A Multi-Feature Based Approach Incorporating Variable Thresholds for Detecting Price Spikes in the National Electricity Market of Australia (original) (raw)
Related papers
Electricity market price spike analysis by a hybrid data model and feature selection technique
Electric Power Systems Research, 2010
In a competitive electricity market, energy price forecasting is an important activity for both suppliers and consumers. For this reason, many techniques have been proposed to predict electricity market prices in the recent years. However, electricity price is a complex volatile signal owning many spikes. Most of electricity price forecast techniques focus on the normal price prediction, while price spike forecast is a different and more complex prediction process. Price spike forecasting has two main aspects: prediction of price spike occurrence and value. In this paper, a novel technique for price spike occurrence prediction is presented composed of a new hybrid data model, a novel feature selection technique and an efficient forecast engine. The hybrid data model includes both wavelet and time domain variables as well as calendar indicators, comprising a large candidate input set. The set is refined by the proposed feature selection technique evaluating both relevancy and redundancy of the candidate inputs. The forecast engine is a probabilistic neural network, which are fed by the selected candidate inputs of the feature selection technique and predict price spike occurrence. The efficiency of the whole proposed method for price spike occurrence forecasting is evaluated by means of real data from the Queensland and PJM electricity markets.
A new prediction strategy for price spike forecasting of day-ahead electricity markets
Applied Soft Computing, 2011
Price spikes are distinctive aspects of electricity price impacting its forecast accuracy. Electricity price spikes can also have serious economical effects on the market participants. However, prediction of electricity price spikes is a complex task and most of current electricity price forecast methods focus on prediction of normal prices. In this paper, a new forecast strategy for prediction of both occurrence and value of electricity price spikes is presented. The proposed strategy has a novel feature selection technique based on information theoretic criteria to select a minimum subset of the most informative features for the forecast process. Also, the strategy includes a new closed loop prediction mechanism composed of probabilistic neural network (PNN) and hybrid neuro-evolutionary system (HNES) forecast engines. The effectiveness of the proposed forecast strategy for the prediction of both price spike occurrence and value is extensively evaluated by the real-life data of PJM (Pennsylvania-New Jersey-Maryland) electricity market. The obtained results confirm the validity of the developed approach.
The logistic regression in predicting spike occurrences in electricity prices
2018
Electricity supply and demand are subject to weather conditions (temperature, wind speed, precipitation) as well as daily, weekly or yearly seasonality due to e.g. an intensity of business activities. These features have a significant impact on the market and price behaviour. As a result of the lack of storage capacity sharp movements of electricity prices are often observed. An ability of modelling and forecasting jumps and spikes plays the crucial role in risk management. In the paper, the logistic regression is employed to predict spike occurrences. We investigate the impact of fundamental variables such as demand, weather and seasonal factors, on spikes occurrences. The point and interval theoretical probabilities are calculated. The classification accuracy is assessed by means of the sensitivity, specificity, accuracy and AUC measures. In our research we detect spikes using a quantile technique and a Bayesian DEJD model. We state that the logistic regression is a quite good too...
Modeling and forecasting multivariate electricity price spikes
Energy Economics, 2016
We consider the problem of forecasting the occurrence of extreme prices in the Australian electricity markets from high frequency spot prices. In particular, we are interested in the simultaneous occurrence of such so-called spikes in two or more markets. Our approach is based on a novel dynamic model for multivariate binary outcomes, which allows the latent variables driving these observed outcomes to follow a vector autoregressive process. Furthermore the model is constructed using a copula representation for the joint distribution of the resulting innovations. This has several advantages over the standard multivariate probit model. First, it allows for nonlinear dependence between the error terms. Second, the distribution of the latent errors can be chosen freely. Third, the computational burden can be greatly reduced making estimation feasible in higher dimensions and for large samples. The model is applied to spikes in half-hourly electricity prices in four interconnected Australian markets. The multivariate model provides a superior fit compared to single-equation models and generates better forecasts for spike probabilities. Furthermore, evidence of spillover dynamics between the markets is revealed.
Data-Driven Method to Detect the Events of Interest in an Electricity Market
Participants in an electricity market expect to have a fair, transparent, and open competition. Independent System Operator (ISO) is responsible to monitor the market outcomes to investigate if they are consistent with fundamentals of electricity market. It can be an intensely time-taking process with high levels of computations in an electricity market with huge number of participants. Besides, a manual review of market operation may be deceitful as human is part of decision making process. If this anomaly detection procedure can be done automatically then it can be a great aid to market surveillance process for having an unbiased and prompt tool to monitor the market. In this paper, an anomaly detection algorithm is proposed to identify the events of interest in an electricity market. This algorithm provides the ISO with a tool to detect the instances in the electricity market which electricity price behavior deviates from normal expected regime. These anomaly hours then can be analyzed further in order to diagnose the reason.
A Data-Driven Method to Detect the Abnormal Instances in an Electricity Market
2015 IEEE 14th International Conference on Machine Learning and Applications (ICMLA), 2015
Participants in an electricity market expect to have a fair, transparent, and open competition. Market Surveillance Administrators (MSA) are responsible for monitoring the market outcomes to investigate if they are consistent with the fundamentals of the electricity markets. It can be an immensely timeconsuming process with high amounts of computations in an electricity market with huge numbers of participants. Besides, a manual review of market operations may be biased by involving humans in the decision making process. If this anomaly detection procedure can be done automatically then it can be a great aid to the market surveillance process for having an unbiased and prompt tool to monitor the market. In this paper, an anomaly detection algorithm is proposed to identify the events of interest in an electricity market. This algorithm provides the MSA with a tool to detect the instances in the electricity market when electricity price behavior deviates from the normal expected regime. These anomalous hours can then be analyzed further in order to diagnose the reason.
Energy Economics, 2013
An important issue in fitting stochastic models to electricity spot prices is the estimation of a component to deal with trends and seasonality in the data. Unfortunately, estimation routines for the long-term and short-term seasonal pattern are usually quite sensitive to extreme observations, known as electricity price spikes. Improved robustness of the model can be achieved by (a) filtering the data with some reasonable procedure for outlier detection, and then (b) using estimation and testing procedures on the filtered data. In this paper we examine the effects of different treatment of extreme observations on model estimation and on determining the number of spikes (outliers). In particular we compare results for the estimation of the seasonal and stochastic components of electricity spot prices using either the original or filtered data. We find significant evidence for a superior estimation of both the seasonal short-term and long-term components when the data have been treated carefully for outliers. Overall, our findings point out the substantial impact the treatment of extreme observations may have on these issues and, therefore, also on the pricing of electricity derivatives like futures and option contracts. An added value of our study is the ranking of different filtering techniques used in the energy economics literature, suggesting which methods could be and which should not be used for spike identification.
Modeling Price Spikes in Electricity Markets—The Impact of Load, Weather, and Capacity
Energy Pricing Models, 2014
We examine the impact of explanatory variables such as load, weather and capacity constraints on the occurrence and magnitude of price spikes in regional Australian electricity markets. We apply the so-called Heckman correction, a two-stage estimation procedure that allows us to investigate the impact of the considered variables on extreme price observations only, while correcting for a selection bias due to non-random sampling in the analysis. The framework is applied to four regional electricity markets in Australia and it is found that for these markets, load, relative air temperature and reserve margins are significant variables for the occurrence of price spikes, while electricity loads and relative air temperature are significant variables to impact on the magnitude of a price spike. The Heckman selection model is also found to outperform standard OLS regression models with respect to forecasting the magnitude of electricity price spikes.
Prediction Algorithm & Learner Selection for European Day-Ahead Electricity Prices
2020
The prediction of day-ahead electricity prices with higher accuracy is always helpful for the market players of the power exchange. This study was intended in the first place to find out the best time series prediction method for the selected 14 European countries. The test results of four time-series methods show that the next day prices were more in line with the previous day prices in 87% of the selected countries; Later, a classification approach is followed by 33 different features of each country to answer the question of which method would be the best for the other countries, that were not studied in this paper, would be? As a result, the support vector machine algorithm results in 57% accuracy in classifying an unknown European country to determine the best prediction method. Therefore, this paper focuses now on two correlated studies to find out the best time series prediction methods and a classification approach for selected countries.