Master of DonNTU Ekaterina Klimova

Ekaterina Klimova

Faculty: Computer Science and Technology

Speciality:Computer Ecological and Economic Monitoring


Theme of master's work:

Mathematical Model of Evolution of Meteoparameters: Developing and Forecast

Scientific adviser: Valeriy Belovodskiy

 

Abstract

Qualification Masters work

"Mathematical Model of Evolution of Meteoparameters: Developing and Forecast"



Actuality and motivation

"Tomorrow we expect blizzard, snow, and anomalous cold. At times stormy wind will reach 15 meters per second." Predictions like this give vital information concerning the weather changing.

In severe weather conditions, short-term forecasts can save lives and reduce material losses. It is important that weather forecasts were as accurate as possible. Each year the world's natural disasters kill nearly 250,000 lives and total losses from them are up to 50-100 billion U.S. dollars. But reliable weather forecast may prevent from 30 to 40% of the losses and completely avoid human casualties [1].

Weather forecasts are important for agriculture too. Analysis of the available data shows that at least 50% of the variability of productivity is due to weather. Harvest losses can be reduced substantially with the use of timely and accurate weather forecasts [2].

Weather forecasting is also very important for many areas of human activity, such as: aviation, shipping, fishing, etc. Increasing the accuracy of weather forecast during last decades is a direct consequence of technological progress, new approaches in forecasting, contemporary means of observations and data collection. Great improvement was made largely due to the appearance of high speed computers that are needed for the huge amounts of computation.

Aims and Tasks

The aim of the master’s work is to study the methods used for weather forecasting, to develop a software project, including the short-term weather forecasts method.

The object of the study is time series of meteorological parameters.

The subject of the study is forecasting models, the quality and range of their results.

Tasks:

  • Analysis of existing methods is used to predict the meteorological parameters using the time series;
  • Development of a mathematical model that allows to predict the weather parameters;
  • Assessment of sufficiency and adequacy of the developed model.

Alleged scientific innovation

Scientific innovation of the research lies in: use of differential equations, where the right sides are the neural network for forecasting the weather parameters.

A review of the research carried out in DonNTU

At the CSM Department for the past two years Masters Gritsenko A.V. and A.S. Sivyakov have developed software system Fcomplex, designed for short-term weather parameters forecasting based on time series using finite mathematical models. Fcomplex is a software system for compiling short-term forecasts of weather parameters values "Temperature", "Pressure", "Humidity" and "Wind speed" with 1, 3, 6 and 9 hour’s advance [3, 4].

Approbation

The results of the work were presented at the II Ukrainian Scientific Conference of Students and Young Scientists "Information control systems and computer monitors" on 11-13 April 2011 and published.

Current and planned results

In this paper, the calculation of prediction will be performed by implementation of the following steps:

  1. obtaining time series from a database (DB) Vantage Proweather station 2;
  2. data analysis. Leveling of the time series is carried out in order to reduce the error of the input data. It’s supposed to achieve that using a moving average, since the experiments showed that this method had the best result.
  3. definition of model dimension;

    4. There are different methods to determine the dimensions of the model. These include: methods of false neighbors, the main components, Grassberger-Procaccia, well-adjusted basis. Let us consider the method of Grassberger- Procaccia. This method is one of the most common ways to determine correlation dimensionality. It shows the number of dots couples necessary to determine the state of the dynamic system [1, 8].

  4. forecasting;

    5. Forecast is supposed to be carried out on the basis of differential equations. Right-hand sides of these equations are the neural network. It is supposed to use non-linear neural networks to forecast. Nonlinear network is a single hidden layer (from 1 to 10 neurons) with hyperbolic tangent activation function and output layer containing a neuron with linear activation function. Figure 1 shows the structure of the nonlinear network [10, 11].

    Neural network

    Figure 1 –Artificial neural network architecture (Animation: volume – 126.0KB; size – 500x277; number of frames – 7; an infinite number of cycles of repetition, delay between shots – 0,5 s; delay between the last and first frame – 0.5 s)

  5. Checking the effectiveness of the model. The best way to test the efficiency is to compare the predictive values obtained to the true values of meteorological parameters.

Conclusion

The result of master’s work is the meteorological forecasting model based on the use of time series. Data and calculations of research may be useful for meteorologists and be used for further research and improvements in this area.

The work is not completed currently and is planned to be finished by December 1st, 2011. The paper and relevant materials can be obtained either form the author or her supervisor after that date.

References

  1. Гриценко А.В. Реконструкция уравнений и прогнозирование метеопараметров по их временным рядам. [Текст] – Донецк, ДонНТУ, 2010. – 149 с.
  2. Importance of weather forecasting. [Electronic resourse] / Интернет-ресурс. – Режим доступа:http://vasat.icrisat.org/?q=node/300
  3. Климова Е.А., Беловодский В.Н. Дифференциальная математическая модель эволюции метеопараметров. [Текст] – Компьютерный мониторинг и информационные технологии 2011 / Материалы I всеукраинской научно-технической конференции, аспирантов и молодых ученых. – Донецк, ДонНТУ – 2011.
  4. Сивяков А. С. Построение прогностического комплекса и внедрения его в электронную сеть университета. [Текст] – Донецк, ДонНТУ, 2011. – 125 с.
  5. Исследователь. [Electronic resourse] / Интернет-ресурс. – Режим доступа: http://ligis.ru/effects/stat/modules/sttimser.html – Анализ временных рядов.
  6. Чечурин А.В. Исследование алгоритмов оценки размернсоти реконструкции аттракторов [Текст] / Чечурин А.В. // Задачи системного анализа, управления и обработки информации. Межвузовский сборник научных трудов. Вып. 1.— М.: МГУП, 2006. - С.15-23
  7. Гудков Г.В. Диагностические возможности определения детерменированного хаоса в структуре вариабельности ритма сердца плода [Electronic resourse] / Интернет-ресурс. – Режим доступа:http://vestnik.kmldo.ru/pdf/08/01/02.pdf
  8. Антипов О.И., Неганов В.А. Фрактальный анализ нелинейных систем и построение на его основе прогнозирующих нейронных сетей [Текст] / О. И. Антипов, В. А. Неганов // Физика волновых процессов и радиотехнические системы. - 2010. - Т. 13, N 3. - С. 54-63
  9. Козлов Д.А. Методы нелинейной динамики в моделировании макро-экономических процессов [Electronic resourse] / Интернет-ресурс. – Режим доступа: http://www.ecfor.ru/pdf.php?id=books/kor001/09
  10. Безручко, Б.П. Математическое моделирование и хаотические временные ряды [Текст] / Б.П. Безручко, Д.А. Смирнов // Саратов: ГосУНЦ «Колледж», 2005. – С. 320.
  11. Медведев В.С. Нейронные сети. MaTLab 6/ Медведев В.С., Потемкин В.Г. [Под общ. ред. к.т.н. В.Г. Потемкина]. – М.: ДИАЛОГ-МИФИ, 2002. 496 с.