Final Report Summary - SOPRIS (Stochastic Optimal Planning for Renewable energy sources Integration in power Systems)
The high penetration of renewable power generation is rendering the power system operation vulnerable. Renewable Energy Sources (RES) are typically non-dispatchable, variable and uncertain which is making difficult the decision-making process of the power system operator.
The Stochastic Optimal Planning for Renewable Energy Sources Integration in Power Systems (SOPRIS) project focused on the development of a unified stochastic framework for optimal decision making, considering uncertainty inputs such as generation from RES while incorporating various controllable network components.
Objectives
The final outcome of the project is a toolbox that provides analyses and solutions of various stochastic power system operational problems. To achieve this goal the workflow of the project had the following objectives:
• Development of models for the uncertain elements in the network.
• Corrective control: modeling certain network elements that are treated as controllability assets.
• Construction of demand response mechanisms for reserve provision and deployment.
• Development of a stochastic, optimization-based framework to address unit commitment, reserve provision and energy scheduling problems, while taking network and reliability constraints into account. Demand response and corrective control actions are integrated in this framework.
• Distributed implementation of the developed algorithms to facilitate their application on large scale
networks.
Results
The main results of the project are summarized in the frame of the organized project tasks:
Task 1: Modelling
• Task 1.1: Uncertainty modelling
The statistical properties of the uncertainty source (e.g. wind power) are analyzed and appropriate models to generate synthetic time series of possible realizations of the uncertainty in the time scale that serves hourly planning problems is built. Those uncertainty realizations or scenarios is the input to the main core of the project; the stochastic optimization programs. The statistical model characterized by advanced accuracy and computational efficiency that was more appropriate to serve as an input for the following tasks was based on a Markov chain Monte Carlo (MCMC) method. This model considers the time correlation of the wind power errors and their dependency on the wind power forecasts. We included also the spatial correlation of wind power generation, increasing though the computational burden of the construction of the transition probability matrix.
• Task 1.2: Network modeling and corrective control.
Integration of RES in the power network increases the level of uncertainty in the system and hence the
amount of reserves needed to counteract the variability in the production. To mitigate this, we proposed the introduction of corrective control schemes that exploit the controllability of certain network components. Adequate models for the corrective control actions offered by controllable components represent the steady state behavior of these network elements by piecewise affine decision function of the uncertainty serving as disturbance compensators.
• Task 1.3: Demand response - Concepts and models.
In this task we focused on the mechanisms for corrective control and reserve provision offered by flexible loads and specifically Thermostatically Controlled Loads (TCLs). We built an accurate model of the TCLs appropriate for the integration in the designed optimization algorithms and a market setup is proposed for a realistic implementation. An analysis concerning the pricing of the demand response for providing reserves is also conducted. The latter resulted in sufficient rules concerning the prices of generators and loads reserves, highlighting under which conditions loads will be preferred to be used as the reserve resource and when the generation cost profile will not allow it.
Task 2: Optimal decision making in power networks with uncertainty
Task 2.1: Problem formulation - Task 2.2: Algorithms
The study in this task is the core of SOPRIS project and required the integration of the models developed
within Task 1. Here we developed stochastic optimization formulations that allows us to optimize simultaneously with respect to unit commitment and reserve decisions, taking both network and reliability constraints into account, but also utilizing the services that various resources can provide. Moreover, scenario-based optimization techniques were investigated to appropriately formulate the stochastic variant of the above-mentioned problem. Towards a unified decision-making tool, in the course of the first reporting period, four independent studies are carried out.
The main results include the following formulations:
1. A reserve scheduling optimal power flow incorporating piecewise affine policies for the control of HVDC lines and generation as a function of the wind power forecast error. The optimal policies were identified using multi-parametric optimization.
2. An accurate reserve scheduling mechanism including demand response. Sufficient pricing rules relating the generation and load reserves costs and the final procurement preference of reserve resource were obtained.
3. A stochastic AC optimal power flow using an AC-QP algorithm using a scenario-based optimization technique that accompanies the solution with a-posteriori probabilistic guaranties for non-convex problems. The resulting formulation provides a probabilistically robust AC feasible solution over the uncertainty and shows promising scalability properties.
4. A stochastic AC-QP OPF algorithm in a planning context determining the maximum wind penetration that can be integrated in a network while maintaining reliability standards.
Task 3: Distributed implementation and computational aspects
Task 3.1: Distributed implementation
The ADMM method has been applied in the problem of the chance constrained OPF and the convergence and the scalability performance of the technique with respect to the network size and the number of scenarios is analyzed.
Task 3.2: Toolbox and user interface
We embedded in a single platform the various steps of algorithms developed in the project so as to provide a toolbox that could assist the power system operator. We are providing a graphical user interface in the form of a Matlab app and the user have the option to choose among the different algorithmic alternatives and automatically get information regarding the robustness properties of the resulting solution. The final release of the toolbox in the project webpage is under way.
Task 4: Validation
Validation of the stochastic algorithms is conducted via Monte Carlo simulations using aggregated wind power forecast error time series of Germany over 2006-2011 by Fraumhofer IWES in Kassel and also using wind power data based on the Eastern Wind Power Dataset provided by NREL that contain both spatial and temporal correlations of different wind plants.
Conclusions
The final results of the SOPRIS project include the development of a novel mechanism for making optimal decisions in power networks with RES. This mechanism integrates demand response actions and exploits the controllability of certain network components. Emphasis is also given on the algorithmic side, proposing the use of recent stochastic optimization techniques to account for the uncertainty in the design phase, and distributed optimization to facilitate the application of the developed methodologies to large scale systems.
The analysis of the results will provide directions on the investments on RES installation on the grid. Directions for appropriate placement of controllable network components that characterizes the level of this penetration can also be given. Moreover, insights for policy makers that aim in integrating in the market thermostatically controlled loads as reserve resources are given. The integration of the developed toolbox in the decision process of the Transmission System Operator (TSO) will enhance the reliability of the operation in the most economical way.
More information on the SOPRIS project can be found in https://soprisproject.wordpress.com/(odnośnik otworzy się w nowym oknie)
The Stochastic Optimal Planning for Renewable Energy Sources Integration in Power Systems (SOPRIS) project focused on the development of a unified stochastic framework for optimal decision making, considering uncertainty inputs such as generation from RES while incorporating various controllable network components.
Objectives
The final outcome of the project is a toolbox that provides analyses and solutions of various stochastic power system operational problems. To achieve this goal the workflow of the project had the following objectives:
• Development of models for the uncertain elements in the network.
• Corrective control: modeling certain network elements that are treated as controllability assets.
• Construction of demand response mechanisms for reserve provision and deployment.
• Development of a stochastic, optimization-based framework to address unit commitment, reserve provision and energy scheduling problems, while taking network and reliability constraints into account. Demand response and corrective control actions are integrated in this framework.
• Distributed implementation of the developed algorithms to facilitate their application on large scale
networks.
Results
The main results of the project are summarized in the frame of the organized project tasks:
Task 1: Modelling
• Task 1.1: Uncertainty modelling
The statistical properties of the uncertainty source (e.g. wind power) are analyzed and appropriate models to generate synthetic time series of possible realizations of the uncertainty in the time scale that serves hourly planning problems is built. Those uncertainty realizations or scenarios is the input to the main core of the project; the stochastic optimization programs. The statistical model characterized by advanced accuracy and computational efficiency that was more appropriate to serve as an input for the following tasks was based on a Markov chain Monte Carlo (MCMC) method. This model considers the time correlation of the wind power errors and their dependency on the wind power forecasts. We included also the spatial correlation of wind power generation, increasing though the computational burden of the construction of the transition probability matrix.
• Task 1.2: Network modeling and corrective control.
Integration of RES in the power network increases the level of uncertainty in the system and hence the
amount of reserves needed to counteract the variability in the production. To mitigate this, we proposed the introduction of corrective control schemes that exploit the controllability of certain network components. Adequate models for the corrective control actions offered by controllable components represent the steady state behavior of these network elements by piecewise affine decision function of the uncertainty serving as disturbance compensators.
• Task 1.3: Demand response - Concepts and models.
In this task we focused on the mechanisms for corrective control and reserve provision offered by flexible loads and specifically Thermostatically Controlled Loads (TCLs). We built an accurate model of the TCLs appropriate for the integration in the designed optimization algorithms and a market setup is proposed for a realistic implementation. An analysis concerning the pricing of the demand response for providing reserves is also conducted. The latter resulted in sufficient rules concerning the prices of generators and loads reserves, highlighting under which conditions loads will be preferred to be used as the reserve resource and when the generation cost profile will not allow it.
Task 2: Optimal decision making in power networks with uncertainty
Task 2.1: Problem formulation - Task 2.2: Algorithms
The study in this task is the core of SOPRIS project and required the integration of the models developed
within Task 1. Here we developed stochastic optimization formulations that allows us to optimize simultaneously with respect to unit commitment and reserve decisions, taking both network and reliability constraints into account, but also utilizing the services that various resources can provide. Moreover, scenario-based optimization techniques were investigated to appropriately formulate the stochastic variant of the above-mentioned problem. Towards a unified decision-making tool, in the course of the first reporting period, four independent studies are carried out.
The main results include the following formulations:
1. A reserve scheduling optimal power flow incorporating piecewise affine policies for the control of HVDC lines and generation as a function of the wind power forecast error. The optimal policies were identified using multi-parametric optimization.
2. An accurate reserve scheduling mechanism including demand response. Sufficient pricing rules relating the generation and load reserves costs and the final procurement preference of reserve resource were obtained.
3. A stochastic AC optimal power flow using an AC-QP algorithm using a scenario-based optimization technique that accompanies the solution with a-posteriori probabilistic guaranties for non-convex problems. The resulting formulation provides a probabilistically robust AC feasible solution over the uncertainty and shows promising scalability properties.
4. A stochastic AC-QP OPF algorithm in a planning context determining the maximum wind penetration that can be integrated in a network while maintaining reliability standards.
Task 3: Distributed implementation and computational aspects
Task 3.1: Distributed implementation
The ADMM method has been applied in the problem of the chance constrained OPF and the convergence and the scalability performance of the technique with respect to the network size and the number of scenarios is analyzed.
Task 3.2: Toolbox and user interface
We embedded in a single platform the various steps of algorithms developed in the project so as to provide a toolbox that could assist the power system operator. We are providing a graphical user interface in the form of a Matlab app and the user have the option to choose among the different algorithmic alternatives and automatically get information regarding the robustness properties of the resulting solution. The final release of the toolbox in the project webpage is under way.
Task 4: Validation
Validation of the stochastic algorithms is conducted via Monte Carlo simulations using aggregated wind power forecast error time series of Germany over 2006-2011 by Fraumhofer IWES in Kassel and also using wind power data based on the Eastern Wind Power Dataset provided by NREL that contain both spatial and temporal correlations of different wind plants.
Conclusions
The final results of the SOPRIS project include the development of a novel mechanism for making optimal decisions in power networks with RES. This mechanism integrates demand response actions and exploits the controllability of certain network components. Emphasis is also given on the algorithmic side, proposing the use of recent stochastic optimization techniques to account for the uncertainty in the design phase, and distributed optimization to facilitate the application of the developed methodologies to large scale systems.
The analysis of the results will provide directions on the investments on RES installation on the grid. Directions for appropriate placement of controllable network components that characterizes the level of this penetration can also be given. Moreover, insights for policy makers that aim in integrating in the market thermostatically controlled loads as reserve resources are given. The integration of the developed toolbox in the decision process of the Transmission System Operator (TSO) will enhance the reliability of the operation in the most economical way.
More information on the SOPRIS project can be found in https://soprisproject.wordpress.com/(odnośnik otworzy się w nowym oknie)