Study On The Load Forecasting Algorithm Based On Trajectory Tracking

Electric power industry is an important basic industry in our country. Power is the necessary energy in daily life, is about the national economy and people's livelihood,and is the lifeblood of the national economy. In recent years, electricity has become more and more important production factor. Load forecasting not only can be used in the planning and distribution of high side load of power grid, and also can provide basis for energy electricity monitoring, planning, and effective use of energy electricity for low voltage user. High accuracy of power load forecasting is the main basis of electric power system planning, market trading, marketing and other departments work, and for the economic optimization of power generation planning, rationally distributing the electric energy, reasonable arrangement of the unit operation, getting a feed-in tariff advantages, achieving maximum economic benefits and social benefits, it has very important significance. So the high precision and strong practicability of power load forecasting method for electricity market and the development of smart grid are very necessary.Although there are a lot of prediction algorithm, for most of them, most existing forecasting methods are essentially model based in that the corresponding forecasting algorithms are derived from the specific load models. And for most these existing models, they have no theoretical basis to support their method can guarantee the convergence of the prediction error to achieve sufficient accuracy. In the paper, we first introduce the trajectory tracking theory into the power load forecasting, according to the trajectory tracking theory, different kinematics models are established, and this paper proposes two kinds of load forecasting algorithm, i.e., the short-term load forecasting method based on prediction error convergence inspection and trajectory correction and the short-term load forecasting based on trajectory tracking control. In this paper, the stability of the system is proved by using Lyapunov stability theory and make the prediction error converge to zero theoretically so as to guarantee the convergence of the prediction error, and thus the prediction algorithms of this paper have universality and robustness. In the simulation of this paper, the different data sets(European intelligence network(EUNITE Competition and open Load Load 1998 sample data and the Global Energy Forecasting Competition 2012 data sets) are adopted for the experimental simulation, by changing different prediction horizons, and compared with well-knownmodels(ARMA model, Auto Regressive Moving Average and BPNN- the back propagation neural network), the result shows that the proposed algorithms have very good load forecasting performance. Because the proposed algorithms in the paper need a little large of the sample data, the calculation speed is faster than the other methods.