This paper presents a stochastic approach for the simulation of particle agglomeration, which is addressed as a two-step process: first, particles are transported by the flow toward each other (collision step) and, second, short-ranged particle-particle interactions lead either to the formation of an agglomerate or prevent it (adhesion step). Particle collisions are treated in the framework of Lagrangian approaches where the motions of a large number of particles are explicitly tracked. The key idea to detect collisions is to account for the whole continuous relative trajectory of particle pairs within each time step and not only the initial and final relative distances between two possible colliding partners at the beginning and at the end of the time steps. The present paper is thus the continuation of a previous work (Mohaupt M., Minier, J.-P., Taniere, A. A new approach for the detection of particle interactions for large-inertia and colloidal particles in a turbulent flow, Int. J. Multiphase Flow, 2011, 37, 746-755) and is devoted to an extension of the approach to the treatment of particle agglomeration. For that purpose, the attachment step is modeled using the DLVO theory (Derjaguin and Landau, Verwey and Overbeek) which describes particle-particle interactions as the sum of van der Waals and electrostatic forces. The attachment step is coupled with the collision step using a common energy balance approach, where particles are assumed to agglomerate only if their relative kinetic energy is high enough to overcome the maximum repulsive interaction energy between particles. Numerical results obtained with this model are shown to compare well with available experimental data on agglomeration. These promising results assert the applicability of the present modeling approach over a whole range of particle sizes (even nanoscopic) and solution conditions (both attractive and repulsive cases).