ShortTerm Load Forecasting Using SystemType Neural Network Architecture

Short-Term Load Forecasting Using System-Type Neural Network Architecture Shu Du, Graduate Student Mentor: Kwang Y. Lee, Professor and Chair Department of Electrical and Computer Engineering Baylor University

Outline Introduction and Background Objectives Ø Load Forecasting Categories Ø Load Forecasting Methods Ø Proposed Approach Regression and Rearrangement Ø System-Type Neural Network Method Ø Learning Algorithm of System-Type Neural Network Ø Extrapolation and Interpolation Ø Simulation Results Rearrangement Ø Output of Semigroup Channel Ø Extrapolation Ø Conclusions

Introduction and Background Ø Objective § Electric power generation, transmission, distribution, security Increase or decrease output of generators ü Interchange power with neighboring systems ü Prevent overloading and reduce occurrences of equipment failures ü § Electric power market Price settings ü Schedule spinning reserve allocation properly ü

Introduction and Background Ø Load Forecasting Categories § Short-term load forecasting One hour ~ One week ü Control and schedule power system in everyday operations ü § Medium-term and Long-term load forecasting One week ~ longer than one year ü Determine capacity of generation, transmission, distribution systems, type of facilities required in transmission expansion planning, development of power system infrastructure, etc. ü

Introduction and Background Ø Load Forecasting Methods § Parametric methods Regression method ü Time series ü Autoregressive Moving Average (ARMA) Spectral expansion technique (Fourier Series) State equations § Artificial intelligence methods ü Artificial neural networks Feedforward network Recurrent network Fuzzy logic ü Expert systems ü

Proposed Approach Ø Regression and Rearrangement § Regression ü Objective Represent given load with respect to two major variables— time and temperature ü Load Form -----Base load component (time factor) -----Weather sensitive load component (weather factor) -----Load component (other factors)

Proposed Approach Ø Regression and Rearrangement § Rearrangement ü Objective Minimize the fluctuation caused by hourly temperature Obtain the smoothness of the given load data Day Temperature üImplementation Align given load based upon magnitudes of hourly temperatures Rearrangem ent 1 2 24 Load before Rearrangement Hou r 1 2 24 Load after Rearrangement Hou r

Proposed Approach Ø System-Type Neural Network Method § Algebraic Decomposition ü Objective Form an approximation load data ü to Implementation Reorganize given load into a parameterized set Select elements and orthonormalize them to a basis set by Gram-Schmidt process Determine the linear combination of basis set for each element Combine the coefficient vector and the basis set to

Proposed Approach Ø System-Type Neural Network Method §Function Channel üStructure— §Semigroup RBF Channel üStructure—Simple networks Recurrent Network üEach üSmoothen network implements one of vector and orthonormal basis functions the coefficient Function Channel (NN 1) Semigroup Channel (NN 2) Realize semigroup property

Proposed Approach Ø Learning Algorithm of System-Type Neural Network § Function Channel ü ü § RBF network can be designed rather than trained RBF networks emulate selected basis functions Semigroup Channel ü ü ü Primary Objective – Replicate and smoothen the vector with a vector which has the semigroup property Secondary Objective – Acquire a semigroup property in the weight space which is the basis for extrapolation The entire trajectory is sliced into a nested sequence of trajectories

Proposed Approach Ø Extrapolation and Interpolation § Extrapolation ü § Interpolation ü Temperat ure Extrapolation is needed only when temperature forecast at a given hour exceeds the historical bounds at the same time Interpolation is needed when temperature forecast at a given hour falls into the historical temperature range at the same time Extrapolat ed Coefficient 4 Decompose & Smoothen 3 Temperat ure 4 Decompose & Smoothen 3 2 Interpolate d Coefficient 2 1 1 2 Hou r 24 Load after Rearrangement 1 4 5 Extrapolation of Coefficient 1 2 Hou r 24 Load after Rearrangement 3 4 Interpolation of Coefficient

Simulation Results Ø Forecasting Procedure § Data Source ü § New England Independent System Operator Historical Data Load – load for the year 2002 ü Temperature – weighted average hourly temperature of 8 stations in the New England area ü § Pattern ü § Weekday pattern (Mon ~ Fri) and Weekend pattern (Sat, Sun) Next Day Forecasting

Simulation Results Ø Simulation of Forecasting A Weekday Load § Rearrangement Rearrange

Simulation Results Ø Simulation of Forecasting A Weekday Load § Output of Semigroup Channel

Simulation Results Ø Simulation of Forecasting A Weekday Load § Extrapolation

Simulation Results Ø Regression Load Forecasting Results

Conclusions Ø Next Day Load Forecasting based upon Weather Forecast A mathematical approach referred to as algebraic decomposition is investigated The system-type neural network architecture combining Radial Basis Function Networks and a Simple Recurrent Network is proposed A new training algorithm in the SRN is proposed Regression and Rearrangement are performed to guarantee smoothness of coefficient vector Interpolation and Extrapolation are implemented based on temperatures Much better results with respect to actual load and

THANK YOU