Forecasting of solar irradiance with just‐in‐time modeling

Abstract

PV power output mainly depends on solar irradiance, which is affected by various meteorological factors. Thus, it is required to predict solar irradiance in the future for the efficient operation of PV systems. In this paper we develop a novel approach for solar irradiance forecasting, in which we combine the black-box model (JIT modeling) with the physical model (GPV data). We investigate the predictive accuracy of solar irradiance over a wide controlled area of each electric power company by utilizing measured data from 44 observation points throughout Japan provided by JMA and 64 points around Kanto provided by NEDO. Finally, we propose an applied forecasting method of solar irradiance to deal with difficulties in compiling databases. We also consider the influence of different GPV default times on solar irradiance prediction.We use MSM-GPV data as the input data for JIT modeling. MSM-GPV [12, 13] is a meteorological numerical forecast model based on a physical model. It is calculated in intervals of 3 hours, and consists of hourly physical predicted values such as air pressure, temperature, cloudiness, and so on for 33 hours in advance at each node of a 5-km grid throughout Japan. We then perform prediction of the hourly integrated next-day solar irradiance by using the MSM-GPV calculated at midnight (00:00). We also per-

Improvement of prediction interval estimation algorithm with just-in-time modeling for PV system operation

“…Neighbors are selected based on the Euclidean distance information between query and past data point. The distance is calculated based on the weight of the each input shown in reference [3]. The nwnber of neighbors is detennined hourly using JIT modeling and the past database as references [2] and [5] indicates.…”

Effect of PV Output and Load Power Forecast Error on Operation Design of PV System with Storage Battery

“…There is the case in which various grid point value (GPV) data [2] (temperature, atmosphere, and cloud amount and so on) are used. If the data are recorded simultaneously with the forecast objective data (PV output and load power), we can derive past forecast values and validate the quantitative effects of forecast errors on battery operation.…”

“…PV system cost + Battery system cost Annual electricity purchase reduction + Annual electricity sales (1) CO 2 PT = CO 2 emissions generated in manufacture of PV and battery Annual electricity purchase reduction × CO 2 emission coefficient (2) Both objective functions are calculated as annual simulation results, and we aim to minimize these functions. We seek Pareto-optimal solutions by using a multiobjective genetic algorithm (MOGA) [8].…”

Estimation Prediction Interval of Solar Irradiance Based on Just‐in‐Time Modeling for Photovoltaic Output Prediction

“…The distance measure, input selection method, and input weighting method were the same as in the previous investigation [3]. In this method, an input vector at prediction time is first specified with respect to a database that contains past input/output pairs C⃝ 2016 Wiley Periodicals, Inc. of a target system (history database).…”