Optimization Framework for Short-Term Control of Energy Storage Systems

2019 IEEE Milan PowerTech

Werner-Kytölä

Gümrükcü

2019

Self Cite

House and building energy management systems (HEMS) are becoming key when it comes to assure grid stability and to offer flexibility. At the same time, energy systems technology has evolved to enable energy storage systems and electric vehicles to be managed together with local generated energy taking into consideration the preferences of the household owner. Contributing to this tendency, this work presents a stochastic optimization platform (SOFW) for optimal control using dynamic programming and stochastic optimization models. A stochastic optimization model involving a household composed of photovoltaics, energy storage system and an electric vehicle is designed and tested within SOFW. The uncertainties of the plugin time and state of charge of the battery of the electric vehicle are modeled using a Markovian process and a Monte-Carlo simulation. The results showed that the proposed stochastic optimization model can be solved using dynamic programming and deployed as a continuous optimal control within SOFW. The system will be deployed shortly in Italy within one use case of the Storage4Grid (S4G) project.

Section: Use Case Specificationmentioning

confidence: 99%

Stochastic optimization framework for online scheduling of an EV charging station in a residential place with photovoltaics and energy storage system

2019 IEEE Milan PowerTech

Werner-Kytölä

Gümrükcü

2019

Self Cite

“…Even though the usage of RNNs for regression is not new on its own, to the best of our knowledge, so far there is no previous work related to the problem of exploiting LSTMs for incremental learning in the context of continuous energy load prediction as input for optimization problems. Furthermore, the upcoming model is additionally used as an input for a subsequent optimization framework as presented in [4].…”

Section: Rnn (Lstm) For Energy Load Predictionmentioning

confidence: 99%

“…In this context, EMS provide short-term control of ESS and renewable generation management by calculating the optimal control action for the power flow in the next time step. The open-source optimization framework (OFW), that we presented in a previous work [4] uses optimization models as basis for the energy management and links predictions to these models. However, prediction uncertainties can affect the performance of the optimization depending on how effective and accurate they are.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Incremental Deep-Learning for Continuous Load Prediction in Energy Management Systems

2019 IEEE Milan PowerTech

Puri

Graß

et al. 2019

Self Cite

In this work, we introduce load prediction as continuous input for optimization models within an optimization framework for short-term control of complex energy systems. In this context, we investigated long short-term memory (LSTM) models for load prediction, because they allow incremental training in an application with continuous real-time data and have not been used in other works for continuous load prediction to our knowledge. The test and evaluation were realized using data sets of real residential data from different locations in different time resolution-hourly and minutely. Accordingly, we tested different recurrent neural network (RNN) parameters of the model such as the number of layers, the number of hidden nodes, the inclusion of regularization, and dropout in order to find the optimal LSTM configuration for our continuous load prediction application. Besides, we analyzed the quality of the LSTM algorithm by comparing it in continuous mode with the baseline model and in batch mode with the statistical model ARIMA. Training and prediction time, as well as the error stabilization time were parameters used for the evaluation. The results showed that LSTM algorithms are highly promising for integrating continuous load prediction with incremental learning.

“…The optimization for the test scenario is implemented within the open‐source optimization framework created within the EU‐Storage4Grid project and registered under the LinkSmart brand by Fraunhofer . The framework enables prototyping and flexible definition of optimization problems according to changing conditions, and clustering of components in either real‐time or discrete applications.…”

Section: System Modeling and Implementationmentioning

confidence: 99%

Distributed model predictive control strategies for coordination of electro‐thermal devices in a cooperative energy management concept

Stoyanova

Gümrükcü

Optim Control Appl Methods

et al. 2019

Self Cite

Summary This work discusses three promising strategies for the compensation of deviations within the online phase of a cooperative energy management concept. Whereas in the planning phase, hourly schedules for the active electro‐thermal devices are negotiated on a daily basis, the second phase tracks the schedule, including forecast updates, and compensates the deviations. The latter online compensation is realized with a distributed and hierarchical‐distributed model predictive control strategies, which are compared with continuous rescheduling with receding horizon and evaluated in terms of computational and optimization performance and scalability. The optimization is implemented and deployed using an open‐source optimization framework, which automatically manages optimization control, integration of renewable generation and load forecasts, multi‐instantiation and communication over an internet of things communication protocol. We address the case of significant deviations from the day‐ahead forecast and the tracking of an obsolete schedule and suggest a combined method, which combines the advantages of rescheduling and a distributed MPC strategy. The method applies a dynamic threshold to evaluate the deviation trend over time and to decide if compensation is still feasible or if a rescheduling should be triggered. The methods are evaluated based on several performance indicators such as residual deviation, number of switching events and computation and communication requirements and scalability, and offer a methodology for the control design of systems of different dimensions.