Pemanfaatan energi terbarukan merupakan suatu hal yang sangat penting dalam upaya mengurangi penggunaan  energi fosil yang semakin menipis. Sumber energi baru dan terbarukan ramah lingkungan yang dimaksud adalah sistem Pembangkit Listrik Tenaga Surya (PLTS) yang terintegrasi dengan sistem grid atau on grid pada rumah tinggal. Dengan sistem PLTS on grid pasokan listrik untuk rumah tinggal berasal dari dua sumber yaitu PLTS dan PLN.

Peramalan daya listrik PLTS on grid sangat diperlukan untuk mengetahui berapa besar daya listrik yang akan diproduksi PLTS on grid . Penggabungan tiga metode dalam penelitian ini terdiri dari metode k-Nearest Neighbor, metode decomposition dan metode feed forward neural network atau bisa disebut dengan metode hybrid k-Nearest Neighbor decomposition feed forward neural network (k-NNDcNN) digunakan dalam peramalan jangka sangat pendek. Penelitian ini bertujuan membahas tentang peramalan daya listrik yang diproduksi PLTS on grid selama lima jam ke depan. PLTS on grid pada penelitian ini berkapasitas 1000 Watt.

Hasil penelitian menunjukkan nilai rata-rata MSE metode k-Nearest Neighbour Decomposition adalah 7,726531231 Watt sedangkan metode Hybrid k-Nearest Neighbour Decomposition Feed Forward adalah 0,627315211 Watt dan nilai MAPE rata-rata metode k-Nearest Neighbour Decomposition adalah 0,025386522% sedangkan metode Hybrid k-Nearest Neighbour Decomposition Feed Forward adalah 0,004340221%. Sehingga, dapat disimpulkan bahwa metode Hybrid k-Nearest Neighbour Decomposition Feed Forward mendapatkan hasil yang lebih baik.

Kata Kunci : Peramalan, PLTS on grid, k-Nearest Neighbor, decomposition, feed forward neural network.

Utilization of renewable energy is a very important thing in an effort to reduce the use of fossil energy which is getting thinner. The new environmentally friendly and renewable energy source referred to is the Solar Power Plant which is integrated with a grid or on grid system in a residential home. Solar Power Plant system on grid electricity supply for residential homes comes from two sources, Solar Power Plant and PLN (State Electricity Company).

Forecasting Solar Power Plant on grid is needed to find out how much electric power will be produced on the grid. The combination of the three methods in this study consists of the k-Nearest Neighbor method, the decomposition method and the feed forward neural network method or can be called the hybrid k-Nearest Neighbor decomposition feed forward neural network (k-NNDcNN) method used in very short-term forecasting. This study aims to discuss the forecasting of electric power produced by PLTS on grid for the next five hours. Solar Power Plant on grid in this research has a capacity of 1000 Watt.

The results showed average value of MSE k-Nearest Neighbor method Decomposition is 7.726531231 Watt while the Hybrid method k-Nearest Neighbor Decomposition Feed Forward is 0.627315211 Watt and the average value of MAPE k-Nearest Neighbor method Decomposition is 0.025386522 % while the Hybrid k-Nearest Neighbor Decomposition Feed Forward method is 0.004340221%. So, it can be concluded that the Hybrid k-Nearest Neighbor Decomposition Feed Forward method gets better results.

Keywords: Forecasting, Solar Power Plant on grid, k-Nearest Neighbor, decomposition, feed forward neural network.