Akihiko Kawashima

Model Predictive Charging Control of In-Vehicle Batteries for Home Energy Management Based on Vehicle State Prediction

Thanks to recent development of reciprocal communication networks and electric power management infrastructure, an energy management system, which can automatically regulate supply–demand imbalances under conditions of the users’ convenience and economy, is attracting great attention. On the other hand, finding of new usage of the batteries employed in electric vehicles and plug-in hybrid vehicles is recognized as one of key issues to realize the sustainable society. In addition, development of vehicle to X technology enables us to use the electric power of in-vehicle batteries for various purposes. Based on these backgrounds, this paper presents an integrated strategy for charging control of in-vehicle batteries that optimizes the charge/discharge of in-vehicle batteries in a receding horizon manner exploiting the predicted information on home power load and future vehicle state in the household. The prediction algorithm of future vehicle state is developed based on semi-Markov model and dynamic programming. In addition, it can also be implemented in receding horizon manner, i.e., the predicted vehicle state is updated at every control cycle based on the new observation. Thus, the harmonious combination of stochastic modeling/prediction and MPC in real-time home energy management system is one of the main contributions of this paper. Effectiveness of the proposed charging control is demonstrated by using an experimental testbed.

Energy management systems based on real data and devices for apartment buildings

Aggregators in smart grid are business operators providing services to aggregate and visualize electricity information, balance demand and supply, reduce energy consumption, and do the other activities with Energy Management Systems (EMSs). An aggregator for Apartment Building Energy Management Systems (Apartment-BEMSs) balances demand and supply considering all households in its building. The buildings with the BEMS are generally equipped with stationary batteries, PV generators, or the other devices to reduce energy cost and wasteful consumption. On the other hand, energy storages such as high capacity batteries are necessary to balance demand and supply in power grid installed a large amount of renewable energy such as solar power or wind power. However, these installation cost is expensive. Therefore alternative storages are required in some cases. As an approach, Vehicle-to-X (V2X) is carried out promptly. The V2X is a utilization of in-vehicle batteries installed in Electric Vehicles (EVs) and Plug-in Hybrid Electric Vehicles (PHEVs) instead of stationary batteries. Then, the authors target constructing an effective EMS for apartment buildings with Vehicle-to-Home (V2H) systems. As the latest work, the authors propose an improved apartment BEMS considering characteristics of actual devices. Through computational experiments based on real data observed and measured in actual society, the authors show the effectiveness of the improved BEMS.

Vehicle Fleet Prediction for V2G System - Based on Left to Right Markov Model.

The regulations for internal combustion vehicles, CO2 or NOx emission or noise and so on, are strengthened. Therefore EV (electric vehicle)s market is expanding. The amount of EV get more, the amount of electric get more and the impact for grid that are voltage fluctuation and frequency fluctuation is concerned. V2G (Vehicle to Grid) can solve this problem, but it has a constraint that EV’s battery can be used during it parked. So as the basic technology, the prediction the vehicles’ state that is driving or parked is important. In this research, machine learning algorithm for predicting vehicle fleets states is developed. The data for study and test is obtained by person-trip survey. The algorithm is based on left to right Markovmodel. The states are stay or drive from an area to an area. Future state probability is predicted using the latest observed state and state transition probability. As the result, the prediction error of stay is less than the prediction error of drive. Therefore study data and test data are separated into sunny day and rainy day, the prediction error becomes less.

Electric Vehicle Charge-Discharge Management for Utilization of Photovoltaic by Coordination between Home and Grid Energy Management Systems

This paper proposes an electric vehicle (EV) charge-discharge management framework for the effective utilization of photovoltaic (PV) output through coordination based on information exchange between home energy management system (HEMS) and grid energy management system (GEMS). In our proposed framework, the HEMS determines an EV charge-discharge plan for reducing the residential operation cost and PV curtailment without disturbing EV usage for driving, on the basis of voltage constraint information in the grid provided by the GEMS and forecasted power profiles. Then, the HEMS controls the EV charge-discharge according to the determined plan and real-time monitored data, which is utilized for mitigating the negative effect caused by forecast errors of power profiles. The proposed framework was evaluated on the basis of the Japanese distribution system simulation model. The simulation results show the effectiveness of our proposed framework from the viewpoint of reduction of the residential operation cost and PV curtailment.

Identification of time-varying parameters in Gipps model for driving behavior analysis

This paper proposes a new method to analyze driver behavior. Analysis of the behavior is done through the observation of the time-evolution of parameters of simple driver models. The behavior analysis is decomposed in two steps. First the driver model have to be selected or designed to represent the average behavior of a large sample of drivers. Then personal drivers behavior evolution can be analyzed over the time. To be able to identify time-varying non-linear hybrid model parameters, an iterative metaheuristic method based on particle optimization and moving average filtering has been created. This method enables to identify parameters of any model type while filtering the parameter time-variation based on the possible parameter dynamics. This methods also enables to interpolate parameters values while model output values are occluded. Demonstration of the identification algorithm efficiency with Gipps car-following driver model is done based on theoretical examples, and time-evolution of parameter are identified from real-world measured data.

Generator of learning data for the TSPs based on the visiting order of the cities on convex hull

The optimal tours of the traveling salesman problems( TSPs) in two dimensional Euclidean space have the characteristics in the visiting order of the cities on the convex hull. Based on this characteristics, the TSPs can be replaced into some shortest Hamiltonian path problems(SHPPs) of which solutions assemble a tour. This reduction is enabled by the calculation of convex hull and the classification of the cities not on the convex hull into the subsets of cities which construct SHPPs. This procedure means that the TSPs are equivalent to the classification problems, which leads to be able to apply existing methods of machine learning to the TSPs. We show that the teaching data for machine learning are available as the optimal classifications in the instances of which the optimal tour has been found.