Renewable energy systems (RES) are increasingly becoming crucial to negate the adverse effects of conventional fossil fuel-based energy production approaches on global climate, such as climate change. State-of-the-art (SOTA) artificial intelligence (AI) tools and techniques can play a key role in the control, management, and maintenance of RES. This article discusses the importance and applications of AI in the renewable energy sector.
Importance of AI in RES
AI has gained significant attention for different applications in the renewable energy industry, such as short-term electric power and load forecasting, solar and wind power forecasting and modeling, electrical load prediction of supermarkets and the city, and sizing the photovoltaic systems, as the technology can improve efficiency, reduce costs, improve sustainability, and increase reliability.
Moreover, integrating the Internet of Things (IoT) and AI into RES has made these systems more robust and responsive. Currently, AI technologies are used in all RES, including solar, wind, hydro, and geothermal. AI has also increased the efficiency and cost-effectiveness of RE technologies, such as wind turbine systems and solar photovoltaics, accelerating the shift from fossil fuel-based energy production to RES.
Among the AI techniques, the artificial neural network (ANN) is used extensively in RES owing to its higher generalization capabilities, shorter computing time, and greater accuracy over alternative modeling techniques, which results in significantly improved efficiency.
Other AI techniques used in RES include support vector machines (SVM), autoregressive integrated moving average (ARIMA), radial basis function (RBF), group method of data handling neural network (GMDHNN), transient system simulation tool (TRNSYS), HIstorical SImilar Mining (HISIMI), seasonal autoregressive integrated moving average (SARIMA), radial basis function neural networks (RBFNN), genetic algorithm (GA), naïve Bayes (NB), adaptive neuro-fuzzy inference system (ANFIS), and feed-forward backpropagation neural network (BPNN).
AI Techniques Used in Renewable Energy
Solar Energy: Several AI techniques are used in several solar energy applications, including solar radiation forecasting, modeling, and prediction. For instance, BPNN is used for solar irradiance prediction, solar radiation prediction, solar water heating system performance assessment, solar beam radiation prediction, daily ambient temperature and solar irradiation prediction, maximum high concentrator photovoltaics (HCPV) power prediction, global solar irradiation prediction, solar energy and hot water quantity prediction, and building energy prediction.
BPNN-Ångström linear methods and BPNN-regression methods are utilized for global solar radiation prediction. Similarly, ANFIS is employed for photovoltaics power supply modeling, hourly global irradiance prediction, clearness index determination, radiation prediction, solar power prediction, and solar chimney power plant (SCPP) performance prediction.
Solar water heating system design and solar tracking can be performed using GA. GA-perturb and observe (PO) can be used for maximum power point tracking (MPPT) of a photovoltaic array. TRNSYS and ANN can be used to predict integrated collector storage (ICS) performance. GA-HISIMI/GA-BPNN and GA-GMDHNN can be employed for solar power prediction and solar system optimization, respectively.
BPNN and batch learning ANN and BPNN and empirical models can be used for mean temperature prediction and diffuse solar radiation prediction, respectively. SVM-RBFNN-autoregressive (AR) model can be utilized for solar power prediction.
RBF and infinite impulse response (IIR)/BPNN and IIR, wavelet transform (WT) and BPNN, and WT and RBFNN can be employed for photovoltaic system size optimization, solar radiation value estimation, and photovoltaic energy prediction, respectively.
Hybrid approaches based on ANN-GA-particle swarm optimization (PSO)/BPNN-genetical swarm optimization (GSO), SARIMA-SVM/SVM-firefly algorithm (FFA), and autoregressive and moving average (ARMA)-time delay neural network (TDNN) can be utilized for photovoltaic power prediction, solar power prediction, and solar radiation prediction, respectively.
Several AI techniques are also used in solar microgrids for different applications. For instance, bacterial foraging optimization (BFO) and PSO are used to tune the key parameters in microgrids' automatic generation control (AGC). Similarly, the ARIMA time series algorithm uses continuous data to predict electric consumption in a solar photovoltaic microgrid village.
A machine learning (ML) algorithm using the regression tree model can predict solar microgrid power generation output. ANN-based prediction model and Sugeno type-fuzzy system and Q-learning algorithm can be utilized for solar energy output prediction and energy management, respectively.
Wind Energy: Wind power prediction can be obtained using BPNN, recurrent high order ANN-NB, ANN-RBFNN-fuzzy methods, the probabilistic method, ANFIS, PSO-BPNN, and a hybrid method based on PSO and ant colony optimization (ACO).
Similarly, BPNN-RBFNN-adaptive linear element (ADALINE), BPNN-trigonometric point cumulative semivariogram (TPCSV), ADALINE-BPNN-RBFNN, ARMA-neural logic networks (NLN)-ANN, BPNN-single exponential smoothing (SES), BPNN and fuzzy methods, SVM and BPNN, ANFIS, empirical mode decomposition based feed-forward neural network (EMD-FNN), ANN-Markov chain (MC), ensemble EMD (EEMD)-SVM, ARIMA-BPNN, fifth-generation mesoscale model (MM5)-ANN, WT-SVM-GA, and support vector regression (SVR)-PSO can be utilized for wind speed prediction.
BPNN and fuzzy-GA can be employed for both wind speed and power prediction, while the fuzzy method and ANFIS can be utilized for wind power generation system design. ANFIS can also be used for missing wind data interpolation. Wind turbine fault diagnosis can be performed using BPNN and WT. A hybrid approach based on ANFIS-PSO-WT can be used for risk optimization in wind energy trading.
Hydro Energy: BPNN-differential dynamic programming (DDP)-k-nearest neighbors (KNN) and GA can be employed for hydropower plant scheduling. BPNN can be used for rainfall-runoff process modeling, stream flow prediction, and power discharge estimation, while BPNN and AR can be utilized for river flow prediction.
River flow prediction can be obtained using ANFIS-ANN-multivariable regression (MR), and HC-fast Fourier transform (FFT)-ANN-based approaches. Optimal material for hydropower projects can be selected using fuzzy logic, while water release can be predicted using ANFIS-GA.
Hybrid approaches based on learning vector quantization (LVQ)-adaptive resonance theory (ART)-multiple adaptive perceptrons (MAP), fuzzy logic controller (FLC)-PSO/FLC-GA and ANN-artificial bee colony (ABC) can be employed for acoustic and maintenance prediction, FLC design for AGC, and hydraulic energy prediction, respectively.
Geothermal Energy: BPNN is used extensively in geothermal energy applications, including vertical ground coupled heat pump (VGCHP) system performance prediction, static formation temperature (SFT) prediction of geothermal well, geothermal map and power prediction, ammonia-nitrogen prediction, geothermal plant modeling, site location modeling, conductivity map generation, geothermal plant pressure prediction, proportional integral derivative (PID) controller efficiency prediction, geothermal district heating system performance prediction, and void fraction prediction.
Similarly, fuzzy logic can be utilized for geothermal recirculating aquaculture system design. A hybrid approach based on GMDH-GA-singular value decomposition (SVD) can be used for geothermal reservoir temperature modeling and prediction.
Recent Studies
Smart grids that efficiently integrate RE and conventional energy systems are crucial for ensuring sustainable power generation due to the intermittent nature of power generation by RES. Although smart meters allow demand prediction in real-time, models that can accurately predict the electricity generated by RES are required as prediction models (PMs) can precisely ensure grid stability, effective energy management, and efficient scheduling. For instance, the smart grid must be transformed into a conventional energy source smoothly to ensure the generated electricity meets the predicted demand when the model predicts a loss of RE supply.
In a study published in the journal Sustainability, researchers developed a model that can accurately replicate a microgrid, predict power supply and demand, seamlessly schedule electricity supply to meet demand, and provide actionable insights into the operation of the smart grid system.
Researchers developed the demand response program (DRP) using improved incentive-based payment as cost suggestion packages. They implemented optimal demand-side management for the smart grid by considering demand response as the compensation for uncertainty produced by solar power and wind power generation in an optimization function with two different goals.
The test results were reviewed in various cases to optimize the operating costs through a multi-objective ant colony optimization algorithm (MOACO) without and with the input of the DRP. The simulations demonstrated that the operating costs will be reduced if the customers employ DR to address the loss of electricity generation caused by uncertainty in solar and wind power.
References and Further Reading
Arumugham, V., Ghanimi, H. M., Pustokhin, D. A., Pustokhina, I. V., Ponnam, V. S., Alharbi, M., Krishnamoorthy, P., Sengan, S. (2023). An Artificial-Intelligence-Based Renewable Energy Prediction Program for Demand-Side Management in Smart Grids. Sustainability, 15(6), 5453. https://doi.org/10.3390/su15065453
Jha, S. K., Bilalovic, J., Jha, A., Patel, N., Zhang, H. (2017). Renewable energy: Present research and future scope of Artificial Intelligence. Renewable and Sustainable Energy Reviews, 77, 297-317. https://doi.org/10.1016/j.rser.2017.04.018
Dellosa, J. T., Palconit, E. C. (2021). Artificial Intelligence (AI) in Renewable Energy Systems: A Condensed Review of its Applications and Techniques. 2021 IEEE International Conference on Environment and Electrical Engineering and 2021 IEEE Industrial and Commercial Power Systems Europe (EEEIC / I&CPS Europe). https://doi.org/10.1109/EEEIC/ICPSEurope51590.2021.9584587
Zhou, Y. (2022). Artificial intelligence in renewable systems for transformation towards intelligent buildings. Energy and AI, 10, 100182. https://doi.org/10.1016/j.egyai.2022.100182