Software development services

Tender description

Challenge — develop a system that forecasts the most economic cost of power system outages taking account multiple inputs varying from power flow analysis, economic generation despatch, system faults and maintenance, weather and system operation criteria (defined in the NETS SQSS). Requirements — phased development and delivery of a machine learning model that will output a risk type view (e.g. Monte Carlo distribution) of forecast outage cost against differing scenarios. The cost output will be for an individual project/circuit intervention (given a background of scenarios or outage configurations defined by the user). This PIN is an expression of interest at this time. Providers maybe invited to tender at the sole discretion of National Grid. Any subsequent tender event will not be subject to the Utilities Contract Regulations 2016 in accordance with Regulation (16). Develop a system that forecasts the most economic cost of power system outages taking account multiple inputs varying from power flow analysis, economic generation despatch, system faults and maintenance, weather and system operation criteria (defined in the NETS SQSS). (a) Note 1: a single outage in isolation could be economically placed however when considered against key influencing factors (e.g. multiple outages across England and Wales, generator constraints, system faults/maintenance, power flows, fault levels, voltage, system stability, etc.) the outage duration may then raise significant costs. (b) Note 2: recording the counterfactual cost options is also necessary to demonstrate the effect of not taking the outage ... this recognises that constraint costs aren't automatic and are influenced on a second by second basis by e.g. parties on a control desk undertaking discrete actions. 2) Benefit – the ability to forecast costs enables ESO and transmission owners to understand the cost of works enabling parties to prioritise works economically ultimately reducing system management costs passed to consumers. 3) Current process — exceptional cost forecasting currently occurs with variable accuracy however, there are existing systems that if they could be combined could drive towards the desired outputs. Example power simulation software (Power Factory) + BID3 (generation despatch scenarios) + Economic despatch builder + NETS SQSS + system info. Note: we do not expect to be able to fully model the outage costs as the electricity system operator retains some knowledge of how outage costs have been generated. 4) Requirements — phased development and delivery of a machine learning model that will output a risk type view (e.g. Monte Carlo distribution) of forecast outage cost against differing scenarios. The cost output will be for an individual project/circuit intervention (given a background of scenarios or outage configurations defined by the user.). 5) Key interfaces: (a) National Grid Portfolio Power System and Plan Optimisation Team, (b) National Grid ESO (TBA), (c) National Grid Capital Delivery (Construction), (d) External – Elexon data team (TBA). 6) Example of key data sources (to ‘mine’ for trends/insights): (a) Elexon data [API downloads] https://www.nationalgrideso.com/balancing-data/system-constraints (b) maintenance data – typical works/durations/failure modes, (c) NG network data of the power system. Phase 1* — Problem definition — scoping/discovery phase. Phase 1a – Collect and analyse — data gathering from key data sources incl. requests for future NG data formatting e.g. defining future Outage datasets)/identify any other data sources to support the prototype development. Phase 2* — Prepare data — definition direction of travel. Outline next steps including timescales and approach to integrating (or not) the system with NG’s working practices. Phase 2a — Evaluate algorithms — prototype system development. Phase 3 — Test and improve results — Historic information capture (Elexon) for model validation. Phase 4 – Live comparison — future system outage cost forecasting.