A Multi-Layer Data-Driven Security Constrained Unit Commitment Approach with Feasibility Compliance

Security constrained unit commitment is an essential part of the day-ahead energy markets. The presence of discrete and continuous variables makes it a complex, mixed-integer, and time-hungry optimization problem. Grid operators solve unit commitment problems multiple times daily with only minor changes in the operating conditions. Solving a large-scale unit commitment problem requires considerable computational effort and a reasonable time. However, the solution time can be improved by exploiting the fact that the operating conditions do not change significantly in the day-ahead market clearing. Therefore, in this paper, a novel multi-layer data-driven approach is proposed, which significantly improves the solution time (90% time-reduction on average for the three studied systems). The proposed approach not only provides a near-optimal solution (<1% optimality gap) but also ensures that it is feasible for the stable operation of the system (0% infeasible predicted solutions). The efficacy of the developed algorithm is demonstrated through numerical simulations on three test systems, namely a 4-bus system and the IEEE 39-bus and 118-bus systems, and promising results are obtained.

A Review on the Unit Commitment Problem: Approaches, Techniques, and Resolution Methods

Optimizing the schedule of thermal generators is probably the most important task when the operation of power systems is managed. This issue is known as the unit commitment problem in operational research. It has been profoundly studied in the literature, where several techniques have been proposed to address a computationally tractable solution. In turn, the ongoing changes of paradigms in energy markets focus the attention on the unit commitment problem as a powerful tool to handle new trends, such as the high renewable energy sources penetration or widespread use of non-conventional energy-storage technologies. A review on the unit commitment problem is propo- sed in this paper. The easy understanding of the diverse techniques applied in the literature for new researchers is the main goal of this state-of-art as well as identifying the research gaps that could be susceptible to further developments. Moreover, an overview of the evolution of the Mixed Integer Linear Programming formulation regarding the improvements of commercial solvers is presented, according to its prevailing hegemony when the unit commitment problem is addressed. Finally, an accurate analysis of modeling detail, power system representation, and computational performance of the case studies is presented. This characterization entails a significant development against the conventional reviews, which only offer a broad vision of the modeling scope of their citations at most.

Decision Support Application for Energy Consumption Forecasting

Energy consumption forecasting is crucial in current and future power and energy systems. With the increasing penetration of renewable energy sources, with high associated uncertainty due to the dependence on natural conditions (such as wind speed or solar intensity), the need to balance the fluctuation of generation with the flexibility from the consumer side increases considerably. In this way, significant work has been done on the development of energy consumption forecasting methods, able to deal with different forecasting circumstances, e.g., the prediction time horizon, the available data, the frequency of data, or even the quality of data measurements. The main conclusion is that different methods are more suitable for different prediction circumstances, and no method can outperform all others in all situations (no-free-lunch theorem). This paper proposes a novel application, developed in the scope of the SIMOCE project (ANI|P2020 17690), which brings together several of the most relevant forecasting methods in this domain, namely artificial neural networks, support vector machines, and several methods based on fuzzy rule-based systems, with the objective of providing decision support for energy consumption forecasting, regardless of the prediction conditions. For this, the application also includes several data management strategies that enable training of the forecasting methods depending on the available data. Results show that by this application, users are endowed with the means to automatically refine and train different forecasting methods for energy consumption prediction. These methods show different performance levels depending on the prediction conditions, hence, using the proposed approach, users always have access to the most adequate methods in each situation.

A New Approach in Short-Term Prediction of the Electrical Charge with Regression Models A Case Study

The accuracy of forecasting of electrical load for the electricity industry has a vital significance in the renewal of economic structure as well as various equations including: purchasing and producing energy, load fluctuation, and the development of infrastructures. Its short-term forecasting has a significant role in designing and utilizing power systems and in the distribution systems and having a variety of systems used to maintain security potentials for the system. In this paper, we attempted to carry out a short-term forecasting of electrical distribution company in west Azerbaijan state in Iran's electricity in a few days on the basis of regression multi linear model. This forecasting which was done during a three-day period is and categorized weekdays into three groups including working days, weekends, and holidays was carried out in an hourly manner. This model regardless of parameters like humidity, wind velocity, daylight time, etc. by minimizing the forecasting error managed to maximize the reliability of the results as well as the safety potential of the system. In this model the only influential parameter on the forecasting was the reliance of the forecasting day on previous days. The main purpose of the present study was to maximize the accuracy and reliability of forecasting for certain days (religious holidays, national holidays …). In this paper, the authors managed to decrease the error of forecasting for particular and regular off days to a great extent.

Neural-network prediction with noisy predictors

Very often the input variables for neural-network predictions contain measurement errors. In particular, this may happen because the original input variables are often not available at the time of prediction and have to be replaced by predicted values themselves. This issue is usually ignored and results in nonoptimal predictions. This paper shows that under some general conditions, the optimal prediction using noisy input variables can be represented by a neural network with the same structure and the same weights as the optimal prediction using exact input variables. Only the activation functions have to be adjusted. Therefore we can achieve optimal prediction without costly retraining of the neural network. We explicitly provide an exact formula for adjusting the activation functions in a logistic network with Gaussian measurement errors in input variables. This approach is illustrated by an application to short-term load forecasting.