Anang Tjahjono, Anang
Program Studi Teknik Elektro Industri, Departemen Teknik Elektro, Politeknik Elektronika Negeri Surabaya

Published : 5 Documents
Articles

Found 5 Documents
Search

FEED FORWARD NEURAL NETWORK SEBAGAI ALGORITMA ESTIMASI STATE OF CHARGE BATERAI LITHIUM POLYMER Dwi Prasetyo, Mohammad Imron; Tjahjono, Anang; Windarko, Novie Ayub
KLIK- KUMPULAN JURNAL ILMU KOMPUTER Vol 7, No 1 (2020)
Publisher : Lambung Mangkurat University

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.20527/klik.v7i1.290

Abstract

Estimasi State Of Charge (SOC) baterai merupakan parameter terpenting dalam Battery Management System (BMS), terlebih sebagai aplikasi dari mobil listrik dan smart grid. SOC tidak dapat dilakukan pengukuran secara langsung, sehingga diperlukan metode estimasi untuk mendapatkan nilai tersebut. Beberapa metode yang pernah diusulkan adalah coloumb counting dan open circuit voltage. Akan tetapi coloumb counting memiliki kelemahan dalam hal inisialisasi SOC awal dan memiliki ketergantungan terhadap sensor arus. Sedangkan metode open circuit voltage hanya dapat digunakan pada baterai dalam kondisi idel. Pada penelitian ini diusulkan metode algoritma Feed Forward Neural Network (FFNN) untuk estimasi SOC baterai lithium polymer. Algoritma ini dapat menyelesaikan sistem nonlinier seperti yang dimiliki oleh baterai lithium polymer. Arsitektur FFNN dibangun dua kali (dual neural) untuk estimasi OCV dan SOC. FFNN pertama dengan input tegangan, arus,  dan waktu charging maupun discharging untuk estimasi OCV. OCV hasil training neural pertama digunakan sebagai input FFNN kedua untuk estimasi SOC. Hasil dari estimasi ini didapatkan dengan nilai hidden neuron 11 pada neural pertama dan hidden neuron 4 pada neural kedua.Keywords: SOC, BMS, Coloumb Counting, OCV, FFNNEstimasi State Of Charge (SOC) baterai merupakan parameter terpenting dalam Battery Management System (BMS), terlebih sebagai aplikasi dari mobil listrik dan smart grid. SOC tidak dapat dilakukan pengukuran secara langsung, sehingga diperlukan metode estimasi untuk mendapatkan nilai tersebut. Beberapa metode yang pernah diusulkan adalah coloumb counting dan open circuit voltage. Akan tetapi coloumb counting memiliki kelemahan dalam hal inisialisasi SOC awal dan memiliki ketergantungan terhadap sensor arus. Sedangkan metode open circuit voltage hanya dapat digunakan pada baterai dalam kondisi idel. Pada penelitian ini diusulkan metode algoritma Feed Forward Neural Network (FFNN) untuk estimasi SOC baterai lithium polymer. Algoritma ini dapat menyelesaikan sistem nonlinier seperti yang dimiliki oleh baterai lithium polymer. Arsitektur FFNN dibangun dua kali (dual neural) untuk estimasi OCV dan SOC. FFNN pertama dengan input tegangan, arus,  dan waktu charging maupun discharging untuk estimasi OCV. OCV hasil training neural pertama digunakan sebagai input FFNN kedua untuk estimasi SOC. Hasil dari estimasi ini didapatkan dengan nilai hidden neuron 11 pada neural pertama dan hidden neuron 4 pada neural kedua.Kata kunci: SOC, BMS, Coloumb Counting, OCV, FFNN
Pendeteksian Harmonisa Arus Berbasis Feed Forward Neural Network Secara Real Time Wahjono, Endro; Anggriawan, Dimas Okky; Satriawan, Achmad Luki; Firdaus, Aji Akbar; Prasetyono, Eka; Sudiharto, Indhana; Tjahjono, Anang; Budikarso, Anang
Jurnal Rekayasa Elektrika Vol 16, No 1 (2020)
Publisher : Universitas Syiah Kuala

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (864.753 KB) | DOI: 10.17529/jre.v16i1.15093

Abstract

The development of power electronics converters has been widespread in the industrial, commercial, and home applications. The device is considered to produce harmonics in non-linear loads. Harmonics cause a decrease in power quality in the electric power system. To prevent a decrease in power quality caused by harmonics in the power system, the detection of harmonics has an important role. Therefore, this paper proposed feed forward neural network (FFNN) for harmonic detection. The design of harmonic detection device is designed with a feed forward neural network method that it has two stages of information processing, namely the training stage and the testing stage. FFNN has input harmonics and THDi as output. To detect harmonics, frst training is conducted to recognize waveform patterns and calculate the fast fourier transform (FFT) process offline. Prototype using the AMC1100DUB current sensor, microcontroller and display. To validate the proposed algorithm, compared by standard measurement tool and FFT. The results show the proposed algorithm has good performance with the average percentage error compared by standard measurement tool and FFT of 5.33 %.
Load Identification Using Harmonic Based on Probabilistic Neural Network Anggriawan, Dimas Okky; Amsyar, Aidin; Prasetyono, Eka; Wahjono, Endro; Sudiharto, Indhana; Tjahjono, Anang
EMITTER International Journal of Engineering Technology Vol 7, No 1 (2019)
Publisher : Politeknik Elektronika Negeri Surabaya (PENS)

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (529.473 KB) | DOI: 10.24003/emitter.v7i1.330

Abstract

Due to increase power quality which are caused by harmonic distortion it could be affected malfunction electrical equipment. Therefore, identification of harmonic loads become important attention  in the power system. According to those problems, this paper proposes a Load Identification using harmonic based on probabilistic neural network (PNN). Harmonic is obtained by experiment using prototype, which it consists of microcontroller and current sensor. Fast Fourier Transform (FFT) method to analyze of current waveform on loads become harmonic load data. PNN is used to identify the type of load. To load identification, PNN is trained to get the new weight. Testing is conducted To evaluate of the accuracy of the PNN from combination of four loads. The results demonstrate that this method has high accuracy to determine type of loads based on harmonic load
Load Identification Using Harmonic Based on Probabilistic Neural Network Anggriawan, Dimas Okky; Amsyar, Aidin; Prasetyono, Eka; Wahjono, Endro; Sudiharto, Indhana; Tjahjono, Anang
EMITTER International Journal of Engineering Technology Vol 7 No 1 (2019)
Publisher : Politeknik Elektronika Negeri Surabaya (PENS)

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (529.473 KB) | DOI: 10.24003/emitter.v7i1.330

Abstract

Due to increase power quality which are caused by harmonic distortion it could be affected malfunction electrical equipment. Therefore, identification of harmonic loads become important attention  in the power system. According to those problems, this paper proposes a Load Identification using harmonic based on probabilistic neural network (PNN). Harmonic is obtained by experiment using prototype, which it consists of microcontroller and current sensor. Fast Fourier Transform (FFT) method to analyze of current waveform on loads become harmonic load data. PNN is used to identify the type of load. To load identification, PNN is trained to get the new weight. Testing is conducted To evaluate of the accuracy of the PNN from combination of four loads. The results demonstrate that this method has high accuracy to determine type of loads based on harmonic load
Maximum Power Point Tracking Menggunakan Algoritma Artificial Neural Network Berbasis Arus Hubung Singkat Panel Surya Habibi, Muhammad Nizar; Jati, Mas Sulung Wisnu; Windarko, Novie Ayub; Tjahjono, Anang
Jurnal Rekayasa Elektrika Vol 16, No 2 (2020)
Publisher : Universitas Syiah Kuala

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.17529/jre.v16i2.14860

Abstract

The conversion of solar energy into electrical can be utilized by using the solar panel, but the energy conversion ratio is still low. Maximum Power Point Tracking (MPPT) is a method used to increase energy production in the process of converting electrical to the solar panel. Artificial Neural Network (ANN) is one of the soft-computing methods that can be applied as MPPT with the advantage of having a learning process, very stable, fast, doesn’t require complicated mathematical modeling, and has good performance. ANN is proposed with input from the short circuit current of the solar panel and is used as a reference for the ANN to reach the maximum power. The process of detecting a short circuit current is indicated by a momentary decrease of the power by the solar panel. The results show the proposed algorithm can reach the maximum power operating point of the solar panel despite the change of radiation. When at maximum power operating point, ANN can hold the value, so the resulting value doesn’t change and doesn’t generate ripple. At radiation of 1000 W/m2 and using 100 WP, ANN can produce a maximum power of 99.97 Watts with a time of 0.063 seconds.