This dissertation introduces a new operational tool for integrating a photovoltaic (PV) system into the utility's generation mix. It is recognized at the outset, that much of the existing research concentrated on the central PV system and its operations have concluded that technical problems in PV operation will override any value or credit that can be earned by a PV system, and that penetration of a PV plant in the utility will be severely limited. These are real problems and their solutions are sought in this dissertation. Judging from the drawbacks of the static approach, it is felt that a new approach or methodology needs to be developed which would give a central station PV plant its due share of credit. This dissertation deals mainly with the development and implementation of this new approach -- a dynamic rule-based dispatch algorithm which takes into account the problems faced by the dispatch operator during a dispatch interval and channels those into a knowledge base. The new dynamic dispatch requires forecasts of photovoltaic generations at the beginning of each dispatch interval. A Box-Jenkins time-series method is used to model the sub-hourly solar irradiance. The irradiance data at any specific site is stripped of its periodicities using a pre-whitening process which involves parameterization of certain known atmospheric phenomena. The pre-whitened data series is considered stationary, although some non-stationarity might be introduced by the discontinuities in the data collection during night hours. This model is extended to yield forecast equations which are then used to predict the photovoltaic output expected to occur at certain lead times coinciding with the economic dispatch intervals. A rule-based (RB) dispatch algorithm is developed in this dissertation. The RB is introduced to operate as a substitute for the dispatch operator. Some of the dispatcher's functions are routine jobs, while some require specialized knowledge or experience. The RB is given these two qualities through a number of rules. This algorithm works in tandem with a conventional economic dispatch algorithm. The functions of the two are coordinated by another algorithm which oversees the now of information and records them. The RB gives one of 16 possible solutions as and when required. These solutions are written as rules which manipulate the non-committable generation to achieve an optimal solution. The RB system during its operation supervises the fact that the PV generation are kept at the maximum level possible under all constraints. The case study revealed that the thermal generating units which are scheduled by the unit commitment are able to absorb most of the small to medium variations present in the PV generations. In cases of large variations during a single interval, the thermal generators reach their response limits before they can reach their maximum or minimum generation, thus causing mismatches in the load and generation. The mismatches are then picked up by the non-committable sources of generation, comprised of pumped storage units, hydro generation plant, or by interconnection tie-lines. If none of these are sufficient, changes are made in the PV generation schedule. It is concluded that results depend on the time of the year and the specific utility. The time of the year information is reflected in the load demand profile. Most utilities in the U.S. have single peaks in summer and double peaks in winter. Also, the time of the peak load occurrence varies with season. The utility generating capacity mix influences the results greatly. / Ph. D.