Вы находитесь на странице: 1из 6

International Journal of Advances in Science and Technology, Vol. 4, No.

1, 2012

Day-Ahead Price Forecasting In Energy Market Using Neural Networks


J.Vasanthi1 and K.Sarada2
1

Department of Electrical Engineering, KL University, Andhra Pradesh, India vasanthiee53@gmail.com

Department of Electrical Engineering, KL University, Andhra Pradesh, India saradak@kluniversity.in

Abstract
Price forecasting is become increasingly relevant to producers and consumers in the new competitive electric power markets. Both for spot markets and long term contractors, price forecasts are necessary to develop bidding strategies. In this paper, we have proposed a neural network (NN) approach to forecast short term hourly electricity prices. The proposed method is examined on the PJM electricity market. The results obtained through the simulation show that the proposed model provides accurate and better results.

Keywords: Electricity market, price forecasting, neural network 1. Introduction


The electricity Industry throughout the world, which has long been dominated by vertically integrated utilities, is undergoing enormous changes. The electricity industry is evolving into a distributed and competitive industry in which market forces drive the price of electricity and reduce the net cost through increased competition. There are two objectives for establishing an electricity market: ensuring a secure operation and facilitating an economical operation. Security is the most important aspect of the power system operation be it a regulated operation or a restructured power market [2, 4, 5]. The economical operation of the electricity market would reduce the cost of electricity utilization. This is a primary motive for restructuring and a way to enhance the security of a power system through its economics. The Independent System Operator (ISO) has the authority to commit and dispatch some or all system resources and to curtail loads for maintaining the system security. The structure for an ISO is based on an optimal power flow dispatch model. Market participants must provide extensive data, such as cost data for every generator, and daily demand for every consumer or load. With these extensive data, the ISO obtains the unit commitment and dispatch that maximizes social welfare, and sets transmission congestion prices, The PJM ISO and the National Grid Company (NGC) in the United Kingdom are the examples of this structure having a wide-ranging of authority and control. The Introduction of restructuring into the electricity power industry, the price of electricity has become the focus of all activities in the power market. It can be categorized into short-term (few days), mid-term (few months) and long-term (few years). In this we are mainly concentrated on short term price forecasting of electricity in the restructured power market. Electricity has distinct characteristics from other commodities. For example, electricity cannot be stored economically, and transmission congestion may prevent a free exchange among control areas. Thus, an electricity price movement shows great volatility among all commodities. In the USA, PJM Interconnection [3] plays a vital role in the U.S. electric system and responsible for the operation of largest centrally dispatched electric system in North America. This market is coordinate with an ISO. It can provide a secured, economical & efficient operation by determining locational marginal prices (LMP). In the PJM market, it is observed that daily power demand curves have similar patterns, where as daily LMP curves are fluctuating. A lot of methods have been reported

January

Page 35 of 101

ISSN 2229 5216

International Journal of Advances in Science and Technology, Vol. 4, No.1, 2012 for price prediction of the electricity markets in last decade. In recent years, forecasting method based on Artificial Intelligence techniques have been reported and especially, NN have received much attention & it provide good solutions to model complex non-linear relationships. In this paper, new approach is used to forecast hourly electricity prices in the PJM market using Neural Network (NN). The advantage of the proposed price forecasting technique with respect to other are: better accuracy and efficiency. To evaluate the performance of the model, mean absolute percentage error (MAPE) is calculated.

2. Proposed Neural Network for Price Forecasting


The classical methods for forecasting include regression and state-space methods. The more modern methods include expert systems, evolutionary programming [1], NN and various combinations of these tools. NN has gained more attention among the existing tools because of its clear model, easy implementation and good performance. A Neural Network (NN) may be defined as an information processing model that is inspired by the way biological nervous systems, such as the brain, process information. This model tries to replicate only the most basic functions of the brain. The key element of NN is the novel structure of its information processing system. An NN is composed of a large number of highly interconnected processing elements (neurons) working in unison to solve specific problems. These are a class of flexible non-linear models that can discover patterns adaptively from the data. In power systems, NNs have been used to solve problems such as load forecasting, Security assessment, Machinery control, Unit commitment etc. In price forecasting applications, the main function of NN is to predict price for the next halfhour(s), day(s) or week(s).In this feed forward model based Back Propagation (BP) trained algorithm is used for predicting the prices. In this paper we proposed a three-layer feed forward model NN for forecasting the next 24th hour electric prices. The network consists of one input layer, one hidden layer and one output layer as shown in fig.1. The input variables of the NN are price data, i.e., actual hourly price data and actual hourly load data and time factor and also the previous past price data is also used for training the network.

Figure 1. A Single Layer Feed Forward Neural Network Model NN forecasting procedure has two steps: Training and Simulation. In the training step, NN finds the relationship between inputs and outputs according to the set of inputs and the corresponding outputs that we give to it. This set is called training set. In learning process, a NN constructs an input-output mapping, adjusting the weights and biases at each iteration based on the minimization of error measure

January

Page 36 of 101

ISSN 2229 5216

International Journal of Advances in Science and Technology, Vol. 4, No.1, 2012 between the output produced and the desired output. So for this most common learning algorithm Back Propagation algorithm is used for error minimization process. In simulation step, a testing set which includes new data given to network and the network acquires through the training step will be tested here.

3. Neural Network Design


According to aim of this paper which is simulation of the 24 hours electricity prices by NN, our NN should have 24 outputs. It has three layers, an input layer, one hidden layer and an output layer. In the present network sigmoid function has been used as activation function. The neurons in the hidden layer that allow neural networks to detect the feature, to capture the pattern in the data and to perform complicated non-linear mapping between input and output variable.

4. Results and Discussion


In the proposed design of the network model has been applied to predict electricity process of dayahead PJM market [3]. The testing is carried out by considering the data of July, August, October, November and December of the year 2003 from the PJM electricity market. Sets of data include hourly price and load data. One day per month (July-December) of year 2003 is studied to validate the performance of the proposed Neural Network (NN) Model. To assess the prediction accuracy of the model, the mean absolute percentage error (MAPE) is used. The MAPE value is computed as follow:

where P represents the Price. Figs. 2-7 show the variation in predicted prices of PJM energy market [3] data during hour to hour, day to day and month to month in a year obtained from the proposed Neural Network model with actual price. By observing the Figs. 2 & 3, the price reaches maximum during morning and evening hours on same day of different month in the year 2003. Similarly, from Fig. 4 i.e., on weekend Sunday the PJM prices increases at 12p.m in the afternoon and during midnight of the predicted day and coming to Fig. 5 & Fig. 6 we can observe that the market prices during morning hours should vary slightly morning hours in and during evening hours the prices should almost equal in the middle of a week on two preceding days of a same month i.e., in the month of November 12 & 13 in a year 2003. During starting of week, the prices should vary from hour to hour and constant prices during morning hours are observed in Fig.7. The performance of the predictions resulting from the proposed neural network model is evaluated by computing Mean Absolute percentage Error (MAPE).

Figure 2. Forecasted Prices in Day- Ahead PJM Market on Saturday (July 19th, 2003)

January

Page 37 of 101

ISSN 2229 5216

International Journal of Advances in Science and Technology, Vol. 4, No.1, 2012

Figure 3. Forecasted Prices in Day-Ahead PJM Market on Saturday (August 23rd, 2003)

Figure 4. Forecasted Prices in Day-Ahead Market on Sunday (October 18th, 2003)

Figure 5. Forecasted Prices in Day-Ahead Market on Wednesday (November 12th, 2003)

January

Page 38 of 101

ISSN 2229 5216

International Journal of Advances in Science and Technology, Vol. 4, No.1, 2012

Figure 6. Forecasted Prices in Day-Ahead PJM Market on Thursday (November 13th, 2003)

Figure 7. Forecasted Prices in Day-Ahead PJM Market on Monday (December 8th, 2003) Table 1. Forecasting Performance of the Proposed NN Model Proposed study

Weekday

Weekend day

MAPE(%)

9.00 9.55

7.55 8.56

5. Conclusion
This paper proposed a technique based on Neural Network to forecast hourly prices in the PJM electricity market. The different factors impacting electricity price forecasting are load factor, time factor and price factor are considered. In all these factors demand factor is most important factor to varies the price factor. For the selected days of the month of the year of PJM electricity market price data should simulate the given data with proposed NN model and the model results obtained through the simulation demonstrate the efficiency, accuracy and high adequacy of the proposed technique.

January

Page 39 of 101

ISSN 2229 5216

International Journal of Advances in Science and Technology, Vol. 4, No.1, 2012

6. References
[1] Gareta, R. Romeo, L., Gil, A., 2006, Forecasting of electricity prices with neural networks, Energy Conversion and management, 1770-1778. [2] Steven Soft, Power System Economics: Designing Markets for Electricity. John Wiley & Sons Inc., New York, 2002. [3] PJM Web Site, http://www.pjm.com, Active March 2003. [4] M. Shahidehpour, H. Yamin, and Z. Li, Market operations in electric power systems: forecasting, scheduling, and risk assessment. John Wiley & Sons, Inc., New York, 2002. [5] Loi Lei Lai, Power System Restructuring and Deregulation John Wiley & sons, Inc., New York, 2002.

Authors Profile

J.Vasanthi received B-Tech degree in Electrical and Electronics Engineering from V.R.Siddhartha Engineering College, Andhra Pradesh, India in 2009 and currently pursuing M-Tech Power Systems in KL University, Andhra Pradesh, India. Her research interests in De-regulation, Power Systems and Power System Stability.

K.Sarada received B-Tech degree in Electrical and Electronics Engineering from Sri Vidya Niketan Engineering College, Andhra Pradesh, India in 2001and M-Tech degrees in Electrical and Electronic Engineering from Sri Venkateswara University, Andhra Pradesh, India in 2004 and currently she is an Associate Professor in KL University, Andhra Pradesh, India. Her research interests in De-regulation, power system stability.

January

Page 40 of 101

ISSN 2229 5216

Вам также может понравиться