🔥🔥 Don’t Miss Out on Our End of Autumn Sale. If you have any questions, feel free to click the bottom right corner to contact our customer service..

Concat us[email protected]
Shopping Cart

No products in the cart.

🔍
BSI PD IEC TR 63043:2020

BSI PD IEC TR 63043:2020

$215.11

Renewable energy power forecasting technology

Published By Publication Date Number of Pages
BSI 2020 142
PDF Format Searchable Printable Preview Now
Guaranteed Safe Checkout
Category:

If you have any questions, feel free to reach out to our online customer service team by clicking on the bottom right corner. We’re here to assist you 24/7.
Email:[email protected]

This Technical Report, which is informative in its nature, describes common practices and state of the art for renewable energy power forecasting technology, including general data demands, renewable energy power forecasting methods and forecasting error evaluation. For the purposes of this document, renewable energy refers to variable renewable energy, which mainly comprises wind power and photovoltaic (PV) power – these are the focus of the document. Other variable renewable energies, like concentrating solar power, wave power and tidal power, etc., are not presented in this document, since their capacity is small, while hydro power forecasting is a significantly different field, and so not covered here.

The objects of renewable energy power forecasting can be wind turbines, or a wind farm, or a region with lots of wind farms (respectively PV systems, PV power stations and regions with high PV penetration). This document focuses on providing technical guidance concerning forecasting technologies of multiple spatial and temporal scales, probabilistic forecasting, and ramp event forecasting for wind power and PV power.

This document outlines the basic aspects of renewable energy power forecasting technology. This is the first IEC document related to renewable energy power forecasting. The contents of this document will find an application in the following potential areas:

  • support the development and future research for renewable energy power forecasting technology, by showing current state of the art;

  • evaluation of the forecasting performance during the design and operation of renewable energy power forecasting system;

  • provide information for benchmarking renewable forecasting technologies, including methods used, data required and evaluation techniques.

PDF Catalog

PDF Pages PDF Title
2 undefined
4 CONTENTS
9 FOREWORD
11 INTRODUCTION
12 1 Scope
2 Normative references
3 Terms, definitions and abbreviated terms
13 3.1 Terms and definitions
15 3.2 Abbreviated terms
17 4 General introduction to renewable energy power forecasting
4.1 History of RPF
4.1.1 General
18 4.1.2 Development of wind power forecasting
19 4.1.3 Development of PV power forecasting
4.2 Use of RPF
4.2.1 General
20 4.2.2 RPF for system operations
4.2.3 RPF for power trading
21 4.2.4 RPF for operations and maintenance
4.3 Methods for forecasting renewable power
4.3.1 General
4.3.2 Classification of forecasting methods
Tables
Table 1 – Classification of RPF methods
23 4.3.3 Classification based on time scale
Figures
Figure 1 – Forecasting of PV power at different spatial and temporal scales
Figure 2 – Introduced data for PV power forecastingat different spatial and temporal scales
24 4.3.4 Classification based on spatial range
4.3.5 Classification based on the forecasting model
26 4.3.6 Classification based on the forecasting form
27 4.4 Summary
5 NWP technology
5.1 General
5.2 Concept and characteristics of NWP
28 Figure 3 – Typical process for running a regional model
29 5.3 Influence on RPF accuracy
5.3.1 Sensitivity analysis
Figure 4 – Power curve of typical wind turbines
30 5.3.2 Error source analysis
Figure 5 – Characteristics of three kinds of forecasting errors
31 5.4 Technology progress for improving NWP
5.4.1 General
5.4.2 Global model
32 Figure 6 – Evolution of ECMWF’s forecasting skillsfor the 500 hPa potential height [35], [54]
Table 2 – Features of global NWP models
33 5.4.3 Regional model
5.5 Key techniques for improving the forecast accuracy of regional models
5.5.1 Improve the accuracy of the initial conditions
34 5.5.2 Ensemble prediction systems
35 Figure 7 – Ensemble forecasting sketch [54]
39 Table 3 – Comparison of different ensemble predictionmethodologies and their attributes [46], [73]
40 5.5.3 Establish regional customized forecasting model
Figure 8 – Illustration of parameterization schemesfor sub-grid physical processes [54]
41 5.5.4 NWP post-processing
5.6 Summary
6 Statistical methods
6.1 General
42 6.2 Methods
44 6.3 Applications
6.3.1 General
6.3.2 Time series models
46 Figure 9 – MAE (% of capacity) versus look-ahead time for 0 h to 3 h forecasts of the 15 min average wind power production from the TWRA aggregate over the one-year period from October 2015 to September 2016 for each of 5 source-dependent sets of predictors employed in the predictor source category experiment [96]
47 Figure 10 – Percentage MAE reduction over persistence by look-ahead time achieved by each source-dependent set of predictors for 0 h to 3 h forecasts of the 15 min average TWRA aggregate (capacity of 2 319 MW) power production over the one-year period from October 2015 to September 2016 [96]
48 Figure 11 – Percentage MAE reduction by look-ahead time achieved by building forecasting models with the XGBoost method versus MLR for the “Add existing external data” (set #4) and “Add targeted sensors” (set #5) predictor sets for 0 h to 3 h forecasts of the 15 min average TWRA aggregate (capacity of 2 319 MW) power production over the one year period from October 2015 to September 2016 [96]
49 6.3.3 Model output statistics (MOS)
Figure 12 – Percentage MAE reduction by look-ahead time achieved by using the “rate of change” (indirect forecasting) versus “the 15 min average power generation” (direct forecasting) as the target predictand for the XGBoost model for 0 h to 3 h forecasts of the 15 min average TWRA aggregate (capacity of 2 319 MW) power production over the one year period from October 2015 to September 2016 [96]
50 Figure 13 – Mean absolute error (MAE) in m/s of two 0 h to 18 h NWP-MOS forecasts of the maximum wind gust in a 15 min period for 33 sites over a 32-case sample of high wind events as a function of training sample size
51 Figure 14 – Percentage reduction in the mean absolute error of NWP-based 0 h to 15 h wind power forecasts for the Tehachapi Wind Resource Area (TWRA) over a one-year period resulting from the application of 26 statistical forecasting methods to the output from the United States National Weather Service’s High Resolution Rapid Refresh (HRRR) model [96]
53 6.3.4 Ensemble composite models (ECM)
Figure 15 – Percentage reduction in the mean absolute error (MAE) of wind power forecasts relative to a baseline of a raw NWP forecast for three NWP models when a MOS procedure is applied to the NWP output (larger percentages are better)
55 6.3.5 Power output models
56 7 Wind power forecasting (WPF) technology
7.1 General
7.2 Short-term WPF
7.2.1 Relationship between wind power output and meteorological elements
Figure 16 – Input and output parameters of the three-days-ahead WPF
57 Figure 17 – Wind power output at different wind speeds underair density of 1,225 kg/m3 (a typical 2 MW wind turbine)
58 Figure 18 – EC distribution of a wind farm at different wind speeds and directions
59 7.2.2 Framework of short-term WPF
Figure 19 – Wind speed and wind power curvesof wind turbines at different air densities
60 7.2.3 Short-term WPF methods
Figure 20 – Typical framework of short-term WPF
61 Figure 21 – Principle of short-term WPF based on physical approaches
62 Figure 22 – Flowchart of short-term WPF based on statistical approaches
Figure 23 – Short-term WPF model based on ANN
64 7.3 Ultra-short-term WPF
Figure 24 – Input and output parameters of the 4 h ultra-short-term WPF
65 Figure 25 – Flowchart of ultra-short-term WPF
66 Figure 26 – Generalized combination methods of ultra-short-term WPF
67 7.4 Probabilistic WPF
7.4.1 General
7.4.2 Basic concepts and model framework definition
Figure 27 – Methods used for probabilistic forecasting
68 7.4.3 Uncertainty modeling approaches
Figure 28 – Overview of probabilistic wind power forecasting
69 7.4.4 Probabilistic WPF model
Figure 29 – Wind power probability distribution forecasting results
Table 4 – Output modes of probabilistic forecasting
70 Figure 30 – Filtering approach with ensemble NWP as input
71 Figure 31 – Dimension reduction approach with ensemble NWP as input
Figure 32 – Direct approach with ensemble NWP as input
73 7.5 Wind power ramp event forecasting
7.5.1 General
7.5.2 Quantitative description of wind power ramp events
74 Figure 33 – Two ramp events of a wind farm
75 Table 5 – Advantages and disadvantages of ramp events definitions
76 7.5.3 Forecasting methods of wind power ramp events
77 7.6 WPF for wind farm clusters
7.6.1 General
7.6.2 Basic concepts of WPF for wind farm clusters
78 7.6.3 Overall framework of the WPF for wind farm clusters
79 Figure 34 – Overall framework of the WPF system for wind farm clusters
Table 6 – Data sources of WPF for wind farm clusters
80 7.6.4 Physical hierarchy of WPF for wind farm clusters
Figure 35 – Physical levels of WPF for wind farm clusters
81 7.6.5 WPF methods of wind farm clusters
Figure 36 – Flow chart of the accumulation method
82 Figure 37 – Flow chart of the statistical upscaling method
83 Figure 38 – Flow chart of the space resource matching method
Table 7 – Comparison of WPF methods for wind farm clusters.
84 7.7 Other WPF techniques
7.7.1 Medium-term and long-term WPF
7.7.2 WPF for offshore wind farms
85 7.8 Summary
8 PV power forecasting technology
8.1 General
8.2 Short-term PVPF
8.2.1 General
8.2.2 Meteorological influence factors of PV power generation
86 Figure 39 – Volt-ampere characteristic curve of PV modulescorresponding to different irradiance
87 Figure 40 – Volt-ampere characteristics of PV modules at different temperatures
88 8.2.3 Basic concepts for short-term PVPF
89 8.2.4 Short-term PVPF model
Figure 41 – Short-term forecasting models of PV power generation
91 8.2.5 Trends in PVPF development and key technical issues
8.3 Ultra-short-term PVPF
8.3.1 General
Figure 42 – PV short-term power physical forecasting method technical route
92 8.3.2 Basic concepts for ultra-short-term PVPF
8.3.3 Ultra-short-term PVPF models
93 Figure 43 – Basic technology roadmap for pv power ultra-short-term forecasting
Figure 44 – Ultra-short-term PVPF based on machine learning model
94 8.3.4 Trends in development and key technical issues
8.4 Minute-time-scale PVPF
95 8.4.1 Basic concepts for minute-time-scale solar power forecasting
8.4.2 Technique routine of minute-time-scale solar power forecasting
96 8.4.3 Trends in development and key technical issues
Figure 45 – Minute-time-scale solar power forecasting technique process
97 8.5 Probabilistic PVPF
8.5.1 Basic concepts of PV power probabilistic forecasting
98 8.5.2 Probabilistic PVPF model
Figure 46 – Example of probabilistic PV model
Figure 47 – Forecasting process of physical PV power probabilistic forecasting model
99 Figure 48 – Forecasting process of statistical probabilistic PVPF model
100 8.5.3 Trends in development and key technical issues
8.6 Distributed PVPF
8.6.1 General
101 8.6.2 Basic concepts for distributed PVPF
8.6.3 Distributed PVPF methods
102 Figure 49 – Framework of clustering statistical forecastingmethod for distributed PVPF
103 Figure 50 – Framework of grid forecasting method for distributed PVPF
104 8.6.4 Trends in development and key technical issues
8.7 Summary
Figure 51 – Comparison between the forecasting resultsof the clustering statistical method and the grid forecast method
105 9 Renewable energy power forecasting (RPF) evaluation
9.1 General
106 9.2 Deterministic forecasts of continuous variables
9.2.1 General
9.2.2 Metrics
9.2.3 Mean bias error
107 9.2.4 Mean absolute error
9.2.5 Root mean square error
108 9.2.6 Skill score
9.2.7 Correlation coefficient
109 9.2.8 Maximum prediction error
9.2.9 Pass rate
110 9.2.10 95 % QDR
111 9.2.11 Customized metrics
9.3 Deterministic forecasts of categorical (event) variables
9.3.1 General
112 9.3.2 Occurrence/non-occurrence metrics
9.3.3 Frequency bias
9.3.4 Probability of detection
Table 8 – Contingency table for forecasts ofthe occurrence/non-occurrence of an event
113 9.3.5 False alarm ratio
9.3.6 Critical success index
9.3.7 Equitable threat score
9.3.8 Heidke skill score
114 9.4 Probabilistic forecasts of categorical (event) variables
9.4.1 General
9.4.2 Overall performance
118 9.4.3 Reliability
119 9.4.4 Resolution
Figure 52 – Example of a reliability diagram for two probabilistic forecasts(Forecast A and Forecast B) of a binary event
120 9.5 Probabilistic forecasts of continuous variables
9.5.1 General
9.5.2 Overall performance
121 9.5.3 Reliability
9.5.4 Resolution
9.6 Sources of forecast error
122 9.7 Comparison of forecast performance
123 Table 9 – A summary of recommended metrics for frequently used forecast types
124 9.8 Selection of an optimal forecast solution
125 10 Conclusions and recommendations
128 Bibliography

Related products

BSI PD IEC TR 63043:2020
$215.11