Operative Forecast of Node Loads in the Electric Power System for the Purposes of State Identification

D.A. Nikolov (Bulgaria)


Modeling, forecasting, electrical load;nodes; Kalman filter; control.


When calculating load-flow in the electrical power system, the values of electrical loads in main nodes are required. The missing information about loads can be substituted by their forecast values. For this purpose, components decomposition is proposed. The optimal estimation of the main components for a specified period can be obtained using Kalman filter. To increase forecasting accuracy, the correction of the matrix in front of main components in each step of filtering using recurrent dependence of Kachmaj is required. The selected main components are required to retain their orthogonality during the filtering process. The orthogonality check is carried out using the value of non diagonal elements in noise covariance matrix when forecasting the electrical loads by means of Kalman filter. In case of significant correlation between the components, the performance of a new component decomposition is required. The proposed method is applied in practice by forecasting the active loads in the five nodes. The average daily mean-squared errors in the operative forecasting of node load is about 23%. Kalman filtering process allows efficient establishment of state vector components dependence on its significance at a previous moment in time, assuming that random process of load development in the EPS is Markov' process in nature.

Important Links:

Go Back