Bath PhD Studentship Competition: Causal AI for Resilient Low-Carbon Energy Systems

Bath PhD Studentship Competition: Causal AI for Resilient Low-Carbon Energy Systems

Quick Decision

Strong but verify fit. This is a competitive University of Bath Electronic and Electrical Engineering PhD, led by Chenghong Gu. It suits applicants who want to combine power-system understanding with causal inference, uncertainty and scientific computing. The opportunity is research into interpretable resilience models, rather than a finished autonomous grid-control product. The most important early questions concern usable data, credible causal assumptions and a manageable first-year problem.

The advertised deadline is 29 November 2026 without a time. The current Engineering and Design URSA policy independently specifies 23:59 UK/GMT on that date. The expected start is 4 October 2027; an alternative requires agreement with the supervisor and Doctoral College, with funding retention confirmed separately. Overseas tuition is eligible for the competition, so a Chinese applicant has a positive route in principle. Neither an individual award nor its 2027 maintenance rate and funded months is confirmed. Non-UK applicants who are not currently studying in the UK must satisfy English before the application deadline, with no exceptions.

Research Problem and Doctoral Work

Connecting Evidence to Causal Explanations

The proposed work has a clear sequence. A knowledge graph would connect heterogeneous operational, environmental and cyber-physical information. Structural causal models would incorporate physical laws and engineering constraints to investigate cascading failures and counterfactual scenarios. AI agents would then use the proposed causal digital twin to explore adaptive resilience strategies. These are advertised research objectives. A completed graph, identifiable causal model, validated twin or operational agent is not supplied with the position.

The doctoral difficulty lies in the transition between those stages. A knowledge graph can record relationships and provenance without establishing which relationships are causal. A predictive model can perform well on familiar observations while giving an incorrect answer to a counterfactual question. Engineering constraints can rule out physically implausible outputs without, by themselves, resolving hidden confounding, incomplete measurements or a changing system. Those distinctions are central to judging this project.

A sensible first discussion should identify the initial decision problem, available evidence and boundary of the model. Ask what data the group can actually provide, which causal assumptions are defensible, how competing explanations will be compared and what independent evidence can test an inferred mechanism. Historical event analysis and synthetic scenarios can be informative, but agreement with a known event is weaker evidence if that event also shaped the model. The advertisement does not name an allocated dataset, approved data-sharing agreement, computing quota or operational partner for this student's thesis.

What the Accident Evidence Actually Supports

The advertisement uses major outages to motivate the topic. Official findings support a more careful causal account than its short summaries. Ofgem's final investigation of the 2019 Great Britain outage describes a lightning event followed by nearly simultaneous responses at Hornsea and Little Barford, together with embedded-generation losses. Calling the initiating events simply unrelated obscures their common trigger and subsequent interactions. The evidence here covers the official executive summary and conclusions; it does not include an independent reconstruction of the accident.

For North Hyde and Heathrow in March 2025, NESO's final findings identify a catastrophic transformer-bushing failure, most likely associated with moisture. They also discuss an earlier elevated moisture reading without appropriate mitigating action, the airport's internal distribution dependencies and cross-sector preparedness. A transformer fire alone is therefore an incomplete account of the disruption. The evidence covers the official findings release; the complete linked investigation is outside that evidence range.

The March 2026 ENTSO-E final findings on the April 2025 Iberian blackout explicitly describe interacting factors: oscillations, voltage/reactive-power control gaps, differing regulation practices, output reductions, generator disconnections and uneven stabilisation. They do not support a unique, settled explanation that solar-generation voltage instability alone caused the blackout. These examples establish the value of studying interdependencies and evidence quality. They do not establish that this proposed AI framework would have prevented any of the events.

Value and Limits of the Research Direction

The useful contribution could be a transparent account of which causal conclusions follow from particular evidence and which remain uncertain. A counterfactual resilience claim should be compared with credible engineering and statistical alternatives, not only a weak predictive baseline. Ask how uncertainty, missing observations and distribution changes will be reported, and how the project will distinguish a plausible explanation from a tested causal mechanism.

Agents can initially be evaluated as research tools in bounded models. The announcement does not grant live-grid access, authority to change operating decisions or a pathway to certify a safety-critical controller. Such permissions should not be assumed from the language of autonomy. Equally, a narrow, well-tested offline contribution can make a strong thesis without completing every part of the advertised vision. The balance between causal methodology and energy-engineering applications needs to be agreed with Gu.

The transferable training could include critical modelling, uncertainty analysis, research software and communication across engineering and data science. Network planning, energy analytics and academic research are plausible directions, not verified placement outcomes. KERAUnIC shows a genuine connection to an industry problem; it does not establish a job offer, internship allocation, salary premium or completion record for this student.

School, Programme and Field Context

The degree belongs to the University of Bath, in the Faculty of Engineering and Design and the Department of Electronic and Electrical Engineering. It is not a Mechanical Engineering registration or a joint degree with the group's collaborators. The current full-time EEE programme describes a usual duration of four years, with a minimum of two and a normal maximum of four. Probationary progress includes written and oral confirmation; the doctorate culminates in a thesis and viva. Academic duration must be separated from scholarship duration.

Bath's institutional research environment is credible, but broad rankings are secondary to the proposed supervision and data. QS 2027 places Bath 125th overall, compared with 132nd in the 2026 table. This is an institutional ranking, not a ranking of causal energy AI or Gu's group. Bath's stated REF 2021 proportion of world-leading/internationally excellent research is also institution-wide. Neither statistic establishes individual student outcomes or the availability of a particular resource.

Gu's profile connects him to Smart Energy Systems and wider Bath centres. The EEE programme says a student joins an appropriate centre as agreed on acceptance; every affiliation is not an automatic entitlement. University doctoral development offers research and transferable-skills support. Ask which technical training, causal-inference expertise, power-system modelling support and research-computing access this student would actually receive. A published collaborator or centre membership should not be silently turned into a named co-supervisor.

Supervisor and Lab Assessment

Current Role and Research Setting

The current institutional research profile lists Chenghong Gu as Professor in Smart Energy Systems, accepting doctoral students, with ORCID 0000-0002-3306-767X and C.Gu@bath.ac.uk. The project gives cg277@bath.ac.uk for enquiries; retain that advertised contact and use the current profile to cross-check it. An older departmental list still showing Reader is weaker evidence for his present title. His interests include multi-vector energy systems, robust optimisation, machine learning, extreme-weather resilience and energy-market design.

The ENA record for KERAUnIC provides concrete industry context: Bath received a lightning-related fault use case and combined it with environmental datasets. ENA marks its network project complete, March 2024–September 2025, while the university profile lists a differently dated academic project record. These records need not represent identical contract scopes. Neither establishes 2027 funding or permission for the new student to reuse restricted data. Ask about daily supervision, formal co-supervision, data/IP conditions and the proportion of time available for methodological development.

Recent Original Research and Reading Boundaries

Xiang et al.: hierarchical planning under uncertainty

Xiang and colleagues, Communications Engineering, 2026, with Gu as co-author, combine hierarchical planning and learning under uncertainty. The original institutional abstract reports a Stackelberg structure, upper-level Monte Carlo tree search and lower-level multi-agent reinforcement learning, evaluated on benchmark and practical distribution cases. It supports relevant planning/learning experience, not causal identification. The evidence is limited to the complete original abstract and publication/data metadata; usable publisher main text was unavailable. Detailed methods and independent reproduction are unverified.

Gao et al.: digital-twin aggregation and distributed control

Gao and colleagues, IEEE Transactions on Smart Grid, 2026, first online in November 2025, connect digital-twin aggregation with distributed model predictive control. The evidence includes the accepted primary paper's architecture, case-study and conclusion sections. Its evaluation uses a modified IEEE 33-node system with some industrial-park parameters. Reported cost/emissions improvements remain model comparisons, not field deployment or proof of causal inference. Table III and the accompanying percentage prose are not fully consistent, so this report does not adopt a precise effect size without checking the underlying calculations.

Liu et al.: coupled transport and electricity resilience

Liu and colleagues, IEEE Transactions on Industry Applications, 2024, first online in April 2023, examine coupled road, electricity and EV resilience under tropical-cyclone uncertainty. The complete original abstract describes multitask forecasting and distributionally robust optimisation on modified 14-node power/transport examples. It is relevant to interdependent risk modelling, but does not demonstrate this project's knowledge graph or counterfactual system. Full methods, figures and discussion remain outside the evidence range. These are three published originals with different evidence depths, not three full-paper replications.

Funding, Application and Living Cost

Competitive Funding and the Completion Gap

The studentship competition can cover Home or Overseas tuition, maintenance and up to £1,000 per year for training support. Training money is not additional personal living income. The advertised £21,805 refers to 2026/27, not a confirmed 2027/28 stipend. International awards are capped, but a numerical proportion was not published in the evidence reviewed. No 30% cap from another scheme should be imported.

The current URSA Engineering policy gives up to 3.5 years as its general support limit. That is not a personal promise of 42 funded months, and a usual four-year PhD does not become four funded years. Obtain the individual award's duration, maintenance rate, fee-indexation coverage and arrangements for any completion period. A gap could arise even when the initial offer covers international fees. The offer should also clarify whether an alternative start affects eligibility or award availability.

The official 2027/28 Engineering Overseas PhD fee is £31,350 per full-time year, with continuing fee increases of up to 8% described by the university. This is the exposure if support does not cover a period; it should not be added as a second tuition bill during an award that fully covers the applicable fee. Home fees were still awaiting the relevant rate. Extra project-specific charges were not established and should be checked, rather than assumed absent or automatically charged.

Budgeting Without a False Surplus

Bath's 2025/26 budget is a historical planning reference. It gives £405 per week/£21,060 per year, but includes a £37 utilities line whose application to bills-included halls is ambiguous. Removing that line produces an analyst's £368 weekly comparison, not another official quote. Average listed halls rent is £230 weekly with a 51-week contract; the off-campus example is £750 monthly before bills. Neither is a guaranteed 2027 room or price.

The scholarship does not cover international visa and immigration health surcharge costs. The current references reviewed give a £558 Student visa application fee and a student IHS rate of £776 yearly; actual totals depend on permission duration and future rules. Deposits, relocation and cash needed before stipend payments also matter. Because 2027 maintenance, personal funded months and housing are unknown, a fixed monthly surplus cannot responsibly be calculated. University housing is not generally guaranteed to every doctoral student.

Submit Through the Correct Route

Apply online for Electronic and Electrical Engineering PhD, selecting University of Bath URSA in the funding section. Quote this exact project title and Gu's name in the relevant first-choice fields. At most two projects may be selected; a second choice requires its own project-title and intended-supervisor fields. Incorrect entries can prevent consideration. Use the current form's fields rather than assuming an informal email is a formal application.

The general programme requires two references, at least one from the most recent academic setting, and a personal statement explaining motivation, relevant experience and career intentions. Naming one or two intended supervisors matters. A proposal of around 1,000 words is specified for an own-idea application; it should not automatically be imposed on this advertised project. General advice to apply three months before entry from overseas does not extend the November funding deadline. Confirm any project-specific additional documents with admissions.

The University encourages applications from under-represented groups. Applicants may explain circumstances that affected their educational attainment on the application form, preferably in a short paragraph at the end of the personal statement. If applicants disclose a disability in the application, Bath's Disability Service will contact them to discuss support needs.

Applicant Fit and Preparation

Academic Fit and a Chinese Applicant's Route

The project requires an obtained or imminent UK first/2:1 honours degree or international equivalent; a master's is advantageous. The general EEE programme also describes a relevant Merit-level master's route, but the project-specific assessment still applies. Power engineering, applied mathematics, computing or adjacent quantitative experience may be useful depending on preparation. A particular programming language, existing knowledge-graph platform or prior operational-grid role is not stated as compulsory.

China is not excluded by the opportunity. The positive basis is the EEE programme's equivalent-qualification route, Overseas tuition eligibility and the international student immigration route. Bath's China research-degree wording found in official indexing refers to a first-class bachelor's with research experience or desirable research components at master's level, while the directly available country page did not reproduce a definitive current doctoral threshold. Do not apply the taught-master's percentage range to a PhD. Send the actual institution, degree, transcript and research background for an individual equivalence decision; a master's is not an automatic override.

English, ATAS and Timing

EEE requires IELTS 6.5 overall with every component at least 6.0. Its TOEFL thresholds distinguish test dates: before 21 January 2026, 90 overall/21 each; on or after that date, 4.5 overall/4 each on the new scale. Home Edition is accepted, with scores normally within 24 months of entry. The programme also lists Duolingo 120 overall, speaking/writing/reading 110 and listening 105. These date-specific rules should not be mixed across scales.

A qualifying English-medium degree may support an exemption, generally within two years or under a relevant residence exception, but admissions must confirm it. A Chinese medium-of-instruction letter does not automatically waive the requirement. For a non-UK applicant not currently studying in the UK, the project's explicit English-by-deadline rule has no exceptions. General language-course or conditional-offer possibilities should not be used to promise later compliance.

The EEE programme is ATAS-covered. China is not nationality-exempt under the reviewed rules, so the actual subject classification and offer/CAS instructions need checking. This is an individual clearance process, not a categorical nationality ban or a guarantee of approval. The Student visa route requires the relevant offer, CAS, English and financial evidence. Allow time for those steps before the expected October start; future fees and processing periods can change.

Next Actions and Questions to Verify

Prepare a concise account of one quantitative project: what question it answered, what assumptions it required, how evidence tested it and what remained uncertain. This is more useful than claiming every advertised AI technique. Ask Gu which first-year problem is realistic, whether the thesis primarily develops causal methodology or energy applications, and what mentoring covers the weaker side of that combination.

Before relying on an offer, obtain written clarification of dataset availability and permissions, independent validation, supervision arrangements, computing/software provision and publication/IP constraints. Also confirm the individual 2027 award rate and term, international fee increases and completion costs, English satisfaction or approved exemption before the hard deadline, and ATAS/CAS timing. The opportunity is worth an informed application; these unresolved personal conditions should be settled before presenting it as an unconditional full-duration financial guarantee.

Evidence and Sources

This assessment draws on information available in October 2026. Current vacancy availability and individual award terms have not been reconfirmed; each paper and incident account retains its stated evidence limits.

Programme, Supervisor and Scientific Evidence

• Current supervisor profile
• EEE PhD programme
• KERAUnIC industry outcomes
• 2026 hierarchical-planning original
• 2026 digital-twin control original
• 2024 transport-energy resilience original
• Ofgem 2019 final investigation
• NESO North Hyde final findings
• ENTSO-E Iberian final findings

Funding, Fees and Research Environment

• Current URSA competition
• Engineering 2027/28 tuition
• Historical 2025/26 budget
• Doctoral admissions FAQ
• Doctoral professional development
• QS2027 Bath ranking
• Bath postgraduate research
• UKRI current stipend reference

Qualifications, English and Immigration

• Doctoral English requirements
• International qualification guidance
• Student visa
• Immigration health surcharge
• ATAS rules