Search

Large events with many attendees cause congestion in the traffic network around the venue. To avoid accidents or delays due to this kind of unexpected congestion, it is important to predict the level of congestion in advance of the event. This study aimed to forecast congestion triggered by large events. However, historical congestion information alone is insufficient to forecast congestion at large venues when non-recurrent events are held there. To address this problem, we utilize microblog posts that refer to future events as an indicator of event attendance. We propose a regression model that is trained with microblog posts and historical congestion information to accurately forecast congestion at large venues. Experiments on next 24-hour congestion forecasting using real-world traffic and Twitter data demonstrate that our model reduces the prediction errors over those of the baseline models (autoregressive and long short term memory) by 20% – 50%.

Electric tree is a phenomenon of local dielectric breakdown with partial discharge (PD). It is also the final stage of insulation deterioration. The electrical treeing has been discussed by using conventional PD parameters such as the generation phase, charge amount and so on. Since the obtained PD waveforms were processed by an analog filter designed according to the frequency characteristics of the measurement instrument such as an oscilloscope, its waveform is a convolution of the original PD waveform and the transfer function of the measurement system. However, the feature values such as a rise time, fall time and frequency spectrum in the original PD waveform should show the behavior of electrons and ions in the discharge space. In this paper, the original PD waveforms generated in micro-silica/epoxy composite with the electrical tree were acquired, and the feature values of the original PD waveform were analyzed. In addition, the behavior of the electric charges in the discharge space was assessed from the change of the feature values.

Mica epoxy insulation as used for generator stator windings is required to be improved as regards withstand electrical stress and heat dissipation. Replacement of epoxy resins by epoxy nanocomposites is a useful method to fulfil a purpose of such improvement. This paper is aimed to evaluate PD-initiated breakdown (BD) performances as one of the most crucial characteristics using simulated model specimens. Two constituent materials, i.e. mica tapes and resin layers were set as about 40 and 10 µm in thickness, respectively. EDX analysis shows that the model specimens include silica nanofillers properly dispersed in narrow nanocomposite regions and are suitable for PD initiated BD experiments. It was found that PD initiated BD lifetime was 2.8 times in nanocomposites than in neat epoxy under a rather high electric stress condition.

Polymer nanocomposites have attracted attention for various superior insulation performances. However, the effect of nanocomposites on the insulation lifetime with an enclosed void defect is unclear. Also, the insulation lifetime extension mechanism has rarely been reported. This paper experimentally investigates the lifetime extension mechanism of epoxy/TiO 2 nanocomposites with the void defect. Centrifugation and dilution procedures are used to control the dispersion state of TiO 2 nanofillers. Time evolutions of partial discharge (PD) current are also recorded to investigate the insulation degradation procedure. Consequently, it is clarified that the lifetime becomes longer with increasing the maximum agglomerate size of the TiO 2 nanofillers. Results of the PD current monitoring suggest that the lifetime extension of the nanocomposite is achieved due to delaying the electrical tree propagation in the sample.

Epoxy nanocomposites (NCs) are known to have advantages when used in the electrical insulation of power equipment. Agglomerates of nanoparticles, which inevitably form during the fabrication process, can affect the insulating properties of the resultant nanocomposites. Our previous research has clarified that the impulse breakdown strength of the solid part, rather than the surface, of epoxy–TiO 2 NCs with agglomerates removed was higher than that of unfilled epoxy resin. Here, we investigated the dependence of the TiO 2 filler volume fraction on the AC breakdown strength of NCs whose agglomerate size was controlled. The volume fraction and the maximum agglomerate size were controlled using a simple dilution procedure and a centrifugation technique. The AC breakdown strength of the NCs was found to increase with decreasing volume fraction of TiO 2 nanoparticles when the maximum agglomerate size was controlled at the same level. The variation of the AC breakdown strength of the NCs decreased with decreasing maximum agglomerate size, suggesting that the existence of agglomerates larger than a critical size could affect the breakdown.

Agglomerates of nanoparticles which inevitably occur in the fabrication process of epoxy nanocomposite can affect the insulation properties of epoxy nanocomposite. This paper presents an attempt to clarify the effect of agglomerate size on two important insulation properties of epoxy/TiO 2 nanocomposite (NC). One is intrinsic AC breakdown strength and the other is long-term partial discharge lifetime. We used a centrifugation technique to control the maximum size of the agglomerates in the range from 5 μm down to 0.05 μm in each sample. As the results, AC intrinsic breakdown strength was increased with decreasing the maximum agglomerate size and became even higher than the neat epoxy resin when the maximum size is less than around 0.05um. On the other hand, partial discharge lifetime decreased with decreasing the maximum agglomerate size. It should be noted that the PD lifetime distinctively decreased when the agglomerate size is less than around 0.2 um indicating the existence of a critical agglomerate size which contribute to the suppression of partial discharge development resulting in the extension of the partial discharge lifetime.

SiC-based nonlinear conductive stress grading tape (SGT) has been used at an end-turn stress grading system of a form-wound rotating machine. Although previous researches reported that high-frequency repetitive pulses cause electric field enhancement and local heating at the SG system, few paper deals with the detailed electrical and thermal behaviors under pulse width modulation (PWM) waveforms, which contain not only the repetitive high-frequency, i.e., the carrier frequency, component but also the fundamental frequency component. In this study, the surface potential, the electric field, the time-averaged power densities as well as the resulting temperature rise under the 3-level PWM waveforms were investigated by the transient electric field and thermal coupled FEM simulation. Furthermore, the behavior of the accumulated charges within the SGT was analyzed, based on the simulation results. Consequently, it has been revealed that the charge behavior and the power densities by the carrier-frequency component and those by the fundamental frequency component could be separated individually.

Electric field grading system (SG system) is introduced to prevent partial discharge at the stator coil end of a rotating machine. The SG system consists of corona armor tape (CAT) and stress grading tape/paint (SGT/SGP). The CAT is a carbon-loaded highly-conductive tape and is wound over the main insulation. A SGT/SGP is a silicon carbide powder-loaded tape/paint with field-dependent conductivity and is wound over partially the end of CAT. In general, the conductivity of SGT/SGP also has temperature dependence, but has little field-dependence under higher temperature condition. The coil temperature under operating condition exceeds 100°C and this means the electric field on the coil end may increase due to the less non-linearity of SGT/SGP. When the temperature is high, the density of air decreases and surface discharge may occur easily even under the same field distribution. To examine this issue, light emission of partial discharge (PD) on a model bar system under 1-kHz AC voltage was observed by a photomultiplier tube and a digital camera equipped with an image intensifier. The temperature on SG system was set to 30, 70, and 100°C. Unexpectedly, light emission from PD became weaker and the light emission area became smaller as the temperature increased. Some reasonable explanations for this are as follows: the change in the occurrence phase of PD: and the amount of water contained in the vicinity of SG system surface. Although further investigation is required, but the unexpected increase in the PD pulses has not been observed so far within the measured range.

In this study, transient analysis of Stress Grading(SG) system under PWM waveform is conducted to reveal the effects of fundamental frequency on potential distribution and power density. Risetime, level and jump voltage of PWM waveform used in an analysis is 1μs, 3 and 5 kV, respectively. Fundamental and carrier frequencies are changed in the range from 50–500 Hz and 0.5-6 kHz, respectively. For comparison, analyses under sinusoidal and square voltages which correspond to fundamental and carrier frequencies are conducted. Analytical results show that maximum surface field on SG system under PWM waveform is 1.1 times higher than that under square voltage regardless of fundamental and carrier frequencies. Power density under PWM waveform increases than that under square voltage. The value of power density under PWM waveform is estimated by the sum of power density under sinusoidal voltage which corresponds to fundamental frequency and square voltage which corresponds to carrier frequency. The error of this estimated value is in the range from 8–15 %. This shows that estimated value is accurate enough to estimate temperature rise under PWM waveform.

In using reinforcement learning [1], [2], there is a problem that states are often provided as continuous rather than discrete values. In this study, we applied continuous state space to reinforcement learning based on the success probability and return (EQ) [3] of conventional research. In addition, we demonstrated that EQ learns more safely than a conventional method in reinforcement learning problem with a continuous state space.

Polymer nanocomposite material, i.e., polymer insulating material with nano-sized inorganic additives, has recently received much attention for their superior dielectric performance as insulating media of high voltage power apparatus. The long-term insulation lifetime is one of the important factors, which determines the insulation reliability of the apparatus. In this research, the insulation lifetimes of epoxy/TiO2 nanocomposite and neat epoxy resin were experimentally investigated. The long-term experiments were carried out with a test electrode system with an artificial enclosed-void, which simulated air cavities contained within the insulation of the actual-scale apparatus. The time evolutions of partial discharge (PD) intensities were also recorded during the long-term experiment. Consequently, the epoxy/TiO2 nanocomposite showed more than 16 times longer lifetime compared to the neat epoxy resin. Finally, the superiority of the epoxy/TiO2 nanocomposite was discussed based on the PD energies.

Additives and fillers are traditionally employed in polymer insulating materials like an epoxy resin (EP) to improve the mechanical, thermal, and sometimes even the electrical properties. To understand the influence of filler on the electrical properties of EP/nano-sized titanium oxide (nTiO 2 ) nanocomposites, treeing breakdown tests were performed. The breakdown voltage of the EP/nTiO 2 nanocomposite with a 1-vol% nTiO 2 content was higher than that of the EP at the same electrode distance. The nTiO 2 might capture electrons, leading to an electron avalanche. The improvement effect by adding the nTiO 2 was also observed in the nanocomposite–mica sheet multi-insulation system. The addition of an appropriate amount of nTiO 2 to EP may suppress tree initiation and/or tree propagation.