Using multivariate time series to estimate location and climate change effects on historical temperatures employed in future electricity demand simulation (original) (raw)
Related papers
Australian & New Zealand Journal of Statistics, 2017
Long-term historical daily temperatures are used in electricity forecasting to simulate the probability distribution of future demand but can be affected by changes in recording site and climate. This paper presents a method of adjusting for the effect of these changes on daily maximum and minimum temperatures. The adjustment technique accommodates the autocorrelated and bivariate nature of the temperature data which has not previously been taken into account. The data are from Perth, Western Australia, the main electricity demand centre for the SouthWest of Western Australia. The statistical modelling involves a multivariate extension of the univariate time series 'interleaving method', which allows fully efficient simultaneous estimation of the parameters of replicated Vector Autoregressive Moving Average processes. Temperatures at the most recent weather recording location in Perth are shown to be significantly lower compared to previous sites. There is also evidence of long-term heating due to climate change especially for minimum temperatures.
Analyzing the Impact of Weather Variables on Monthly Electricity Demand
IEEE Transactions on Power Systems, 2005
The electricity industry is significantly affected by weather conditions both in terms of the operation of the network infrastructure and electricity consumption. Following privatization and deregulation, the electricity industry in the UK has become fragmented and central planning has largely disappeared. In order to maximize profits, the margin of supply has decreased and the network is being run closer to capacity in certain areas. Careful planning is required to manage future electricity demand within the framework of this leaner electricity network. There is evidence that the climate in the UK is changing with a possible 3 0 C average annual temperature increase by 2080. This paper investigates the impact of weather variables on monthly electricity demand in England and Wales. A multiple regression model is developed to forecast monthly electricity demand based on weather variables, gross domestic product and population growth. The best model is able to explain approximately 95% of the variability in monthly electricity demand from 1983 to 1995 and approximately 91% over the period 1999 to 2003, though the accuracy of forecasting during winter and summer is lower. This may reflect the non-linear dependence of demand on temperature at the hot and cold temperature extremes, however, the inclusion of the effects of relative humidity seem to improve the demand forecast during the summer months.
A new approach to modeling the effects of temperature fluctuations on monthly electricity demand
Energy Economics, 2016
This paper proposes a novel approach to measure and analyze the effect of temperature on electricity demand. This temperature effect is specified as a function of the density of temperatures observed at a high frequency with a functional coefficient, which we call the temperature response function. This approach contrasts with the usual approach to model the temperature effect as a function of heating and cooling degree days. We further investigate how non-climate variables, which include the price of electricity relative to that of substitutable energy and latent variables such as preferences and technology that we proxy by a linear time trend, affect the demand response to temperature changes. Our approach and methodology are demonstrated using Korean electricity demand data for residential and commercial sectors.
Seasonality and Weather Effects on Electricity Loads: Modeling and Forecasting
2001
We wish to investigate climate change-driven effects on electricity demand and production. We model hourly loads for the electricity region and the individual electric utilities of the Pennsylvania, New Jersey, and Maryland Interconnection (PJM) from January 1 st , 1998 through April 30 th , 2001. We create a database of hourly electricity loads for PJM, by individual utilities and in aggregate. We then estimated a set of hourly forecasting models incorporating autoregressive components, heating and cooling degree temperature effects and trading day variation for holidays and weekends. We use the models' short-run elasticities to perform a simulation of a 2°F increase in daily temperature, finding a small but positive impact on electricity demand.
Load forecasting under changing climatic conditions for the city of Sydney, Australia
Energy
In the current context, climate change has become an unequivocal phenomenon. Although it primarily encompasses change in temperature, nevertheless other weather variables such as rainfall, wind speed, evaporation and humidity can also be affected as a result of climate change. Addressing the impacts of climate change on electricity demand is essential for predicting the future demand. For example, cooling and heating requirements change significantly with respect to climate change that may result to the change in electricity load demand. In this paper, a backward elimination based multiple regression approach is proposed for analyzing the influence of climatic variables on load forecasting. A correlation analysis has been carried out using Pearson's correlation coefficient to examine the interdependency between different climatic variables in the context of Sydney, one of the most densely populated cities in Australia. Regression based analysis has been performed to examine the relationship between per capita electricity demand and associated climatic variables. 'Degree Days' concept has been utilized to determine balance point temperature. Backward elimination based multiple regression is used to exclude non-significant climatic variables and evaluate the sensitivity of significant variables related to the load demand. Average change in future per capita electricity demand has been predicted using the proposed approach for the city of Sydney, Australia. Results indicate that the demand for Sydney will increase by 6%
The impact of climate change on electricity demand in Australia
Energy & Environment, 2018
This study estimates the short- and long-term impacts of climate change on electricity demand in Australia. We used an autoregressive distributed lag (ARDL) model with monthly data from 1999 to 2014 for six Australian states and one territory. The results reveal significant variations in electricity demand. We then used long-term coefficients for climatic response to simulate future electricity demand using four scenarios based on the representative concentration pathways (RCPs) of the Intergovernmental Panel on Climate Change (IPCC). Our results show a gradual increase in electricity consumption due to warmer temperatures with the possibility of peak demand in winter; however, demand tends to decrease in the middle of the twenty-first century across the RCPs, while the summer peak load increases by the end of the century. Finally, we simulated the impact of policy uncertainty through sensitivity analysis and confirmed the potential benefits of climate change adaptation and mitigation.
Applied Stochastic Models in Business and Industry, 2013
The problem of forecasting the electricity load has been widely discussed in the literature. The underlying phenomenon is known to be a complex one, and a number of models have been proposed to meet this challenge. The authors go one step further by introducing a broad class of dynamic models, which includes many of those proposals. The resulting structure takes into account most of the relevant factors affecting consumption; it is complex but flexible. It is certainly promising, and the authors are to be congratulated for this achievement. It has been recognized that electricity load series are the result of many different superimposed causes such as economic activity, population dynamics, the seasons of the year, meteorological conditions, temperature patterns, the type of day (weekday, weekend or holiday), the hour of the day and the observance of daylight saving time, for example. Some of these components have an impact on the trend of the series, others produce cyclic patterns (with different frequencies) and some others have a combined effect (trend and cycle). It is clear that a model able to explicitly cope with each one of these effects cannot be simple. Indeed, the authors rely on several simplifying assumptions to make the analyses feasible. An important aspect to consider concerns the use of the forecasts. For the purpose of planning, long-term forecasts are relevant, and so general trend and main effects are the most relevant components. In contrast, as pointed out by the authors, optimal operation of the electricity network requires short-term and even very short-term predictions of consumption ([1]). In fact, in this latter setting, the general trend or cyclic patterns can be ignored in favor of more local effects. As far as we understand, the proposed class of models is intended to provide support to the operation of the system, and so the focus is on short-time forecasts. The authors discuss and compare multivariate and univariate approaches. In the former case, they explore joint modeling of hourly measurements using a structural factor model, which includes terms for temperature as well as some dummy variables. As a first simplification, they estimate the decay factor parameters and subsequently treat them as known. In the latter case, they consider separate univariate models for hourly measurements, each of which takes into account the trend, temperature, seasonality and dummy variables. As a further simplification, the authors assume a common discount factor. The paper clearly shows the complexity involved in the implementation of these models. Not only the computational effort is considerable, specially in the multivariate case, but also a number of non-automatic inputs (including the priors) are required to feed the forecast procedure. It would be interesting if the authors could comment on the way they think the operators of the electricity network could implement and use these models as a routine exercise. The other important issue is that of the comparison of results. Intuitively, taking into account the correlation between hourly measurements should give rise to better forecasts. However, the proposed multivariate and univariate models are not directly comparable. Both may provide similar point-wise forecasts, but differences in precision may be explained not only by the correlation between hourly measurements but also by the different location structures assumed for each type of model (compare Figures 8 and 9). On the other hand, we understand that, because the new models generalize many of the existing proposals, their results cannot be worse than those obtained by their predecessors. Nevertheless, it would be useful to have a benchmark model to compare with, both in terms of accuracy of the forecasts and difficulty of implementation. In this regard, we would suggest a simple model such as that introduced by Mendoza and de Alba [2] to deal with short-time series. This could be appropriate if we recall that, in the case of short-term forecasts, not only the general trend is ignored but also usually a small running window is defined so as to only use the most recent data. In contrast with the highly elaborated models in this paper, which explicitly account for every short-term cause to explain the consumption, the oversimplified proposal by Mendoza and de Alba [2] only relies on the assumption of a stable seasonal pattern of a positive random variable, which can be regarded as the result of an accumulation process.
Using Weather Patterns to Forecast Electricity Consumption in Sri Lanka: An ARDL Approach
International Energy Journal, 2021
It is crucial to plan the electricity supply to match the future demand since electricity has become a dominant utility. Sri Lanka as a developing country, has over 98% of households electrified, which sometimes suffer from interruptions in supply. This study aims at forecasting monthly electricity consumption in Sri Lanka by considering the influence of weather patterns. Rainfall, humidity, and temperature are the three main weather parameters found to affect the electricity demand. We compared eight forecasting approaches including four econometric models and four algorithmic forecasting methods in forecasting monthly electricity consumption. Twenty meteorological stations were considered to spatially interpolate the weather data using the Inverse Distance Weighted (IDW) interpolation method. Results revealed that Autoregressive Distributed Lag (ARDL) model which incorporates the weather patterns as predictors outperforms in forecasting the monthly electricity consumption compared with all other forecasting approaches.
Short-Term Electricity Demand Forecasting: Impact Analysis of Temperature for Thailand
Energies, 2020
Accurate electricity demand forecasting for a short horizon is very important for day-to-day control, scheduling, operation, planning, and stability of the power system. The main factors that affect the forecasting accuracy are deterministic variables and weather variables such as types of days and temperature. Due to the tropical climate of Thailand, the marginal impact of weather variables on electricity demand is worth analyzing. Therefore, this paper primarily focuses on the impact of temperature and other deterministic variables on Thai electricity demand. Accuracy improvement is also considered during model design. Based on the characteristics of demand, the overall dataset is divided into four different subgroups and models are developed for each subgroup. The regression models are estimated using Ordinary Least Square (OLS) methods for uncorrelated errors, and General Least Square (GLS) methods for correlated errors, respectively. While Feed Forward Artificial Neural Network...