IoT, AIというキーワードが毎日の生活の中で耳にすることが多くなりました。これらの技術の実用化という観点では，情報処理分野，ロボット分野等で様々な適用例が発表されています。一方，本特集号の対象とするエネルギー分野においては他の分野と比較して，アプリケーションレベ

This paper describes the latest trends in AI research and the challenges and possibilities for applications in the energy field. As for AI research trends, the latest technological trends such as E2E learning and the fusion of inductive and deductive reasoning are introduced, and new issues such as AI life-cycle and fairness issues are described. In the second half, I will discuss the possibility of utilizing AI in four important fields: equipment maintenance, smart grid operation, energy market, and resilience.

There is a growing need to use photovoltaic (PV) technology to mitigate global warming and the depletion of fossil fuels. However, the high network penetration of PVs potentially lessen the stability and the reliability of electrical power systems in various ways. Under its high network penetration, monitoring of the unmeasurable power outputs of PVs is needed for the proper operations of an electrical power system. To suffice this need, a novel method was proposed to estimate the PV power output of the system using the measured power flow and solar radiation intensity estimated from satellite image in our previous work. However, it occasionally causes large estimation errors of the PV power output based on the wrong estimation of parameters induced by a large data sampling interval. To address this issue, in this paper, we improved our proposed method by introducing an parameter estimation procedure to exclude outliers of estimated values and increase the number of data used for estimation. We confirmed its enhanced accuracy using data observed from The Kansai Electric Power Co., Inc.

This paper proposes an energy disaggregation method for the total energy consumption of a commercial building which only has a limited number of facility status monitor points. The method can estimate and identify the energy consumption each monitored and unmonitored facilities utilize through the statuses of monitored facilities. In order to solve a linear regression model with unmonitored facilities, the linear basis function model is applied to the unmonitored facilities. The total energy consumption of unmonitored facilities can be expressed by linear combination of daily periodic Gaussian basis functions. The regression parameters of monitored facilities and the basis functions are determined by the total energy consumption of the building and the operation statuses of monitored facilities. The method is applied to the commercial building which has six monitored air conditioning outdoor units as well as unmonitored facilities including lighting and office automation equipment. The experiment result shows the usefulness of the proposed method. The variable energy consumption of unmonitored facilities can be derived by the basis functions and the estimate accuracy of disaggregated energy consumption is applicative for the energy management. This paper also considers the effectiveness of regularization technique such as Lasso and Ridge.

In order to operate heat source systems effectively, it is necessary to make clear their optimum operations which based on the actual performances of the equipment. In past study, the target is limited to a part of the system and also the study hasn't enough about the influence that the change of the actual performance on the optimal operation. Therefore, this paper propose regression modeling methods based on actual operation data for each component of heat source systems first. The model for refrigerators is secured versatility by applying piecewise linear regression method. Then, optimum operations including the start / stop of refrigerators are shown through numerical simulations using the proposed models. It showed that the optimum cooling water temperature and the start / stop changed due to the actual performance change of the refrigerators. Finally, this paper describes the test results applying the proposed method to actual machine controls, and 3.6% energy saving effect and the usefulness of the proposed method are verified.

The purpose of this paper is to propose a power management method in a hospital for a combination of emergency generators (EGs) with photovoltaic power generation (PV). The power balance of the grid not only influences the droop control for the generator but also the output fluctuations of the PV. Frequency control and a load control by a load prediction are necessary for the system grid combined with EGs and PV in an islanded operation mode. When the PV system is installed in the grid, the EG system should distribute power to small generators, the reason is because when the EG is too large, the power balance cannot be maintained to stabilize the system frequency in all power ranges. Since the distributed generation system needs the demand for each generator, it is important to predict the load. This paper proposes a new method for power energy management for stabilization with the islanded operation mode in a hospital power grid with load prediction using deep learning. The proposed method can realize operation by using a power emulator with the hospital power grid model. The verification of results show that the power emulator is effective in the energy management strategies.

This study shows performance evaluation of a data-driven prediction model for energy consumption of electric vehicles (EVs) on expressways by using data obtained from a demonstration experiment where the authors and testers from the public actually drove EVs with the proposed EV charging navigation system in 2018. By comparing actual energy consumption and predicted one, we have found that the predicted model has a good prediction accuracy, whereas predicted results tend to deviate from actual ones in the region where actual energy consumption is 15 kWh or more. We also confirmed a possibility that high velocity may degrade prediction performance. With respect to error of state of charge (SOC), most of the SOCs calculated by the prediction model do not have more than 20% error, which may lead to run-out of electricity of EV. Although the prediction model is effective in respect of SOC error, the result indicates a necessity that automatic grouping and prediction model based on such groups should be studied as a future work.

In order to cope with the shortage of electrical technicians, it is expected to the improvement of work efficiency by the introduction of advanced technology, such as the Internet of Things (IoT) and Artificial Intelligence (AI). In the safety inspection of electrical facilities, insulation monitoring is expected to systemize correspondence judgment based on data such as measured leakage current.

In this study, from the data of the security business core system such as leakage current value measured at the periodic inspection and weather data, we created some models to predict the leakage current measured when the abnormality warning was issued and per customer and the presence or absence of alarms on the next day. The combination of the best explanatory variables makes the model more accurate. Variable importance analysis using Random Forest (RF) was performed to find variables that are important for each objective variable. This analysis shows that the accuracy of the prediction model of the leakage current is the highest when the explanation variable is the data of the security business core system, the weather data, the presence / absence of alarms on the previous day. Other predictive models need further verification. As a result of variable importance analysis, We found out that the leakage current value at the periodic inspection and the time of alarm are important for all purpose variables.

Speckle reduction method for laser diode based displays was newly proposed. The method was based on temporal change of the active layer temperature which could be controlled by optimizing the driving condition of the laser diode under pulse. The red LD consisted of AlGaInP has low thermal conductivity and small energy difference between the active layer and the cladding one in the conduction band, resulting in the large temporal change of the lasing spectrum under the adequate driving condition in pulse. It indicates the speckle contrast could be controlled by changing the condition. The speckle contrast dependence on the driving condition of the high-power 638-nm lateral single mode laser diode was experimentally studied. In the experiment, not only the conventional square waveform but also the saw one were used. The speckle contrast under the adequate conditions for both waveforms was reduced approximately 25% compared with that under CW. In particular, the saw one might be expected as a very robust driving condition for the speckle reduction. This can be used with other speckle reduction method such as moving screen, moving diffuser and so on. In this sense, the newly proposal may be the speckle reduction method for “last one mile”.

In this paper, we propose a new adiabatic logic circuit which adds a feedback section using diode to existing EE-SPFAL. In order to investigate the influence on the evaluation index, we simulate an S-box using proposed circuit, and the result is compared with those of the other adiabatic logic circuits. From the simulation results, the evaluation index NED and NSDe meaning scatter of energy consumption are improved compared with the SPGAL and EE-SPFAL.

C-2C D/A converters (DAC) can be designed with fewer unit capacitors than binary weighted capacitive DAC (CDAC), which are advantages in power consumption, operating speed, and circuit area. However, in applications with medium or higher resolution, it needs larger circuit area because an accurate non-integer capacitance ratio is required in order to compensate the influence of the parasitic capacitance at the floating node. In this paper, by configuring C-2C DAC with only a simple integer capacitance ratio, the proposed circuit can be achieved with a smaller circuit area than split CDAC in 9-bit or higher. It also can compensate the variation of parasitic capacitance at floating node by applying the conventional digital correction technique.

Detecting abnormal heart sounds through auscultation is a necessary part of diagnosing heart disease. This paper presents a method for detecting abnormal heart sounds using the Mahalanobis-Taguchi Method. Six features of heart sounds are selected to find changes in cardiac activity. Reference data are created based on these six features calculated from normal cardiac sounds. Mahalanobis distance (MD) from the reference data is used as a measure to estimate abnormalities. The effectiveness of this method is evaluated with 24 clinical cases and yielding correct results in 100% for cardiac murmur except abnormal heart sounds with extra heartbeats. This method may also be used to find a symptom of heart disease by monitoring the trend of MD values in everyday healthcare management.

This paper proposes a controller design method to reduce deterioration of control quality such as instability and degradation of response performance (transient and steady-state responses, etc.) due to packet loss in wireless control systems. In our proposed method, the influence of packet loss to control systems is considered as a disturbance which is an external input causing deterioration of control quality. The proposed method can keep stability and reduce degradation of response performance even if packet loss occurs by considering the characteristics from a disturbance (an external input) to outputs in control design. In particular, this paper clarifies that the influence of packet loss can be modeled as a control-dependent disturbance, and proposes a controller design method that uses this dependency in controller design and does not depend on packet loss rate. Numerical simulation shows advantages of the proposed controller through comparisons with conventional controllers of existing researches.

Model predictive control (MPC) in process control has shown its superior performance for tracking control. However, MPC has been applied mainly on larger scale plants because it requires high-performance computing environment. In this paper, we propose a novel approach for MPC based on formula manipulation and a finite-step prediction filter in order to achieve fast predictive control, which we can implement on edge-devices with low computing resources. We also propose a new approach to enhance regulatory control performance under unobservable disturbance, where a past prediction error filter and robust gain design are introduced. We discuss a convergence conditions and show numerical examples to validate the effectiveness of the proposed method.

The aim of functional electrical stimulation (FES) is restoring a disabled patient's upper motor neuron by directly applying electrical currents to the paralyzed muscles. This paper considers a FES alternate knee bending and stretching trike system with RISE-based control for rehabilitation. The alternate knee bending and stretching mechanism and a rider with two muscle groups (Quadriceps, Hamstrings) are modeled as an Euler Lagrange system based on knee bending and stretching force direction. A chain and pulley gear mechanism and a linear motion conversion mechanism are loaded on a recumbent trike to keep disabled patient's motivation for rehabilitation and to use as transportation. The chain and pulley gear mechanism achieves alternate knee bending and stretching motion, and the linear motion conversion mechanism achieves converting linear motion to rotary motion. Experimental results in healthy participants are shown to confirm the validity of the proposed system and the potential of use as transportation.

The number of cases of plagiarism is increasing as it becomes easier for students to obtain well-written reports from the Internet or to copy and paste the contents of their classmates' reports into their own. Consequently, student plagiarism is becoming a primary issue interfering with fair grading by teachers. Academic reports tend to contain common expressions or academic terms. To write a good report, students try to use popular expressions for academic reports. Thus, it is important for teachers to detect plagiarism through careful attention to coincidental similarities. There is another important issue to be addressed: Plagiarism detection causes psychological burdens for both teachers and students. In this study, we introduce a plagiarism detection method for academic reports written in Japanese involving different types of characters. We train a number of Hidden Markov Models called writing models and identify authors by their writing style.

For multiagent environments, a centralized reinforcement learner can find optimal policies, but it is time-consuming. A method is proposed for finding the optimal policies acceleratingly. The method basically uses the centralized learner and supplementarily uses independent learners in the former phase. The independent learners transfer their learning results to the centralized learner, but excessive transfers cause the failure of learning. Therefore the independent learners should stop according to an appropriate condition. However, it is difficult for this method to find optimal policies for environments in which initial states are far from termination states. In order to find the optimal policies acceleratingly for such environments, this paper proposes multiagent reinforcement learning methods introducing new stop conditions.

In this research, we examine how the news on Japanese listed companies affects the stock market. Specifically, we use a lot of news information to clarify what kind of attribute news influences the stock price of the company. Our results confirm that the sentiment that judged from the news by the polar dictionary has a certain influence on the stock market. We also confirm effectiveness of the investment strategy using the news sentiment. By handling a large amount of information mechanically, investors will be able to conduct more efficient investment behavior.

Bus transportation service is more influenced than other public transport by various factors such as traffic congestion, weather condition, number of passengers, traffic signals. These factors often cause delay and the users may feel inconvenience while waiting at the bus stop. In the case of snowfall event, a large delay occurs, which greatly reduces the convenience of the bus. This paper aims at highly accurate arrival time prediction for each bus stop section in snow event in urban area. We investigate vulnerability of bus operations to snowfall and incorporate into predictions using geographical characteristics. In each bus stop section, we estimate geographical characteristics (gradient angle and gradient direction) and snow accumulation amount with detailed spatial resolution as factors affecting bus delay. Then, we evaluate a prediction accuracy using the arrival time prediction model with multiple regression analysis and the Kalman filter. As a result of the multiple regression analysis, it was found that the geographical characteristics of each bus stop section were the explanatory variables that greatly affect the bus delay at snowfall event. Furthermore, we predicted the bus arrival time using actual bus operation data. Of the 29 routes, 18 routes showed improvement in the predicted arrival time.

Optimization plays an important role in various disciplines of engineering. Multi-objective optimization is usually characterized by a set of trade-off points, which are commonly referred to as Pareto optimal points. In this paper, we study parallel search strategies for multi-objective constraint optimization problems. We propose to combine two strategies, and result of computational experiments shows the effectiveness of the proposed method.

Towels used in restaurants, hotels, medical facilities, etc. are repeatedly reused. It is necessary to inspect whether contaminations are mixed in washed towels. In this process, missing of hair attaching towels is a serious problem. To solve this problem, the system which detects hair automatically is required. In this paper, we propose an image processing method to detect hair considering implementation on hardware for real-time operation.