Residential electrical demand forecasting in very small scale: An evaluation of forecasting methods

Abstract

Abstract-Applications such as generator scheduling, household smart device scheduling, transmission line overload management and microgrid islanding autonomy all play key roles in the smart grid ecosystem. Management of these applications could benefit from short-term load prediction, which has been successfully achieved on large-scale systems such as national grids. However, the scale of the data for analysis is much smaller, similar to the load of a single transformer, making prediction difficult. This paper examines several prediction approaches for day and week ahead electrical load of a community of houses that are supplied by a common residential transformer, in particular: artificial neural networks; fuzzy logic; auto-regression; autoregressive moving average; auto-regressive integrated moving average; and wavelet neural networks. In our evaluation, the methods use pre-recorded electrical load data with added weather information. Data is recorded from a smart-meter trial that took place during 2009-2010 in Ireland, which registered individual household consumption for 17 months. Two different scenarios are investigated, one with 90 houses, and another with 230 houses. Results for the two scenarios are compared and the performances of the evaluated prediction methods are discussed.

Multiple energy carrier optimisation with intelligent agents

“…Deviation requests 10 were initiated by the Commercial Aggregator agent, based on typical forecasting errors. An average load forecasting 11 error of 3.455% is reported in [44]. For the studied system, this equates to 3.42 kW for electricity and 6.08 kW for gas.…”

“…The optimisation is performed by varying the power inputs and/or 43 the dispatch factors. The objective function of the optimisation problem is given in (2) below, where 44 indicates a factor relating to the objective, as described in (a), (b) and (c) below: 45 Three different objectives are considered in the 1 optimization process: 2 a) Minimize total energy input: Minimize total input 3 energy consumption for the given loads, improving 4 overall hub efficiency. In this case, = 1 at all 5 times.…”

Energy storage day-ahead scheduling to reduce grid energy export and increase self-consumption for micro-grid and small power park applications

“…These are an ARIMA model with lags of 48 samples, and feed-forward Neural Network with 48 output nodes and 49 input nodes as used in [16], model based on the Gaussian Process (GP) and Gradient Boost Machine (GBM) used by [17], Ensemble Forecast (combination of all used demand forecasting methods by simple unweighted average) [18,19] and finally Persistent forecasting.…”

A hybrid approach to very small scale electrical demand forecasting

“…On such a small scale, previous results have shown the advantages of different forecasting methods, such as artificial neural networks (ANN), neuro-fuzzy (NF), auto-regressive integrated moving average (ARIMA); neither of the approaches clearly outperformed the others [11]. In the particular test case neural networks have proven good at evening peak estimation, fuzzy logic performed well on the morning peak estimation, and auto-regressive techniques have shown good overall results without particular accuracy in critical areas when compared to the previous approaches.…”

“…According to the work presented in [11], four methods have proven superior at different intervals of the day: ANNs, wavelet neural networks (WNNs), NF and ARIMA. While all methods rely on historical load information, some of them include weather information such as temperature and humidity for the day-ahead electrical load prediction.…”