Biochemical oxygen demand (BOD) predictive model based on water quality characteristics with deep learning neural network / Azhar Jaffar

• Main Author: Jaffar, Azhar
• Format: Thesis
• Language: English
• Published: 2023
• Subjects: Central nervous system; Qualities of water. Water quality
• Online Access: https://ir.uitm.edu.my/id/eprint/106817/1/106817.pdf; https://ir.uitm.edu.my/id/eprint/106817/

Water is an essential element on this earth in which humans, animals, and plants depend entirely on water sources for a living. Unfortunately, human activities have polluted water sources. Water pollution is typically caused by organic or chemical impurities. Organic matter is contaminated by microorganisms such as bacteria and viruses that are formed by human, animal, and plant waste. Pesticides such as nitrates and phosphates, industrial acids, hydrocarbons, home products, and heavy metals all contribute to chemical contamination. Water is often examined for pollution in its physical, biological, and chemical properties. In Malaysia, the Klang River experiences high water pollution levels every year and takes a long time to rectify the problem. As a solution, each river must have a smart system capable of assessing water quality and monitoring the water quality index readings. In redressing this situation, water quality must be tested for pollution. However, to test and determine the reasonable acceptance level for all water parameters takes some time. Some of the value of the parameter is not available impromptu. The most critical indicator parameter in the assessment of water quality is the Biochemical Oxygen Demand (BOD). The amount of oxygen required to eliminate waste organic matter from water is measured as BOD. When assessing water quality, the most important factor is BOD. BOD readings can be obtained through time-consuming laboratory tests that are influenced by environmental factors at the time the test is performed. Therefore, the emphasis is on implementing environmentally friendly techniques that do not necessitate the use of physical tools for measurement. The deep learning neural network has emerged over the past several years as a promising and versatile method for data prediction in many fields. A deep learning neural network is used by making predictions using samples and previous readings of the water parameters being tested. Water quality data collected over a seven-year period was utilized to validate the deep learning neural network model's accuracy. Nevertheless, a problem arises where the water quality data recorded throughout the period is not uniform due to various human and environmental factors. This suggests that the data gathered by the personnel does not adhere to a set timetable and that there may be variations in the water's quality because of heat, rainfall, and turbidity. The uneven gap found in the water quality data is restructured by using interpolation techniques to reorganize the data. Interpolation is a form of mathematics for predicting the values between known data points. This procedure aids in filling in the gaps. To compare the prediction results made by deep learning neural network techniques, three models were used, namely LSTM, bi-LSTM and GRU.