Neural Predictive PV Interruption Control for Battery-Backed Microgrids
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems lack an efficient control mechanism to smooth solar photovoltaic power fluctuations while optimizing battery state of charge and reducing ramp rate under practical constraints, which is crucial for stable grid integration of intermittent renewable energy sources.
Innovation Solution
A neural network-based predictive controller (NNPC) system that combines model predictive control with neural networks and a battery energy storage system to predict and smooth solar PV power fluctuations, using inputs from solar irradiance, temperature, and other environmental factors to manage battery charging/discharging and optimize power delivery to the grid.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Stability of the object's composition
If traditional control methods (fuzzy logic controllers, low pass filters) are used to smooth solar PV power fluctuations, then power delivery stability is improved, but the ability to optimize battery state of charge and extend battery life deteriorates
Solution Approach 1:
The neural network predictor performs preliminary action by forecasting future PV power output before fluctuations occur. This predictive capability allows the controller to proactively adjust battery charging/discharging strategies, optimizing state of charge in advance rather than reactively responding to fluctuations, thereby simultaneously achieving power stability and battery optimization
Solution Approach 2:
The system implements feedback by continuously monitoring actual PV power output, battery state of charge, and comparing them with predicted values. The neural network controller adjusts control parameters based on this feedback, creating a closed-loop system that optimizes both power delivery stability and battery management dynamically
2Duration of action of stationary object
If model predictive control is used to optimize battery operation, then battery lifespan is improved, but computational complexity and system response time worsen
Solution Approach 1:
The neural network predictor performs preliminary computation by forecasting PV power output in advance. This pre-computed prediction is then used by the controller to determine optimal battery operation strategies, shifting computational burden to the prediction phase and enabling faster real-time control decisions that extend battery life without excessive computational delay
Solution Approach 2:
The system uses a neural network model that copies and learns from historical PV power patterns and environmental conditions. This learned model provides rapid predictions without requiring complex real-time optimization calculations, thus extending battery lifespan through intelligent control while maintaining fast response times
3Stability of the object's composition
If neural network prediction is used to forecast PV power output, then power fluctuation smoothing is improved, but system complexity and computational requirements worsen
Solution Approach 1:
The neural network predictor serves multiple functions simultaneously: it forecasts PV power output, identifies fluctuation patterns, and provides input for battery control optimization. This multi-functionality achieves effective power smoothing without requiring separate dedicated systems for each function, thereby reducing overall system complexity despite the advanced prediction capability
Solution Approach 2:
The neural network model is trained offline using historical data and then operates autonomously to predict PV power output based on current environmental conditions. This self-service capability eliminates the need for complex real-time data processing and manual intervention, achieving power fluctuation smoothing through the predictor's inherent learning ability while keeping the online system relatively simple
Data Source
AI summary
An intermittent power system to provide smoothed electric power into a power grid that includes an intermittent power source, a neural network-based predictive controller (NNPC) and a low pass filter (LPF) connected to the power grid to provide the smoothed electric power. The LPF provides a smoothed power reference for the NNPC. The system further includes a neural network predictor connected between the intermittent power source and the NNPC, and a power grid connection. The neural network predictor takes electric power from the intermittent power source as an input and makes a prediction of unsmoothed electric power. A power grid connection provides the smoothed electric power of the NNPC into the power grid. The NN model solves issues related to the mathematical complexity of a conventional MPC model that arises due to the increasing complications in an intermittent power plant, including a PV power plant.