Spatial-Wind-Power-Forecasting

Spatial and temporal wind power prediction

Files:

1. Final Project : Analysis and Plots of Zones 1-10
2. Final Project : Analysis and Plots of Zones 11-20
3. Final Project : Analysis and Plots of Zones 21-30
4. Final Project : Analysis and Plots of Zones 31-40
5. Final Project : Analysis and Plots of Zones 41-50
6. Final Project : Analysis and Plots of Zones 51-64
7. Final-Project - 2 hour forecast : Forecasting of next 2 hours for Zones 1-10
8. Final-Project - 3 hour forecast : Forecasting of next 3 hours for Zones 1-10
9. Final-Project - 4 hour forecast : Forecasting of next 4 hours for Zones 1-10

Steps:

1. Creating 64 geographic zones on US maps, each of about 100km x 180km.

2. Using latitude and longitude to divide wind turbine locations into the 64 zones.

3. For each site the following calculations are performed to select the right data: a) Automatically reading files by zones. b) Finding Capacity Factor of each wind power plant. c) Selecting one power plant for each site to represent wind power for that location.

4. Autoregressive Integrated Moving Average (ARIMA) model is used to forecast wind power using the following steps: a) 15-minute data is converted to hourly data. b) The analysis is performed for the months of April and July. April shows high standard deviation in wind speeds. July shows high wind speed fluctuation. c) The model takes wind power data from past 72 hours to train the model and determine the wind speed of the next hour. d) In cases of 2-hour, 3-hour and 4-hour forecasts, the past 72 hours' data is used to determine the wind power in next 2, 3 and 4 hours.

5. A power curve is developed.

6. Residuals are generated by comparing actual wind speed data with modeled wind speed data.

7. Actual and ARIMA Model wind power is plotted for each site.

8. Mean Absolute Percentage Error (MAPE) is calculated for each site.

9. The MAPE is grouped by zones to determine the overall reduction in forecasting error in wind power.