Refine
Document Type
Language
- English (4)
Has Fulltext
- no (4)
Publication reviewed
- begutachtet (4)
Keywords
- Machine learning (1)
- Microgrid (1)
- Modeling (1)
- Smart city (1)
- Thermostatic load (1)
- nZEB (1)
Institute
Increasing shares of renewable energy sources in combination with rising popularity of demand response applications and flexibility programs forces higher awareness for production and consumption balancing. Accurate models for forecasting are not just necessary for PV- or wind power sources in smart cities, but also the prediction of loads respectively consumption, which can be based on time series analysis or machine learning methods. Three of those methods, namely a linear regression (LM), a long short-term memory network (LSTM) and a neural network model (NN), have been selected to see their performance on predicting the load of a large smart city on the example of the Estonian electricity consumption data. Hourly data of the year 2019 was used as training data to predict the first 20 days of 2020. For this kind of prediction, the LM showed the lowest root mean square error (RMSE) and had the lowest computational time. The neural network was slightly less accurate. The LSTM showed the worst performance in terms of accuracy and computational time. Thus, LSTM is not the preferred method for this kind of prediction and the recommendation for forecasting such loads would be a LM because the RMSE and computational effort needed are lower than for a NN
Impact of Load Matching Algorithms on the Battery Capacity with different Household Occupancies
(2019)
Due to an increasing use of renewable energy sources in the power grid, it is of high importance to balance supply and demand for grid utilities and microgrid operators. If there are mismatches in the balancing, microgrids with islanded operation capabilities would be preferrable. In islanded mode, nearly zero energy buildings commonly use a stand-alone photovoltaics power supply with a battery storage. A battery storage is expensive and the capacity in case of off-grid operation depends on the electricity consumption of the dwelling's occupants. Using thermostatically controlled appliances like a freezer, water heater and space heating as additional storage systems can reduce the capacity of the battery storage system or increase the operation time in islanded mode for a fixed battery size. This paper analyzes the battery capacity dependency both on the control algorithms for the thermal storages and on the occupancy of the dwelling. Possible battery reductions for different selected occupancies are presented in this work by comparing the simulation results of different load matching algorithms to each other and between the different occupancies. The analysis of those results enables recommendations on the most suitable algorithm for most occupancy scenarios of an existing dwelling with respect to a minimized battery capacity. This can be particularly useful, for example, for dwelling and apartment owners who are renting out dwellings.
Due to an increasing share of renewable energy sources the balancing of energy production and consumption is getting a lot of interest considering future smart grids. In this context, many investigations on demand-response programs are being conducted to achieve flexibility from different energy storages and loads. As space heating is an important schedulable load for flexibility simulations, there are different modelling approaches due to its interdisciplinary nature. Models can be built from the civil engineering or electrical engineering point of view, depending on the computational expense and accuracy level. Scheduling optimizations need a lot of simulations, preferably with computationally light models. Thus, this work will use a computationally light neural network load prediction model for space heating which is based on a detailed civil engineering model. Simulations with different scheduling times were conducted to see the long- and short-term effects of the demand response action. Results show, that applying the same demand response action at different times results in different behaviors of the system resp. energy consumption, which requires further studies for developing optimized scheduling methods.
Balancing the energy production and consumption is a huge challenge for future smart grids. In this context, many demand-side management programs are being developed to achieve flexibility from different loads like space heating. As space heating models for flexibility simulations are an interdisciplinary field of work, complex civil engineering thermal models need to be combined with complex electrical engineering control simulations in different software frameworks. Traditionally used methods have shortcomings in one of those two domains as the publications that provide complex control strategies for demand response are lacking complex thermal models and vice versa. Co-simulations overcome this problem but are computationally expensive and have compatibility limitations. Thus, the aim of this work is to develop a methodology for designing space heating/cooling models, intended for positive energy district- or smart city simulations, which provide high accuracy at low computational expense. This could be achieved by synthesizing neural network object models from IDA-ICE civil engineering models in Matlab. These machine learning models showed improvements of more than 30% in different error metrics and a simulation time reduction of more than 80% compared to other methods, making them suitable for use in microgrid simulations, including flexibility analyses.