The paper presents a model-based predictive control algorithm that uses a limited number of control sequences for on-line simulation of future behaviour of the process. Each control sequence used in simulation generates a predicted sequence of the output signal. The predicted output sequences are analysed and evaluated and then, using a set of rules, the `optimal' control signal is computed. To simulate the future behaviour of the process it is used a process model and also the previous sequences of the input and output signals from the process. The algorithm permits directly use of the nonlinear model of the process. The algorithm is used for simulation of the temperature control in a house and compared with the usual algorithms of type PI.