The price of electricity in the Mibel is very changeable. This creates a lot of uncertainty and risk in market actors. Due to continuous changes in demand and marginal price adjustment, buyers and sellers cannot know in advance the evolution of prices. The study of this uncertainty motivates this work. Unlike other published work, this paper analyzes the perspective of the buyer and not the seller's perspective, as is usual in the literature. The aim of this work is to develop predictive models of electric price to build tools to manage and reduce the risk associated with the volatility of the wholesale electricity market and therefore provide better opportunities for small traders to participate in that market. On the other hand, these models are useful to large industrial consumers by enabling them to design strategies to optimize its production capacity in function to signals of electricity market price and can get better on their production costs. Therefore, this article is based on the prediction of energy prices instead of demand. This paper analyzes the model of energy prices to determine the key variables that define its final value. The proposed model has been applied to Mibel 2012. The results suggest the use of several models based on calendar and taking into account different combinations.

1. Modeling prices in electricity Spanish markets
under uncertainty
G-TeC research group, Complutense University, Madrid
eKergy Technologies, SL Madrid, Spain
ISKE 2013, ShenZhen – China
G-TeC members:
Guadalupe Miñana, Raquel Caro, Beatriz. González, Victoria Lopez
eKergy Technologies members:
Hugo Marrao, Jesús Gil

2. Index
1.
2.
3.
4.
5.
6.
7.
8.
9.
Introduction to the electricity market
Motivation & objectives
Modeling price
Mibel's Day-ahead Market
Variables
The model
Results
Conclusions
Future work

3. The electricity market
•
•
•
Electricity is the main energy source from today society.
Electricity can't be store
Electrical market depends on the distribution grid and it requires that generation
equals consumption at every instant.
•
In general the full electrical system is divided in 4 activities that required a higher
coordination:
o
•
Generation, Transport, Distribution and consumption.
Traditionally, the electrical market prices were regulated by the government. With the
evolution and growing of such good in Europe, the electrical market deregulation led
countries to create electrical markets in order to fulfill their needs.
•
In Spain, Mibel 2009, is the case of Spain and Portugal as response to the
deregulation.
•
Other countries in Europe have created similar markets, for example Nord Pool or
EEX

4. Motivation & Objectives
•
Web platform for statistical modeling and profiling energy consumption
for home users.
•
Mathematical tool for modeling, statistical profiling and consumption
simulation scenarios.
•
Aimed at promoting green energy cooperatives and assistance in
forecasting consumption, purchasing power and customer management

5. Hour price setting evaluation in the
Mibel's Day-ahead Market

6. Modeling price
•
The object of this work has been to analyze and identify the variables
that most influence on the price.
•
To achieve this goal we applied Multiple Linear Regression (MLR) using
SPSS tool.
•
A Linear Dependency Analysis among Electric Market Prices and the
amount of energy produced by each technology, the electric demand and
the wind power generated is due.
•
We aim to discover if is correct and necessary analyze the dataset
according to the season, working or non-working days, and the time
slot.
•
Several techniques are studied to offer different models depending on
these variables.

7. Variables

8. The model

9. Results
•
A new dataset is generated with the inputs and outputs of day-ahead
market of the years 2012 (information obtained from the market
operator).
•
There is one day with 23 hours and other day with 26 hours (due
to the time change that takes place twice a year)

10. Results

11. Results
•
This table shows summary of descriptive statistics for some of the
samples. We can note the variations of the mean price and the standard
deviation by our calendar criterion. For instance, the mean price in the
whole data set is 48,02 units and the standard deviation is 12,22 units.
However, for the sample f(0,0,2) the mean price is 65,38 units and the
standard deviation is 7,21 units. These results confirm the importance of
the calendar effect on the market price.
•
Table 1 Summary of descriptive statistics for some of the samples

12. Results
•
•
This table shows the Pearson correlation coefficients for some of the samples.
Variables WG (wind power generation) and SRV (traded special regimen volume) reduce
marginal price and the other variables increase it.
•
For nuclear energy, the values show that there are samples in which this variable reduces the
price and others in which this variable increases the price. This is because this method does
not reflect the character constant of the traded volume of this technology.

13. Results
•
Another observation we can make is that, in most of the samples, the
variables that most influence on the price are the traded volume of
imported coal, combined cycle and conventional hydraulics.
Also, there are some samples in which the demand, the wind
generated and the traded special regimen volume become
important. All of these variables have different weights depending
on the sample. These results also confirm the importance of the
calendar effect on the marginal price.

14. Results

15. Conclusions and Future Work
•
A Linear Dependency Analysis among Electric Market Prices and the amount of energy
produced by each technology, the electric demand and the wind power generated is
presented in this work.
•
This method has confirmed that is correct and necessary analyze the data set in function
of the season, if the day is working day or non-working day, and the time slot.
Therefore, this technique offers different models depending on these variables.
•
This method reveals deficiencies in interpreting the marginal character of the price of
electricity. Therefore, other modelling and forecasting techniques must take into account
due to the special characteristics of nuclear energy and the energies of special regime, the
marginal character of the electricity market and the uncertainty introduced by demand
and wind power.
•
In conclusion, we can say that this analysis shows us that it is good to have a range of
predictions models and a decision-making algorithms to choose the best model
for each situation  sMeCoop could be a useful tool for electricity market managers.

16. Conclusions and Future Work
•
In order to find the best method to model and predict the electric energy
price for each sample, the next step is to apply different prediction
techniques (Exponential Smoothingand, Moving Average, the nearneighbors, neuronal networks) and make a comparison among them.
•
After the modeling and forecasting, we are going to develop a decision
Making Model using fuzzy logics. This will allow us to choose the best
prediction model for each situation. The codification that we have
presented in this work will be used in order to obtain the logical fuzzy
system that uses the defined Models.
•
In the future work we will work with R language instead of using SPSS.
With R we want to achieve better results and modeling of the problem.

17. Modeling prices in electricity Spanish markets
under uncertainty
G-TeC research group, Complutense University, Madrid
Indizen Technologies, SL Madrid, Spain
ISKE 2013, ShenZhen – China
G-TeC members:
Guadalupe Miñana, Raquel Caro, Beatriz. González, Victoria Lopez
eKergy Technologies members:
Hugo Marrao, Jesús Gil