丽水电网风光水储智慧调度辅助决策平台应用分析数据

采集风光水储基础信息，实现丽水地区智慧调度辅助决策平台的数据全景感知；通过大数据和人工智能搭建新能源发电能力预测算法，支持多时间尺度下的出力预测结果展示；构建风光水储出力预测结果模型，实现集群联合优化调度控制策略；基于径流式与库容式水库出力模型，提升典型场景下出力分配策略能力。1、数据清洗：采用聚类等数据处理手段对缺失值、异常值等进行删补，对各类基础数据中的非规范化数据进行转换。
2、数据转换:基于清洗数据与识别到的模型特征，通过人工智能和数据挖掘算法，对数据集训练，提高出力预测算法结果。
3、数据加工：处理全社会非统调出力环比增减、预测准确率及平均绝对误差等数据。
全社会非统调出力环比增减：（某时刻出力值-上一日某时刻出力值）/上一日某时刻出力值；
平均绝对误差：先计算每点（实际功率-预测功率）/装机容量，各点加和后取均值；
预测曲线最大误差：对所选日期内每天96点的实际功率和预测功率分别进行误差计算，误差=（实际功率-预测功率），选取计算结果绝对值最大的在界面显示。若实际功率或预测功率存在0值时，另一功率不超过3%的装机容量，则为0，否则为100%装机容量；
合格率：先计算每点（实际功率-预测功率）/装机容量，绝对值与25%做比较，若小于25%，则记为一次合格，用合格总次数除以总点数；
相关系数：时刻累计（预测功率-预测功率平均值）×（实际功率-实际功率平均值）÷根号{时刻累计（预测功率-预测功率平均值）²×时刻累计（实际功率-实际功率平均值）²}

This dataset collects basic information of wind, solar, hydropower and energy storage to enable comprehensive data perception for the intelligent dispatching auxiliary decision-making platform in Lishui Region. It constructs new energy generation capacity prediction algorithms via big data and artificial intelligence, supporting the display of output prediction results across multiple time scales. A prediction result model for wind, solar, hydropower and energy storage output is developed to realize cluster joint optimal dispatching and control strategies. Based on the output models of run-of-river and reservoir-based hydropower stations, the capability of output allocation strategies in typical scenarios is enhanced.

1. Data Cleaning: Adopt data processing methods such as clustering to delete and impute missing values, outliers and other abnormal data, and convert non-standardized data in various basic datasets.
2. Data Transformation: Based on the cleaned data and identified model features, train the dataset using artificial intelligence and data mining algorithms to improve the performance of output prediction algorithms.
3. Data Processing: Process indicators including the month-on-month change of non-unified dispatching output across the entire society, prediction accuracy, mean absolute error, etc.
- Month-on-month change of non-unified dispatching output across the entire society: (Output value at a certain moment - Output value at the same moment of the previous day) / Output value at the same moment of the previous day;
- Mean Absolute Error (MAE): Calculate (Actual Power - Predicted Power) / Installed Capacity for each data point, sum the values of all points, then take the average;
- Maximum Error of Prediction Curve: Calculate the error (Error = Actual Power - Predicted Power) for each of the 96 data points per day within the selected date range. The result with the largest absolute value is selected for display on the interface. If either the actual power or predicted power is 0, and the other power does not exceed 3% of the installed capacity, the error is set to 0; otherwise, it is set to 100% of the installed capacity;
- Qualification Rate: Calculate (Actual Power - Predicted Power) / Installed Capacity for each data point, compare the absolute value with 25%. If the value is less than 25%, count it as one qualified instance. Divide the total number of qualified instances by the total number of data points;
- Correlation Coefficient: The formula is Σ[(Predicted Power - Average Predicted Power) × (Actual Power - Average Actual Power)] / √[Σ(Predicted Power - Average Predicted Power)² × Σ(Actual Power - Average Actual Power)²]

搜集汇总

数据集介绍

特点

丽水电网风光水储智慧调度辅助决策平台应用分析数据集包含799条记录，每日更新，涵盖电站信息、功率预测和气象数据等多种字段，用于支持新能源发电能力预测和优化调度控制策略。