Artificial neural network for load forecasting in smart grid

1. ARTIFICIAL NEURAL NETWORK
FOR LOAD FORECASTING IN
SMART GRID
Ehsan Zeraatparvar
RMRP Course

2. Outline
 Introduction and Background
 Objectives
 Load Forecasting Methods
 Why ANN?
 Proposed Approach
 What Are Neural Networks
 Different Types of Neural Networks
 Network Structure
 Training a Neural Network
 Back Propagation
 Simulation Results
 Training Performances
 Result Comparison
 Conclusions

3. Introduction and Background
 Objective
 Electric power generation, transmission, distribution, security
 Increase or decrease output of generators
 Interchange power with neighboring systems
 Prevent overloading
 Electric power market
 Price settings
 Economic operation of power plants

4. Introduction and Background
 Load Forecasting Methods
 Parametric methods
 Artificial intelligence methods
 Artificial neural networks
Feed-forward network
Feedback network
 Fuzzy logic
 Expert systems

5. What Are Neural Networks?
 Massively parallel networks of simple processing elements (neurons)
 Designed to emulate the functions and structure of the brain
 can solve very complex problems
 a new method of programming computers

6. Different Types of Neural Networks
 Feed-forward Network
 Signals travel in one way only; from input to output
 No feedback
 the output of any layer does not affect that same layer
 Feedback Network
 signals traveling in both directions by introducing loops
 Feedback networks are dynamic
 They remain at the equilibrium point until the input changes

7. Network Structure
 Estimate the number of layers and of neurons
 trial and error procedure
 Two types of adaptive algorithms can be used:
 start from a large network
 begin with a small network

8. Network Training
 The process of determining the network parameters to achieve the
desired objective
 Neural networks learn from examples
 The most basic method of training; Trial and Error
 epoch-by-epoch learning
 The Aim: determine a set of weights which minimizes the error

9. Back Propagation
 Most widely and frequently used neural network learning algorithm
 mathematically designed to minimize the error
 and propagate backward the local error terms

10. Simulation Results
 Forecasting Procedure
 Data Source
 Ontario weather stations and dispatching centers
 Historical Data
 Load – load for the year 2008
 Weather – weighted average hourly weather conditions of
stations in Ontario Province, Canada for 2 years

11. Simulation Results
trainlm algorithm performance plot trainbr algorithm performance plot
 Training Performances

12. Result Comparison
 The network test simulation should be made in order to find out the
performance in a real problem:
The simulation result of the trainbr algorithm with 8 neurons

13. Result Comparison
The simulation results of both trainbr and trainlm with 8 neurons
The simulation results of trainlm with 8 neurons and 30 neurons
Test target, Trainlm with 30 neurons, trainlm with 8 neurons

14. Conclusions
 Trainbr is one of the best choices to do load forecast
 More neurons are needed in network structure to obtain
accurate results
 A few of the simulation part didn’t match the real demand
because of lack of information
 More enough information and a precise training give us better
results for load forecasting in a smart grid.

15. Special thanks for your care.