The characteristic feature of a power engineering system is its constant variability. In order to operate a power engineering system, as well as to plan its development it is necessary to carry out forecasts. Carrying out a correct forecast is an uneasy task that requires a lot of experience, knowledge and intuition. In order to be able to control and foresee the processes that occur in a power engineering system it is necessary to carry out research in the field of analyses of power loads. Along with social and political transformations and following economical transformations observed in our country during the past eighteen years, evolution of attitudes towards the role of power engineering forecasting is being noticed. In the paper a forecasting model MRK, based on the canonical distribution of a vector of random variables, has been presented. It is a forecasting method, able to predict short-term and long-term forecasts on demands of either electric energy that influence the development of power engineering system in Poland. The MRK model, it is most suitable in the case when knowledge about the object under study is limited, which in this case is the domestic power engineering system. In this case the feature, that it generates a certain random process, is utilized. If no analytical relationships between explanatory variables, such that would be able to provide a full mathematical description of the object, are known, however if some knowledge about the object is available (e.g. on some factors influencing the considered object) is available, it would be possible to use the canonical distribution of the variables and the conditional distributions of realizations of these variables for a description of the processes under examination.