Stefan Ungureanu

Short-term load forecasting predetermines how power systems operate because electricity production needs to sustain demand at all times and costs. Most load forecasts for the non-residential consumers are empirically done either by a customer’s employee or supplier personnel based on experience and historical data, which is frequently not consistent. Our objective is to develop viable and market-oriented machine learning models for short-term forecasting for non-residential consumers. Multiple algorithms were implemented and compared to identify the best model for a cluster of industrial and commercial consumers. The article concludes that the sliding window approach for supervised learning with recurrent neural networks can learn short and long-term dependencies in time series. The best method implemented for the 24 h forecast is a Gated Recurrent Unit (GRU) applied for aggregated loads over three months of testing data resulted in 5.28% MAPE and minimized the cost with 5326.17 € compared with the second-best method LSTM. We propose a new model to evaluate the gap between evaluation metrics and the financial impact of forecast errors in the power market environment. The model simulates bidding on the power market based on the 24 h forecast and using the Romanian day-ahead market and balancing prices through the testing dataset.

In recent times, globalization and energy efficiency policies have further required temperature control applications in different daily activities, especially with the greenhouse effect. In residential buildings, water and living space heating consumes most of the primary energy. The paper presents a method to keeping the comfort temperature in living space by using an approach based on an artificial intelligence technique, namely fuzzy logic. The support for the proposed method is an electrical hot-water radiator heating system with a tankless heater. The fuzzy logic controller makes a continuous regulation of the indoor temperature by adapting the temperature of the heating agent regarding the indoor and exterior temperature. To validate and test the proposed method, MATLAB-Simulink was employed. The comparison with an On/Off controller that is the most used control device due to financial benefits showed the efficiency of the fuzzy logic controller.

Reliable real-time power measurement data represents a vital role in understanding energy behavior from a consumer standpoint. Energy efficiency is directly influenced by monitoring electric parameters. With collected live data, the challenge becomes how to organize and analyze datasets of information. The purpose is to design and build a low-cost electronic monitoring device using Arduino and IoT for electricity monitoring. The main capabilities are to measure, record, and transmit using wi-fi connection all the essential electric parameters in a web-based application. Monitored electric parameters are voltage, current, power factor, phase angle, power, and energy. The developed prototype is versatile in multiple types of low voltage installations, mainly household appliances. Measurements are transmitted remotely in an online database, where the data can be stored and analyzed in real-time. The solution has permanent storage and facilitates consultation of past and current records. The market approach is possible because of the low-cost of the device. A commercial version would be suitable and could easily integrate Smart Grid applications. Using real-time energy data and monitoring software, consumers can measure high consumption loads, reduce base loads, evaluate peak and off-peak hours and identify irregularities. optimize usage scheduling and track energy-saving targets and goals more accurately.

Electricity is a commodity traded from the market through various products, but this level of accessibility represents a great opportunity and challenge for energy suppliers. Storing electricity in large quantities is not yet possible, meaning that the delivery to the customers needs to happen at the time of production. Energy suppliers must have a reliable strategy to buy or sell on the power market. The strategy efficiency determines if the supplier can optimally balance the portfolio, with the lowest impact on the overall costs. Energy imbalance is the difference between energy forecast and real consumption/ generation per portfolio. The cost of imbalances can negatively impact the margin of a portfolio that aggregates production, consumption, and trading. In this paper, the authors present complex situations for an energy supplier who needs to balance a portfolio consisting of consumption (industrial and commercial consumers), generation (photovoltaic and hydropower plants), and trading. The goal is to minimize market risk through excesses and shortfalls as much as possible and to reduce balancing energy expenses. Each market participant operating in the electricity market is obliged to have a contract with a BRP (Balancing responsible parties) or be one. The greater the number of participants in a group, the lower their balancing energy expenses, and keeping track of fluctuations both within a balanced group and with other balance groups can reduce costs.

The paper presents the results obtained at the end of a study whose main goal was to answer the question: What will happen in a distribution network if there will be a large injection of power from a big number of small PV units? This situation is typical for rural distribution networks where the domestic consumers have their own PV generation units, thus becoming prosumers. The study began with the modeling of the usual domestic consumer from rural regions. After this step being finalized, the implementation of the distribution network in a simulation software followed. For this stage MatLab/Simulink was used. Thus, different situations were simulated in order to determine a full understanding of the possible problems that can appear, and consequently to be able to take the proper actions to minimize the negative impact. The results of the simulations show that in a typical summer day, when the PV units produce electricity at 100% of their capacity and the loads’ consumption is the lowest, the voltage in the point of injection in distribution network can reach values that will determine the faulty operation of the pronsumers due to the protection apparatus’ activation. Various configurations for the grid have been considered. Also, the study points out that integrating PV units by residential consumers can also help the electric grid to better maintain voltage variations in cases of larger distribution lines.

Short-term electricity load forecasting has attracted considerable attention as a result of the crucial role that it plays in power systems and electricity markets. This paper presents a novel hybrid forecasting method that combines an autoregressive model with Gaussian process regression. Mixed-user, hourly, historical data are used to train, validate, and evaluate the model. The empirical wavelet transform was used to preprocess the data. Among the perturbing factors, the most influential predictors that were recorded were the weather factors and day type. The developed methodology is upgraded using a novel closed-loop algorithm that uses the forecasting values and influential factors to predict the residuals. Most performance indicators that are computed indicate that forecasting the residuals actually improves the method’s precision, decreasing the mean absolute percentage error from 5.04% to 4.28%. Measured data are used to validate the effectiveness of the presented approach, making it a suitable tool for use in load forecasting by utility companies.

Integration of industrial consumers into the smart grid concept can be facilitated by optimizing load forecasting for industrial consumers. Minimizing forecast errors can improve the supplier-consumer relationship by reducing balancing costs and anticipate possible network faults. The present paper aims to research the efficiency of machine learning applied for industrial load. The dataset consists of hourly recorded values for electricity consumption generated by a meat processing facility. In the context of installing complex monitoring systems with high frequency recording intervals, huge amounts of data will be generated that require detalied analysis and real time processing, otherwise the investments in the smart grid are not justified, consequently disfavouring the development and digitization of electrical networks. Integration of the industrial consumer into the smart grid concept can be applied in great detail at large industrial consumers through robust forecasting. Forecasting the energy behaviour of a industrial consumer is a difficult task, high forecasting errors have been obtained due to the unpredictability of the consumer.

Integration of industrial consumers into the smart grid concept can be facilitated by optimizing short and very short-term forecasts of load curves for industrial consumers. Minimizing forecast errors can improve the supplier-consumer relationship by reducing supplier balancing costs, which can lower the energy bill, and anticipate possible network faults or network congestion which means improved operation across the entire power system. The present paper aims at demonstrating the usefulness of energy forecasting by using machine learning methods. In the context of installing monitoring systems with load curve recording data at short intervals of time, huge amounts of data are obtained that need to be capitalized in real time (automated), otherwise the investments necessary to modernize the way in which the energy is used are not justified, consequently disfavouring the development and digitization of electrical networks. Integration of the industrial consumer into the smart grid concept can be applied in detail at large industrial consumers where forecasts can be made on certain areas inside factories or electric equipment with high installed electric capacity. The use of smart grid technologies removes the current limits for monitoring, processing and communication between consumption and production. To be able to capitalize on the forecasts, a very important aspect is the automation of their realization and quick access to the decision makers who may be human or other smart grid integrated systems. That is why the present paper proposes the use of machine learning methods in the short and very short forecast, and the results will be compared with the classical forecasting methods. Hourly load curves from the industrial consumers will be used for forecasts.