Document Details
|
Document Type |
: |
Thesis |
|
Document Title |
: |
FORECASTING THE FUTURE DEMAND FOR WATER IN JEDDAH CITY USING DIFFERENT FORECASTING TECHNIQUES التنبؤ بالطلب المستقبلي على المياه في مدينة جدة باستخدام تقنيات التنبؤ المختلفة |
|
Subject |
: |
Faculty of Engineering |
|
Document Language |
: |
Arabic |
|
Abstract |
: |
Water demand in Jeddah city is too high over the years, and that is confirmed by looking at the statistics from Ministry of Environment, Water and Agriculture (MEWA); it is clear that the rate of water consumption is increasing. Forecasting for water consumption is important for water authorities for decision making to meet the growing demand. The objective of this study is to identify factors affecting water consumption, and to forecast water quantities needed in Jeddah city for next years. Statistical forecasting techniques were used on monthly water consumption data for Jeddah City from January 2009 to October 2018 with a total of 118 months to reach the estimated quantities needed for Jeddah by the end of 2022.
Factors that influence water consumption were selected based on literature review that was carried out as follows: The population of Jeddah city and the weather factors: Monthly temperature (large, small and medium), monthly humidity (highest percentage, lowest ratio, average); economic factors are: Gross Domestic Product (GDP), monthly inflation rate, per capita GDP. Pearson Correlation Coefficient was used to determine the correlation between water consumption in Jeddah city and each of those factors separately. Correlation analysis showed that Jeddah city water consumption was significantly correlated with five factors. These factors are: Jeddah population, Saudi Arabia GDP, Saudi Inflation rate, GDP per capita and the monthly average minimum temperature of Jeddah city. The correlation coefficient for these five factors was as follows: 0.806, 0.726, -0.683, 0.452 and 0.396, respectively.
Historical monthly water data of Jeddah city for 118 months was used as a time-series to forecast water demand for Jeddah. Seventeen different forecasting techniques in SPSS were used to forecast Jeddah water demand. Holt and winters’ method gave the best and most accurate model among all the other methods used in this part of the study. The forecasted data showed that water demand in Jeddah will jump to 44.6 million m3 in December 2022, from its current level of 38.7 million m3 in October 2018; and increase representing about 13.5% by the end of the forecasted period. The highest demand was expected to reach to 45.49 million m3 in July 2022.
Jeddah water consumption was forecasted by including the five significant factors that were identified. Multiple regression, ARIMA and Neural Network were used in this part of the study. Multiple regression was used to forecast Jeddah water demand with 3,4 and 5 factors. Based on accuracy measures of these models were compared, the best model among these used was model with 5 factors. The R2 of these models were 76.88, 77.69 and 81.24 respectively
ARIMA forecasting technique was used to forecast Jeddah water demand. Eleven different models were developed by changing different model parameters. The best ARIMA model – based on accuracy measures- was ARIMA (0,1,12) and it was identified based on a recommendation by SPSS software package used in forecasting. This mode gave the best accuracy measures among all the other eleven models.
Neural Network technique was also used to forecast Jeddah water demand. This technique was used with 3,4 and 5 factors, and with 1 and 2 hidden layers with 4,5 and 6 units in a one hidden layer model, and with 2 and 3 units in two hidden layer model. The best mode among all these developed models was with 3 factors, 1 hidden layer and 5 units.
The forecasting was in two stages: the first: time-series forecasting only without considering the factors. The second was using factors affecting consumption. In the first stage, trend analysis techniques (linear, exponential and quadratic) were used to obtain the best and most accurate model for water consumption in Jeddah. The Holt-Winters method provides the best and most accurate model among all the techniques used in the study. Using this technique, we concluded that while monthly water consumption in Jeddah reached 38.7 million m3 in October 2018, it is expected to increase by 10-15% to 44.6 million m3 by December 2022.
In the second stage, methods and techniques that take into account factors related to consumption have been taken into account: multiple regression, ARIMA and MLP-Neural Network.
In the multiple regression technique, the final results showed that 85.72% of the variation in Y can be explained by the regression model we obtained. In ARIMA, the model accuracy may increase by increasing the number of factors to the model. We concluded that the best model among all models tested (in terms of accuracy measures) is the ARIMA model (0,1,12) with 5 factors.
In Multi-Layer Perceptron Neural Network (MLP-NN) we concluded that adding more layers does not necessarily improve the model's accuracy results. On the other hand, some models do not need all five factors to give the most accurate model. This means adding more factors in MLP-NN may not give higher resolution. The relative error of the best model was in neural networks (0.177 in the training set, and 0.092 in the testing set).
Accuracy Measures (MAE, MAPE, MSE, RMSE) were calculated for all techniques used in the study. The results of all models in different techniques that were used were compared based on the smallest errors in the accuracy measures. |
|
Supervisor |
: |
Prof. Dr. Seraj Y. Abed |
|
Thesis Type |
: |
Master Thesis |
|
Publishing Year |
: |
1442 AH
2021 AD |
|
Co-Supervisor |
: |
Dr. Nader Al Sayed |
|
Added Date |
: |
Saturday, June 5, 2021 |
|
Researchers
| حامد عبدالعليم وردك | Wardak, Hamed Abdul Aleem | Researcher | Master | |
|
Back To Researches Page
|