Yuzuru Ueda

The increased deployment of photovoltaic (PV) technology has led to an increased demand for grid-balancing capacity owing to growing short-term variability and forecast uncertainty. Simultaneously, higher PV penetration can lead to daytime energy market oversupply, pushing day-ahead prices toward zero and undermining PV revenues. Against this backdrop, this study investigated a market participation paradigm in which PV power plants supply reserve power themselves while actively absorbing their own uncertainty, rather than merely relying on balancing the services provided by external resources. We propose a risk-aware framework that classifies solar irradiance prediction errors into four risk categories using GPV-GSM numerical weather forecast data, translating the inferred risk level into practical bidding rules for balancing market participation. We adopted a hierarchical classification pipeline consisting of sign determination (stage 1, under- vs. overprediction), followed by degree determination (Stages 2 and 3), implemented with a multi-layer perceptron. To enhance class separability and reduce features, we introduced a stage-wise area under the curve (AUC)-based feature selection and compared AUC-selected and all-features settings under identical training conditions. The proposed strategies substantially reduce shortage events compared with directly using the original predictions as bids, although they increase surplus energy. The AUC-based model achieves comparable imbalance evaluation results, indicating that the selected features are sufficient for practical bidding support.

The variation in solar output due to weather hinders both reliability and profitability, particularly because an increase in photovoltaic (PV) penetration reduces daytime electricity prices. In this study, a method to generate a reserve value by managing prediction errors is developed. A model to estimate the error threshold corresponding to a specific cumulative probability is constructed using a polynomial approximation based on the historical distribution of prediction errors at a single site. This threshold is considered as the headroom required to absorb the prediction errors. The model is then applied to multiple sites to evaluate its overall applicability and the smoothing effect of site aggregation. The output is divided into clipped and nonclipped portions under overloading conditions; the former is modeled with a gamma distribution and the latter with a polynomial surface. Two models are tested: one using overall errors and another incorporating conditional probability with the clearness index. The results indicate that site aggregation reduces imbalance events and increases bid amounts. Under overloading, the optimal DC/AC ratio depends on the specified cumulative probability. Future works must focus on improving characterizing prediction errors under overloading and developing methods for predicting clipping to enhance balancing market participation of PV plants.

The increasing adoption of photovoltaic (PV) systems has intensified the challenges in managing output fluctuations. Although corporate power purchase agreements (CPPAs) offer long-term renewable energy procurement, research on integrated on/off-site energy management, business continuity planning (BCP)-compliant frameworks, and optimal battery energy storage system (BESS) operations within CPPA models remains limited. This study proposes a CPPA business model for three customer types, each equipped with on-site PV systems, a BESS, and access to an off-site solar power plant with additional BESS. An algorithm is developed to generate optimal day-ahead planned power flows (PPFs) while considering BCP requirements and managing real-time BESS operations. The model comprises: 1) on-site CPPA for direct power supply and 2) off-site CPPA, enabling remote energy supply from a solar power plant and facilitating power sharing among customers. The algorithm employs multi-objective optimization to minimize grid dependence, maximize surplus electricity sales, and balance PV utilization with revenue maximization through a weighting coefficient. A lower coefficient increases PV utilization, whereas a higher value increases revenue. Simulation results demonstrate that optimal BESS operations effectively reduce PV generation prediction errors, reduce reverse power flow, fulfill PPF requirements, and ensure BCP compliance with approximately 97.5% backup power reliability, irrespective of weighting coefficient. The optimal coordination of on-site and off-site CPPAs enhances local consumption of PV-generated electricity while aligning power transactions with market prices to maximize revenue. These findings provide insight into the developing advanced business models and algorithms for CPPA projects and PV plant operators, particularly in post-feed-in tariff (FIT) agreements.

The increasing reliance on photovoltaic (PV) generation as a cornerstone of carbon neutrality has led to transformative changes in the energy structure, further impacting electricity market trading mechanisms and price volatility. The electric power system reform also promoted wholesale trading in the Japan Electric Power Exchange (JEPX) spot market. This study explores an effective JEPX spot market price forecasting model that enables PV power suppliers to make informed production decisions and ensure revenue optimization. We found that understanding the net demand (total demand minus PV generation) is crucial for accurate price forecasting, as it allows for a more precise reflection of the gradual evolution of the solar-dominated energy structure of the dynamic electricity demand. We also conducted parameter tests and a comparative analysis of different training loops and periods in the basic form using an artificial neural network (ANN) and support vector regression (SVR) algorithms. The results indicated that the narrow ANN and SVR models with a linear kernel function and training in the continuous loop method performed better in spot market price forecasting than other model settings. Our proposed approach can provide essential insights into future price trends, facilitating informed sustainable energy planning and resource allocation for power generation to guarantee the benefits of achieving solar promotion and net-zero transition.

Load forecasting is an essential task in the operation management of a power system. Electric power companies utilize short-term load forecasting (STLF) technology to make reasonable power generation plans. A forecasting model with low prediction errors helps reduce operating costs and risks for the operators. In recent years, machine learning has become one of the most popular technologies for load forecasting. In this paper, a two-stage STLF model based on long short-term memory (LSTM) and multilayer perceptron (MLP), which improves the forecasting accuracy over the entire time horizon, is proposed. In the first stage, a sequence-to-sequence (seq2seq) architecture, which can handle a multi-sequence of input to extract more features of historical data than that of single sequence, is used to make multistep predictions. In the second stage, the MLP is used for residual modification by perceiving other information that the LSTM cannot. To construct the model, we collected the electrical load, calendar, and meteorological records of Kanto region in Japan for four years. Unlike other LSTM-based hybrid architectures, the proposed model uses two independent neural networks instead of making the neural network deeper by concatenating a series of LSTM cells and convolutional neural networks (CNNs). Therefore, the proposed model is easy to be trained and more interpretable. The seq2seq module performs well in the first few hours of the predictions. The MLP inherits the advantage of the seq2seq module and improves the results by feeding artificially selected features both from historical data and information of the target day. Compared to the LSTM-AM model and single MLP model, the mean absolute percentage error (MAPE) of the proposed model decreases from 2.82% and 2.65% to 2%, respectively. The results demonstrate that the MLP helps improve the prediction accuracy of seq2seq module and the proposed model achieves better performance than other popular models. In addition, this paper also reveals the reason why the MLP achieves the improvement.

There are approximately 200,000 households living in detached houses and gers (yurts) with small coal stoves that burn raw coal in Ulaanbaatar city. A proper heating system and improvement of the energy efficiency of residential dwellings are vitally important for Ulaanbaatar city to reduce air pollution as well as for the operation of the current central energy system. This study shows the experimental results for two gers with two different heating systems and different thermal insulation, for investigating the merits of each. The technical feasibility of the system consisting of an electric thermal storage (ETS) heater with a daytime charging schedule and areal photovoltaic (PV) system was also examined by using a simulation with software developed in MATLAB (R2020a, MathWorks, USA). As a result of the experiment, the indoor comfort level and energy efficiency of the ger with added insulation and an ETS heater with nighttime charging were shown to be enhanced compared with those of the reference ger. The ger with added insulation and the ETS heater consumed 3169 kWh for electric appliances and 5989 kWh for the heating season. The simulation showed that the PV self-consumption rate is 76% for the Ger 2 with the ETS heater because of the daytime charging schedule of the ETS heater. The PV system supplied 31% of the total energy consumed, with the remaining 69% from the main grid.

With a widespread of photovoltaic (PV) systems, a long-term use of the systems is expected. Since the amount of power generation can be decreased due to failures by various causes in the system, a practical way to detect the cause of the failure is needed for long-term uses of the system. This paper proposes a method to determine the cause of the failure in the PV system using convolutional neural network (CNN) by classifying the image of measured I-V curves which might have some failures. As a result, a case of the I-V curves which were caused by a stepped defect in a string unit was detected 100% correctly.

We present modeling and analysis of day-ahead spatio-temporal energy markets in which each competitive aggregator aims at making the highest profit by managing a complex mixture of different energy resources, such as conventional generators, storage batteries, and uncertain renewable resources. First, we develop an energy market model in terms of an adjustable robust convex program. This market modeling is novel in the sense that the prosumption cost function of each aggregator, which evaluates the cost to realize an amount of spatio-temporal energy prosumption, is a multi-variable function resulting from a "parameterized" max-min program, in which the variable of the prosumption cost function is involved as a continuous parameter and the variable of dispatchable resources is involved as an adjustable variable for energy balance. This formulation enables to reasonably evaluate a reward for intertemporal dispatchability enhancement and a penalty for renewable energy uncertainty in a unified way. In addition, it enables to enforce a market regulation in which every aggregator is responsible for absorbing his renewable energy uncertainty by managing his own dispatchable energy resources. Second, in view of social economy as well as personal economy, we conduct a numerical analysis on the premise of several photovoltaic penetration levels. In this numerical analysis, we demonstrate that renewable generators do not always have priority of energy supply higher than conventional generators due to their uncertainty and limited dispatchability, meaning that the merit order of conventional and renewable generators can reverse. Furthermore, we analyze long-term evolution of competitive energy markets demonstrating that there can be found a social equilibrium of battery penetration levels, at which maximum personal profit with respect to battery system enhancement is attained.