In this paper, a Short-Term i o a d Prediction (ST-P) algorithm for dynamic economic dispatch is presented. This algorithm is based on using a finite autoregressive (AR) time series p r w e s s in conjunction with Kalmon filter. I t s objective is t o predict :oad value with five minute intervals up t o one hour into the future, with sufficient accuracy to be used for targeting dynamic economic dispatch.
Abstlact -INTRODUCTIONLoad forecasting is an integral p a r t o f electric power system aperatlon. Long lead time forecast of 5 to 20 years aheod are needed for scheduling construction of new generating capacity as well as the determination o f prices and regulatory policy. intermediate term forecast of few months to 5 years ahead are needed for maintenance scheduling, coordination of power sharing arrangements and setting of prices, so that demand can be met with fixed capacity. Shcr: term forecast of a few hours to a few weeks ahead are needed for economic scheduling of generating capacity, scheduling of fuel purchases, security analysis and short term maintenance scheduling. Very short term forecast of a few minutes to an hour ahead are needed for real-time control and real-time security evaluation.There has been a wide variety of procedures for short term load forecasting reported in the literature. Some of these procedures have included multiple regression, spectral decomposition, exponential smoothing and state space (Kalman filter) methods.One of the most commonly discussed procedures for load forecasting is time series analysis (or system dentification) normally using procediires outlined b y Box and Jenkins or Kashyap and Rao. This paper will ancantrate on the application of time series analysis method t o a short term ioad forecasting.Many of the inherent problems with automatic generation control (AGC) could be solved with an accurate load forecasting algorithm.Load forecast algorithms should utilize as much as possible existing knowledge of the electric demand. in most power systems. weather has an important impact on load and, therefore, load forecast algorithms should be weother sensitive. Changing environments make it essential the inclusion of an adaptive mechanism to continuously update parameter values. Furthermore, model parameters should recover quickly from unusual situations. This requires that anormalous and nonconforming loads be identified and measures can be taken to preserve the quality of the forecasts and to avoid the possibility of contaminating model parameter values. Short term load forecast models should reflect existing lood and weather periodicities and time correlations. In this paper it will be seen that the new method overcomes these difficulties with better performance.
DYNAMIC ECONOMIC DISPATCH WlTH ' LOOK AHEAD' PROPERTYThe basic purpose of the economic dispatch function is to schedule the outputs of the on-line power generators serving a particular area so as to meet the net area load at 'least' cost. The state-of-the-art in methods for economic dispatch (ED) is that 'sta...
