CC BY
24
ing working organs-35 sm. Results of the studies are presented in the table.
Thereby, results of the studies have shown that basic researches are taken correctly. Ensuring the qualitative loosening of ground under minimum depth of the furrow with compacted wall and the least expenses of energy for loosening of ground the form of working surfaces of loosening paws must be protuberant.
As the results of theoretical researches have shown when using protuberant working surfaces of loosening paws the layer is compressed but sprawls in longitude direction. This brings about improvement of the cutting of ground and decreasing tractive resistance of the unit.
The results of carried out studies have shown that increasing the velocity of the moving the unit brings to increasing the
tractive resistance of loosener and perfect the quality of cutting ground, that is fractions in size more than 100 mms decrease, but size less 50 mms increase. The main reason of this is that influence of power increases at increasing the velocity of the moving unit to inertias of ground on working organs of the machines and blow from the working organs on the ground. At increasing the velocity of the moving unit from 5,0 km/hour up to 7,0 km/hour of the tractive resistance of loosening paws with protuberant surface and working organ changed not so, much in contrast to flat and convex working units.
So on, the base of results of theoretical and experimental studies we can confirm with confidence that for qualitative processing of ground with minimum expenses of energy working organs with protuberant working surface are necessary.
References:
1. Sineokov G. N., Panov I. M. Theory and calculation ground cultivating machine. - Moscow: Machine building, - 1977.
2. Tuhtaquziev A., Hushvaktov B., Mamadaliev M. Way spare energy when processing of ground//AGRO ILM - ZH. Agriculture of Uzbekistan - T., - 2007. - No 3.
3. Khudoyorov A., Mamadaliev M. Results of the comparative test of working organ loosener//ZH. Agriculture of Uzbekistan - T., -2008. - No 8.
DOI: http://dx.doi.org/10.20534/ESR-16-11.12-140-142
Khudayarov Muzaffar Burhanovich, candidate of technical Sciences associate Professor, Tashkent State Technical University E-mail: muzaffarhudayarov@rambler.ru Khabibulina Albina Talgatovna, Senior teacher, TGTU E-mail: habibylina-albina@mail.ru Karimkulov Hojiakbar Kholmuradovich, Undergraduate, TSTU E-mail: 007akbar@mail.ru
Energy consumption forecasting methodology of a set of objects
Abstract: This article presents the methodology and questions the use of different types of models for forecasting of energy consumption a set of objects. To improve the forecasting results is carried out procedure to select the best model for each object together.
Keywords: technocenosis, forecasting, data matrix, forecasting vector, forecasting models.
Introduction
One of the most important tasks of effective management of a set of objects (technocenosis [1]) is the task of energy consumption forecasting of individual objects and the whole set.
In this case, the forecasting is a procedure which consists in determining the probable values of energy consumption in the future, for planning purposes. Forecasting is designed to help decision making and planning in the present.
According to [2] forecasting can be done by different methods, which include the G-methods (based on Gaussian mathematical statistics), the Z-method (based on Zipf mathematical statistics) [3]. Also for forecasting are widely used methods based on the use of artificial neural networks [4], neuro-fuzzy models [5] and others.
The main aim of this work is to develop forecasting method and evaluate the different types of models to predict the energy consumption of a set of buildings.
Energy consumption forecasting method
The first step to perform forecasting is collection of monthly data (5 years or more) on the energy consumption of all the objects of a set, allowing you to prepare a database for further use in the calculations.
As it is known, the data which used for further analysis is not always are quite correct. There are zero data, absolutely equal data, and also so-called "outliers" resulting in incorrect database. In some cases, when there is a small volume of data need to simply increase existing database on a few years "back".
Consequently, on the second step, it requires pre-verification database that includes the following procedures: 1) elimination of zero data 2) elimination error data (outliers); 3) elimination of absolutely equal data; 4) recovery of lost data.
It should be noted that the verification procedure is not mandatory, but it must always be applied in that case there is the slightest doubt in correctness of the initial data.
Energy consumption forecasting methodology of a set of objects
Further, in the third step, performed statistical processing ofdata Finally, the fourth step is a forecasting. In Table 1 shows the
on energy consumption ofbuildings. The result of statistical process- structure of the database for forecasting task, where the monthly or
ing is to determine the possibility of further use of the data for energy annual data can be used as a database. management tasks at the system level [6], in particular forecasting.
Table 1. - The structure of the database for forecasting task
Time Series The objects of a set
1 2 3 3 5 n-1 n
t-5 D15 D25 D35 D35 D55 D(n-1)5 D(n)5
t-3 D13 D23 D33 D33 D53 D(n-1)3 D(n)3
t-3 D13 D23 D33 D33 D53 D(n-1)3 D(n)3
t-2 D12 D22 D32 D32 D52 D (n-1)2 D(n)2
t-1 D11 D21 D31 D31 D51 D(n-1)1 D(n)1
t D10 D20 D30 D30 D50 D(n-1)0 D(n)0
t+1 D1 frt D2 frt D3 frt D3 frt D5 frt D(n-1) frt D(n) frt
(n-1)0, D(n)0) are
The data for the current period (D10, D20, D the "Control Vector", using which to check the accuracy of the fore, ..., D
casting models. The data in the next year (D1 frt, d2 frt,
(n-l)frt'
D(n) frt) are defined as "Forecasting Vector". All other data begin from (t-5) to (t-1), form a "Matrix Data".
The forecasting process involves two interrelated stages.
In the first stage as a forecasting base uses the "Matrix Data" to which consistently are realized all forecasting models. Statistical comparison of the forecasting results with the data of "Control Vector" allows for each object to determine the most effective model. The model selection criterion is the minimum value of the relative annual error.
Then, in the second step a "Control Vector" joins the "Matrix Data" and made final forecast, with the model which defined as the most effective for this object in the first stage.
Q Z-methods of forecasting
To forecasting used 12 different models:
• Model based on Decomposition of time series (DTS);
• Models based on principal component analysis of the recurrence (PCAR) and vector (PCAV) prediction;
• Model with division (WithDCZ) and without division into caste zone (WithoutDCZ);
• Model-based on linear regression (REG);
• Model based on artificial neural networks: cascade forward net (CFN) and feed forward net (FFN);
• Model based on fuzzy model type Mamdani (FMM) and Sugeno (FMS);
• Model-based on Fourier series (FS);
• Model based on the sum of Sines series (SS).
Figurel. The results of forecasting using Z-methods
The DTS Model is based on a division of random energy consumption process on deterministic, seasonal, and residual random components. The PCA Models based on converting a one-dimensional series into a multi-dimensional, study it by using principal
component analysis and restoration of a series by chosen components.
The models with and without division into caste areas and REG model take into account system properties.
Models based on artificial neural networks and fuzzy logic take into account the nonlinearity of the process, and forecasting based on FS and SS very efficiently for processes with cyclical.
As an example, in Figure 1 shows the results of forecasting using Z-methods that take into account system properties [7].
n The choice of objects forecasting model
The process of selecting the most effective model is performed based on the minimum value of the relative annual error in the "Control Vector". For each object, select the most appropriate forecasting model (with a minimum error value) from the considered 12 models (Figure 2). Further, according to the selected model for each object, calculates forecast values of energy consumption at the next time step.
Value of the relative annual error in the Control Vector
№ DTS PCAR PCAV WDCZ WOUTD.,. REG I CFN FFN I FMM FMS SS № mod Error
Тиббиёт бирлашмгси 1 11 234788 0 6364 61690 19.1525 60.8578 19.1525 2.1913 26.6418 22.7633 29.5746 2.7904 21.4921 2 0.6364 -
Красногорск шпфсхснсси 2 21.8773 0.7177 10.1117 33.2509 69.0113 31.9377 28.7898 6.3727 6.9765e+07 2.9003 0.7000 2.8866 11 0.7000 1—1
Нуробод шифохонаси 3 30.6154 30 5067 19 0244 19.9598 62.8534 18.1572 33.1941 36.9639 51 6628 100 4.7517 4.2834 12 4 2834
Саркв КВП 4 34.6592 19.5471 1.2029 5.5593 56.3505 6.5720 11.0420 13.7922 50.5927 0.7626 2.3395 3.0475 10 0.7626
Парчакв КВП 5 2.3492 67.3959 22.5543 9.7215 55.1274 5.2577 10.5917 31.6560 6.8941 100 0.9260 4.5230 11 0.9260
Авангард КВП 6 11.9420 42.3313 6.6235 4.46S9 55.6925 3.5424 0.4026 3.3288 433.5103 0.6799 0.3155 1.7356 11 0.3155
Янги хает КВП 7 7.2562 7.5479 19.4591 5.3936 56.1207 4.7472 8.9653 3.2748 4.7309e+05 9.4632 2.0498 2.5215 11 2.0498
Хамза КВП 8 2 5638 3.7472 8 8942 7.1627 56.9416 7.5661 1 5880 2.3953 133.6443 1.0803 2.5297 3 3915 10 1.0803
Дехконобод КВП 9 91.5300 09748 32.7624 5.6909 56.2566 4.6095 71.4509 90.7468 76 0070 100 21.1003 33.0458 2 09748
Хос КВП 10 16.5744 79350 6.6174 8.5118 57.5571 12.4982 466032 19.7131 17.0370 18.1688 8.2257 16.9590 3 6.6174 -
Comparison of the results for objects modeling on the Control Vector
№ FACT DTS PCAR PCAV WDCZ WOUTD,.. REG CFN I FFN I FMM FMS I FS SS C.MOD
Тиббиёт бирлашмаси 1 11 1296378 992004 1304628 1216405 1048090 507431 1048090 1267971 951000 1591477 1679777 1332552 1574997 1304628 -
Красногорск шифохонаси 2 482873 377234 479408 531700 322314 149637 328655 621892 513645 3.3687e+11 496878 479493 496812 479493 ' '
Нуробод шифохонаси 3 65755 45624 45696 53246 52631 24426 53796 43929 41450 99726 0 68860 62939 62939
Саркв КВП 4 1599 2556 1523 1877 1755 529 1769 1690 2161 939 1914 1944 1957 1914
Парчакв КВП 5 2190 2242 3666 2691 1975 915 2009 2429 2884 2341 0 2170 2296 2170
Авангард КВП 6 1722 1517 2451 1537 1646 763 1661 1716 1750 5744 1734 1717 1752 1717
Янги хаёт КВП 7 1938 1798 1786 1561 1834 851 1846 1765 2002 9170452 2122 1899 1890 1899
Хаша КВП 8 1890 1939 1820 2059 1755 814 1747 1860 1845 636 1911 1938 1826 1911
Дехконобод КВП 9 2484 4758 2509 3298 2343 1087 2370 4259 4739 4373 0 1960 1664 2509
Хос КВП 10 6933 5784 7484 6475 6343 2943 6067 3703 5567 5752 5674 6363 5758 6475 T
SUMMARY DATA ON MODELS 150 10406065 5409727 10073273 9935003 5420337 3929539 8127729 11534032 9583255 5.13956+11 9309309 10421957 1070558 10382459
ERROR in % relatively FACT data 0 0 19.1851 3.1958 4.4957 19.0526 62.2355 21.5947 10.8388 7.9075 4.9388e+06 10.5401 0.1520 2.9093 0.2272
Forecasting results of all objects ort next year
Тиббиёт вирлашмаси l| Красногорск шифсхснаси Нуробод шифохонаси Саркв КВП ["парчакв КВП|Ааамгард КВГ||Янги яаёт КВП| Хамаа КВП |д<акомрбрд КВп] Хос КВП Навбааор КВп]у.гуг&ек
Forecasting results for set of objects on next year: SOV_Pr = 10405087.00;
Monthly data
Figure 2. Select the most appropriate forecasting model for objects
Conclusion
The results of the use this methodology for a set of social facilities showed that this approach can improve the quality of forecasting of energy consumption for the entire set of objects
and individual objects together, in relation to the case of the use of one specific approach. Error forecasting for a set of objects was 0.22%, and the maximum value for the objects of error does not exceed 5-6%.
References:
1. Gnatyuk V. I. Law of optimum construction technocenosis: Monography - 2 nd ed., revised. and ext. - Kaliningrad, - 2014. - 475 p.
2. Chuchueva I. A. Model prediction of time series based on a sample of maximum similarity. - Moscow, - 2012 g. - URL: http: www. mbureau.ru/articles
3. Gnatyuk V. I. GZ-rank analysis of parametric distribution electricity needs. - Kaliningrad, - 2010. - URL: http: www.gnatukvi.ru
4. Krug P. G. Neural networks and neurocomputers. Chapter 1.5. Forecasting. - Moscow - 2002 g. - URL: http://lib.tuit.uz
5. Shtovba S. D. Design of fuzzy systems MATLAB tools. - M. - 2007. - 288 p.
6. Salikhov T. P., Khudayarov M. B. Methodology for building energy consumption management of social facilities. Magazine " Energu security and energy efficiency", - Moscow, - No 3, - 2015, - P. 16-21.
7. Salikhov T. P., Khudayarov M. B. The complex of programs for energy management of buildings. Certificate DGU 02932 from 12.26.2014. Agency on Intellectual Property of the Republic of Uzbekistan, Tashkent.
