Wireless monitoring and arima stream analytics system for freshwater lobster farm

• Main Author: Nur Syahirah, Mohd Sabli
• Format: Thesis
• Language: English
• Published: 2021
• Subjects: TA Engineering (General). Civil engineering (General)
• Online Access: http://umpir.ump.edu.my/id/eprint/35727/1/Wireless%20monitoring%20and%20arima%20stream%20analytics%20system%20for%20freshwater%20lobster%20farm.ir.pdf; http://umpir.ump.edu.my/id/eprint/35727/
• Institution: Universiti Malaysia Pahang

Majority of the population are directly or indirectly dependent on aquaculture. Recent development in technology has a great impact on aquaculture. Among the crustacean breeds in Malaysia, Cherax Quadricarinatus species or also known as freshwater lobster has become favourable for farmers to breed them. Water quality monitoring has become a problem to farmers as predictions on water quality were observed conventionally through experience. In this research integration of IoT with forecasting for the freshwater lobsters were developed to do predictions based on real-time data. This IoT system consist of variety of sensors such as Electrical Conductivity (EC), Total dissolved Solid (TDS), Dissolve Oxygen (DO), potential of Hydrogen (pH), temperature and humidity were integrated to Arduino for sensing and transmitting data as End Node Unit. To ensure the reliability of collected data, the sensors have been calibrated with reference to manufacturing datasheet which contribute to the total of 25,920 data collected from August 2020 until January 2021. Those data were transmitted wirelessly from End Node Unit (ENU) and received by gateway and this bundle of data were parallelly uploaded to Cayenne Cloud via MQ Telemetry Transport (MQTT) protocol and saved in database in server through Wi-Fi. The real-time data of ENU in Structured Query Language (SQL) was displayed on the website purposely for remote monitoring. The real-time data query from ENU is streamed through Structured Query Language (SQL) right into R Studio and Autoregressive Integrated Moving Average (ARIMA) predictions were done on the query table. 70% of this stream real-time data query were taken as training dataset meanwhile another 30% were taken as testing dataset. Auto.arima functions are applied in the streaming dataset from SQL as it automatically chooses ARIMA models based on the pattern of the dataset. ARIMA models in this thesis were set to predict 24 hours while updating the real-time and prediction graph were set to one hour which monitored through the developed website. Moreover, the changes of parameter level in lobster’s tank can be notified through SMS in order to help the farmers to do remote monitoring. For DO, ARIMA, Neural Network Autoregressive (NNetAR) and Naïve Bayes accuracy on average are almost similar, with accuracy obtained in the range of 95% to 99%. For pH, ARIMA prediction are in the range of 95 % to 100 % while Naïve Bayes prediction range 89 % to 95 % and NNetAR prediction range are between 85 % to 95 % while for EC, NNetAR and Naïve Bayes indicate that prediction error of these two models are inaccurate by range 10% to 15% compared to error by ARIMA which is below 5%. In conclusion, ARIMA analytics does provide accurate predictions for monitoring water quality in freshwater lobster farms. The efficiency of this system has been proven with a 92.8% mortality rate.