model, continue to use kernel function K ₁w (x) workshop E (y) energy resource consumption to (n+2) month is predicted, support vector machine utilizes grid-search algorithm to find best penalty parameter c and gamma value, obtains M (n+2) _ W (x) _ E (y) _ PV,

6), if M (n+2) _ W (x) _ E (y) _ DV≤5%*M (n+2) _ W (x) _ E (y) _ RV, beer producers energy resource consumption monitor system thinks that the energy-output ratio in (n+2) month is in normal range, the value of cumulative number Times (n+2) is constant, if M (n+2) _ W (x) _ E (y) _ DV>5%*M (n+2) _ W (x) _ E (y) _ RV, the value of cumulative number Times (n+2) increases by 1 time, M (n+2) _ W (x) _ E (y) _ RV is increased to the input value of W (x) workshop E (y) energy forecast model, continue to use kernel function K1 to predict W (x) workshop E (y) energy resource consumption in (n+3) month, support vector machine utilizes grid-search algorithm to find best penalty parameter c and gamma value, obtain M (n+3) _ W (x) _ E (y) _ PV,

7), if M (n+3) _ W (x) _ E (y) _ DV≤5%*M (n+3) _ W (x) _ E (y) _ RV, beer producers energy resource consumption monitor system thinks that the energy-output ratio in (n+3) month is in normal range, the value of cumulative number Times (n+3) is constant, if M (n+3) _ W (x) _ E (y) _ DV>5%*M (n+3) _ W (x) _ E (y) _ RV, the value of cumulative number Times (n+3) increases by 1 time, if Times (n+3) < 3, energy resource consumption monitor system assert that W (x) workshop E (y) energy resource consumption is in normal range, M (n+3) _ W (x) _ E (y) _ RV is increased to the input value of W (x) workshop E (y) energy forecast model, continue to use kernel function K ₁w (x) workshop E (y) energy resource consumption to (n+4) month is predicted, support vector machine utilizes grid-search algorithm to find best penalty parameter c and gamma value, obtains M (n+4) _ W (x) _ E (y) _ PV,

8), according to the method for step 7), judge the value of Times (n+4), if Times (n+4) < 3, continue to judge according to the method for step (6) value of Times (n+5), if Times (n+5) < 3, continue to judge according to the method for step 7) the value of Times (n+6), until Times (n+m) >=3;

9), energy consumption warning and processing unit are reported to the police: if Times (n+m) >=3, energy resource consumption monitor system assert that W (x) workshop E (y) energy resource consumption is in undesired scope, beer producers energy resource consumption monitor system gives the alarm to W (x) workshop E (y) energy resource consumption situation

Require supvr to W (x) workshop, to check that E (y) energy is used, Expenditure Levels, whether investigation W (x) plant personnel there is the situation of E (y) energy dissipation;

10), by the value zero clearing of cumulative number Times (n+3), i.e. Times (n+3)=0, simultaneously according to step 1) and step 2) method again find best kernel function;

Wherein, PK: polynomial kernel function;

LK: linear kernel function;

GK: gaussian kernel function;

SK:sigmoid kernel function;

M (n): in n month, n is positive integer;

M (n+m): in (n+m) month, n, m are positive integer;

W (x): x workshop;

E (y): the y energy;

PV: forecast consumption value;

RV: actual consumption value;

M (n) _ W (x) _ E (y) _ PV: E (y) the energy forecast consumption figures in n month W (x) workshop;

M (n) _ W (x) _ E (y) _ RV: E (y) energy actual consumption value in n month W (x) workshop;

M (n) _ W (x) _ E (y) _ DV: the absolute value of difference between E (y) the energy forecast consumption figures in n month W (x) workshop and actual consumption value, M (n) _ W (x) _ E (y) _ DV=|M (n) _ W (x) _ E (y) _ PV-M (n) _ W (x) _ E (y) _ RV|;

Times (n+m): the difference between E (y) energy forecast consumption figures and the actual consumption value in same month W (x) workshop be greater than actual consumption value 5% before cumulative number in (n+m) individual month.

The energy resource consumption of beer producers packing shop mainly contains and produces water consumption, production power consumption, produces consumption vapour, the power consumption of throwing light on, presses empty power consumption, carbon dioxide power consumption, the energy resource consumption of power houses mainly contains the water consumption that freezes, refrigeration power consumption, injection moulding power consumption, injection moulding water consumption, power and lighting, the empty power consumption of pressure, carbon dioxide power consumption, sewage power consumption, water treatment power consumption, heating consumption vapour, and the energy resource consumption of brewing workshop mainly contains and produces water consumption, production power consumption, production consumption vapour, presses empty power consumption, carbon dioxide power consumption, saccharification illumination, heating consumption vapour.By the various combination of W (x) and E (y), represent the different energy sources in different workshops, that is: the production water consumption that W (1) _ E (1) is packing shop, the production power consumption that W (1) _ E (2) is packing shop, the production consumption vapour that W (1) _ E (3) is packing shop, the illumination power consumption that W (1) _ E (4) is packing shop, the empty power consumption of pressure that W (1) _ E (5) is packing shop, the carbon dioxide power consumption that W (1) _ E (6) is packing shop, the refrigeration water consumption that W (2) _ E (1) is power houses, the refrigeration power consumption that W (2) _ E (2) is power houses, the injection moulding power consumption that W (2) _ E (3) is power houses, the injection moulding water consumption that W (2) _ E (4) is power houses, the power and lighting that W (2) _ E (5) is power houses, the empty power consumption of pressure that W (2) _ E (6) is power houses, the carbon dioxide power consumption that W (2) _ E (7) is power houses, the sewage power consumption that W (2) _ E (8) is power houses, the water treatment power consumption that W (2) _ E (9) is power houses, the heating consumption vapour that W (2) _ E (10) is power houses, the production water consumption that W (3) _ E (1) is brewing workshop, the production power consumption that W (3) _ E (2) is brewing workshop, the production consumption vapour that W (3) _ E (3) is brewing workshop, the empty power consumption of pressure that W (3) _ E (4) is brewing workshop, the carbon dioxide power consumption that W (3) _ E (5) is brewing workshop, the saccharification illumination that W (3) _ E (6) is brewing workshop, heating that W (3) _ E (7) is brewing workshop consumption vapour, utilizes method described in the 1st step to the 10 steps, and beer producers consumes monitor system can judge a following month packing shop, power houses, whether the energy-output ratio separately of brewing workshop exists extremely, for energy management person provides objective evaluation standard, has promoted to improve the energy-conservation subjective initiative of plant personnel simultaneously, reduces the energy dissipation situation causing due to artificial origin and occurs.

Compared with prior art beneficial effect of the present invention have following some:

(1) the beer producers energy resource consumption measure of supervision of a kind of energy performance prediction of the present invention, compared with the existing method that relies on supvr's micro-judgment, has the feature simple, that increased objective evaluation standard of using.

(2) the beer producers energy resource consumption measure of supervision of a kind of energy performance prediction of the present invention utilizes compared with BP neural network, wavelet analysis equal energy source Forecasting Methodology with existing, has that required original energy data sample energy forecast model generalization ability few, that set up is strong, energy forecast has dynamic, utilizes the feature of upgrading kernel function and input data and realize higher precision of prediction.

(3) the beer producers energy resource consumption measure of supervision of a kind of energy performance prediction of the present invention is realized compared with energy forecast method with the existing support vector machine of utilizing, and has energy forecast and have dynamic, utilizes the feature of upgrading kernel function and input data and realize higher precision of prediction.

(4) the beer producers energy resource consumption measure of supervision of a kind of energy performance prediction of the present invention, its energy forecast process is simple, precision of prediction is high, energy resource consumption supervision has objective evaluation standard, has avoided different supvrs to have the defect of different subjective judgement standards, and process is simple, convenient, speed is fast, step is clear, saves time, laborsaving.

Accompanying drawing explanation

Fig. 1 is the beer producers energy consumption system figure of energy performance prediction.

Fig. 2 is the beer producers energy resource consumption measure of supervision schematic flow sheet of energy performance prediction.

Embodiment

Below in conjunction with embodiment, the present invention will be further described:

1. consult Fig. 1, by the mode of complicate statistics, the data of each month different energy sources consumption are input in energy data input cell, this unit is EXCEL table; Data processing module, by [0,1] normalized method, is normalized data unification, and this module is the mapminmax(in matlab) function; Data after energy forecast unit reception & disposal, adopt support vector machine to carry out the prediction of energy resource consumption value; Energy resource consumption is predicted the outcome and of that month actual energy resource consumption value is input in energy management unit simultaneously, analyze energy resource consumption whether in normal range: if energy resource consumption is normal, continue the energy resource consumption prediction of next stage; If energy resource consumption is undesired, carry out energy resource consumption alarm unit and processing unit, by chimes of doom or light emitting diode, send alarm, managerial personnel examine this energy resource consumption situation.

Matlab is matrix experiment chamber.

2. consult Fig. 2, in energy forecast unit, pass through support vector machine, utilize grid-search algorithm to find best penalty parameter c and gamma value, adopt respectively polynomial kernel function, linear kernel function, gaussian kernel function, sigmoid kernel function, E (y) energy in W (x) workshop of beer producers energy resource consumption monitor system to (n+m) month carries out energy performance prediction, energy forecast precision when more each workshop adopts different IPs function, utilization realizes the kernel function that precision of prediction is the highest the future source of energy consumption in this workshop is predicted.According to predicting the outcome, if meet evaluation criterion () Rule of judgment, that is: the difference between W (x) workshop E (y) energy forecast consumption figures and the actual consumption value in prediction month is less than or equal to 5% of actual consumption value., energy monitor system assert that W (x) workshop E (y) energy resource consumption is in normal range; If meet evaluation criterion (two) condition, that is: the difference between W (x) workshop E (y) energy forecast consumption figures and the actual consumption value in prediction month is greater than 5% of actual consumption value, but this difference is greater than 5% cumulative number of actual consumption value and is less than 3 times, energy monitor system assert that W (x) workshop E (y) energy resource consumption is in normal range; If meet evaluation criterion (three) condition, that is: the difference between W (x) workshop E (y) energy forecast consumption figures and the actual consumption value in prediction month is greater than 5% of actual consumption value, and cumulative number is equal to or greater than 3 times, energy monitor system assert that W (x) workshop E (y) energy resource consumption, in undesired scope, needs supvr to investigate use, the Expenditure Levels of W (x) workshop E (y) energy.

Take the production water consumption of packing shop as example, the beer producers energy resource consumption measure of supervision of a kind of energy performance prediction of the present invention is described.

1. add up certain beer producers packing shop in the production water consumption data of 36 totally months in January, 2010 to Dec, in January, 2011 to Dec, in January, 2012 to Dec, as table 1.

Unit: ton

Table 1 packing shop is produced water consumption statistical form

Using in January, 2010 to Dec, in January, 2011 to 12, in January, 2012 to July the production water consumption data of 31 totally months as training set, find the highest kernel function of precision of prediction.(in January, 2010 was to Dec by first 30 months, in January, 2011 is to Dec, in January, 2012 is to June) packing shop produce water consumption actual consumption Value Data as the input data of forecast model, (1.90, 1.83, 1.31), support vector machine utilizes grid-search algorithm to find best penalty parameter c and gamma value, be respectively 2.918 and 12.892, support vector machine adopts respectively linear kernel function, polynomial kernel function, gaussian kernel function, during sigmoid kernel function, the production water consumption consumption forecast value in the 31st month (in the July, 2012) obtaining is as table 2.

Unit: ton

Production water consumption predicted value when table 2 adopts different IPs function

Calculate M (31) _ W (1) _ E (1) _ DV=|M (24) _ W (1) _ E (1) _ PV-M (24) _ W (1) _ E (1) _ RV|, can obtain M (31) _ W (1) _ E (the 1) _ DV=2 under linear kernel function condition, M (31) _ W (1) _ E (1) _ DV=1 under polynomial kernel function condition, M (31) _ W (1) _ E (1) _ DV=0 under gaussian kernel function condition, M (31) _ W (1) _ E (1) _ DV=1 under sigmoid kernel function condition, visible, when support vector machine adopts gaussian kernel function, its precision of prediction is the highest.The actual consumption Value Data 1.31 of the 31st month is increased in the input data of forecast model, (1.90,1.83 ..., 1.31,1.31), support vector machine utilizes grid-search algorithm to find best penalty parameter c and gamma value, is respectively 1.298 and 13.128, adopts gaussian kernel function, production water consumption consumption to the 32nd month (in August, 2012) predicts, obtaining consumption forecast value M (32) _ W (1) _ E (1) _ PV is 1.31.

3. calculate

M(32)_W(1)_E(1)_DV=|M(32)_W(1)_E(1)_PV-M(32)_W(1)_E(1)_RV|=|1.31-1.32|

=0.01,5%*M (32) _ W (1) _ E (1) _ RV=0.066, M (32) _ W (1) _ E (1) _ DV≤5%*M (32) _ W (1) _ E (1) _ RV, therefore beer producers energy resource consumption monitor system assert that the energy-output ratio in (32) month is in normal range, packing shop personnel possess sense of responsibility, do not exist and produce water consumption energy dissipation phenomenon, the cumulative number Times (32)=0 that exceeds standard in front 32 middle of the month.

4. M (32) _ W (1) _ E (1) _ RV is increased to input value, (1.90,1.83,1.31,1.31,1.32), due to cumulative number Times (32)=0, therefore continuing to use gaussian kernel function predicts the production water consumption energy resource consumption in (33) month (in September, 2012), support vector machine utilizes grid-search algorithm to find best penalty parameter c and gamma value, is respectively 3.596 and 20.926, and obtaining M (33) _ W (1) _ E (1) _ PV is 1.40.

5. calculate

M(33)_W(1)_E(1)_DV=|M(33)_W(1)_E(1)_PV-M(33)_W(1)_E(1)_RV|=|1.40-1.62|

=0.22,5%*M (33) _ W (1) _ E (1) _ RV=0.081, M (33) _ W (1) _ E (1) _ DV>5%*M (33) _ W (1) _ E (1) _ RV, therefore beer producers energy resource consumption monitor system assert that the production water consumption energy-output ratio in (33) month is in undesired scope, give a warning, need supvr to investigate use, the Expenditure Levels of the packing shop production water consumption energy, reduce unnecessary production water consumption waste.The cumulative number Times (33) that exceeds standard in front 33 middle of the month increases once, i.e. Times (33)=1.

6. M (33) _ W (1) _ E (1) _ RV is increased to input value, (1.90,1.83,1.31,1.31,1.32,1.62), due to cumulative number Times (33) <3, therefore continue to use gaussian kernel function to predict the production water consumption energy resource consumption in (34) month (in October, 2012), support vector machine utilizes grid-search algorithm to find best penalty parameter c and gamma value, be respectively 1.046 and 18.418, obtaining M (34) _ W (1) _ E (1) _ PV is 1.41.

7. calculate

M(34)_W(1)_E(1)_DV=|M(34)_W(1)_E(1)_PV-M(34)_W(1)_E(1)_RV|=|1.41-1.63|

=0.22,5%*M (34) _ W (1) _ E (1) _ RV=0.0815, M (34) _ W (1) _ E (1) _ DV>5%*M (34) _ W (1) _ E (1) _ RV, therefore beer producers energy resource consumption monitor system assert that the production water consumption energy-output ratio in (34) month is in undesired scope, give a warning, need supvr to investigate use, the Expenditure Levels of the packing shop production water consumption energy, reduce unnecessary production water consumption waste.The cumulative number Times (34) that exceeds standard in front 34 middle of the month increases once, i.e. Times (34)=2.

8. M (34) _ W (1) _ E (1) _ RV is increased to input value, (1.90,1.83,1.31,1.31,1.32,1.62,1.63), due to cumulative number Times (34) <3, therefore continuing to use gaussian kernel function predicts the production water consumption energy resource consumption in (35) month (in November, 2012), support vector machine utilizes grid-search algorithm to find best penalty parameter c and gamma value, is respectively 6.003 and 15.936, and obtaining M (35) _ W (1) _ E (1) _ PV is 1.45.

9. calculate

M(35)_W(1)_E(1)_DV=|M(35)_W(1)_E(1)_PV-M(35)_W(1)_E(1)_RV|=|1.45-1.61|

=0.16,5%*M (34) _ W (1) _ E (1) _ RV=0.0805, M (35) _ W (1) _ E (1) _ DV>5%*M (35) _ W (1) _ E (1) _ RV, therefore beer producers energy resource consumption monitor system assert that the production water consumption energy-output ratio in (35) month is in undesired scope, give a warning, need supvr to investigate use, the Expenditure Levels of the packing shop production water consumption energy, reduce unnecessary production water consumption waste.The cumulative number Times (35) that exceeds standard in front 35 middle of the month increases once, i.e. Times (35)=3.Because cumulative number Times (35) has reached 3 times, therefore again find the kernel function that precision of prediction is the highest, that is: support vector machine is respectively by linear kernel function, polynomial kernel function, gaussian kernel function, sigmoid kernel function is as kernel function, the production water consumption energy actual consumption Value Data (1.90 in (34) individual month before input, 1.83, 1.63), as the input data of forecast model, utilize grid-search algorithm to find best penalty parameter c and gamma value, be respectively 2.198 and 10.391, support vector machine is by the study to input data, obtain forecast model, utilize this forecast model can obtain the production water consumption energy resource consumption predicted value in (35) month, be M (35) _ W (1) _ E (1) _ PV, can obtain M (35) _ W (1) _ E (the 1) _ PV=1.39 under linear kernel function condition, M (35) _ W (1) _ E (1) _ PV=1.46 under polynomial kernel function condition, M (35) _ W (1) _ E (1) _ PV=1.45 under gaussian kernel function condition, M (35) _ W (1) _ E (1) _ PV=1.49 under sigmoid kernel function condition, calculate respectively M (35) _ W (1) _ E (1) _ DV=|M (35) _ W (1) _ E (1) _ PV-M (35) _ W (1) _ E (the 1) _ RV| under different IPs function, can obtain M (35) _ W (1) _ E (the 1) _ DV=0.22 under linear kernel function condition, M (35) _ W (1) _ E (1) _ DV=0.15 under polynomial kernel function condition, M (35) _ W (1) _ E (1) _ DV=0.16 under gaussian kernel function condition, M (35) _ W (1) _ E (1) _ DV=0.12 under sigmoid kernel function condition.Visible, sigmoid kernel function makes M (35) _ W (1) _ E (1) _ DV value minimum, M (35) _ W (1) _ E (1) _ RV is increased in the input data of producing water consumption energy forecast model, (1.90, 1.83, 1.63, 1.61), support vector machine utilizes grid-search algorithm to find best penalty parameter c and gamma value, be respectively 3.285 and 6.189, utilize sigmoid kernel function to predict the production water consumption energy resource consumption in (36) month (in Dec, 2012), obtaining M (36) _ W (1) _ E (1) _ PV is 1.46.Owing to again having found the kernel function of precision of prediction, therefore by exceed standard cumulative number Times (35) zero clearing, the i.e. Times (35)=0 in front 35 middle of the month.

10. calculate

M(36)_W(1)_E(1)_DV=|M(36)_W(1)_E(1)_PV-M(36)_W(1)_E(1)_RV|=|1.46-1.45|

=0.01,5%*M (36) _ W (1) _ E (1) _ RV=0.0725, M (36) _ W (1) _ E (1) _ DV<5%*M (36) _ W (1) _ E (1) _ RV, therefore beer producers energy resource consumption monitor system assert that the energy-output ratio in (36) month is in normal range, packing shop personnel possess sense of responsibility, do not exist and produce water consumption energy dissipation phenomenon, the cumulative number Times (36)=0 that exceeds standard in front 36 middle of the month.

Claims (1)

1. a beer producers energy resource consumption measure of supervision for energy performance prediction, is characterized in that comprising the following steps:

1), the structure of energy consumption system: energy consumption system comprises energy data input cell, data processing unit, energy forecast unit, energy management unit, energy consumption warning and processing unit, described data processing unit is by [0,1] normalized method, data unification is normalized, energy forecast is provided with support vector machine in unit, energy service condition is predicted, if when energy resource consumption is undesired, energy consumption is reported to the police and processing unit sends alarm by chimes of doom or light emitting diode;

2), statistics: the actual consumption Value Data of W (x) workshop E (y) energy by n month before the mode of complicate statistics is input in energy data input cell;

3), data processing unit deal with data: the input data of the mode by complicate statistics using the actual consumption value of the W in front (n-1) individual month (x) workshop E (y) energy as W (x) workshop E (y) energy forecast model, this unit is EXCEL table; Data processing unit, by [0,1] normalized method, is normalized data unification;

4), energy forecast unit is predicted energy service condition: pass through support vector machine, respectively by PK, LK, GK, SK is as kernel function, support vector machine utilizes grid-search algorithm to find best penalty parameter c and gamma value, support vector machine is by the study to input data, obtain W (x) workshop E (y) energy forecast model, utilize energy forecast model can obtain W (x) workshop E (y) energy resource consumption predicted value M (n) _ W (x) _ E (the y) _ PV in n month, calculate M (n) _ W (x) _ E (y) _ DV, relatively PK, LK, GK, the size of M (n) _ W (x) _ E (the y) _ DV value producing under SK kernel function condition, find out and make the minimum kernel function K of M (n) _ W (x) _ E (y) _ DV value ₁, M (n) _ W (x) _ E (y) _ RV is increased in the input data of forecast model, support vector machine utilizes grid-search algorithm to find best penalty parameter c and gamma value, utilizes kernel function K ₁w (x) workshop E (y) energy resource consumption to (n+1) month is predicted, obtain W (x) workshop E (y) the energy resource consumption predicted value in (n+1) month, be M (n+1) _ W (x) _ E (y) _ PV, establish Times (n)=0,

5), if M (n+1) _ W (x) _ E (y) _ DV≤5%*M (n+1) _ W (x) _ E (y) _ RV, beer producers energy resource consumption monitor system thinks that the energy-output ratio in (n+1) month is in normal range, the value of cumulative number Times (n+1) is constant, if M (n+1) _ W (x) _ E (y) _ DV>5%*M (n+1) _ W (x) _ E (y) _ RV, the value of cumulative number Times (n+1) increases by 1 time, M (n+1) _ W (x) _ E (y) _ RV is increased to the input value of W (x) workshop E (y) energy forecast model, continue to use kernel function K1 to predict W (x) workshop E (y) energy resource consumption in (n+2) month, support vector machine utilizes grid-search algorithm to find best penalty parameter c and gamma value, obtain M (n+2) _ W (x) _ E (y) _ PV,

6), if M (n+2) _ W (x) _ E (y) _ DV≤5%*M (n+2) _ W (x) _ E (y) _ RV, beer producers energy resource consumption monitor system thinks that the energy-output ratio in (n+2) month is in normal range, the value of cumulative number Times (n+2) is constant, if M (n+2) _ W (x) _ E (y) _ DV>5%*M (n+2) _ W (x) _ E (y) _ RV, the value of cumulative number Times (n+2) increases by 1 time, M (n+2) _ W (x) _ E (y) _ RV is increased to the input value of W (x) workshop E (y) energy forecast model, continue to use kernel function K1 to predict W (x) workshop E (y) energy resource consumption in (n+3) month, support vector machine utilizes grid-search algorithm to find best penalty parameter c and gamma value, obtain M (n+3) _ W (x) _ E (y) _ PV,

7), if M (n+3) _ W (x) _ E (y) _ DV≤5%*M (n+3) _ W (x) _ E (y) _ RV, beer producers energy resource consumption monitor system thinks that the energy-output ratio in (n+3) month is in normal range, the value of cumulative number Times (n+3) is constant, if M (n+3) _ W (x) _ E (y) _ DV>5%*M (n+3) _ W (x) _ E (y) _ RV, the value of cumulative number Times (n+3) increases by 1 time, if Times (n+3) < 3, energy resource consumption monitor system assert that W (x) workshop E (y) energy resource consumption is in normal range, M (n+3) _ W (x) _ E (y) _ RV is increased to the input value of W (x) workshop E (y) energy forecast model, continue to use kernel function K1 to predict W (x) workshop E (y) energy resource consumption in (n+4) month, support vector machine utilizes grid-search algorithm to find best penalty parameter c and gamma value, obtain M (n+4) _ W (x) _ E (y) _ PV,

8), according to the method for step 7), judge the value of Times (n+4), if Times (n+4) < 3, continue to judge according to the method for step (6) value of Times (n+5), if Times (n+5) < 3, continue to judge according to the method for step 7) the value of Times (n+6), until Times (n+m >=3;

9), energy consumption warning and processing unit are reported to the police: if Times (n+m) >=3, energy resource consumption monitor system assert that W (x) workshop E (y) energy resource consumption is in undesired scope, beer producers energy resource consumption monitor system gives the alarm to W (x) workshop E (y) energy resource consumption situation, require supvr to W (x) workshop, to check that E (y) energy is used, Expenditure Levels, whether investigation W (x) plant personnel there is the situation of E (y) energy dissipation;

10), by the value zero clearing of cumulative number Times (n+3), i.e. Times (n+3)=0, simultaneously according to step 1) and step 2) method again find best kernel function;

Wherein, PK: polynomial kernel function;

LK: linear kernel function;

GK: gaussian kernel function;

SK:sigmoid kernel function;

M (n): in n month, n is positive integer;

M (n+m): in (n+m) month, n, m are positive integer;

W (x): x workshop;

E (y): the y energy;

PV: forecast consumption value;

RV: actual consumption value;

M (n) _ W (x) _ E (y) _ PV: E (y) the energy forecast consumption figures in n month W (x) workshop;

M (n) _ W (x) _ E (y) _ RV: E (y) energy actual consumption value in n month W (x) workshop;

M (n) _ W (x) _ E (y) _ DV: the absolute value of difference between E (y) the energy forecast consumption figures in n month W (x) workshop and actual consumption value, M (n) _ W (x) _ E (y) _ DV=|M (n) _ W (x) _ E (y) _ PV-M (n) _ W (x) _ E (y) _ RV|;

Times (n+m): the difference between E (y) energy forecast consumption figures and the actual consumption value in same month W (x) workshop be greater than actual consumption value 5% before cumulative number in (n+m) individual month.

Images