Replication Data for: A gradient boosting approach for optimal selection of bidding strategies: Simple model - Original variables

Access to an increasing amount of data opens for the application of machine learning models to predict the best combination of models and strategies for bidding of hydro power in a de-regulated market for any given day. This data-set describe the historical performance-gap of two given bidding strategies over several years (2016-2018). Data from two different bidding strategies are presented in the data-set. The first is bidding the expected volume. The expected volumes are found by deterministic optimization against forecasted price and inflow using the SHOP software, and are submitted as fixed hourly bids to the Nord Pool power exchange. The second strategy is stochastic bidding. The stochastic model is based on the deterministic method, but allows for a stochastic representation of inflow to the reservoir and day-ahead market prices. SHOP is a software tool for optimal short-term hydropower scheduling developed by SINTEF Energy Research, used by many hydropower producers in the Nordic market. The total performance-gap for the two strategies in the data-set are calculated as the difference between the optimum value for the relevant bidding date and the value of the investigated strategy. A high number for indicate poor performance. In addition, a set of of relevant variables accessible prior to bidding have been collected and are published in the data-set. Realized- and prognosed prices in the data-set are prices for the NO2 area in Nordpool. The reservoir and watervalue in the data-set are associated with a river system located in south-western Norway

随着数据量的日益增长，为机器学习模型的应用提供了可能，以预测在任何特定一天内，在去监管的电力市场中，水力发电投标的最佳模型组合及策略。该数据集描述了两种给定投标策略在数年（2016-2018年）间的历史性能差距。数据集中展示了两种不同投标策略的数据。第一种是预期体积投标，通过使用SINTEF能源研究开发的SHOP软件，针对预测的价格和入流进行确定性优化，并将固定小时投标提交至Nord Pool电力交易所。第二种策略为随机投标。随机模型基于确定性方法，但允许对水库入流和次日市场价格进行随机表示。SHOP是一款用于最优短期水力发电调度的软件工具，由SINTEF能源研究开发，被北欧市场的众多水力发电企业所采用。数据集中两种策略的总性能差距是通过相关投标日期的最优值与调查策略的值之间的差异来计算的。数值越高，表明性能越差。此外，收集并发布在数据集中的是一组在投标前可获取的相关变量。数据集中的实现价格和预测价格是Nordpool NO2区域的报价。数据集中的水库和水价值与位于挪威西南部的河流系统相关。