The Swiss limit for NOx exhaust gas emissions from decentralised electrical power and heat cogeneration units using stationary I.C. engines operating on natural gas, despite a recent increase from 80 to 250 mg/mN3 (5% O2), remains the most severe in force in the world. The only technology currently able to meet this limit - the use of a stoichiometric mixture and a three-way catalyst - results in a relatively low global engine efficiency and suffers from a limited reliability of the exhaust gas treatment system. In this context, the Laboratory for Energy Systems of the EPFL has undertaken to develop new technologies (mixture, ignition,...) on the basis of a converted diesel engine for natural gas operation and to evaluate their reliability as well as their potential to reduce exhaust gas emissions, particularly NOx. The objective of this project is to perform an experimental study of the influence of the operating mode, either a stoichiometric mixture (l = 1) and a 3-way catalyst for exhaust gas treatment or a lean mixture (l >> 1) without catalyst, on the performance and emissions of a Liebherr G 926 engine, in order to establish a basis of comparison for future evaluation of new technologies. The realisation of this project required the construction of a new test bed especially developed for the study of gas engines. In the case of the stoichiometric operating mode, the results show that at an MBT spark timing of 20 °CABTDC, a mechanical power output of 106.3 ± 0.5 kW at 1502 ± 5 rpm and a global efficiency of 35.9 +1,7/-0.6% is achieved. A series of tests with different sparkplugs showed only a weak influence on the engine behaviour. The use of a 3-way catalyst for exhaust gas treatment reduces the NOX and CO emissions comfortably below the limits prescribed by the Swiss Federal Clean Air Act. In the case of the naturally aspirated engine, lean operation with an air to fuel ratio over 1.6 keeps the NOX emissions under the prescribed limit without a catalyst. However, the power output falls below 69 kW and the global efficiency below 35%. By turbocharging and intercooling the intake air and gas mixture the power output was increased from this low level to 150 kW (pme = 12 bar). As with stoichiometric operation, changing the spark plugs has only a weak influence on the engine behaviour. However, an increase of the ignition coil discharge time led to a perceptible reduction of the cycle-by-cycle variability. The results show that the best method of reducing the amount of energy dissipated by the exhaust manifold is to insulate the manifold, rather than water cool it. The insulation preserves more energy for the turbocharger and a high temperature for a possible catalytic exhaust gas after treatment. The reduction of the dead volume located between the cylinder liner, the cylinder head and the cylinder gasket led to a 40 % decrease of the THC emissions. Further, the use of a new piston, generating a higher level of turbulence during the combustion, raised the lean limit and thus reduced NOX emissions below 150 mg/mN3, 5% O2, while still keeping the global efficiency over 37%; these results are valid for a intake mixture temperatures of both 50 and 80 °C, however in the second case at a somewhat lower power output than 150 kW. Despite these improvements, and because of an extremely low CO raw emissions limit, the use of a oxidation catalyst is needed to fulfil the requirements of the Swiss Federal Clean Air Act. In order to reduce the raw CO emissions and to increase the engine global efficiency, this project will continue with a third part aiming to study the potential of an ignition system based on a prechamber.