• Journal of Electrical Engineering and Technology
• /
• v.10 no.3
• /
• pp.1342-1348
• /
• 2015

The objective of this study is to propose a method to calculate prediction intervals for one-day-ahead hourly forecasts of photovoltaic power generation and to evaluate its performance. One year of data of two systems, representing contrasting examples of forecast’ accuracy, were used. The method is based on the maximum likelihood estimation, the similarity between the input data of future and past forecasts of photovoltaic power, and on an assumption about the distribution of the error of the forecasts. Two assumptions for the forecast error distribution were evaluated, a Laplacian and a Gaussian distribution assumption. The results show that the proposed method models well the photovoltaic power forecast error when the Laplacian distribution is used. For both systems and intervals calculated with 4 confidence levels, the intervals contained the true photovoltaic power generation in the amount near to the expected one.

• Journal of Electrical Engineering and Technology
• /
• v.13 no.4
• /
• pp.1504-1514
• /
• 2018

The objective of this study is to compare the suitability of a non-parametric and 3 parametric distributions in the characterization of prediction intervals of photovoltaic power forecasts with high confidence levels. The prediction intervals of the forecasts are calculated using a method based on recent past data similar to the target forecast input data, and on a distribution assumption for the forecast error. To compare the suitability of the distributions, prediction intervals were calculated using the proposed method and each of the 4 distributions. The calculations were done for one year of day-ahead forecasts of hourly power generation of 432 PV systems. The systems have different sizes and specifications, and are installed in different locations in Japan. The results show that, in general, the non-parametric distribution assumption for the forecast error yielded the best prediction intervals. For example, with a confidence level of 85% the use of the non-parametric distribution assumption yielded a median annual forecast error coverage of 86.9%. This result was close to the one obtained with the Laplacian distribution assumption (87.8% of coverage for the same confidence level). Contrasting with that, using a Gaussian and Hyperbolic distributions yielded median annual forecast error coverage of 89.5% and 90.5%.

• Asia pacific journal of information systems
• /
• v.7 no.1
• /
• pp.67-83
• /
• 1997

The utility of learning techniques in investment analysis has been demonstrated in many areas, ranging from forecasting individual stocks to entire market indexes. To date, however, the application of artificial intelligence to financial forecasting has focused largely on short predictive horizons. Usually the forecast window is a single period ahead; if the input data involve daily observations, the forecast is for one day ahead; if monthly observations, then a month ahead; and so on. Thus far little work has been conducted on the efficacy of long-term prediction involving multiperiod forecasting. This paper examines the impact of alternative procedures for extended prediction using knowledge discovery techniques. One dimension in the study involves temporal granularity: a single jump from the present period to the end of the forecast window versus a web of short-term forecasts involving a sequence of single-period predictions. Another parameter relates to the numerosity of input variables: a technical approach involving only lagged observations of the target variable versus a fundamental approach involving multiple variables. The dual possibilities along each of the granularity and numerosity dimensions entail a total of 4 models. These models are first evaluated using neural networks, then compared against a multi-input jump model using case based reasoning. The computational models are examined in the context of forecasting the S&P 500 index.

• KIEE International Transactions on Power Engineering
• /
• v.4A no.3
• /
• pp.159-166
• /
• 2004

Forecasting prices in electricity markets is critical for consumers and producers in planning their operations and managing their price risk. We utilize the generalized autoregressive conditionally heteroskedastic (GARCH) method to forecast the electricity prices in two regions of New York： New York City and Central New York State. We contrast the one-day forecasts of the GARCH against techniques such as dynamic regression, transfer function models, and exponential smoothing. We also examine the effect on our forecasting of omitting some of the extreme values in the electricity prices. We show that accounting for the extreme values and the heteroskedactic variance in the electricity price time-series can significantly improve the accuracy of the forecasting. Additionally, we document the higher volatility in New York City electricity prices. Differences in volatility between regions are important in the pricing of electricity options and for analyzing market performance.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2021.06a
• /
• pp.131-131
• /
• 2021

Deep learning models, especially those based on long short-term memory (LSTM), have presented their superiority in addressing time series data issues recently. This study aims to comprehensively evaluate the performance of deep learning models that belong to the supervised learning category in streamflow prediction. Therefore, six deep learning models-standard LSTM, standard gated recurrent unit (GRU), stacked LSTM, bidirectional LSTM (BiLSTM), feed-forward neural network (FFNN), and convolutional neural network (CNN) models-were of interest in this study. The Red River system, one of the largest river basins in Vietnam, was adopted as a case study. In addition, deep learning models were designed to forecast flowrate for one- and two-day ahead at Son Tay hydrological station on the Red River using a series of observed flowrate data at seven hydrological stations on three major river branches of the Red River system-Thao River, Da River, and Lo River-as the input data for training, validation, and testing. The comparison results have indicated that the four LSTM-based models exhibit significantly better performance and maintain stability than the FFNN and CNN models. Moreover, LSTM-based models may reach impressive predictions even in the presence of upstream reservoirs and dams. In the case of the stacked LSTM and BiLSTM models, the complexity of these models is not accompanied by performance improvement because their respective performance is not higher than the two standard models (LSTM and GRU). As a result, we realized that in the context of hydrological forecasting problems, simple architectural models such as LSTM and GRU (with one hidden layer) are sufficient to produce highly reliable forecasts while minimizing computation time because of the sequential data nature.

• Journal of Environmental Science International
• /
• v.26 no.10
• /
• pp.1167-1180
• /
• 2017

Numerical Weather Prediction (NWP) models such as the Weather Research and Forecasting (WRF) model are essential for forecasting one-day-ahead solar irradiance. In order to evaluate the performance of the WRF in forecasting solar irradiance over the Korean Peninsula, we compared WRF prediction data from 2008 to 2010 corresponding to weather observation data (OBS) from the Korean Meteorological Administration (KMA). The WRF model showed poor performance at polluted regions such as Seoul and Suwon where the relative Root Mean Square Error (rRMSE) is over 30%. Predictions by the WRF model alone had a large amount of potential error because of the lack of actual aerosol radiative feedbacks. For the purpose of reducing this error induced by atmospheric particles, i.e., aerosols, the WRF model was coupled with the Community Multiscale Air Quality (CMAQ) model. The coupled system makes it possible to estimate the radiative feedbacks of aerosols on the solar irradiance. As a result, the solar irradiance estimated by the coupled system showed a strong dependence on both the aerosol spatial distributions and the associated optical properties. In the NF (No Feedback) case, which refers to the WRF-only stimulated system without aerosol feedbacks, the GHI was overestimated by $50-200W\;m^{-2}$ compared with OBS derived values at each site. In the YF (Yes Feedback) case, in contrast, which refers to the WRF-CMAQ two-way coupled system, the rRMSE was significantly improved by 3.1-3.7% at Suwon and Seoul where the Particulate Matter (PM) concentrations, specifically, those related to the $PM_{10}$ size fraction, were over $100{\mu}g\;m^{-3}$. Thus, the coupled system showed promise for acquiring more accurate solar irradiance forecasts.

• Journal of the Korean Solar Energy Society
• /
• v.39 no.6
• /
• pp.113-125
• /
• 2019

With the opening of the small power brokerage business market in December 2018, the small power trading market has started in Korea. Operators must submit the day-ahead estimates of power output and receive incentives based on its accuracy. Therefore, the accuracy of power generation forecasts is directly affects profits of the operators. The forecasting process for power generation can be divided into two procedure. The first is to forecast solar irradiation and the second is to transform forecasted solar irradiation into power generation. There are two methods for transformation. One is to simulate with physical model, and another is to use regression model. In this study, we found the best-fit regression model by analyzing hourly data of PV output and solar irradiation data during three years for 242 PV plants in Korea. The best model was not a linear model, but a sigmoidal model and specifically a Gompertz model. The combined linear regression and Gompertz curve was proposed because a the curve has non-zero y-intercept. As the result, R² and RMSE between observed data and the curve was significantly reduced.

• International Area Studies Review
• /
• v.14 no.3
• /
• pp.62-84
• /
• 2010

Accurate forecasting of volatility is of considerable interest in financial volatility research, particularly in regard to portfolio allocation, option pricing and risk management because volatility is equal to market risk. So, we attempted to delineate a model with good ability to forecast and identified stylized features of volatility, with a focus on volatility persistence or long memory in the Australian futures market. In this context, we assessed the long-memory property in the volatility of index futures contracts using three conditional volatility models, namely the GARCH, IGARCH and FIGARCH models. We found that the FIGARCH model better captures the long-memory property than do the GARCH and IGARCH models. Additionally, we found that the FIGARCH model provides superior performance in one-day-ahead volatility forecasts. As discussed in this paper, the FIGARCH model should prove a useful technique in forecasting the long-memory volatility in the Australian index futures market.

• International Area Studies Review
• /
• v.14 no.2
• /
• pp.25-40
• /
• 2010

Accurate forecasting of volatility is of considerable interest in financial volatility research, particularly in regard to portfolio allocation, option pricing and risk management because volatility is equal to market risk. So, we attempted to delineate a model with good ability to forecast and identified stylized features of volatility, with a focus on volatility persistence or long memory in the Australian futures market. In this context, we assessed the long-memory property in the volatility of index futures contracts using three conditional volatility models, namely the GARCH, IGARCH and FIGARCH models. We found that the FIGARCH model better captures the long-memory property than do the GARCH and IGARCH models. Additionally, we found that the FIGARCH model provides superior performance in one-day-ahead volatility forecasts. As discussed in this paper, the FIGARCH model should prove a useful technique in forecasting the long-memory volatility in the Australian index futures market.