SciGRID is a project which started in and will be running for three years. GridKit uses spatial and topological analysis to transform map objects from OpenStreetMap into a network model of the electric power system. The abstraction and all additional modules are controlled by a Python-environment. The 2nd version of the Bialek European Model is downloadable as an Excel file and in the format of the proprietary modelling software PowerWorld. The model covers voltages from kV a single line in the Balkans up to kV. It is released under a Public Domain Dedication.
|Published (Last):||8 May 2006|
|PDF File Size:||4.57 Mb|
|ePub File Size:||3.36 Mb|
|Price:||Free* [*Free Regsitration Required]|
Limited data on global power infrastructure makes it difficult to respond to challenges in electricity access and climate change. Although high-voltage data on transmission networks are often available, medium- and low-voltage data are often non-existent or unavailable.
This presents a challenge for practitioners working on the electricity access agenda, power sector resilience or climate change adaptation. Using state-of-the-art algorithms in geospatial data analysis, we create a first composite map of the global power system with an open license. The results from this study pave the way for improved efforts in electricity modelling and planning and are an important step in tackling the Sustainable Development Goals. Reliable infrastructure networks form the backbone of a prosperous modern economy, with electricity networks providing a central role in helping to deliver health, education, and other infrastructure services.
In response, Sustainable Development Goal 7 SDG 7 calls for universal access to affordable, reliable, and modern energy services by 2. Achieving this goal is similarly important in achieving other SDGs 3 , 4 , 5 and as such requires a coordinated effort — not only in the targeted expansion of electricity networks, but also towards increases in renewable energy share of the global energy mix and improved network resilience.
To measure the joint global progress on achieving SDG 7, detailed spatial information on the current locations and properties of electricity infrastructure is essential.
Physical assets, such as generation plants, transmission networks typically higher voltage lines used for the bulk movement of electricity , distribution networks typically lower voltage lines for distributing electricity from transmission networks to end consumers , substations for transforming between voltages , and related assets are important to assess the practicalities of connecting isolated communities to wider transmission networks.
This makes it difficult for government bodies and utility companies to plan for extending, strengthening, and modifying their electricity networks.
Even in advanced economies, where this network data may exist, it can be difficult to find it in an open and standardized way.
Projects such as OpenStreetMap have greatly democratized access to geospatial data generally 7 , and while there have been significant efforts on collating data on electricity generation infrastructure 8 , and high voltage transmission lines in OpenGridMap 9 , a globally consistent database of medium and lower voltage distribution networks still lag. In addition to the physical assets, the geospatial locations of electrified settlements within each country is often still unknown.
To fill this gap, several studies have developed methods to map these locations and compute accessibility metrics when combined with high-resolution population maps. However, applications have largely been focused on country or sub-continental scales 10 , 11 , While these methods are showing great promise, a globally consistent approach and map of accessibility is still lacking which can ultimately aid development and planning efforts in attaining SDG 7.
This tool applies multiple filtering algorithms to night-time light imagery to identify locations most likely to be producing light from electricity. These light sources target-locations are then connected to known electricity networks through a least-cost routing algorithm following roads and known distribution lines adopted from OpenStreetMap. The dataset is validated two-fold. Firstly, against 16 electricity networks across 14 countries representing the range of World Bank income groupings: High, Upper-Middle, Lower-Middle, and Low.
Secondly, we evaluate the effectiveness of the dataset in predicting accessibility with respect to income groups, human development index HDI , and investment requirements as commonly reported in the literature and used as key reporting metrics. The small geographic scale and prevalence of buried LV cables push the boundaries of satellite imagery detecting abilities.
To create this dataset we therefore turn to spatial heuristics to create a weighted layer of electricity access rate by applying an algorithm calibrated to national urban and rural accessibility statistics.
Results of this study pave the way for improved efforts in electricity modelling and planning. Although only predictive, this standardized global dataset of transmission, distribution and low-voltage lines will be a valuable starting point for electrification planners and researchers in several fields, such as assessing social inequalities 14 , estimating exposure to natural hazards 15 , and quantifying electricity infrastructure roll-out requirements 12 , An important limitation is that this dataset does not attempt to replicate actual network configurations or precise structures, as would be needed for electrical modelling such as power flow modelling; this was seen as out of scope, and is unlikely to be possible at a global scale with current data sources.
The use of open data now allows anyone to freely use or develop this tool see Code Availability further to model the power network for any location, as we have demonstrated for the entire globe. A schematic overview of the computational process for generating a map of the global power grid is illustrated in Fig. Simplified overview of methodology. Data inputs in green, filtering steps in yellow, intermediate results in red; outputs in blue. To develop the initial high-voltage network layer, there are several existing power network datasets, both open and proprietary.
Chief among these is OpenStreetMap, which is relatively comprehensive for transmission lines. Machine learning methods for feature detection of transmission networks from satellite imagery have made some progress, but still rely heavily on manual tagging In some areas, it is comprehensively traced from satellite imagery and ground truth, while in others particularly developing countries accuracy and completeness is variable, depending on whether someone has taken the effort to map an area.
Although it is often unsuitable for power flow analysis, it is enough for the goal of mapping approximate infrastructure. The goal of this research is to produce a single global dataset that is valuable to researchers and practitioners and easy to replicate, and to serve as a starting point for future improvements and region-specific efforts. Focusing on a specific country or region allows the inclusion of locally valuable data sources e.
In addition, smaller context or additional effort could be used to better understand temporal changes in electricity access and infrastructure. The first step in predicting distribution lines is predicting the points that they connect. The definition of electrification targets as locations likely to be connected to a medium-voltage network informed the approach described below. To filter transient values such as fires and reflections, twelve months of night-time lights imagery from VIIRS 19 are merged, using for each pixel the 70 th percentile value.
This value is chosen based on manual testing over several sample countries covering different development levels and land types; higher values above 80 did not sufficiently filter transient events and extraneous sources such as fires, gas flaring, snow, and lunar reflections , while lower values below 60 erased significant sources of real light-emissions.
In order to highlight locations that are brighter than their immediate surroundings, thus accentuating even dim points that nonetheless stand out above background light, a two-dimensional filter is convolved over the image with. The goal of this filter is to find pixels with a higher value than their neighborhood, biased towards closer areas.
To achieve this bias, a non-linear function is needed, and a cubic function was found to achieve better results than a square function. A threshold is then applied to create a binary raster of electrification targets. A more lenient threshold value provides higher granularity in detecting small or marginal locations, but creates many more false positives. As this analysis focuses on distribution-connected locations, we settled on a value of 0.
A different filtering function would require this threshold to be re-examined, and, as above, a closer inspection of a specific geographical areas might reveal a more appropriate location-specific threshold. The output is a global two-dimensional array of target coordinates that must be connected by network lines.
An example is given in Fig. Night-time lights as a proxy for electricity access in four countries with different income levels. Several studies have validated this approach with ground-truth data 20 , 21 , 22 , but there is yet no standardized methodology. In addition, validations have been done with limited geographical and socio-cultural scope, and with small datasets.
Thus, there is significant uncertainty extrapolating these techniques globally. Given most indoor lights do not show up on night-time light imagery; in effect the images measure output from street lights and other large industrial light source. Since our primary purpose is to identify locations very likely connected to the distribution network and not to assess the connection status of individual buildings or households, these issues are of less concern. Attempting to highlight every area with access at a smaller scale would necessitate a closer examination of this issue and more detailed data.
Previous efforts have used Voronoi diagrams and minimum spanning trees, but these require highly detailed data on substations and demand points 9. This process repeats until all cells have been visited. The algorithm attempts to connect every location with the shortest possible distance, while following roads preferring larger where possible. In other words, it attempts to predict distribution networks based on the assumptions of optimum network topology and network lines tending to follow roads.
Although both assumptions are frequently false, for a variety of historical reasons, our work indicates that they hold true most of the time: while for cost reasons transmission lines are often built in straight lines, distribution lines generally follow roads making them easier to build and maintain. The algorithm relies exclusively on open and easily accessible datasets but could be improved with proprietary data sources and produces medium-voltage network data for any given area and model parameters.
The algorithm is weighted by a cost function based on existing roads from OpenStreetMap. Existing grid lines from OSM are assigned a cost of 0. Areas with no road are assigned a cost of 1. This creates a two-dimensional array representing the cost of traversing each cell between the target points. The result is post-processed by removing lines that replicate existing OSM lines, so that the result only shows additions to the OSM data. The final result is therefore disaggregated between lines directly modelled from OSM data, and new lines discovered by the algorithm.
The output is an estimate for each cell of whether it contains a medium-voltage line. There was a similar process for predicted submarine lines, for example in Indonesia. Moreover, several administrative areas had to be split or removed, such as exclaves, foreign island territories, or spread-out islands. The wider resulting dataset is given in Fig.
Results of the distribution line modelling. This highlights an important issue in energy access — network extension is a necessary but non-sufficient factor of electrification. This is an important topic to examine if the Sustainable Development Goals are to be reached. Regional electrical infrastructure proximity and investment value, by country and income level. The investment values see Fig.
They show a large discrepancy in resources between different regions and income levels. Geographically spread-out and energy-intensive populations such as Sweden have a large amount of infrastructure, while more compact countries can get by on less.
These regional differences are likely even bigger since the gridfinder tool only connects locations with a single line when, there will frequently be multiple lines, either on the same or different towers to provide redundancy in the network. Therefore, these investment values may significantly under predict the divergence, as more developed areas, particularly those at greater risk to natural hazards could be more likely to have such redundancies built in. In countries with universal access to electricity, accessibility maps mirror population distribution.
However, in countries with lower access rates, we turn to night-time lights, urban extents, the output of our gridfinder analysis, and national statistics to attempt to quantify the number of people in each grid cell with meaningful electricity access. The goal of this algorithm is to match national statistics on urban and rural access rates, and to heuristically apply these rates based on population density, overall brightness, and brightness per capita.
This algorithm works by splitting each country into eight groups with the urban-rural split on one dimension and four quantiles of population density on the other. We therefore aim to capture variations in brightness per capita between and within each of these groups.
This process is iterated, modifying the factor connecting brightness and access for each group, until a result is achieved satisfactorily close to national statistics.
This is limited by being based purely on national statistics as opposed to sub-nationally disaggregated data , and subsequent research for specific countries or regions could improve on this. The resulting geospatial database is used to estimate the length of lines in each cell based on population, demand and household size 16 , according to:. This equation is derived from a consideration of the number of lines needed to serve a given capacity, as well as the physical length needed to reach all buildings in an area Parameters were varied by region and calibrate against detailed low-voltage network data.
Maps & Data
The implementation of the network codes, for example, is resulting in closer collaboration between TSOs and other actors of the energy sector. To stay fast and efficient, business models and related IT infrastructures must be adapted across regional dimensions and physical- and market-related issues. It will also extend to IT, data architecture, or data exchange standard settings for the entire industry to which ENTSO-E contributes and aims at addressing the increasingly interlinked IT needs of TSOs and other players, such as regional security coordinators, the Joint Allocation Office, capacity calculation regions or power exchanges. Grid operators use computer models of their network to simulate its behaviour to make decisions. Up to now, each TSO has been using its own grid model. As networks become more interconnected and as European electricity markets are getting increasingly integrated, there is a need to develop a common grid model that will ease the cooperation between TSOs and will result in an even more secure and cost-efficient European grid. The common grid model is a prerequisite for any joint regional security evaluation and capacity calculation among several TSOs and is specified in the CACM Regulation and in the System Operations Guideline.
Predictive mapping of the global power system using open data
ENTSO-E Electricity infrastructure projects map
The synchronous grid of Continental Europe also known as Continental Synchronous Area ; formerly known as the UCTE grid is the largest synchronous electrical grid by connected power in the world. It is interconnected as a single phase-locked 50 Hz mains frequency electricity grid that supplies over million customers in 24 countries , including most of the European Union. In , GW of production capacity was connected to the grid, providing approximately 80 GW of operating reserve margin. Albania is operating the national grid synchronously with the synchronous grid of Continental Europe. In April , the grid of Turkey was synchronized with the European grid.