Energy value estimation of silages for substrate in biogas plants using an artificial neural network

Abstract The typical tests of biogas efficiency require a great deal of time and are quite expensive. Thus, there is a necessity to develop tools for estimating the energy value of silage more quickly. This paper describes the application of a prediction model based on artificial neural networks to estimate the methane production from various substrates in the form of silages. For this prediction, basic silage parameters were used. The learning file contained input data such as the kind of silage, pH, dry matter, organic dry matter, conductivity and fermentation time. The output data in the database sheet contained the cumulative methane production. The prediction model of methane production that was created was a Radial Basis Function (RBF) with 5 inputs, 2 neurons in a hidden layer and 1 output. However, the resulting optimal prediction model has 73% of the quality of the network with a Root Mean Square Error (RMSE) less than 3%. This is a satisfying result, which can be increased significantly by enhancing the model with the addition of a new analysis of silages. The RBF model created can help to estimate the energy value of different silages quickly and without the necessity of expensive, long-term analysis.