Application-Driven Learning: A Closed-Loop Prediction and Optimization Approach Applied to Dynamic Reserves and Demand Forecasting

Read original: arXiv:2102.13273 - Published 4/9/2024 by Joaquim Dias Garcia, Alexandre Street, Tito Homem-de-Mello, Francisco D. Mu~noz

🔮

Overview

This paper presents a new "closed-loop" framework for forecasting and decision-making, where these two processes are merged and co-optimized.
The proposed methodology, called "application-driven learning," is designed to converge to the best estimator in terms of the expected cost of the selected application.
The paper includes two solution methods: an exact method based on the KKT conditions of the second-level problem, and a scalable heuristic approach suitable for decomposition methods.
The methodology is applied to the problem of defining dynamic reserve requirements and conditional load forecasts, offering an alternative to current ad hoc industry practices.
The proposed approach is benchmarked against the standard sequential least-squares forecast and dispatch planning process, and is shown to be scalable and consistently outperform the open-loop approach.

Plain English Explanation

In the real world, forecasting (predicting future events) and decision-making (choosing the best course of action) are often treated as separate, sequential steps. This "open-loop" approach can lead to suboptimal results, as there is no feedback from the decision-making process back to the forecasting process.

This paper presents a new "closed-loop" framework called "application-driven learning" that merges the forecasting and decision-making processes. The key idea is to co-optimize these two steps, so that the forecasting model is tailored to the specific application or decision-making task at hand.

For example, imagine you're trying to forecast electricity demand and set appropriate reserve levels for a power grid. In the traditional approach, you would first forecast demand using historical data, then make decisions about reserve levels. In the new framework, the forecasting model would be optimized specifically to minimize the expected cost of the reserve decisions, rather than just minimizing the forecast error.

The paper presents two methods for solving this "bilevel optimization problem" - one exact approach and one scalable heuristic. The methods are tested on realistic power grid scenarios, and the results show that the new closed-loop approach consistently outperforms the traditional open-loop approach.

Technical Explanation

The paper presents a new closed-loop framework for forecasting and decision-making, where these two processes are merged and co-optimized through a bilevel optimization problem.

The first level of the optimization problem involves selecting the forecasting model parameters to minimize the expected cost of the downstream decision-making task. The second level involves making the optimal decisions given the forecasts from the first level.

The authors prove that the solution to this bilevel problem converges to the best forecasting estimator in terms of the expected cost of the selected application. They then propose two solution methods:

An exact method based on the Karush-Kuhn-Tucker (KKT) conditions of the second-level problem.
A scalable heuristic approach suitable for decomposition methods.

The proposed methodology is applied to the problem of defining dynamic reserve requirements and conditional load forecasts in power systems. This offers an alternative to the ad hoc procedures currently used in industry.

The authors benchmark their methodology against the standard sequential least-squares forecast and dispatch planning process. They test the approach on a range of system sizes, from dozens of buses to large-scale realistic systems with thousands of buses. The results demonstrate that the proposed application-driven learning framework is scalable and consistently outperforms the traditional open-loop approach.

Critical Analysis

The paper presents a promising new framework for integrating forecasting and decision-making, but there are a few potential limitations and areas for further research:

The paper focuses on a specific power systems application, but the framework is general and could potentially be applied to other domains. Further research is needed to explore its generalizability.
The paper does not address the computational complexity of solving the bilevel optimization problem, especially for large-scale systems. The scalable heuristic approach helps, but more work is needed to fully understand the scalability and tractability of the proposed methods.
The paper does not discuss the sensitivity of the results to the choice of forecasting model, decision-making objective, or other key assumptions. Understanding the robustness of the approach to these factors would be an important area for further investigation.
The paper does not compare the proposed framework to more advanced forecasting and decision-making techniques, such as those based on machine learning or robust optimization. Exploring these comparisons could provide further insights.

Overall, the paper presents a novel and promising approach to the fundamental challenge of integrating forecasting and decision-making. Further research to address the limitations and explore the broader applicability of the framework could lead to significant advancements in this important area.

Conclusion

This paper introduces a new "closed-loop" framework for integrating forecasting and decision-making, called "application-driven learning." The key idea is to co-optimize these two processes, rather than treating them as sequential steps, in order to converge to the best forecasting model for a given application.

The authors demonstrate the effectiveness of their approach on the problem of defining dynamic reserve requirements and conditional load forecasts in power systems. Their results show that the proposed methodology consistently outperforms the standard open-loop approach, while also being scalable to large-scale realistic systems.

The framework presented in this paper has the potential to significantly improve decision-making in a wide range of domains, by ensuring that forecasting models are tailored to the specific applications they are intended to support. Further research to explore the broader applicability and robustness of this approach could lead to important advancements in the field of integrated forecasting and decision-making.

This summary was produced with help from an AI and may contain inaccuracies - check out the links to read the original source documents!

Follow @aimodelsfyi on 𝕏 →

Related Papers

🔮

Application-Driven Learning: A Closed-Loop Prediction and Optimization Approach Applied to Dynamic Reserves and Demand Forecasting

Joaquim Dias Garcia, Alexandre Street, Tito Homem-de-Mello, Francisco D. Mu~noz

Forecasting and decision-making are generally modeled as two sequential steps with no feedback, following an open-loop approach. In this paper, we present application-driven learning, a new closed-loop framework in which the processes of forecasting and decision-making are merged and co-optimized through a bilevel optimization problem. We present our methodology in a general format and prove that the solution converges to the best estimator in terms of the expected cost of the selected application. Then, we propose two solution methods: an exact method based on the KKT conditions of the second-level problem and a scalable heuristic approach suitable for decomposition methods. The proposed methodology is applied to the relevant problem of defining dynamic reserve requirements and conditional load forecasts, offering an alternative approach to current ad hoc procedures implemented in industry practices. We benchmark our methodology with the standard sequential least-squares forecast and dispatch planning process. We apply the proposed methodology to an illustrative system and to a wide range of instances, from dozens of buses to large-scale realistic systems with thousands of buses. Our results show that the proposed methodology is scalable and yields consistently better performance than the standard open-loop approach.

4/9/2024

📶

Probabilistic Load Forecasting Based on Adaptive Online Learning

Ver'onica 'Alvarez, Santiago Mazuelas, Jos'e A. Lozano

Load forecasting is crucial for multiple energy management tasks such as scheduling generation capacity, planning supply and demand, and minimizing energy trade costs. Such relevance has increased even more in recent years due to the integration of renewable energies, electric cars, and microgrids. Conventional load forecasting techniques obtain single-value load forecasts by exploiting consumption patterns of past load demand. However, such techniques cannot assess intrinsic uncertainties in load demand, and cannot capture dynamic changes in consumption patterns. To address these problems, this paper presents a method for probabilistic load forecasting based on the adaptive online learning of hidden Markov models. We propose learning and forecasting techniques with theoretical guarantees, and experimentally assess their performance in multiple scenarios. In particular, we develop adaptive online learning techniques that update model parameters recursively, and sequential prediction techniques that obtain probabilistic forecasts using the most recent parameters. The performance of the method is evaluated using multiple datasets corresponding with regions that have different sizes and display assorted time-varying consumption patterns. The results show that the proposed method can significantly improve the performance of existing techniques for a wide range of scenarios.

8/16/2024

📉

Decision-Focused Forecasting: Decision Losses for Multistage Optimisation

Egon Perv{s}ak, Miguel F. Anjos

Decision-focused learning has emerged as a promising approach for decision making under uncertainty by training the upstream predictive aspect of the pipeline with respect to the quality of the downstream decisions. Most existing work has focused on single stage problems. Many real-world decision problems are more appropriately modelled using multistage optimisation as contextual information such as prices or demand is revealed over time and decisions now have a bearing on future decisions. We propose decision-focused forecasting, a multiple-implicitlayer model which in its training accounts for the intertemporal decision effects of forecasts using differentiable optimisation. The recursive model reflects a fully differentiable multistage optimisation approach. We present an analysis of the gradients produced by this model showing the adjustments made to account for the state-path caused by forecasting. We demonstrate an application of the model to an energy storage arbitrage task and report that our model outperforms existing approaches.

5/24/2024

Learning Joint Models of Prediction and Optimization

James Kotary, Vincenzo Di Vito, Jacob Cristopher, Pascal Van Hentenryck, Ferdinando Fioretto

The Predict-Then-Optimize framework uses machine learning models to predict unknown parameters of an optimization problem from exogenous features before solving. This setting is common to many real-world decision processes, and recently it has been shown that decision quality can be substantially improved by solving and differentiating the optimization problem within an end-to-end training loop. However, this approach requires significant computational effort in addition to handcrafted, problem-specific rules for backpropagation through the optimization step, challenging its applicability to a broad class of optimization problems. This paper proposes an alternative method, in which optimal solutions are learned directly from the observable features by joint predictive models. The approach is generic, and based on an adaptation of the Learning-to-Optimize paradigm, from which a rich variety of existing techniques can be employed. Experimental evaluations show the ability of several Learning-to-Optimize methods to provide efficient and accurate solutions to an array of challenging Predict-Then-Optimize problems.

9/10/2024