Iot Based Light Control System Using Adaptive Neuro-Fuzzy System

Abstract

In The Past, Few Works Have Been Done To Automate Home Appliances Using Iot And Machinelearning Technologies. About 20% Of A Home?�?S Use Of Energy Goes To Lighting, Therefore, Improvingthe Energy Consumption Of Buildings For A Large Saving Can Be Achieved By Reducing The Electricityusage And Improving The Way To Control Them, The Need For High-Quality And Intelligent Systemsthat Can Monitor And Control Home Devices Such As Light Control Is Required.The Most Efficient Way To Keep Down The Electricity Use And Improve The Energy Consumption Is Toimplement Control Of Electrical Lights Based On The Presence Of Occupants, In Manual Operationpeople Always Forget To Switch Off Their Light Whenever Power Is Gone While Using It, To Solve This,An Adaptive Neuro-Fuzzy Inference System (ANFIS) Which Is A Hybrid Combination Of Neuralnetwork And Fuzzy Logic, Thus Having The Advantage Of Both, As A Neural Network Which Is Good Atcomputing A Model To Be Adaptive By Learning From The Data, And, Fuzzy Logic Is Good At Reasoningand Decision Making. ANFIS Is Proposed In This Thesis To Model The Parameter Such As Lightintensity, Temperature And, Motion Detection As An Input To The Model And Light Chance As Output,Each Having Three Membership Function Such As High, Medium, And Low (3 3 3), And, MF Typeof Trimf, Trapmf, Gbellmf, Gaussmf, And Digmf With An Output Type Of Linear. The Model Is Trainedusing A Hybrid Of Feedforward And Backpropagation Optimization Algorithm, And Which Consists Ofsensors That Collect Data From The Sensor Where Fuzzy Logic Converts The Raw Data In A Linguisticvariable That Is Trained In An Adaptive Neuro-Fuzzy Inference System To Get Trained Model With 33 =27 Rules.In The Proposed System We Used A Data Set Collected Using Sensors Such As PIR, For Detectingmotion, Lm35 For Collecting Temperature And, LDR For Collecting Light Data In The Room, A Totaldata Of 1506, Which Is Collected From Sensors Embedded With Arduino Micro-Controller, Out Ofwhich 70% The Dataset Was Used For The Training Set, The Remaining 30% (15% The Data Was Usedfor The Testing Set, And 15% Was Used For Checking Set). The Designed Controller Is Based Onadaptive Neuro-Fuzzy Inference System To Control The Light In The Room. Based On Evaluation Testcases Results Were Analyzed And Achieved Training Error In RMSE = 0.07434, And Average Testingerror = 0.273674 Was Achieved, To Achieve The Goal And The Objective Of The System.MATLAB/Simulink-Based Simulation Is Used For The Experiments And The Results Show Asatisfactory Output.