WIND SPEED FORECASTING IN THE PRESENCE OF SOME METEOROLOGICAL AGENCIES USING THE VECTOR AUTOREGRESSIVE MODEL AND MULTINOMIAL LOGISTIC REGRESSION

Introduction

1.1 Background to the Study

With the advent of science and technology, the demand for electrical energy becomes inevitable. Nigeria is a country endowed with abundant energy resources like coal, solar, water (Dam), wind and so on which can be used as a form of electricity generation. Despite the abundance of these energy resources, there is inconsistent supply of electricity, which may be due to underutilization of the potentials.

Wind energy is the fastest growing renewable source of energy. With respect to this, the need for wind speed modelling and forecasting becomes paramount. Energy generation by wind is of great advantage because wind turbines do not produce any form of pollution when sited strategically. Moreover, it blends with the natural landscape. The utilization of wind energy will ensure the growth of socio-economic development and improvement in the quality of life of the citizens. The demand for more sustainable energy sources is on the increase in order to address the growing needs of humans. It is also in line with taking care of the environment and the minimal use of natural resources, which means there is an urgent need for developing renewable energy.

Wind energy is now becoming the current trend in renewable energy as it addresses rising energy demand while being nature friendly at the same time. in the long run, electricity generated from the wind turbines cost less than the conventional power plants since it does not consume fossil fuel. Researches on the potentials of wind energy in some major cities in Nigeria show high wind speed in Lagos, Maiduguri, Enugu, Jos, Kano, Funtua and Sokoto (Idris et al., 2012).

The gathering of wind data is important for the wind farm beginning from its feasibility to its actual operation. Prior to the construction of a wind farm, at least one year of meteorological study is necessary and a detailed verification of the specific on-site wind conditions are necessary. Meteorological values, more specifically wind speed and wind direction are necessary for the calculation of the wind farm’s yearly electrical generation profile. The harnessing of Kinetic energy through the wind has been used for centuries, be it in form of powering sail boats, wind mills, or furnaces (Aliyu and Mohammed 2014). But it was not until 1979 that the modern wind power industry began in earnest with the production of wind turbines. The use of wind energy as a form of renewable energy gained momentum in the 80s and 90s and there are now thousands of wind turbines operating all over the world (Minh et al., 2011). The modern and most commonly used wind turbine has a horizontal axis with two or more aerodynamic blades mounted on the shaft. These blades can travel at over several times the wind speed, generating electricity which is captured by a medium voltage power collection system and fed through the power transmission network (Garba and Al-Amin, 2014).

Wind energy is unarguably the most economic renewable energy apart from hydropower, its usage, versatility and ability to use it as a decentralized energy form make its applications possible in rural areas where it is technically and economically feasible in the country. Winds are caused by the uneven heating of the atmosphere by the sun, irregularities of the earth’s surface, and rotation of the earth. Wind flow patterns are modified by the earth’s terrain, water bodies and vegetative cover (Reddy et al., 2015). The major challenge of using Wind as a source of energy is that winds are intermitted, and it is not available always when electricity is needed.

Wind speed forecasting is essential for a secured and reliable power system for a particular site. This research forecasts wind speed in Samaru, Zaria using the Vector Auto Regressive (VAR) model. The explanatory variables used are rainfall, relative humidity, air temperature, earth temperature, sunshine and pressure. In addition, the wind directions were also studied in relation to wind speed using the Multinomial Logistic Regression model.

---

1.2 Statement of the Problem

The growing demands of energy supply by mankind has triggered the potentials of using wind energy due to its constant availability and nature-friendly environment. The importance of wind energy cannot be over emphasized. Wind energy can be seen as the energy of the future for Nigerians if properly harnessed and utilized. For example, the 10MW Katsina wind farm project which is owned by the Federal Ministry of Power is a pioneer project in Nigeria aiming to generate 10MW of power via wind turbine with the Federal Government’s desire to boost electricity generation and have constant power supply. Wind speed modelling and forecasting are necessary for a reliable and secured power system. And it is therefore of paramount importance to know the impact of meteorological variables on wind speed. Cherie et al., (2014) used the Vector Autoregressive Model in modelling wind speed in the presence of some meteorological variables which are humidity, temperature and pressure. However, we applied the Vector Autoregressive Model and Multinomial Logistic Regression in studying wind speed and its direction respectively with some selected meteorological variables which are rainfall, air temperature, earth temperature, humidity, evaporation, sunshine and pressure.

---

1.3 Aim and Objectives of the Study

The aim of this study is to forecast wind speed using the Vector Autoregressive and the Multinomial Logistic Regression Model. This aim will be achieved through the following objectives;

1. Determination of the co-integrating variables and order of co-integrating VAR model.
2. Fitting of the Cointegrated Vector Autoregressive Model.
3. Model adequacy check and forecasting.
4. Fitting of the Multinomial Logistic Regression Model.
5. To make recommendation or possible implication of the result.

---

1.4 Significance of the Study

This study will serve as a contribution to the existing literatures on modelling of wind speed in the scientific world. In addition, the result of this study will be of paramount importance to the Kaduna State Government, the Federal Government and also to the Nigerian Meteorological Agency as it creates awareness on the importance of wind speed, can provide valuable information on the expected daily and seasonal load for a project. The model which will be adopted can be used as a tool by the government in making forecasts on wind speed in order to facilitate proactive planning.

---

1.5 Scope of the Study

The data used for this study is a 30-year daily data from 1984 to 2015 obtained from the Department of Soil Science, Ahmadu Bello University, Zaria. The data consist of the daily amount of rainfall, relative humidity, pressure, sunshine, air temperature and earth temperature in Zaria, Kaduna state.

---

1.6 Definition of Terms

• Wind: This refers to the movement of air.
• Air temperature: This is the degree of hotness or coldness of the air.
• Earth temperature: This is the degree of hotness or coldness of the earth’s surface.
• Humidity: This is the presence of water vapor in the atmosphere.
• Pressure: This refers to the force exerted onto a surface by the weight of the air.
• Evaporation: This is the process in which a substance in its liquid state changes into gaseous state.
• Rainfall: This refers to the amount of precipitation falling in a given area at a particular period of time.
• Sunshine: This refers to the intensity of the light and heat coming from the sun.
• Time series: This is a collection of data on a particular phenomenon or variable over a period of time. The time period can be hourly, daily, weekly or yearly.
• Stationary Series: These are series with constant mean and variance. These series have a mean that is defined in which it can fluctuate around with a constant variance.
• Nonstationary Series: These are series with no constant mean and variance over a period of time. These series have different mean at different time period and with variance increasing with respect to the sample size.
• Cointegration: Two or more time series variables are said to be cointegrated if the linear combination of the series results into a stationary series. Cointegration is therefore a linear combination of nonstationary time series variables whose linear combination gives a stationary series.
• Autocorrelation: This is the correlation of a variable with itself over a successive time interval.
• Heteroscedasticity: This is a situation in which the variance of the errors is not constant.
• Cointegration Rank: This is referred to as the number of independent linearly cointegrating vectors (Johansen, 2000).