For drilling optimization, the Rate of Penetration (ROP) is crucial; optimizing the ROP will reduce high boost costs considerably. This work uses the standard severe learning system and an effective test model (USA) for the ROP prediction. The shape of the structure, rock mechanical properties, hydraulic characteristics, bit size and characteristics (bit weight and rotary speed), and mud characteristics, are the most important ROP input parameters, and these are known as ROP predictions. The prediction model was built using industrial data collected in an oil reservoir at Bohai Bay, China. The prediction model has been designed and tested for its prediction accuracy and contrasted with the widely used traditional Artificial Neural Network (ANN). The results show that of all ROP prediction models, such as ANN, ELM and USA, have the benefit of higher learning rates and improved generalization vision both in the ELM and USA models. In the area of ROP forecasts throughout modern oil and gas explorations, the simulation findings are very good for ELM and the USA models as they surpass the ANN pattern. Meantime, this work offers ROP prediction to drilling engineers, more choices according to their calculation and precision requirements.