Data-Driven Photovoltaic Generation Forecasting Based on a Bayesian Network With Spatial–Temporal Correlation Analysis

As the cost of the residential solar system decreases, rooftop photovoltaic (PV) has been widely integrated into distribution systems. Most rooftop PV systems are installed behind-the-meter (BTM), i.e., only the net demand is metered, while the native demand and PV generation are not separately recorded. Under this condition, the PV generation and native demand are invisible to utilities, which brings challenges for optimal distribution system operation and expansion. In this paper, we have come up with a novel two-layer approach to disaggregate the unknown PV generation and native demand from the known hourly net demand data recorded by smart meters: 1) At the aggregate level, the proposed approach separates the aggregate PV generation time series from the aggregate net demand time series for customers with PVs. 2) At the customer level, the separated aggregate-level PV generation is allocated to individual PVs. These two layers leverage the spatial correlations of native demand and PV generation, respectively. One primary advantage of our proposed approach is that it is more independent and practical compared to previous works because it does not require PV array parameters, meteorological data and previously recorded solar power exemplars. This paper has verified our proposed approach using real native demand and PV generation data.

Solar photovoltaic (PV) technology is now a key contributor worldwide in the transition towards low‐carbon electricity systems. To date, PV commonly receives subsidies in order to accelerate adoption rates by increasing investor returns. However, many aleatory and epistemic uncertainties exist with regard to these potential returns. In order to manage these uncertainties, an innovative probabilistic approach using Bayesian networks has been applied to the techno‐economic analysis of domestic solar PV. Empirical datasets from over 600 domestic PV systems, together with national domestic electricity usage datasets, have been used to generate and calibrate prior probability distributions for PV yield and domestic electricity consumption, respectively, for typical urban housing stock. Subsequently, conditional dependencies of PV self‐consumption with regard to PV generation and household electricity consumption have been simulated via stochastic modelling using high temporal resolution demand and PV generation data. A Bayesian network model is subsequently applied to deliver posterior probability distributions of key parameters as part of a discounted cash flow analysis. The results illustrate the sensitivity of PV investment returns to parameters such as PV self‐consumption, PV degradation rates and geographical location and quantify inherent uncertainties when evaluating the impact of sector‐specific PV adoption upon economic indicators. The outcomes are discussed in terms of the value and impact of this new Bayesian approach in terms of supporting robust and rigorous policy and investment decision‐making, especially in post‐subsidy contexts globally. © 2016 The Authors. Progress in Photovoltaics: Research and Applications published by John Wiley & Sons Ltd.

Several utilities are offering, or plan to offer, customers the choice between standard and “Time-of-Use” (ToU) rates for electricity. ToU rates for energy are based on a tiered pricing structure as opposed to standard flat rates. ToU rates are typically less than the standard rate during off-peak times and higher than the standard rate at peak times. This pricing structure can incentivize combining demand response (DR) with photovoltaic (PV) generation which can shift a customer's net demand from peak to off-peak times. This paper presents a demand response scheme designed for implementation with photovoltaic generation within a ToU pricing environment. The DR algorithm is designed to require minimal sensors, does not require forecasting, and is suitable for automation. Simulation results validating the DR algorithm and quantifying potential savings based on the level of DR are presented based on PV generation data and residential load consumption data in New York.

Photovoltaic (PV) systems are swiftly expanding within the building sector, offering significant benefits such as renewable energy integration, yet introducing challenges due to mismatches between local generation and demand. With the increasing availability of data and advanced modeling tools, stakeholders are increasingly motivated to adopt energy management and optimization techniques, where accurate forecasting of PV generation is essential. While the existing literature provides valuable insights, a comprehensive review of methodologies specifically tailored for the forecast of PV generation in buildings remains scarce. This study aims to address this gap by analyzing the forecasting methods, data requirements, and performance metrics employed, with the primary objective of providing an in-depth review of previous research. The findings highlight the critical role of improving PV energy generation forecasting accuracy in enhancing energy management and optimization for individual buildings. Additionally, the study identifies key challenges and opportunities for future research, such as the limited exploration of localized environmental and operational factors (such as partial shading, dust, and dirt); insufficient data on building-specific PV output patterns; and the need to account for variability in PV generation. By clarifying the current state of PV energy forecasting methodologies, this research lays essential groundwork for future advancements in the field.

With the quick incorporation of distributed photovoltaic (PV) systems into the contemporary energy grid, guaranteeing consistent power output and system reliability has become critical. These systems are susceptible to a variety of operational anomalies, including dust accumulation, and shading which can reduce energy efficiency without being detected. Traditional anomaly detection methods frequently fail to capture the internal dynamics and latent factors that affect PV system performance over time. There is an urgent need for a robust, real-time, and dynamic monitoring strategy that can accurately detect abnormal behavior in distributed PV systems. The goal of this research is to create and execute a state space model (SSM)-based anomaly detection framework that dynamically estimates system conditions and alerts to deviations from expected power output in real time. A dataset of 5000 records and 10 attributes was used, which contained time-series data from distributed PV systems, such as actual and expected power output, irradiance, temperature, and other operational parameters. Model matrices (A, B, and C) were estimated using linear regression and enhanced by Kalman filtering. The system estimated power output at each time step and calculated the deviation from actual values. A threshold-based approach was employed to classify anomalies. The model was trained on data labeled as normal and validated against both normal and abnormal entries. The proposed SSM-based anomaly detection model had an accuracy of 94.70%, a precision of 93.85%, a recall of 94.30%, an F1-score of 94.07%, and a Matthews Correlation Coefficient (MCC) of 89.32%, indicating strong predictive capability in detecting abnormal performance in PV systems.

Based on the nonlinear dynamic modeling of photovoltaic power generation system output time sequences, parsing the kinetic characteristics of the output power of photovoltaic power generation system, the chaos characteristics of the photovoltaic power generation system output power is put forward. According to the phase space reconstruction theory, adding weight first order local prediction method is used for photovoltaic output power chaotic prediction. The actual measurement data of the output power acquired from the photovoltaic power generation system of a printing plant roof is used to verify the prediction method. Results show that the photovoltaic power generation system output power has chaotic characteristics, and it is feasible that the chaos prediction method apply to photovoltaic output power prediction, the method has a high accuracy for photovoltaic power ultra-short term prediction, it provide a new way for photovoltaic output power prediction.

In order to remedy problems encompassing large-scale data being collected by photovoltaic (PV) stations, multiple dimensions of power prediction mode input, noise, slow model convergence speed, and poor precision, a power prediction model that combines the Candid Covariance-free Incremental Principal Component Analysis (CCIPCA) with Long Short-Term Memory (LSTM) network was proposed in this study. The corresponding model uses factor correlation coefficient to evaluate the factors that affect PV generation and obtains the most critical factor of PV generation. Then, it uses CCIPCA to reduce the dimension of PV super large-scale data to the factor dimension, avoiding the complex calculation of covariance matrix of algorithms such as Principal Component Analysis (PCA) and to some extent eliminating the influence of noise made by PV generation data acquisition equipment and transmission equipment such as sensors. The training speed and convergence speed of LSTM are improved by the dimension-reduced data. The PV generation data of a certain power station over a period is collected from SolarGIS as sample data. The model is compared with Markov chain power generation prediction model and GA-BP power generation prediction model. The experimental results indicate that the generation prediction error of the model is less than 3%.

The control of a photovoltaic system with a battery for the object needs according to the forecast of photovoltaic generation with the possibility of increasing the load power above the limit for consumption from the grid is considered. At the same time, the goal is to limit the depth of battery discharge and energy consumption during peak demand hours. A system has been proposed for regulating the power consumed from the grid based on the deviation of the battery state of charge relative to a given schedule, which is determined by forecast using an additional variable that sets a limit on power consumption. This ensures the minimum possible value of the discharge depth and peak consumption in accordance with photovoltaic generation for the degree of power increase from 1 to 1.5. An algorithm and control structure has been developed. The system is not critical by the load schedule and discharge depth value are maintained even when photovoltaic generation is below forecast by up to 20%. The possibilities for reducing peak demand are limited. Key words: hybrid photovoltaic system, battery discharge depth, power regulation, battery state of charge formation, increasing power over the limit of consumption from the grid, photovoltaic generation forecast, daily cycle modeling

This paper presents a novel two-stage approach for tracking the maximum power point (MPP) of a large photovoltaic (PV) generator under non-uniform irradiance. The non-uniform irradiance in large PV arrays is attributed to partially shaded PV modules and may significantly increase the complexity of the MPP tracking problem. In Stage 1, the method makes use of real-time irradiance measurements in certain regions of the generator to estimate the power versus voltage (P-V) curve and to deduce an estimate of the global peak region. This is followed in Stage 2 by a single peak tracking approach starting at the estimated global peak, which combines ripple correlation and incremental conductance, to accurately locate the exact global power point. The system consists of a PV generator, with pyranometers distributed across it, a DC-DC converter feeding a battery, and a controller implementing the tracking algorithm. The system simulation model was developed in MATLAB, taking into consideration the implementation of bypass and blocked diodes together with the pyranometer irradiance inputs. The paper first presents a modeling and simulation scheme suitable for studying the I-V and P-V characteristic curves of a PV array with partial shading configuration given the exact irradiance values. This is followed by a method of global peak region estimation knowing the location and readings of the distributed pyranometers. Tracking results of the two-staged approach are demonstrated under various shading configuration cases.

<p>As Europe strives for a decarbonization of electricity supply, many countries proclaim ambitious targets for capacity additions of wind and photovoltaic (PV) power. By increasing renewable capacities, countries seek to reduce their dependency on fossil fired electricity generation. For instance, Germany recently announced new capacity targets that mean a triplication of wind and PV power capacities while at the same time aiming for a full coal exit until 2030. However, the potential of renewable energies is not equally distributed over the European continent. Due to higher irradiance levels with decreasing latitude, PV power achieves higher full load hours and thus higher generation per installed capacity in southern Europe.  Wind power reaches higher capacity factors close to shorelines, where near-surface wind speeds are higher.</p><p>In light of the unequal distribution of PV and wind power generation potentials, national capacity targets may not lead to optimal utilization of capacities, i.e. maximum renewable generation. We analyse the effects of such national capacity targets for the case of PV generation. To do so, we compare two scenarios: A scenario with PV generation capacity distributed over Europe according to national political targets and a scenario with a distribution of newly installed capacities chosen such that it optimizes the renewable power potentials. Next to PV generation potentials, our assessment considers the European power grid, because line capacities limit the transmission and hence the trading between different countries. Line limits can have a negative impact on the overall power generation since renewable generation has to be curtailed if power can neither be locally used nor transmitted to another region.  We employ a linear optimization-based electricity market model with inter-European transmission constraints and hourly resolution weather data for 35 historic weather years (1982-2016). We analyse the scenarios if terms of differences in power generation and curtailment, greenhouse gas emissions, and system operational cost. We pay special attention to extreme weather situations and the effects of a sub-optimal PV power allocation on security of supply.</p>

This paper presents a data acquisition and visualization system, with storage in the cloud, applied to photovoltaic (PV) systems. The proposed system uses an embedded computer to perform communication with PV inverters via the RS485 standard. The system developed also performs the acquisition of meteorological data, the computer embedded obtain data by a microcontroller that performs reading of the radiation and temperature sensors. The result was a cloud database with weather and PV generation data, to view the data more quickly and in a friendly way, a visualization systems have been developed. A visual system on the web and an application for Android smartphones have been developed. The developed system was applied to a real PV system with 42 PV modules of 235W connected in two PV inverters of 2.5kW and 6kW.

The analysis presented in this paper focuses on the energetic aspects of the use of photovoltaic (PV) systems in office buildings. Energy generation from PV system has been simulated on an hourly time scale and compared with the energy demand of two office buildings located in Krakow (southern Poland). The buildings’ annual load exceeded 1 GWh in both cases. The analysis dealt with estimating how much energy generated by a PV system can be utilized on-site (self-consumption) and how big the energy surpluses will be (energy generation greater than demand). Capacities of PV systems ranging from 0.1 to 1.0 MW have been considered. Also, the impact of PV generation on the residual load parameters and changes in the maximal monthly and hourly energy demand were investigated. The results show that although the building energy demand is similar (in terms of annual volume of energy consumed) the potential of PV systems to cover it is different. The 100-kW PV system can reduce the observed hourly peak energy demand by 1% in December but by over 30% in June (respectively, from 171 kWh to 169 kWh, and from 333 kWh to 255 kWh). Considering the annual patterns of the office buildings’ energy demand and PV generation (both have their respective peaks in summer), the application of a PV system changes the office building energy demand pattern significantly from the perspective of the power system operator. After installing 500 kW PV in an office building consuming 1 GWh annually, the months with highest demand are no longer in summer but in winter.

Photovoltaic (PV) generation system is an important way to promote energy transformation and to cope with environmental challenges, firstly, based on the steady-state power flow equations of the PV system AC side connected to weak grid, the influence of power grid structure parameters on their maximum transmission power is studied, which provides a theoretical basis for the design of subsequent virtual impedance; secondly, a time-domain small-signal model for PV system connected to weak grid is established, and based on the proposed small-signal model, eigenvalue analysis is employed to study the stability of PV system connected to weak grid within different grid strength. The results show that: in the case of a given SCR, the smaller the impedance angle is, the greater the maximum transmission power of the system is; when the output power of PV system is rated, the smaller grid impedance angle makes the PV system can operate under the weaker grid conditions; with the decreasing of the grid strength, the PV system prone to power oscillation Finally a virtual resistance control strategy is proposed to improve the operation stability of PV system The validity of the theoretical analysis and the feasibility of the control strategy are verified by time domain simulation.

The precise forecasting of photovoltaic energy generation holds paramount importance in refining scheduling and ensuring safe operation of extensive photovoltaic power stations. However, the inherent instability and volatility of photovoltaic power generation pose significant challenges to prediction accuracy. To address this, this article conducts a thorough analysis of the seasonal characteristics of photovoltaic power generation and introduces a hybrid prediction model based on the ensemble empirical mode decomposition (EEMD)-improved whale optimization algorithm (IWOA)-bidirectional long short-term memory network (BiLSTM) algorithm. This model leverages multi-seasonal meteorological features to enhance forecasting accuracy. First, EEMD is used to decompose and reconstruct photovoltaic power generation data to eliminate its instability and volatility. Second, three improved strategies are proposed for the position update in different stages of the IWOA, and a multi-seasonal prediction model based on IWOA-optimized Bidirectional LSTM is established. Finally, the operational data of a photovoltaic power station in the northwest region of China are used as a case study to evaluate the prediction performance of the model in detail. The results show that the model's accuracy rate ranges from 97.1% to 98.7%, which can accurately predict photovoltaic power generation and improve the utilization rate of renewable energy.

In order to reduce the impact of photovoltaic power output uncertainty on the economic operation of power system, this paper studies the problem of controllable output power of photovoltaic power generation system. In grid connected photovoltaic energy storage system, typical photovoltaic power generation capacity curves obtained by statistical analysis, using particle swarm optimization algorithm to optimize the economic model under different power system output control, when the lowest total cost is the optimal selection of the battery capacity. The battery capacity is determined, the photovoltaic power generation pre measurement, using particle swarm optimization algorithm to optimize the system economic operation model, each time the amount of output of photovoltaic power plant is expected to system debugging, the specific numerical output. Combined with the output of photovoltaic power generation, the output power of the battery is calculated by the analysis and calculation unit, and the economic operation of the grid connected photovoltaic energy storage power generation system is realized through the DC/DC control of the battery charging and discharging.